From 4d018a14c3aa398d11db5e88cad75964ab5ae421 Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 03:01:43 +0800 Subject: [PATCH 01/20] Add provenance-aware paper metadata storage and PMLR collection --- .gitea/workflows/check.yml | 35 ++ README.md | 22 +- collect_pmlr.py | 123 +++++++ import_arxiv.py | 78 +++++ import_records.py | 69 ++++ index_core.py | 120 +++++++ query.py | 52 +++ requirements.txt | 1 + schema.sql | 92 +++++ test_import_arxiv.py | 57 ++++ test_index_core.py | 56 ++++ venues.json | 664 +++++++++++++++++++++++++++++++++++++ 12 files changed, 1365 insertions(+), 4 deletions(-) create mode 100644 .gitea/workflows/check.yml create mode 100644 collect_pmlr.py create mode 100644 import_arxiv.py create mode 100644 import_records.py create mode 100644 index_core.py create mode 100644 query.py create mode 100644 requirements.txt create mode 100644 schema.sql create mode 100644 test_import_arxiv.py create mode 100644 test_index_core.py create mode 100644 venues.json diff --git a/.gitea/workflows/check.yml b/.gitea/workflows/check.yml new file mode 100644 index 0000000..5fbeb95 --- /dev/null +++ b/.gitea/workflows/check.yml @@ -0,0 +1,35 @@ +name: Metadata integrity checks +on: + push: + branches: [main, 'codex/**'] +jobs: + metadata: + runs-on: ces-control + env: + ARCHIVE_SHA: ${{ github.sha }} + PYTHONDONTWRITEBYTECODE: '1' + steps: + - name: Verify metadata storage and query commands + run: | + python3 -B - <<'PY' + import io, os, pathlib, re, subprocess, tarfile, tempfile, urllib.request + sha = os.environ['ARCHIVE_SHA'] + if not re.fullmatch(r'[0-9a-fA-F]{40,64}', sha): + raise SystemExit('Invalid source commit') + # Public repository archive: no owner credentials or runner SSH keys. + url = 'https://gitea.entroia.com/Zayden-Jung/paper-atlas/archive/' + sha + '.tar.gz' + with urllib.request.urlopen(url, timeout=60) as response: + archive = response.read(16*1024*1024+1) + if len(archive)>16*1024*1024: + raise SystemExit('Unexpectedly large source archive') + with tempfile.TemporaryDirectory(prefix='paper-atlas-ci-') as tmp: + root = pathlib.Path(tmp) + with tarfile.open(fileobj=io.BytesIO(archive), mode='r:gz') as tar: + tar.extractall(root, filter='data') + candidates = list(root.glob('*/test_import_arxiv.py')) + if len(candidates)!=1: + raise SystemExit('Unexpected repository archive layout') + src = candidates[0].parent + subprocess.run(['python3','-B','-m','unittest','discover','-v'],cwd=src,check=True) + subprocess.run(['python3','-B','query.py','--help'],cwd=src,check=True) + PY diff --git a/README.md b/README.md index 17edc73..162d6b2 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 目标包括：完整历史元数据导入、逐年会议目录、跨平台论文去重、公开拒稿/撤稿信息、平台标签与自定义标签隔离、可恢复更新和覆盖率报告。 -**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、DOI/标识合并和只读查询脚本。服务器上进行过合成样本验证；全量快照下载与历史导入尚未完成。会议适配器、跨平台模糊匹配复核流程、完整覆盖率核验与 CI 仍待实现。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 +**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR 官方元数据采集、带清单计数核验的会议导入和只读查询脚本。服务器上六项合成回归检查通过；完整 ICML 2025 元数据已采集 3,330 条，尚未作为全量交付。arXiv 全量快照下载与历史导入尚未完成。其他会议适配器、跨平台疑似重复复核流程、完整覆盖率核验仍待实现。已配置 ces-control:host 对应的远程 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 ## 数据范围与来源 @@ -41,16 +41,16 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 | `edition_sources` | 每届每个来源的覆盖状态、预计/已导入数量、错误与断点 | | `submissions` / `decisions` | 多次投稿及带证据的决定历史，支持改投、撤稿和决定变更 | | `tags` / `paper_tags` | 平台标签与人工标签，使用命名空间隔离 | -| `match_candidates` | 证据不充分的潜在重复记录及人工复核状态 | +| `match_candidates` / `paper_merges` | 潜在重复复核状态、确定合并审计 | | `ingest_runs` | 导入批次、状态、错误和恢复位置 | 决定状态为 `accepted`、`rejected`、`desk_rejected`、`withdrawn`、`under_review`、`unknown`。一篇论文可以对应多次投稿和多个决定，不能用单一“最终会议”字段覆盖历史。查询中按观察时间选择最近状态；后续分析需要同时检查来源覆盖率。 -当前确定性去重使用 arXiv ID 和单个 DOI。保留各平台源记录，避免合并后丢失出处。规范化标题只建检索索引，尚未实现模糊匹配流程；**现阶段不保证跨平台去重已经完成**。不同平台标签不自动等同，自定义标签应使用 `custom:*` 命名空间。 +当前确定性去重使用 arXiv ID、单个 DOI 和 OpenReview forum ID；跨平台同标题还需至少两位作者吻合，或作者吻合且摘要相同，才自动合并。保留各平台源记录，避免合并后丢失出处。规范化标题只建检索索引，尚未实现模糊匹配流程；**现阶段不保证跨平台去重已经完成**。不同平台标签不自动等同，自定义标签应使用 `custom:*` 命名空间。 ## 运行与查询 -依赖 Python 3 标准库和带 FTS5 的 SQLite，目前无第三方 Python 依赖，无项目构建步骤。服务器已检查为 Python 3.14.4。所有运行数据放在仓库外，不提交数据库、下载包或访问凭据。 +基础导入和查询依赖 Python 3 标准库和带 FTS5 的 SQLite；PMLR 采集另需 `requirements.txt` 中锁定的 PyYAML 6.0.3。无项目构建步骤。服务器已检查为 Python 3.14.4。所有运行数据放在仓库外，不提交数据库、下载包或访问凭据。 服务器暂存目录：`/data/paper-index/`，快照目录：`/data/paper-index/raw/`，运行代码目录：`/data/paper-index/code/`。这些路径不是对外访问的 API。 @@ -82,6 +82,20 @@ python3 -B query.py --db /data/paper-index/papers.sqlite --venue COLM --year 202 查询连接使用 SQLite 只读模式，结果输出 JSON。当前没有公共 Web 页面或网络查询服务。 +## 官方会议元数据采集 + +```sh +# 单卷示例：ICML 2025，读取官方 gh-pages 元数据归档，不下载 PDF +python3 -B collect_pmlr.py --cache /data/paper-index/raw/http-cache \ + --out /data/paper-index/collected --volume 267 + +# 待主库完成 arXiv 导入、且没有其他写入进程后，合并会议记录 +python3 -B import_records.py --db /data/paper-index/papers.sqlite \ + /data/paper-index/collected/pmlr-v267.jsonl +``` + +采集输出 JSONL 和对应 manifest，只有显式官方录用记录才产生 accepted 决定。清单数目不一致时整个文件导入回滚，记录失败批次，不能标为覆盖完成。多个历史卷的全量覆盖仍需单独核验。 + ## 长期维护与更新 1. 每次采集先核对来源、许可、可访问性、空闲空间和现有运行批次。保持低并发，遵守来源限流、重试与人机验证要求。 diff --git a/collect_pmlr.py b/collect_pmlr.py new file mode 100644 index 0000000..1e9da16 --- /dev/null +++ b/collect_pmlr.py @@ -0,0 +1,123 @@ +"""Collect official PMLR metadata, never the PDFs linked by the metadata.""" +import argparse +import hashlib +import io +import json +import re +import tarfile +import time +import urllib.error +import urllib.request +from pathlib import Path + +import yaml + + +def fetch(url,cache,limit=64*1024**2): + cache = Path(cache) + cache.mkdir(parents=True,exist_ok=True) + path = cache/hashlib.sha256(url.encode()).hexdigest() + if path.exists(): + return path.read_bytes() + for attempt in range(3): + try: + request = urllib.request.Request(url,headers={'User-Agent':'PaperAtlas/0.1 (public scholarly metadata research)'}) + with urllib.request.urlopen(request,timeout=90) as response: + if response.status != 200: + raise RuntimeError('Expected complete HTTP 200 response') + data = response.read(limit+1) + if len(data)>limit: + raise RuntimeError('Metadata response exceeds byte limit') + part = path.with_suffix('.part') + part.write_bytes(data) + part.replace(path) + path.with_suffix('.json').write_text(json.dumps({'url':url,'sha256':hashlib.sha256(data).hexdigest(),'bytes':len(data),'retrieved_at':time.strftime('%Y-%m-%dT%H:%M:%SZ',time.gmtime())})) + time.sleep(0.5) + return data + except urllib.error.HTTPError as exc: + if exc.code not in (429,500,502,503,504) or attempt==2: + raise + time.sleep(min(30,max(2**attempt,int(exc.headers.get('Retry-After','1'))))) + + +def collect_volume(volume,cache,out,allowed=None): + config_url = f'https://raw.githubusercontent.com/mlresearch/v{volume}/gh-pages/_config.yml' + config = yaml.safe_load(fetch(config_url,cache)) + venue = str(config.get('shortname','')).upper() + if allowed and venue not in allowed: + return None + year = int(config['year']) + title = str(config.get('booktitle','')) + track = 'workshop' if 'workshop' in title.lower() else 'main' + archive = fetch(f'https://codeload.github.com/mlresearch/v{volume}/tar.gz/refs/heads/gh-pages',cache) + path = Path(out)/f'pmlr-v{volume}.jsonl' + path.parent.mkdir(parents=True,exist_ok=True) + part = path.with_suffix('.part') + count = 0 + with tarfile.open(fileobj=io.BytesIO(archive),mode='r:gz') as tar,part.open('w') as output: + for member in tar: + if not member.isfile() or '/_posts/' not in member.name or not member.name.endswith(('.md','.markdown','.html')): + continue + if member.size>8*1024**2: + raise RuntimeError('Unexpectedly large metadata member') + content = tar.extractfile(member).read().decode('utf-8-sig') + parts = content.split('---',2) + if len(parts)<3 or parts[0].strip(): + raise RuntimeError('Missing YAML frontmatter: '+member.name) + row = yaml.safe_load(parts[1]) + if row.get('layout') not in ('inproceedings','article') or not row.get('title'): + continue + external = f"v{volume}/{row.get('id') or Path(member.name).stem}" + ids = {} + for key in ('doi','openreview','arxiv'): + value = row.get(key) + if value: + value = str(value) + if key=='arxiv': + value = re.sub(r'^https?://arxiv.org/(abs|pdf)/','',value).removesuffix('.pdf') + if key=='openreview': + value = value.split('id=')[-1].split('&')[0] + ids[key] = value + record = {'source':'pmlr','external_id':external,'url':f"https://proceedings.mlr.press/v{volume}/{row.get('id')}.html",'title':row['title'],'abstract':row.get('abstract'),'authors':row.get('author',[]),'identifiers':ids,'venue':venue,'year':year,'track':track,'status':'accepted','list_url':f'https://proceedings.mlr.press/v{volume}/','metadata':{key:row[key] for key in ('id','volume','date','page','doi','openreview','arxiv','publisher') if key in row}} + output.write(json.dumps(record,ensure_ascii=False,default=str)+'\n') + count += 1 + if count==0: + part.unlink(missing_ok=True) + raise RuntimeError('No papers parsed; refusing an empty successful collection') + part.replace(path) + result = {'source':'pmlr','venue':venue,'year':year,'track':track,'volume':volume,'count':count,'list_url':f'https://proceedings.mlr.press/v{volume}/','file':str(path),'status':'collected'} + path.with_suffix('.manifest.json').write_text(json.dumps(result,indent=2)) + print(json.dumps(result),flush=True) + return result + + +def main(): + parser = argparse.ArgumentParser() + parser.add_argument('--cache',required=True) + parser.add_argument('--out',required=True) + parser.add_argument('--volume',type=int,action='append') + parser.add_argument('--registry',default=str(Path(__file__).with_name('venues.json'))) + args = parser.parse_args() + allowed = {v['acronym'].upper() for v in json.loads(Path(args.registry).read_text())['venues']} + volumes = args.volume + if not volumes: + index = fetch('https://proceedings.mlr.press/',args.cache).decode() + volumes = sorted(set(int(v) for v in re.findall(r'href=["\'](?:https?://proceedings\.mlr\.press)?/v(\d+)/?["\']',index))) + if not volumes: + raise RuntimeError('PMLR index yielded no volumes') + errors = [] + for volume in volumes: + try: + collect_volume(volume,args.cache,args.out,None if args.volume else allowed) + except Exception as exc: + error = {'volume':volume,'status':'blocked','error':str(exc)} + errors.append(error) + print(json.dumps(error),flush=True) + Path(args.out).mkdir(parents=True,exist_ok=True) + (Path(args.out)/'pmlr-errors.json').write_text(json.dumps(errors,indent=2)) + if errors: + raise SystemExit(1) + + +if __name__=='__main__': + main() diff --git a/import_arxiv.py b/import_arxiv.py new file mode 100644 index 0000000..8e71dfd --- /dev/null +++ b/import_arxiv.py @@ -0,0 +1,78 @@ +"""Stream the official metadata ZIP into SQLite. No PDFs, source files or builds.""" +import argparse +import json +import time +import shutil +import zipfile +from pathlib import Path + + +from index_core import Store + + +def ingest(db, archive): + store = Store(db) + conn = store.conn + conn.execute("INSERT OR IGNORE INTO sources(name,homepage) VALUES('arxiv','https://arxiv.org')") + source = conn.execute("SELECT id FROM sources WHERE name='arxiv'").fetchone()[0] + previous = conn.execute('SELECT id,checkpoint_json FROM ingest_runs WHERE source_id=? AND status IN (\'running\',\'failed\') ORDER BY id DESC LIMIT 1', (source,)).fetchone() + fingerprint = {'path':str(Path(archive).resolve()), 'bytes':Path(archive).stat().st_size, 'mtime_ns':Path(archive).stat().st_mtime_ns} + start = 0 + if previous: + checkpoint = json.loads(previous[1] or '{}') + if checkpoint.get('archive') == fingerprint: + run = previous[0] + start = checkpoint.get('line', 0) + conn.execute("UPDATE ingest_runs SET status='running',error=NULL WHERE id=?", (run,)) + else: + raise RuntimeError('Pending run belongs to a different archive; resolve it explicitly') + else: + run = conn.execute('INSERT INTO ingest_runs(source_id,checkpoint_json) VALUES(?,?)', (source,json.dumps({'archive':fingerprint,'line':0}))).lastrowid + conn.commit() + line_number = start + last = time.monotonic() + try: + with zipfile.ZipFile(archive) as zipped: + # Reading the member to EOF also checks its CRC. + with zipped.open('arxiv-metadata-oai-snapshot.json') as stream: + for line_number, line in enumerate(stream, 1): + if line_number <= start: + continue + row = json.loads(line) + external_id = row['id'] + title = ' '.join(row['title'].split()) + abstract = row.get('abstract', '').strip() + authors = row.get('authors_parsed') or row.get('authors') or [] + metadata = {k:v for k,v in row.items() if k not in ('title','abstract','authors','authors_parsed')} + paper,record = store.record('arxiv',external_id,'https://arxiv.org/abs/'+external_id,title,abstract,authors,{'arxiv':external_id,'doi':row.get('doi')},metadata,(row.get('versions') or [{}])[0].get('created')) + conn.execute("DELETE FROM paper_tags WHERE source_record_id=? AND evidence_key='arxiv:categories'", (record,)) + for category in (row.get('categories') or '').split(): + conn.execute("INSERT OR IGNORE INTO tags(namespace,value) VALUES('arxiv:category',?)",(category,)) + tag = conn.execute("SELECT id FROM tags WHERE namespace='arxiv:category' AND value=?",(category,)).fetchone()[0] + conn.execute('INSERT OR IGNORE INTO paper_tags VALUES(?,?,?,?)',(paper,tag,record,'arxiv:categories')) + if line_number % 500 == 0: + if shutil.disk_usage(Path(db).parent).free < 4*1024**3: + raise RuntimeError('Stopped before exhausting disk: less than 4 GiB free') + conn.execute('UPDATE ingest_runs SET records_seen=?,records_written=?,checkpoint_json=? WHERE id=?',(line_number,line_number,json.dumps({'archive':fingerprint,'line':line_number}),run)) + conn.commit() + if time.monotonic()-last >= 20: + print(json.dumps({'run':run,'rows':line_number}),flush=True) + last = time.monotonic() + conn.execute("UPDATE ingest_runs SET status='complete',finished_at=CURRENT_TIMESTAMP,records_seen=?,records_written=?,checkpoint_json=? WHERE id=?",(line_number,line_number,json.dumps({'archive':fingerprint,'line':line_number}),run)) + conn.commit() + print(json.dumps({'status':'complete','rows':line_number}),flush=True) + except BaseException as exc: + conn.rollback() + conn.execute("UPDATE ingest_runs SET status='failed',error=? WHERE id=?",(str(exc),run)) + conn.commit() + raise + finally: + conn.close() + + +if __name__ == '__main__': + parser = argparse.ArgumentParser() + parser.add_argument('--db', required=True) + parser.add_argument('--archive', required=True) + args = parser.parse_args() + ingest(args.db,args.archive) diff --git a/import_records.py b/import_records.py new file mode 100644 index 0000000..ea66aee --- /dev/null +++ b/import_records.py @@ -0,0 +1,69 @@ +"""Import normalized conference JSONL files with coverage evidence.""" +import argparse +import json +import time +from pathlib import Path + +from index_core import Store + + +def ingest_file(db,path): + path = Path(path) + manifest_path = path.with_suffix('.manifest.json') + manifest = json.loads(manifest_path.read_text()) if manifest_path.exists() else {} + store = Store(db) + c = store.conn + source_name = manifest.get('source') + if not source_name: + raise ValueError('A source manifest is required') + source_id = store.source(source_name,manifest['list_url']) + run = c.execute('INSERT INTO ingest_runs(source_id,checkpoint_json) VALUES(?,?)',(source_id,json.dumps({'file':str(path)}))).lastrowid + c.commit() + count = 0 + seen = set() + editions = set() + try: + with path.open() as stream: + for line in stream: + row = json.loads(line) + if row['source'] != source_name: + raise ValueError('Source differs from manifest') + external = row['external_id'] + if external in seen: + raise ValueError('Duplicate external ID in one collected file: '+external) + seen.add(external) + paper,record = store.record(source_name,external,row['url'],row['title'],row.get('abstract'),row.get('authors',[]),row.get('identifiers'),row.get('metadata')) + if row.get('status') == 'accepted': + edition = store.accepted(paper,record,row['venue'],row['year'],row['track'],external,row['url']) + editions.add(edition) + else: + raise ValueError('This importer requires an explicit supported decision status') + count += 1 + if not count: + raise ValueError('Empty metadata file') + if len(editions) != 1: + raise ValueError('A manifest must describe exactly one conference edition and track') + edition = next(iter(editions)) + expected = manifest.get('count') + if expected is not None and expected != count: + raise ValueError(f'Manifest count {expected} differs from records {count}') + c.execute('INSERT INTO edition_sources(edition_id,source_id,list_url,status,expected_count,imported_count,last_success_at) VALUES(?,?,?,?,?,?,?) ON CONFLICT(edition_id,source_id,list_url) DO UPDATE SET status=excluded.status,expected_count=excluded.expected_count,imported_count=excluded.imported_count,last_success_at=excluded.last_success_at,last_error=NULL',(edition,source_id,manifest['list_url'],'complete' if expected==count else 'partial',expected,count,time.strftime('%Y-%m-%dT%H:%M:%SZ',time.gmtime()))) + c.execute("UPDATE ingest_runs SET status='complete',finished_at=CURRENT_TIMESTAMP,records_seen=?,records_written=? WHERE id=?",(count,count,run)) + c.commit() + print(json.dumps({'file':str(path),'status':'imported','records':count}),flush=True) + except BaseException as exc: + c.rollback() + c.execute("UPDATE ingest_runs SET status='failed',finished_at=CURRENT_TIMESTAMP,error=? WHERE id=?",(str(exc),run)) + c.commit() + raise + finally: + c.close() + + +if __name__=='__main__': + parser = argparse.ArgumentParser() + parser.add_argument('--db',required=True) + parser.add_argument('files',nargs='+') + args = parser.parse_args() + for filename in args.files: + ingest_file(args.db,filename) diff --git a/index_core.py b/index_core.py new file mode 100644 index 0000000..2bb2898 --- /dev/null +++ b/index_core.py @@ -0,0 +1,120 @@ +"""Shared identity resolution and provenance-preserving metadata storage.""" +import json +import re +import sqlite3 +import unicodedata +from pathlib import Path + + +def normalized(value): + return ''.join(c for c in unicodedata.normalize('NFKD',value).casefold() if c.isalnum()) + + +def author_keys(authors): + if not isinstance(authors,list): + return set() + result = set() + for author in authors: + if isinstance(author,list): + author = ' '.join(str(x) for x in author[1:2]+author[:1]) + elif isinstance(author,dict): + author = ' '.join(str(author.get(k,'')) for k in ('given','family')) + key = normalized(str(author)) + if key: + result.add(key) + return result + + +class Store: + def __init__(self,path): + self.conn = sqlite3.connect(path,timeout=60) + self.conn.executescript(Path(__file__).with_name('schema.sql').read_text()) + self.conn.execute('PRAGMA synchronous=NORMAL') + self.conn.execute('PRAGMA cache_size=-32768') + + def source(self,name,url): + self.conn.execute('INSERT OR IGNORE INTO sources(name,homepage) VALUES(?,?)',(name,url)) + return self.conn.execute('SELECT id FROM sources WHERE name=?',(name,)).fetchone()[0] + + def merge(self,target,other,reason): + c = self.conn + if target == other: + return + c.execute('INSERT INTO paper_merges(former_paper_id,surviving_paper_id,reason) VALUES(?,?,?)',(other,target,reason)) + c.execute('UPDATE identifiers SET paper_id=? WHERE paper_id=?',(target,other)) + c.execute('UPDATE source_records SET paper_id=? WHERE paper_id=?',(target,other)) + c.execute('UPDATE submissions SET paper_id=? WHERE paper_id=?',(target,other)) + c.execute('INSERT OR IGNORE INTO paper_tags SELECT ?,tag_id,source_record_id,evidence_key FROM paper_tags WHERE paper_id=?',(target,other)) + c.execute('DELETE FROM paper_tags WHERE paper_id=?',(other,)) + c.execute('UPDATE paper_merges SET surviving_paper_id=? WHERE surviving_paper_id=?',(target,other)) + # Reattach pending candidate evidence without losing colliding rows. + candidates = c.execute('SELECT left_paper_id,right_paper_id,method,score,evidence_json,status FROM match_candidates WHERE left_paper_id=? OR right_paper_id=?',(other,other)).fetchall() + c.execute('DELETE FROM match_candidates WHERE left_paper_id=? OR right_paper_id=?',(other,other)) + for left,right,method,score,evidence,status in candidates: + left = target if left == other else left + right = target if right == other else right + if left != right: + c.execute('INSERT OR IGNORE INTO match_candidates VALUES(?,?,?,?,?,?)',(min(left,right),max(left,right),method,score,evidence,status)) + c.execute('DELETE FROM papers WHERE id=?',(other,)) + + def record(self,source,external_id,url,title,abstract,authors,identifiers=None,metadata=None,published=None): + c = self.conn + identifiers = dict(identifiers or {}) + # External IDs are scoped to platforms; an OpenReview ID is a forum ID. + identifiers[source] = external_id + ids = [] + for scheme,value in identifiers.items(): + if not value: + continue + value = str(value).strip() + if scheme == 'doi': + value = re.sub(r'^https?://(?:dx\.)?doi\.org/','',value,flags=re.I).lower() + if not re.fullmatch(r'10\.\d{4,9}/\S+',value): + continue + if scheme == 'arxiv': + value = re.sub(r'v\d+$','',value) + ids.append((scheme,value)) + matches = set() + for scheme,value in ids: + found = c.execute('SELECT paper_id FROM identifiers WHERE scheme=? AND value=?',(scheme,value)).fetchone() + if found: + matches.add(found[0]) + old = c.execute('SELECT paper_id FROM source_records WHERE source_id=? AND external_id=?',(source_id:=self.source(source,url),external_id)).fetchone() + if old: + matches.add(old[0]) + key = normalized(title) + title_matches = c.execute('SELECT id,authors_json,abstract FROM papers WHERE title_key=?',(key,)).fetchall() + authors_set = author_keys(authors) + for paper,old_authors,old_abstract in title_matches: + overlap = authors_set & author_keys(json.loads(old_authors)) + # Cross-platform title matching needs independent author corroboration. + same_source = c.execute('SELECT 1 FROM source_records WHERE paper_id=? AND source_id=? LIMIT 1',(paper,source_id)).fetchone() + if not same_source and len(key)>=25 and (len(overlap)>=2 or (overlap and abstract and abstract.strip()==(old_abstract or '').strip())): + matches.add(paper) + if matches: + paper = min(matches) + for other in sorted(matches-{paper}): + self.merge(paper,other,'explicit identifier or exact title corroborated by authors') + c.execute("UPDATE papers SET abstract=CASE WHEN abstract IS NULL OR abstract='' THEN ? ELSE abstract END WHERE id=?",(abstract,paper)) + else: + paper = c.execute('INSERT INTO papers(title,abstract,title_key,authors_json,first_published) VALUES(?,?,?,?,?)',(title,abstract,key,json.dumps(authors,ensure_ascii=False),published)).lastrowid + for scheme,value in ids: + c.execute('INSERT OR IGNORE INTO identifiers VALUES(?,?,?)',(scheme,value,paper)) + c.execute('INSERT INTO source_records(source_id,external_id,paper_id,url,title,abstract,authors_json,metadata_json) VALUES(?,?,?,?,?,?,?,?) ON CONFLICT(source_id,external_id) DO UPDATE SET paper_id=excluded.paper_id,title=excluded.title,abstract=excluded.abstract,authors_json=excluded.authors_json,metadata_json=excluded.metadata_json,retrieved_at=CURRENT_TIMESTAMP',(source_id,external_id,paper,url,title,abstract,json.dumps(authors,ensure_ascii=False),json.dumps(metadata or {},ensure_ascii=False,default=str))) + record = c.execute('SELECT id FROM source_records WHERE source_id=? AND external_id=?',(source_id,external_id)).fetchone()[0] + for other,_,_ in title_matches: + if other != paper and c.execute('SELECT 1 FROM papers WHERE id=?',(other,)).fetchone(): + c.execute('INSERT OR IGNORE INTO match_candidates VALUES(?,?,?,?,?,?)',(min(paper,other),max(paper,other),'exact_title_unconfirmed',None,json.dumps({'title':title}), 'pending')) + return paper,record + + def accepted(self,paper,record,venue,year,track,external_id,url): + c = self.conn + c.execute("INSERT OR IGNORE INTO venues(acronym,name,field,venue_type) VALUES(?,?,?,'conference')",(venue,venue,'unclassified')) + venue_id = c.execute('SELECT id FROM venues WHERE acronym=?',(venue,)).fetchone()[0] + c.execute('INSERT OR IGNORE INTO editions(venue_id,year,track) VALUES(?,?,?)',(venue_id,int(year),track)) + edition = c.execute('SELECT id FROM editions WHERE venue_id=? AND year=? AND track=?',(venue_id,int(year),track)).fetchone()[0] + c.execute('INSERT OR IGNORE INTO submissions(paper_id,edition_id,source_record_id,source_submission_id) VALUES(?,?,?,?)',(paper,edition,record,external_id)) + submission = c.execute('SELECT id FROM submissions WHERE source_record_id=? AND edition_id=? AND source_submission_id=?',(record,edition,external_id)).fetchone()[0] + if not c.execute("SELECT 1 FROM decisions WHERE submission_id=? AND status='accepted' AND evidence_url=?",(submission,url)).fetchone(): + c.execute("INSERT INTO decisions(submission_id,status,raw_value,evidence_url,observed_at) VALUES(?,'accepted','listed in official proceedings',?,strftime('%Y-%m-%dT%H:%M:%fZ','now'))",(submission,url)) + return edition diff --git a/query.py b/query.py new file mode 100644 index 0000000..b293291 --- /dev/null +++ b/query.py @@ -0,0 +1,52 @@ +"""Read-only command-line queries for the paper index.""" +import argparse +import json +import sqlite3 +from pathlib import Path + + +def main(): + parser = argparse.ArgumentParser() + parser.add_argument('--db', required=True) + parser.add_argument('--query', help='SQLite FTS5 query over title and abstract') + parser.add_argument('--tag', help='Filter as namespace=value, e.g. arxiv:category=cs.AI') + parser.add_argument('--venue', help='Conference acronym') + parser.add_argument('--year', type=int) + parser.add_argument('--status', choices=['accepted','rejected','desk_rejected','withdrawn','under_review','unknown']) + parser.add_argument('--limit', type=int, default=20) + parser.add_argument('--stats', action='store_true') + args = parser.parse_args() + conn = sqlite3.connect(Path(args.db).resolve().as_uri()+'?mode=ro', uri=True) + conn.row_factory = sqlite3.Row + if args.stats: + result = {name:conn.execute('SELECT count(*) FROM '+name).fetchone()[0] for name in ('papers','source_records','submissions','decisions','tags','match_candidates')} + result['coverage'] = [dict(r) for r in conn.execute('SELECT status,count(*) AS sources FROM edition_sources GROUP BY status')] + result['runs'] = [dict(r) for r in conn.execute('SELECT * FROM ingest_runs ORDER BY id DESC LIMIT 10')] + else: + where = []; params = [] + if args.query: + where.append('p.id IN (SELECT rowid FROM papers_fts WHERE papers_fts MATCH ?)'); params.append(args.query) + if args.tag: + namespace,value = args.tag.split('=',1) + where.append('EXISTS(SELECT 1 FROM paper_tags pt JOIN tags t ON t.id=pt.tag_id WHERE pt.paper_id=p.id AND t.namespace=? AND t.value=?)'); params.extend([namespace,value]) + if args.venue or args.year or args.status: + clause = ['s.paper_id=p.id'] + if args.venue: + clause.append('v.acronym=?'); params.append(args.venue) + if args.year: + clause.append('e.year=?'); params.append(args.year) + if args.status: + clause.append('EXISTS(SELECT 1 FROM decisions d WHERE d.submission_id=s.id AND d.status=? AND NOT EXISTS(SELECT 1 FROM decisions newer WHERE newer.submission_id=d.submission_id AND (newer.observed_at>d.observed_at OR (newer.observed_at=d.observed_at AND newer.id>d.id))))'); params.append(args.status) + where.append('EXISTS(SELECT 1 FROM submissions s JOIN editions e ON e.id=s.edition_id JOIN venues v ON v.id=e.venue_id WHERE '+' AND '.join(clause)+')') + sql = 'SELECT p.* FROM papers p'+(' WHERE '+' AND '.join(where) if where else '')+' ORDER BY p.id LIMIT ?' + params.append(max(1,min(args.limit,200))) + result = [] + for row in conn.execute(sql,params): + paper = dict(row) + paper['identifiers'] = [dict(r) for r in conn.execute('SELECT scheme,value FROM identifiers WHERE paper_id=?',(row['id'],))] + result.append(paper) + print(json.dumps(result,ensure_ascii=False,indent=2)) + + +if __name__ == '__main__': + main() diff --git a/requirements.txt b/requirements.txt new file mode 100644 index 0000000..f62ce0c --- /dev/null +++ b/requirements.txt @@ -0,0 +1 @@ +PyYAML==6.0.3 diff --git a/schema.sql b/schema.sql new file mode 100644 index 0000000..8a93f24 --- /dev/null +++ b/schema.sql @@ -0,0 +1,92 @@ +PRAGMA foreign_keys=ON; +PRAGMA journal_mode=WAL; +CREATE TABLE IF NOT EXISTS sources ( + id INTEGER PRIMARY KEY, name TEXT NOT NULL UNIQUE, homepage TEXT NOT NULL +); +CREATE TABLE IF NOT EXISTS papers ( + id INTEGER PRIMARY KEY AUTOINCREMENT, title TEXT NOT NULL, abstract TEXT, + title_key TEXT NOT NULL, authors_json TEXT NOT NULL DEFAULT '[]', + first_published TEXT, created_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP +); +CREATE INDEX IF NOT EXISTS papers_title_key ON papers(title_key); +CREATE TABLE IF NOT EXISTS identifiers ( + scheme TEXT NOT NULL, value TEXT NOT NULL, paper_id INTEGER NOT NULL REFERENCES papers(id), + PRIMARY KEY(scheme,value) +); +CREATE INDEX IF NOT EXISTS identifiers_paper ON identifiers(paper_id); +CREATE TABLE IF NOT EXISTS source_records ( + id INTEGER PRIMARY KEY, source_id INTEGER NOT NULL REFERENCES sources(id), + external_id TEXT NOT NULL, paper_id INTEGER NOT NULL REFERENCES papers(id), + url TEXT NOT NULL, title TEXT NOT NULL, abstract TEXT, authors_json TEXT NOT NULL, + metadata_json TEXT NOT NULL, retrieved_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP, + UNIQUE(source_id,external_id) +); +CREATE INDEX IF NOT EXISTS source_records_paper ON source_records(paper_id); +CREATE TABLE IF NOT EXISTS venues ( + id INTEGER PRIMARY KEY, acronym TEXT NOT NULL UNIQUE, name TEXT NOT NULL, + field TEXT NOT NULL, venue_type TEXT NOT NULL CHECK(venue_type IN ('conference','journal','review_platform')) +); +CREATE TABLE IF NOT EXISTS editions ( + id INTEGER PRIMARY KEY, venue_id INTEGER NOT NULL REFERENCES venues(id), + year INTEGER NOT NULL, track TEXT NOT NULL DEFAULT 'main', + UNIQUE(venue_id,year,track) +); +CREATE TABLE IF NOT EXISTS edition_sources ( + edition_id INTEGER NOT NULL REFERENCES editions(id), source_id INTEGER NOT NULL REFERENCES sources(id), + list_url TEXT NOT NULL, status TEXT NOT NULL DEFAULT 'pending' + CHECK(status IN ('pending','running','complete','partial','blocked','unavailable','not_yet_published')), + expected_count INTEGER, imported_count INTEGER NOT NULL DEFAULT 0, + cursor TEXT, last_success_at TEXT, last_error TEXT, + PRIMARY KEY(edition_id,source_id,list_url) +); +CREATE TABLE IF NOT EXISTS submissions ( + id INTEGER PRIMARY KEY, paper_id INTEGER NOT NULL REFERENCES papers(id), + edition_id INTEGER NOT NULL REFERENCES editions(id), + source_record_id INTEGER NOT NULL REFERENCES source_records(id), + source_submission_id TEXT NOT NULL, submitted_at TEXT, + UNIQUE(source_record_id,edition_id,source_submission_id) +); +CREATE INDEX IF NOT EXISTS submissions_paper ON submissions(paper_id); +CREATE TABLE IF NOT EXISTS decisions ( + id INTEGER PRIMARY KEY, submission_id INTEGER NOT NULL REFERENCES submissions(id), + status TEXT NOT NULL CHECK(status IN ('accepted','rejected','desk_rejected','withdrawn','under_review','unknown')), + raw_value TEXT NOT NULL, evidence_url TEXT NOT NULL, observed_at TEXT NOT NULL, + decision_date TEXT, UNIQUE(submission_id,status,raw_value,evidence_url,observed_at) +); +CREATE TABLE IF NOT EXISTS tags ( + id INTEGER PRIMARY KEY, namespace TEXT NOT NULL, value TEXT NOT NULL, + description TEXT, UNIQUE(namespace,value) +); +CREATE TABLE IF NOT EXISTS paper_tags ( + paper_id INTEGER NOT NULL REFERENCES papers(id), tag_id INTEGER NOT NULL REFERENCES tags(id), + source_record_id INTEGER REFERENCES source_records(id), + evidence_key TEXT NOT NULL, PRIMARY KEY(paper_id,tag_id,evidence_key) +); +CREATE TABLE IF NOT EXISTS match_candidates ( + left_paper_id INTEGER NOT NULL REFERENCES papers(id), right_paper_id INTEGER NOT NULL REFERENCES papers(id), + method TEXT NOT NULL, score REAL, evidence_json TEXT NOT NULL, + status TEXT NOT NULL DEFAULT 'pending' CHECK(status IN ('pending','merged','distinct')), + PRIMARY KEY(left_paper_id,right_paper_id,method) +); +CREATE TABLE IF NOT EXISTS paper_merges ( + id INTEGER PRIMARY KEY, former_paper_id INTEGER NOT NULL, + surviving_paper_id INTEGER NOT NULL REFERENCES papers(id), + reason TEXT NOT NULL, merged_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP +); +CREATE TABLE IF NOT EXISTS ingest_runs ( + id INTEGER PRIMARY KEY, source_id INTEGER NOT NULL REFERENCES sources(id), + started_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP, finished_at TEXT, + status TEXT NOT NULL DEFAULT 'running', records_seen INTEGER NOT NULL DEFAULT 0, + records_written INTEGER NOT NULL DEFAULT 0, error TEXT, checkpoint_json TEXT +); +CREATE VIRTUAL TABLE IF NOT EXISTS papers_fts USING fts5(title,abstract,content='papers',content_rowid='id'); +CREATE TRIGGER IF NOT EXISTS papers_insert AFTER INSERT ON papers BEGIN + INSERT INTO papers_fts(rowid,title,abstract) VALUES(new.id,new.title,new.abstract); +END; +CREATE TRIGGER IF NOT EXISTS papers_delete AFTER DELETE ON papers BEGIN + INSERT INTO papers_fts(papers_fts,rowid,title,abstract) VALUES('delete',old.id,old.title,old.abstract); +END; +CREATE TRIGGER IF NOT EXISTS papers_update AFTER UPDATE OF title,abstract ON papers BEGIN + INSERT INTO papers_fts(papers_fts,rowid,title,abstract) VALUES('delete',old.id,old.title,old.abstract); + INSERT INTO papers_fts(rowid,title,abstract) VALUES(new.id,new.title,new.abstract); +END; diff --git a/test_import_arxiv.py b/test_import_arxiv.py new file mode 100644 index 0000000..4c2c6e8 --- /dev/null +++ b/test_import_arxiv.py @@ -0,0 +1,57 @@ +import json +import sqlite3 +import tempfile +import unittest +import zipfile +from pathlib import Path + +from import_arxiv import ingest + + +class ImportTests(unittest.TestCase): + def setUp(self): + self.tmp = tempfile.TemporaryDirectory() + self.addCleanup(self.tmp.cleanup) + self.root = Path(self.tmp.name) + self.db = self.root/'papers.sqlite' + self.archive = self.root/'metadata.zip' + + def write(self, rows): + with zipfile.ZipFile(self.archive,'w') as z: + z.writestr('arxiv-metadata-oai-snapshot.json','\n'.join(json.dumps(r) for r in rows)) + + def row(self, identifier, doi=None): + return dict(id=identifier,title='A shared title',abstract='Searchable abstract',authors_parsed=[['Doe','Jane','']],categories='cs.AI cs.LG',doi=doi,versions=[]) + + def test_identity_provenance_and_idempotency(self): + self.write([self.row('2601.00001','10.0000/example'),self.row('2601.00002','10.0000/example')]) + ingest(self.db,self.archive) + ingest(self.db,self.archive) + with sqlite3.connect(self.db) as c: + self.assertEqual(c.execute('SELECT count(*) FROM papers').fetchone()[0],1) + self.assertEqual(c.execute('SELECT count(*) FROM source_records').fetchone()[0],2) + self.assertEqual(c.execute("SELECT count(*) FROM papers_fts WHERE papers_fts MATCH 'Searchable'").fetchone()[0],1) + self.assertEqual(c.execute('SELECT count(*) FROM decisions').fetchone()[0],0) + self.assertEqual(c.execute('PRAGMA foreign_key_check').fetchall(),[]) + self.assertEqual(c.execute('PRAGMA integrity_check').fetchone()[0],'ok') + + def test_title_alone_does_not_merge(self): + self.write([self.row('2601.00001'),self.row('2601.00002')]) + ingest(self.db,self.archive) + with sqlite3.connect(self.db) as c: + self.assertEqual(c.execute('SELECT count(*) FROM papers').fetchone()[0],2) + + def test_platform_and_custom_tags_remain_distinct(self): + self.write([self.row('2601.00001')]) + ingest(self.db,self.archive) + with sqlite3.connect(self.db) as c: + paper = c.execute('SELECT id FROM papers').fetchone()[0] + tag = c.execute("INSERT INTO tags(namespace,value) VALUES('custom:topic','cs.AI')").lastrowid + c.execute('INSERT INTO paper_tags VALUES(?,?,NULL,?)',(paper,tag,'manual:1')) + ingest(self.db,self.archive) + with sqlite3.connect(self.db) as c: + self.assertEqual(c.execute("SELECT count(*) FROM paper_tags pt JOIN tags t ON pt.tag_id=t.id WHERE t.value='cs.AI'").fetchone()[0],2) + + +if __name__ == '__main__': + unittest.main() diff --git a/test_index_core.py b/test_index_core.py new file mode 100644 index 0000000..1d7c0f6 --- /dev/null +++ b/test_index_core.py @@ -0,0 +1,56 @@ +import json +import sqlite3 +import tempfile +import unittest +from pathlib import Path + +from index_core import Store +from import_records import ingest_file + + +class StoreTests(unittest.TestCase): + def setUp(self): + self.tmp = tempfile.TemporaryDirectory() + self.addCleanup(self.tmp.cleanup) + self.db = Path(self.tmp.name)/'papers.sqlite' + self.store = Store(self.db) + self.addCleanup(self.store.conn.close) + + def record(self,source,identifier,ids=None,authors=None): + return self.store.record(source,identifier,'https://example.org/'+identifier,'A sufficiently long example research paper title','Example abstract',authors or ['Jane Doe','John Smith'],ids) + + def test_bridge_identifiers_preserves_submission_and_tags(self): + first,record = self.record('arxiv','2601.00001',{'arxiv':'2601.00001'},['First Author']) + second,other_record = self.record('pmlr','v1/other',{'doi':'10.0000/example'},['Other Author']) + self.store.accepted(second,other_record,'ICML',2025,'main','v1/other','https://example.org/other') + tag = self.store.conn.execute("INSERT INTO tags(namespace,value) VALUES('custom:topic','test')").lastrowid + self.store.conn.execute('INSERT INTO paper_tags VALUES(?,?,?,?)',(second,tag,other_record,'manual')) + resolved,_ = self.record('openreview','forum1',{'arxiv':'2601.00001','doi':'10.0000/example'},['Anonymous']) + self.assertEqual(resolved,first) + c=self.store.conn + self.assertEqual(c.execute('SELECT count(*) FROM papers').fetchone()[0],1) + self.assertEqual(c.execute('SELECT paper_id FROM submissions').fetchone()[0],first) + self.assertEqual(c.execute('SELECT paper_id FROM paper_tags').fetchone()[0],first) + self.assertEqual(c.execute('PRAGMA foreign_key_check').fetchall(),[]) + self.assertEqual(c.execute('SELECT count(*) FROM paper_merges').fetchone()[0],1) + + def test_cross_platform_exact_title_needs_author_corroboration(self): + paper,_=self.record('arxiv','2601.00001') + same,_=self.record('pmlr','v1/paper') + other,_=self.record('acl','2025.acl.1',authors=['Unrelated Person']) + self.assertEqual(same,paper) + self.assertNotEqual(other,paper) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM match_candidates').fetchone()[0],1) + + def test_manifest_count_mismatch_rolls_back_records(self): + path=Path(self.tmp.name)/'source.jsonl' + path.write_text(json.dumps({'source':'pmlr','external_id':'v1/paper','url':'https://example.org/paper','title':'Paper','abstract':'Abstract','authors':[],'venue':'ICML','year':2025,'track':'main','status':'accepted'})+'\n') + path.with_suffix('.manifest.json').write_text(json.dumps({'source':'pmlr','list_url':'https://example.org/list','count':2})) + with self.assertRaisesRegex(ValueError,'differs'): + ingest_file(self.db,path) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM source_records').fetchone()[0],0) + self.assertEqual(self.store.conn.execute('SELECT status FROM ingest_runs').fetchone()[0],'failed') + + +if __name__=='__main__': + unittest.main() diff --git a/venues.json b/venues.json new file mode 100644 index 0000000..5f7adc6 --- /dev/null +++ b/venues.json @@ -0,0 +1,664 @@ +{ + "scope_note": "Broad field coverage, not a claim of objective ranking relative to COLM. Main tracks, Findings, workshops, journals remain distinct. Missing editions must be reported, not silently skipped.", + "sources": { + "arxiv": "https://info.arxiv.org/help/bulk_data/index.html", + "openreview": "https://docs.openreview.net/getting-started/using-the-api", + "acl_anthology": "https://github.com/acl-org/acl-anthology/tree/master/data/xml", + "pmlr": "https://proceedings.mlr.press/", + "neurips": "https://proceedings.neurips.cc/", + "cvf": "https://openaccess.thecvf.com/", + "ecva": "https://www.ecva.net/papers.php", + "aaai": "https://ojs.aaai.org/index.php/AAAI/issue/archive", + "ijcai": "https://www.ijcai.org/all_proceedings/", + "dblp": "https://dblp.org/db/conf/", + "colm": "https://proceedings.colm.cc/" + }, + "venues": [ + { + "acronym": "NeurIPS", + "field": "machine_learning", + "venue_type": "conference", + "discovery_sources": [ + "openreview", + "pmlr", + "neurips" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ICML", + "field": "machine_learning", + "venue_type": "conference", + "discovery_sources": [ + "openreview", + "pmlr", + "neurips" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ICLR", + "field": "machine_learning", + "venue_type": "conference", + "discovery_sources": [ + "openreview", + "pmlr", + "neurips" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "COLM", + "field": "machine_learning", + "venue_type": "conference", + "discovery_sources": [ + "openreview", + "pmlr", + "neurips" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "COLT", + "field": "machine_learning", + "venue_type": "conference", + "discovery_sources": [ + "openreview", + "pmlr", + "neurips" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "AISTATS", + "field": "machine_learning", + "venue_type": "conference", + "discovery_sources": [ + "openreview", + "pmlr", + "neurips" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "UAI", + "field": "machine_learning", + "venue_type": "conference", + "discovery_sources": [ + "openreview", + "pmlr", + "neurips" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "CLeaR", + "field": "machine_learning", + "venue_type": "conference", + "discovery_sources": [ + "openreview", + "pmlr", + "neurips" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ACL", + "field": "language", + "venue_type": "conference", + "discovery_sources": [ + "acl_anthology", + "openreview" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "EMNLP", + "field": "language", + "venue_type": "conference", + "discovery_sources": [ + "acl_anthology", + "openreview" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "NAACL", + "field": "language", + "venue_type": "conference", + "discovery_sources": [ + "acl_anthology", + "openreview" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "EACL", + "field": "language", + "venue_type": "conference", + "discovery_sources": [ + "acl_anthology", + "openreview" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "COLING", + "field": "language", + "venue_type": "conference", + "discovery_sources": [ + "acl_anthology", + "openreview" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "CoNLL", + "field": "language", + "venue_type": "conference", + "discovery_sources": [ + "acl_anthology", + "openreview" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "CVPR", + "field": "vision", + "venue_type": "conference", + "discovery_sources": [ + "cvf", + "ecva", + "openreview" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ICCV", + "field": "vision", + "venue_type": "conference", + "discovery_sources": [ + "cvf", + "ecva", + "openreview" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ECCV", + "field": "vision", + "venue_type": "conference", + "discovery_sources": [ + "cvf", + "ecva", + "openreview" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "AAAI", + "field": "general_ai", + "venue_type": "conference", + "discovery_sources": [ + "aaai", + "ijcai", + "dblp" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "IJCAI", + "field": "general_ai", + "venue_type": "conference", + "discovery_sources": [ + "aaai", + "ijcai", + "dblp" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ECAI", + "field": "general_ai", + "venue_type": "conference", + "discovery_sources": [ + "aaai", + "ijcai", + "dblp" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "KDD", + "field": "retrieval_data", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "publisher" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "SIGIR", + "field": "retrieval_data", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "publisher" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "WWW", + "field": "retrieval_data", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "publisher" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "WSDM", + "field": "retrieval_data", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "publisher" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "RecSys", + "field": "retrieval_data", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "publisher" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "CIKM", + "field": "retrieval_data", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "publisher" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ICDM", + "field": "retrieval_data", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "publisher" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "SDM", + "field": "retrieval_data", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "publisher" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ECML-PKDD", + "field": "retrieval_data", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "publisher" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ICRA", + "field": "robotics", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "pmlr", + "rss" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "IROS", + "field": "robotics", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "pmlr", + "rss" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "RSS", + "field": "robotics", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "pmlr", + "rss" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "CoRL", + "field": "robotics", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "pmlr", + "rss" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ICAPS", + "field": "reasoning_agents", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "proceedings" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "KR", + "field": "reasoning_agents", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "proceedings" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "AAMAS", + "field": "reasoning_agents", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "proceedings" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ICASSP", + "field": "speech_audio", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "isca", + "ismir" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "INTERSPEECH", + "field": "speech_audio", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "isca", + "ismir" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "ISMIR", + "field": "speech_audio", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "isca", + "ismir" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "MICCAI", + "field": "specialist", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "publisher" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + }, + { + "acronym": "FAccT", + "field": "specialist", + "venue_type": "conference", + "discovery_sources": [ + "dblp", + "publisher" + ], + "scope": "candidate_pending_scope_review", + "years": "all_publicly_available", + "tracks": [ + "main" + ], + "coverage": "pending" + } + ], + "additional_tracks": [ + "Findings", + "workshop" + ], + "additional_nonconference_venues": [ + "TMLR", + "JMLR", + "TACL", + "ACL ARR" + ] +} -- 2.52.0 From 4a4272219e77df7d2a735386e2348641cf7d1a29 Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 03:12:44 +0800 Subject: [PATCH 02/20] Collect explicit public ICLR decisions and refresh metadata indexes --- README.md | 31 ++++++++++++++++++-- collect_peerread.py | 71 +++++++++++++++++++++++++++++++++++++++++++++ collect_pmlr.py | 4 +-- import_records.py | 4 +-- index_core.py | 13 ++++++--- test_index_core.py | 7 +++++ test_peerread.py | 33 +++++++++++++++++++++ 7 files changed, 152 insertions(+), 11 deletions(-) create mode 100644 collect_peerread.py create mode 100644 test_peerread.py diff --git a/README.md b/README.md index 162d6b2..e61530e 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 目标包括：完整历史元数据导入、逐年会议目录、跨平台论文去重、公开拒稿/撤稿信息、平台标签与自定义标签隔离、可恢复更新和覆盖率报告。 -**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR 官方元数据采集、带清单计数核验的会议导入和只读查询脚本。服务器上六项合成回归检查通过；完整 ICML 2025 元数据已采集 3,330 条，尚未作为全量交付。arXiv 全量快照下载与历史导入尚未完成。其他会议适配器、跨平台疑似重复复核流程、完整覆盖率核验仍待实现。已配置 ces-control:host 对应的远程 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 +**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR 官方元数据采集、带清单计数核验的会议导入和只读查询脚本。服务器上持续运行合成回归检查；完整 ICML 2025 元数据已采集 3,330 条，尚未作为全量交付。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），全量导入正在服务器运行。新增 PeerRead 公开 ICLR 2017 明确决定采集，其他会议适配器和全量历史导入尚未完成。其他会议适配器、跨平台疑似重复复核流程、完整覆盖率核验仍待实现。已配置 ces-control:host 对应的远程 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 ## 数据范围与来源 @@ -30,6 +30,22 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 - [AAAI](https://ojs.aaai.org/index.php/AAAI/issue/archive) - [DBLP](https://dblp.org/db/conf/) +## 候选会议覆盖目录 + +| 领域 | 会议系列 | 主要发现入口 | +| --- | --- | --- | +| 机器学习、理论 | NeurIPS、ICML、ICLR、COLM、COLT、AISTATS、UAI、CLeaR | OpenReview、PMLR、NeurIPS、COLM 官方论文集 | +| 自然语言 | ACL、EMNLP、NAACL、EACL、COLING、CoNLL | ACL Anthology | +| 视觉 | CVPR、ICCV、ECCV | CVF、ECVA | +| 综合 AI | AAAI、IJCAI、ECAI | 官方论文集、DBLP | +| 检索与数据 | KDD、SIGIR、WWW、WSDM、RecSys、CIKM、ICDM、SDM、ECML-PKDD | 官方出版方、DBLP，具体年份入口待逐项登记 | +| 机器人 | ICRA、IROS、RSS、CoRL | 官方论文集、PMLR、DBLP | +| 规划、知识与智能体 | ICAPS、KR、AAMAS | 官方论文集、DBLP | +| 语音与音频 | ICASSP、INTERSPEECH、ISMIR | 官方论文集、ISCA、DBLP | +| 相关专门领域 | MICCAI、FAccT | 官方出版方、DBLP | + +这些是领域覆盖候选，不构成统一排名，也不能据此声称已全部入库。逐年目录以真实举办届次和正式出版清单为准：双年会议不补造不存在的年份；主会、Findings、workshop 分开。`edition_sources` 保存实际采集状态，不能用候选名单代替覆盖率报告。 + ## 数据模型 | 表 | 用途 | @@ -52,7 +68,7 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 基础导入和查询依赖 Python 3 标准库和带 FTS5 的 SQLite；PMLR 采集另需 `requirements.txt` 中锁定的 PyYAML 6.0.3。无项目构建步骤。服务器已检查为 Python 3.14.4。所有运行数据放在仓库外，不提交数据库、下载包或访问凭据。 -服务器暂存目录：`/data/paper-index/`，快照目录：`/data/paper-index/raw/`，运行代码目录：`/data/paper-index/code/`。这些路径不是对外访问的 API。 +服务器暂存目录：`/data/paper-index/`，快照目录：`/data/paper-index/raw/`，运行代码目录：`/data/paper-index/code/`。这些路径不是对外访问的 API。本地工作区只用于编辑代码、文档和 Git；采集、全量导入和运行检查均在用户确认的服务器执行。 下载完成并验证 ZIP 后，在服务器运行： @@ -94,7 +110,16 @@ python3 -B import_records.py --db /data/paper-index/papers.sqlite \ /data/paper-index/collected/pmlr-v267.jsonl ``` -采集输出 JSONL 和对应 manifest，只有显式官方录用记录才产生 accepted 决定。清单数目不一致时整个文件导入回滚，记录失败批次，不能标为覆盖完成。多个历史卷的全量覆盖仍需单独核验。 +采集输出 JSONL 和对应 manifest，只有显式官方录用记录才产生 accepted 决定。清单数目不一致时整个文件导入回滚，记录失败批次，不能标为覆盖完成。多个历史卷的全量覆盖仍需单独核验。缓存默认有效 24 小时；同一次批次可复用缓存，后续更新会重新读取动态入口。 + +PeerRead 采集仅选择作者公开发布的 ICLR 2017 投稿元数据，固定到源仓库 commit。只依据明确的布尔 `accepted` 标签形成 accepted/rejected；空标签为 unknown，不使用其他数据分区中的 probably-rejected 推断。不会保存 reviews 或 parsed_pdfs 正文。 + +```sh +python3 -B collect_peerread.py --cache /data/paper-index/raw/http-cache \ + --out /data/paper-index/collected +``` + +数据任务以一次性 systemd 服务运行：`paper-atlas-arxiv-import`、`paper-atlas-pmlr-collect`、`paper-atlas-peerread-collect`。它们不是定时任务。主库只允许一个写入进程；会议采集先写独立 JSONL，待 arXiv 主批次结束后再合并。运行日志位于 `/data/paper-index/*-import.log`、`/data/paper-index/*-collect.log`。任何服务启动成功都不等于批次已完成，需检查批次状态、清单与完整性结果。 ## 长期维护与更新 diff --git a/collect_peerread.py b/collect_peerread.py new file mode 100644 index 0000000..ca18641 --- /dev/null +++ b/collect_peerread.py @@ -0,0 +1,71 @@ +"""Collect explicit public ICLR decisions from the authors' PeerRead release. + +Review bodies and parsed full paper text are deliberately not persisted. +""" +import argparse +import json +import time +import urllib.request +from collections import defaultdict +from pathlib import Path + +from collect_pmlr import fetch + + +def collect(cache,out): + root = Path(out) + root.mkdir(parents=True,exist_ok=True) + commit = json.loads(fetch('https://api.github.com/repos/allenai/PeerRead/commits/master',cache))['sha'] + tree = json.loads(fetch(f'https://api.github.com/repos/allenai/PeerRead/git/trees/{commit}?recursive=1',cache)) + if tree.get('truncated'): + raise RuntimeError('Truncated repository tree') + paths = [v['path'] for v in tree['tree'] if v['path'].startswith('data/iclr_2017/') and '/reviews/' in v['path'] and v['path'].endswith('.json')] + if not paths: + raise RuntimeError('No explicit ICLR2017 metadata files found') + cache_dir = root/'peerread-normalized'/commit + cache_dir.mkdir(parents=True,exist_ok=True) + grouped = defaultdict(list) + for number,path in enumerate(paths,1): + external = '/'.join(path.split('/')[1:]) + normalized_cache = cache_dir/(path.replace('/','_')) + if normalized_cache.exists(): + record = json.loads(normalized_cache.read_text()) + else: + raw_url = f'https://raw.githubusercontent.com/allenai/PeerRead/{commit}/{path}' + with urllib.request.urlopen(urllib.request.Request(raw_url,headers={'User-Agent':'PaperAtlas/0.1 metadata research'}),timeout=60) as response: + payload = response.read(3*1024**2+1) + if len(payload)>3*1024**2: + raise RuntimeError('Unexpected metadata size') + data = json.loads(payload) + conference = data.get('conference','') + if 'iclr' not in conference.lower() or '2017' not in conference: + raise RuntimeError('Unexpected conference metadata: '+conference) + accepted = data.get('accepted') + status = ('accepted' if accepted else 'rejected') if isinstance(accepted,bool) else 'unknown' + authors = data.get('authors') or [] + if isinstance(authors,str): + authors = [author.strip() for author in authors.split(',') if author.strip()] + record = {'source':'peerread','external_id':external,'url':f'https://github.com/allenai/PeerRead/blob/{commit}/{path}','title':data['title'],'abstract':data.get('abstract'),'authors':authors,'identifiers':{},'venue':'ICLR','year':2017,'track':'workshop' if 'workshop' in conference.lower() else 'main','status':status,'raw_decision':json.dumps({'accepted':accepted}),'metadata':{'conference':conference,'dataset_commit':commit,'dataset_path':path,'original_id':data.get('id'),'accepted':accepted,'decision_provenance':'PeerRead authors public release; explicit label, not inferred from publication absence'}} + normalized_cache.write_text(json.dumps(record,ensure_ascii=False)) + time.sleep(0.25) + grouped[record['track']].append(record) + if number%25==0: + print(json.dumps({'source':'peerread','collected':number,'expected':len(paths)}),flush=True) + for track,records in grouped.items(): + target = root/f'peerread-iclr2017-{track}.jsonl' + part = target.with_suffix('.part') + with part.open('w') as output: + for record in records: + output.write(json.dumps(record,ensure_ascii=False)+'\n') + part.replace(target) + manifest = {'source':'peerread','venue':'ICLR','year':2017,'track':track,'count':len(records),'list_url':f'https://github.com/allenai/PeerRead/tree/{commit}/data/iclr_2017','dataset_commit':commit,'file':str(target),'decision_counts':{status:sum(r['status']==status for r in records) for status in ('accepted','rejected','unknown')}} + target.with_suffix('.manifest.json').write_text(json.dumps(manifest,indent=2)) + print(json.dumps(manifest),flush=True) + + +if __name__=='__main__': + parser = argparse.ArgumentParser() + parser.add_argument('--cache',required=True) + parser.add_argument('--out',required=True) + args = parser.parse_args() + collect(args.cache,args.out) diff --git a/collect_pmlr.py b/collect_pmlr.py index 1e9da16..09910b0 100644 --- a/collect_pmlr.py +++ b/collect_pmlr.py @@ -17,7 +17,7 @@ def fetch(url,cache,limit=64*1024**2): cache = Path(cache) cache.mkdir(parents=True,exist_ok=True) path = cache/hashlib.sha256(url.encode()).hexdigest() - if path.exists(): + if path.exists() and time.time()-path.stat().st_mtime < 24*3600: return path.read_bytes() for attempt in range(3): try: @@ -102,7 +102,7 @@ def main(): volumes = args.volume if not volumes: index = fetch('https://proceedings.mlr.press/',args.cache).decode() - volumes = sorted(set(int(v) for v in re.findall(r'href=["\'](?:https?://proceedings\.mlr\.press)?/v(\d+)/?["\']',index))) + volumes = sorted(set(int(v) for v in re.findall(r'href=["\'](?:https?://proceedings\.mlr\.press/|\./|/)?v(\d+)/?["\']',index))) if not volumes: raise RuntimeError('PMLR index yielded no volumes') errors = [] diff --git a/import_records.py b/import_records.py index ea66aee..02ed87b 100644 --- a/import_records.py +++ b/import_records.py @@ -33,8 +33,8 @@ def ingest_file(db,path): raise ValueError('Duplicate external ID in one collected file: '+external) seen.add(external) paper,record = store.record(source_name,external,row['url'],row['title'],row.get('abstract'),row.get('authors',[]),row.get('identifiers'),row.get('metadata')) - if row.get('status') == 'accepted': - edition = store.accepted(paper,record,row['venue'],row['year'],row['track'],external,row['url']) + if row.get('status') in ('accepted','rejected','desk_rejected','withdrawn','under_review','unknown'): + edition = store.decision(paper,record,row['venue'],row['year'],row['track'],external,row.get('evidence_url',row['url']),row['status'],row.get('raw_decision','listed in official proceedings' if source_name=='pmlr' else row['status'])) editions.add(edition) else: raise ValueError('This importer requires an explicit supported decision status') diff --git a/index_core.py b/index_core.py index 2bb2898..2c012e1 100644 --- a/index_core.py +++ b/index_core.py @@ -79,7 +79,7 @@ class Store: found = c.execute('SELECT paper_id FROM identifiers WHERE scheme=? AND value=?',(scheme,value)).fetchone() if found: matches.add(found[0]) - old = c.execute('SELECT paper_id FROM source_records WHERE source_id=? AND external_id=?',(source_id:=self.source(source,url),external_id)).fetchone() + old = c.execute('SELECT paper_id,title,abstract FROM source_records WHERE source_id=? AND external_id=?',(source_id:=self.source(source,url),external_id)).fetchone() if old: matches.add(old[0]) key = normalized(title) @@ -95,6 +95,8 @@ class Store: paper = min(matches) for other in sorted(matches-{paper}): self.merge(paper,other,'explicit identifier or exact title corroborated by authors') + if old: + c.execute("UPDATE papers SET title=?,abstract=?,title_key=?,authors_json=? WHERE id=? AND title=? AND COALESCE(abstract,'')=COALESCE(?,'')",(title,abstract,key,json.dumps(authors,ensure_ascii=False),paper,old[1],old[2])) c.execute("UPDATE papers SET abstract=CASE WHEN abstract IS NULL OR abstract='' THEN ? ELSE abstract END WHERE id=?",(abstract,paper)) else: paper = c.execute('INSERT INTO papers(title,abstract,title_key,authors_json,first_published) VALUES(?,?,?,?,?)',(title,abstract,key,json.dumps(authors,ensure_ascii=False),published)).lastrowid @@ -107,7 +109,7 @@ class Store: c.execute('INSERT OR IGNORE INTO match_candidates VALUES(?,?,?,?,?,?)',(min(paper,other),max(paper,other),'exact_title_unconfirmed',None,json.dumps({'title':title}), 'pending')) return paper,record - def accepted(self,paper,record,venue,year,track,external_id,url): + def decision(self,paper,record,venue,year,track,external_id,url,status,raw_value): c = self.conn c.execute("INSERT OR IGNORE INTO venues(acronym,name,field,venue_type) VALUES(?,?,?,'conference')",(venue,venue,'unclassified')) venue_id = c.execute('SELECT id FROM venues WHERE acronym=?',(venue,)).fetchone()[0] @@ -115,6 +117,9 @@ class Store: edition = c.execute('SELECT id FROM editions WHERE venue_id=? AND year=? AND track=?',(venue_id,int(year),track)).fetchone()[0] c.execute('INSERT OR IGNORE INTO submissions(paper_id,edition_id,source_record_id,source_submission_id) VALUES(?,?,?,?)',(paper,edition,record,external_id)) submission = c.execute('SELECT id FROM submissions WHERE source_record_id=? AND edition_id=? AND source_submission_id=?',(record,edition,external_id)).fetchone()[0] - if not c.execute("SELECT 1 FROM decisions WHERE submission_id=? AND status='accepted' AND evidence_url=?",(submission,url)).fetchone(): - c.execute("INSERT INTO decisions(submission_id,status,raw_value,evidence_url,observed_at) VALUES(?,'accepted','listed in official proceedings',?,strftime('%Y-%m-%dT%H:%M:%fZ','now'))",(submission,url)) + if not c.execute('SELECT 1 FROM decisions WHERE submission_id=? AND status=? AND raw_value=? AND evidence_url=?',(submission,status,raw_value,url)).fetchone(): + c.execute("INSERT INTO decisions(submission_id,status,raw_value,evidence_url,observed_at) VALUES(?,?,?,?,strftime('%Y-%m-%dT%H:%M:%fZ','now'))",(submission,status,raw_value,url)) return edition + + def accepted(self,paper,record,venue,year,track,external_id,url): + return self.decision(paper,record,venue,year,track,external_id,url,'accepted','listed in official proceedings') diff --git a/test_index_core.py b/test_index_core.py index 1d7c0f6..91a5979 100644 --- a/test_index_core.py +++ b/test_index_core.py @@ -51,6 +51,13 @@ class StoreTests(unittest.TestCase): self.assertEqual(self.store.conn.execute('SELECT count(*) FROM source_records').fetchone()[0],0) self.assertEqual(self.store.conn.execute('SELECT status FROM ingest_runs').fetchone()[0],'failed') + def test_source_update_refreshes_canonical_text_and_fts(self): + paper,_=self.record('arxiv','2601.00001') + updated,_=self.store.record('arxiv','2601.00001','https://arxiv.org/abs/2601.00001','Updated research title','New searchable content',['Jane Doe']) + self.assertEqual(updated,paper) + self.assertEqual(self.store.conn.execute('SELECT title FROM papers').fetchone()[0],'Updated research title') + self.assertEqual(self.store.conn.execute("SELECT count(*) FROM papers_fts WHERE papers_fts MATCH 'searchable'").fetchone()[0],1) + if __name__=='__main__': unittest.main() diff --git a/test_peerread.py b/test_peerread.py new file mode 100644 index 0000000..40ac185 --- /dev/null +++ b/test_peerread.py @@ -0,0 +1,33 @@ +import io +import json +import tempfile +import unittest +from pathlib import Path +from unittest.mock import patch + +from collect_peerread import collect +from import_records import ingest_file +from index_core import Store + + +class PeerReadTests(unittest.TestCase): + def test_explicit_rejection_and_no_review_body_storage(self): + tree={'sha':'a'*40,'tree':[{'path':'data/iclr_2017/test/reviews/123.json'}]} + paper={'conference':'ICLR 2017 conference submission','id':'123','title':'Public paper','abstract':'Abstract','authors':'Jane Doe, John Smith','accepted':False,'reviews':[{'text':'Do not persist this review body'}]} + with tempfile.TemporaryDirectory() as tmp: + root=Path(tmp) + with patch('collect_peerread.fetch',side_effect=[json.dumps({'sha':'a'*40}).encode(),json.dumps(tree).encode()]),patch('collect_peerread.urllib.request.urlopen',return_value=io.BytesIO(json.dumps(paper).encode())),patch('collect_peerread.time.sleep'): + collect(root/'cache',root/'out') + path=root/'out/peerread-iclr2017-main.jsonl' + record=json.loads(path.read_text()) + self.assertEqual(record['status'],'rejected') + self.assertNotIn('Do not persist',path.read_text()) + ingest_file(root/'db.sqlite',path) + store=Store(root/'db.sqlite') + self.addCleanup(store.conn.close) + self.assertEqual(store.conn.execute('SELECT status FROM decisions').fetchone()[0],'rejected') + self.assertEqual(store.conn.execute('SELECT status FROM edition_sources').fetchone()[0],'complete') + + +if __name__=='__main__': + unittest.main() -- 2.52.0 From 66b98fa3759ec9ae20ee276b968f119b2f4e412e Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 03:19:30 +0800 Subject: [PATCH 03/20] Fetch old PMLR metadata without entire PDF archives --- README.md | 4 ++-- collect_pmlr.py | 61 +++++++++++++++++++++++++++++++++++++++++-------- 2 files changed, 53 insertions(+), 12 deletions(-) diff --git a/README.md b/README.md index e61530e..3da3ff5 100644 --- a/README.md +++ b/README.md @@ -101,7 +101,7 @@ python3 -B query.py --db /data/paper-index/papers.sqlite --venue COLM --year 202 ## 官方会议元数据采集 ```sh -# 单卷示例：ICML 2025，读取官方 gh-pages 元数据归档，不下载 PDF +# 单卷示例：ICML 2025，读取官方 gh-pages 元数据，不下载论文正文 python3 -B collect_pmlr.py --cache /data/paper-index/raw/http-cache \ --out /data/paper-index/collected --volume 267 @@ -110,7 +110,7 @@ python3 -B import_records.py --db /data/paper-index/papers.sqlite \ /data/paper-index/collected/pmlr-v267.jsonl ``` -采集输出 JSONL 和对应 manifest，只有显式官方录用记录才产生 accepted 决定。清单数目不一致时整个文件导入回滚，记录失败批次，不能标为覆盖完成。多个历史卷的全量覆盖仍需单独核验。缓存默认有效 24 小时；同一次批次可复用缓存，后续更新会重新读取动态入口。 +采集输出 JSONL 和对应 manifest，只有显式官方录用记录才产生 accepted 决定。清单数目不一致时整个文件导入回滚，记录失败批次，不能标为覆盖完成。多个历史卷的全量覆盖仍需单独核验。旧卷可能把 PDF 放入归档；采集器对前 50 卷使用 Git partial clone 和 sparse checkout，仅取 `_posts` 和卷配置，不执行出版方项目代码。较新归档若超出 64 MiB 限额，也转为此方式。已对原本超限的 AISTATS 2009（第 5 卷）验证采集 84 条且工作区无 PDF。缓存默认有效 24 小时；同一次批次可复用缓存，后续更新会重新读取动态入口。 PeerRead 采集仅选择作者公开发布的 ICLR 2017 投稿元数据，固定到源仓库 commit。只依据明确的布尔 `accepted` 标签形成 accepted/rejected；空标签为 unknown，不使用其他数据分区中的 probably-rejected 推断。不会保存 reviews 或 parsed_pdfs 正文。 diff --git a/collect_pmlr.py b/collect_pmlr.py index 09910b0..51ea9e9 100644 --- a/collect_pmlr.py +++ b/collect_pmlr.py @@ -4,6 +4,8 @@ import hashlib import io import json import re +import os +import subprocess import tarfile import time import urllib.error @@ -25,6 +27,8 @@ def fetch(url,cache,limit=64*1024**2): with urllib.request.urlopen(request,timeout=90) as response: if response.status != 200: raise RuntimeError('Expected complete HTTP 200 response') + if int(response.headers.get('Content-Length','0'))>limit: + raise RuntimeError('Metadata response exceeds byte limit') data = response.read(limit+1) if len(data)>limit: raise RuntimeError('Metadata response exceeds byte limit') @@ -40,6 +44,49 @@ def fetch(url,cache,limit=64*1024**2): time.sleep(min(30,max(2**attempt,int(exc.headers.get('Retry-After','1'))))) +def metadata_texts(volume,cache): + archive = None + if volume>50: + try: + archive = fetch(f'https://codeload.github.com/mlresearch/v{volume}/tar.gz/refs/heads/gh-pages',cache) + except RuntimeError as exc: + if 'exceeds byte limit' not in str(exc): + raise + if archive is not None: + with tarfile.open(fileobj=io.BytesIO(archive),mode='r:gz') as tar: + for member in tar: + if member.isfile() and '/_posts/' in member.name and member.name.endswith(('.md','.markdown','.html')): + if member.size>8*1024**2: + raise RuntimeError('Unexpectedly large metadata member') + yield member.name,tar.extractfile(member).read().decode('utf-8-sig') + return + # Old volumes may contain PDFs. Partial clone plus sparse checkout retrieves + # trees and metadata blobs only, without checking out or executing publisher code. + checkout = Path(cache)/'metadata-checkouts'/f'pmlr-v{volume}' + checkout.parent.mkdir(parents=True,exist_ok=True) + env = dict(os.environ,GIT_TERMINAL_PROMPT='0') + def git(*args): + result = subprocess.run(['git','-c','credential.helper=',*args],env=env,capture_output=True,text=True,timeout=180) + if result.returncode: + raise RuntimeError('Public metadata Git operation failed: '+result.stderr[-500:]) + return result.stdout + if not checkout.exists(): + git('clone','--filter=blob:none','--depth','1','--single-branch','--branch','gh-pages','--no-checkout',f'https://github.com/mlresearch/v{volume}.git',str(checkout)) + git('-C',str(checkout),'sparse-checkout','set','--no-cone','/_posts/','/_config.yml') + git('-C',str(checkout),'checkout','gh-pages') + else: + if git('-C',str(checkout),'status','--porcelain').strip(): + raise RuntimeError('Metadata checkout has local changes; refusing to overwrite') + git('-C',str(checkout),'fetch','--filter=blob:none','--depth','1','origin','gh-pages') + git('-C',str(checkout),'checkout','--detach','FETCH_HEAD') + posts = checkout/'_posts' + for path in sorted(posts.rglob('*')): + if path.is_file() and path.suffix in ('.md','.markdown','.html'): + if path.stat().st_size>8*1024**2: + raise RuntimeError('Unexpectedly large metadata member') + yield str(path.relative_to(checkout)),path.read_text(encoding='utf-8-sig') + + def collect_volume(volume,cache,out,allowed=None): config_url = f'https://raw.githubusercontent.com/mlresearch/v{volume}/gh-pages/_config.yml' config = yaml.safe_load(fetch(config_url,cache)) @@ -49,25 +96,19 @@ def collect_volume(volume,cache,out,allowed=None): year = int(config['year']) title = str(config.get('booktitle','')) track = 'workshop' if 'workshop' in title.lower() else 'main' - archive = fetch(f'https://codeload.github.com/mlresearch/v{volume}/tar.gz/refs/heads/gh-pages',cache) path = Path(out)/f'pmlr-v{volume}.jsonl' path.parent.mkdir(parents=True,exist_ok=True) part = path.with_suffix('.part') count = 0 - with tarfile.open(fileobj=io.BytesIO(archive),mode='r:gz') as tar,part.open('w') as output: - for member in tar: - if not member.isfile() or '/_posts/' not in member.name or not member.name.endswith(('.md','.markdown','.html')): - continue - if member.size>8*1024**2: - raise RuntimeError('Unexpectedly large metadata member') - content = tar.extractfile(member).read().decode('utf-8-sig') + with part.open('w') as output: + for member_name,content in metadata_texts(volume,cache): parts = content.split('---',2) if len(parts)<3 or parts[0].strip(): - raise RuntimeError('Missing YAML frontmatter: '+member.name) + raise RuntimeError('Missing YAML frontmatter: '+member_name) row = yaml.safe_load(parts[1]) if row.get('layout') not in ('inproceedings','article') or not row.get('title'): continue - external = f"v{volume}/{row.get('id') or Path(member.name).stem}" + external = f"v{volume}/{row.get('id') or Path(member_name).stem}" ids = {} for key in ('doi','openreview','arxiv'): value = row.get(key) -- 2.52.0 From 5d9d13a27218f4463e3d6ae29f269d1a74c32372 Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 11:34:58 +0800 Subject: [PATCH 04/20] Verify historical conference metadata and expose consistent paper statistics --- README.md | 12 ++++++--- collect_pmlr.py | 56 ++++++++++++++++++++++++++++++++++------- merge_conferences.py | 60 ++++++++++++++++++++++++++++++++++++++++++++ query.py | 23 ++++++++++++++--- schema.sql | 1 + test_pmlr.py | 25 ++++++++++++++++++ test_query.py | 39 ++++++++++++++++++++++++++++ 7 files changed, 200 insertions(+), 16 deletions(-) create mode 100644 merge_conferences.py create mode 100644 test_pmlr.py create mode 100644 test_query.py diff --git a/README.md b/README.md index 3da3ff5..3362078 100644 --- a/README.md +++ b/README.md @@ -96,7 +96,9 @@ python3 -B query.py --db /data/paper-index/papers.sqlite --tag arxiv:category=cs python3 -B query.py --db /data/paper-index/papers.sqlite --venue COLM --year 2025 --status rejected ``` -查询连接使用 SQLite 只读模式，结果输出 JSON。当前没有公共 Web 页面或网络查询服务。 +查询连接使用 SQLite 只读模式，结果输出 JSON。`--stats` 中 `papers` 是按当前规则去重后的论文实体数，`records_by_source` 是各平台已入库记录数，`decision_events_by_status` 是决定事件数。采集文件 manifest 的 `count` 是已采集数量，不能等同已入库数量，也不能把各平台记录数相加当成独立论文数。 + +用 `--paper 531782` 可查看单篇论文的所有已知来源、平台标签、投稿届次及带证据的决定历史。会议筛选不区分大小写，状态筛选使用该次投稿的最新已知决定。当前没有公共 Web 页面或网络查询服务。 ## 官方会议元数据采集 @@ -110,7 +112,7 @@ python3 -B import_records.py --db /data/paper-index/papers.sqlite \ /data/paper-index/collected/pmlr-v267.jsonl ``` -采集输出 JSONL 和对应 manifest，只有显式官方录用记录才产生 accepted 决定。清单数目不一致时整个文件导入回滚，记录失败批次，不能标为覆盖完成。多个历史卷的全量覆盖仍需单独核验。旧卷可能把 PDF 放入归档；采集器对前 50 卷使用 Git partial clone 和 sparse checkout，仅取 `_posts` 和卷配置，不执行出版方项目代码。较新归档若超出 64 MiB 限额，也转为此方式。已对原本超限的 AISTATS 2009（第 5 卷）验证采集 84 条且工作区无 PDF。缓存默认有效 24 小时；同一次批次可复用缓存，后续更新会重新读取动态入口。 +采集输出 JSONL 和对应 manifest，只有明确来源标签才产生相应决定：官方论文集记录为 accepted；PeerRead 使用其明确的 accepted 布尔标签，缺失标签保留 unknown。清单数目不一致时整个文件导入回滚，记录失败批次，不能标为覆盖完成。多个历史卷的全量覆盖仍需单独核验。旧卷可能把 PDF 放入归档；采集器对前 50 卷使用 Git partial clone 和 sparse checkout，仅取 `_posts` 和卷配置，不执行出版方项目代码。较新归档若超出 64 MiB 限额，也转为此方式。已对原本超限的 AISTATS 2009（第 5 卷）验证采集 84 条且工作区无 PDF。旧 YAML 的 `!binary` 文本、空文档尾标记会安全解析；带年份的 AISTATS/COLT 等历史 shortname 会规范化，避免名称别名漏届；缺失 year 时仅从官方会议 start 日期提取年份。`pmlr-volume-status.json` 逐卷保存结果和候选范围归属，`pmlr-last-run.json` 记录本批实际请求范围，局部重试不能证明全历史完成。缓存默认有效 24 小时；同一次批次可复用缓存，后续更新会重新读取动态入口。 PeerRead 采集仅选择作者公开发布的 ICLR 2017 投稿元数据，固定到源仓库 commit。只依据明确的布尔 `accepted` 标签形成 accepted/rejected；空标签为 unknown，不使用其他数据分区中的 probably-rejected 推断。不会保存 reviews 或 parsed_pdfs 正文。 @@ -119,7 +121,7 @@ python3 -B collect_peerread.py --cache /data/paper-index/raw/http-cache \ --out /data/paper-index/collected ``` -数据任务以一次性 systemd 服务运行：`paper-atlas-arxiv-import`、`paper-atlas-pmlr-collect`、`paper-atlas-peerread-collect`。它们不是定时任务。主库只允许一个写入进程；会议采集先写独立 JSONL，待 arXiv 主批次结束后再合并。运行日志位于 `/data/paper-index/*-import.log`、`/data/paper-index/*-collect.log`。任何服务启动成功都不等于批次已完成，需检查批次状态、清单与完整性结果。 +数据任务以一次性 systemd 服务运行：`paper-atlas-arxiv-import`、`paper-atlas-pmlr-collect`、`paper-atlas-peerread-collect`；`paper-atlas-pmlr-verify` 用于全索引复核，`paper-atlas-conference-merge` 用于单写入合并。它们不是定时任务。主库只允许一个写入进程；会议采集先写独立 JSONL，待 arXiv 主批次结束后再合并；也可以使用 `merge_conferences.py` 在保存断点处暂停本项目的 arXiv unit、顺序合并会议清单并恢复。该操作应作为独立的一次性 systemd 任务运行，避免 SSH 断开影响恢复；不能同时启动多个合并任务。运行日志位于 `/data/paper-index/*-import.log`、`/data/paper-index/*-collect.log`。任何服务启动成功都不等于批次已完成，需检查批次状态、清单与完整性结果。 ## 长期维护与更新 @@ -135,4 +137,6 @@ python3 -B collect_peerread.py --cache /data/paper-index/raw/http-cache \ 2026-11-06 从目标服务器测试：Kaggle 元数据 GET 可下载；OpenReview 会议 group 接口可读，但 notes 接口返回 `ChallengeRequiredError`。拒稿记录导入尚受此限制，不能用录用名单替代全部投稿，也不能声称已覆盖全部拒稿。需通过平台支持的访问方式或有合法来源的公开数据导出继续补全。 -全量导入、会议适配器、正式 runner 配置与 PR 交付状态应随实现及时更新本 README。 +已验证的阶段性结果：ICML 2025 的 3,330 条记录和 PeerRead ICLR 2017 的 427 条记录已合并主库，合计 3,502 条 accepted 决定、255 条 rejected 决定；arXiv 曾在 529,000 条断点暂停合并，已验证恢复。这些是阶段性值，不是全量交付数据量。 + +代码在 `codex/metadata-index` 工作分支交付，[Draft PR #1](https://gitea.entroia.com/Zayden-Jung/paper-atlas/pulls/1) 尚未合并，等待全量就绪后用户手动合并。已推送的前三个开发提交均有成功的远程 CI；后续提交仍需检查其各自 CI。全量导入、其他会议适配器、去重复核和覆盖率核验继续进行，应随实际结果更新此说明。 diff --git a/collect_pmlr.py b/collect_pmlr.py index 51ea9e9..f43cfe3 100644 --- a/collect_pmlr.py +++ b/collect_pmlr.py @@ -15,6 +15,29 @@ from pathlib import Path import yaml +class MetadataLoader(yaml.SafeLoader): + pass + + +def binary_text(loader,node): + return loader.construct_yaml_binary(node).decode('utf-8') + + +MetadataLoader.add_constructor('!binary',binary_text) +MetadataLoader.add_constructor('tag:yaml.org,2002:binary',binary_text) + + +def load_metadata(value): + documents=[doc for doc in yaml.load_all(value,Loader=MetadataLoader) if doc is not None] + if len(documents)!=1 or not isinstance(documents[0],dict): + raise ValueError('Expected exactly one nonempty metadata mapping') + return documents[0] + + +def venue_key(shortname): + return re.sub(r'[\s_-]*(?:19|20)\d{2}$','',str(shortname).strip().upper()) + + def fetch(url,cache,limit=64*1024**2): cache = Path(cache) cache.mkdir(parents=True,exist_ok=True) @@ -89,11 +112,16 @@ def metadata_texts(volume,cache): def collect_volume(volume,cache,out,allowed=None): config_url = f'https://raw.githubusercontent.com/mlresearch/v{volume}/gh-pages/_config.yml' - config = yaml.safe_load(fetch(config_url,cache)) - venue = str(config.get('shortname','')).upper() + config = load_metadata(fetch(config_url,cache)) + venue = venue_key(config.get('shortname','')) if allowed and venue not in allowed: - return None - year = int(config['year']) + return {'source':'pmlr','volume':volume,'status':'excluded_by_registry','raw_shortname':venue,'booktitle':str(config.get('booktitle',''))} + year_value = config.get('year') + if not year_value: + year_value = str(config.get('start',''))[:4] + if not re.fullmatch(r'\d{4}',str(year_value)): + raise ValueError('No explicit year or conference start date') + year = int(year_value) title = str(config.get('booktitle','')) track = 'workshop' if 'workshop' in title.lower() else 'main' path = Path(out)/f'pmlr-v{volume}.jsonl' @@ -102,10 +130,10 @@ def collect_volume(volume,cache,out,allowed=None): count = 0 with part.open('w') as output: for member_name,content in metadata_texts(volume,cache): - parts = content.split('---',2) + parts = re.split(r'(?m)^---\s*$',content,maxsplit=2) if len(parts)<3 or parts[0].strip(): raise RuntimeError('Missing YAML frontmatter: '+member_name) - row = yaml.safe_load(parts[1]) + row = load_metadata(parts[1]) if row.get('layout') not in ('inproceedings','article') or not row.get('title'): continue external = f"v{volume}/{row.get('id') or Path(member_name).stem}" @@ -139,23 +167,33 @@ def main(): parser.add_argument('--volume',type=int,action='append') parser.add_argument('--registry',default=str(Path(__file__).with_name('venues.json'))) args = parser.parse_args() - allowed = {v['acronym'].upper() for v in json.loads(Path(args.registry).read_text())['venues']} + allowed = {venue_key(v['acronym']) for v in json.loads(Path(args.registry).read_text())['venues']} volumes = args.volume if not volumes: index = fetch('https://proceedings.mlr.press/',args.cache).decode() volumes = sorted(set(int(v) for v in re.findall(r'href=["\'](?:https?://proceedings\.mlr\.press/|\./|/)?v(\d+)/?["\']',index))) if not volumes: raise RuntimeError('PMLR index yielded no volumes') + Path(args.out).mkdir(parents=True,exist_ok=True) + catalogue_path=Path(args.out)/'pmlr-volume-status.json' + catalogue=json.loads(catalogue_path.read_text()) if catalogue_path.exists() else {} errors = [] for volume in volumes: try: - collect_volume(volume,args.cache,args.out,None if args.volume else allowed) + result=collect_volume(volume,args.cache,args.out,None if args.volume else allowed) + result['selection_mode']='explicit' if args.volume else 'registry' + result['selected_by_registry']=result.get('venue',result.get('raw_shortname')) in allowed + catalogue[str(volume)]=result except Exception as exc: error = {'volume':volume,'status':'blocked','error':str(exc)} errors.append(error) + catalogue[str(volume)]=error print(json.dumps(error),flush=True) - Path(args.out).mkdir(parents=True,exist_ok=True) + part=catalogue_path.with_suffix('.part') + part.write_text(json.dumps(catalogue,indent=2)) + part.replace(catalogue_path) (Path(args.out)/'pmlr-errors.json').write_text(json.dumps(errors,indent=2)) + (Path(args.out)/'pmlr-last-run.json').write_text(json.dumps({'requested_volumes':volumes,'selection_mode':'explicit' if args.volume else 'registry','errors':errors},indent=2)) if errors: raise SystemExit(1) diff --git a/merge_conferences.py b/merge_conferences.py new file mode 100644 index 0000000..a70098a --- /dev/null +++ b/merge_conferences.py @@ -0,0 +1,60 @@ +"""One-shot server operation: merge collected files at an arXiv checkpoint. + +Run as a transient systemd job so an SSH disconnect cannot strand the importer. +Only the Paper Atlas arXiv unit is controlled; business services are untouched. +""" +import argparse +import json +import sqlite3 +import subprocess +from pathlib import Path + +from import_records import ingest_file + + +UNIT='paper-atlas-arxiv-import.service' + + +def property_value(name): + return subprocess.check_output(['systemctl','show',UNIT,'-p',name,'--value'],text=True).strip() + + +def merge(db,archive,files): + # Validate inputs before stopping the existing writer. + for filename in files: + path=Path(filename) + if not path.is_file() or not path.with_suffix('.manifest.json').is_file(): + raise ValueError('Missing collected file or manifest: '+filename) + if not Path(archive).is_file(): + raise ValueError('Missing arXiv archive') + if property_value('ActiveState')!='active': + raise RuntimeError('Expected the existing arXiv unit to be active; inspect it before proceeding') + subprocess.run(['systemctl','stop',UNIT],check=True) + try: + with sqlite3.connect(db) as c: + checkpoint=c.execute("SELECT records_written FROM ingest_runs WHERE source_id=(SELECT id FROM sources WHERE name='arxiv') ORDER BY id DESC LIMIT 1").fetchone() + print('Saved arxiv checkpoint:',checkpoint,flush=True) + for filename in files: + ingest_file(db,filename) + with sqlite3.connect(db) as c: + errors=c.execute('PRAGMA foreign_key_check').fetchall() + if errors: + raise RuntimeError('Foreign key violations after merge') + print('Decision totals:',c.execute('SELECT status,count(*) FROM decisions GROUP BY status').fetchall(),flush=True) + finally: + if property_value('LoadState')=='loaded': + subprocess.run(['systemctl','start',UNIT],check=True) + else: + subprocess.run(['systemd-run','--unit='+UNIT,'--property=WorkingDirectory='+str(Path(__file__).parent),'--property=CPUQuota=50%','--property=MemoryMax=512M','--property=Nice=15','--property=IOWeight=10','--property=StandardOutput=append:/data/paper-index/arxiv-import.log','--property=StandardError=append:/data/paper-index/arxiv-import.log','/usr/bin/env','PYTHONDONTWRITEBYTECODE=1','python3','-B',str(Path(__file__).with_name('import_arxiv.py')),'--db',db,'--archive',archive],check=True) + if property_value('ActiveState')!='active': + raise RuntimeError('arXiv unit did not resume; inspect the logs') + print('ARXIV_RESUMED',flush=True) + + +if __name__=='__main__': + parser=argparse.ArgumentParser() + parser.add_argument('--db',required=True) + parser.add_argument('--archive',required=True) + parser.add_argument('files',nargs='+') + args=parser.parse_args() + merge(args.db,args.archive,args.files) diff --git a/query.py b/query.py index b293291..c322f43 100644 --- a/query.py +++ b/query.py @@ -9,6 +9,7 @@ def main(): parser = argparse.ArgumentParser() parser.add_argument('--db', required=True) parser.add_argument('--query', help='SQLite FTS5 query over title and abstract') + parser.add_argument('--paper',type=int,help='Paper ID; include sources, tags and submission history') parser.add_argument('--tag', help='Filter as namespace=value, e.g. arxiv:category=cs.AI') parser.add_argument('--venue', help='Conference acronym') parser.add_argument('--year', type=int) @@ -18,34 +19,50 @@ def main(): args = parser.parse_args() conn = sqlite3.connect(Path(args.db).resolve().as_uri()+'?mode=ro', uri=True) conn.row_factory = sqlite3.Row + conn.execute('BEGIN') # One consistent read snapshot while ingestion continues. if args.stats: result = {name:conn.execute('SELECT count(*) FROM '+name).fetchone()[0] for name in ('papers','source_records','submissions','decisions','tags','match_candidates')} + result['records_by_source'] = [dict(r) for r in conn.execute('SELECT s.name,count(r.id) AS imported_records FROM sources s LEFT JOIN source_records r ON r.source_id=s.id GROUP BY s.id ORDER BY s.name')] + result['decision_events_by_status'] = [dict(r) for r in conn.execute('SELECT status,count(*) AS events FROM decisions GROUP BY status')] result['coverage'] = [dict(r) for r in conn.execute('SELECT status,count(*) AS sources FROM edition_sources GROUP BY status')] result['runs'] = [dict(r) for r in conn.execute('SELECT * FROM ingest_runs ORDER BY id DESC LIMIT 10')] else: where = []; params = [] + if args.paper is not None: + where.append('p.id=?'); params.append(args.paper) if args.query: where.append('p.id IN (SELECT rowid FROM papers_fts WHERE papers_fts MATCH ?)'); params.append(args.query) if args.tag: namespace,value = args.tag.split('=',1) where.append('EXISTS(SELECT 1 FROM paper_tags pt JOIN tags t ON t.id=pt.tag_id WHERE pt.paper_id=p.id AND t.namespace=? AND t.value=?)'); params.extend([namespace,value]) if args.venue or args.year or args.status: - clause = ['s.paper_id=p.id'] + clause = ['1=1'] if args.venue: - clause.append('v.acronym=?'); params.append(args.venue) + clause.append('v.acronym=? COLLATE NOCASE'); params.append(args.venue) if args.year: clause.append('e.year=?'); params.append(args.year) if args.status: clause.append('EXISTS(SELECT 1 FROM decisions d WHERE d.submission_id=s.id AND d.status=? AND NOT EXISTS(SELECT 1 FROM decisions newer WHERE newer.submission_id=d.submission_id AND (newer.observed_at>d.observed_at OR (newer.observed_at=d.observed_at AND newer.id>d.id))))'); params.append(args.status) - where.append('EXISTS(SELECT 1 FROM submissions s JOIN editions e ON e.id=s.edition_id JOIN venues v ON v.id=e.venue_id WHERE '+' AND '.join(clause)+')') + where.append('p.id IN (SELECT s.paper_id FROM submissions s JOIN editions e ON e.id=s.edition_id JOIN venues v ON v.id=e.venue_id WHERE '+' AND '.join(clause)+')') sql = 'SELECT p.* FROM papers p'+(' WHERE '+' AND '.join(where) if where else '')+' ORDER BY p.id LIMIT ?' params.append(max(1,min(args.limit,200))) result = [] for row in conn.execute(sql,params): paper = dict(row) + paper['authors'] = json.loads(paper.pop('authors_json')) paper['identifiers'] = [dict(r) for r in conn.execute('SELECT scheme,value FROM identifiers WHERE paper_id=?',(row['id'],))] + if args.paper is not None: + paper['tags'] = [dict(r) for r in conn.execute('SELECT t.namespace,t.value,pt.evidence_key FROM paper_tags pt JOIN tags t ON t.id=pt.tag_id WHERE pt.paper_id=?',(row['id'],))] + paper['sources'] = [dict(r) for r in conn.execute('SELECT s.name,r.external_id,r.url,r.title,r.abstract,r.retrieved_at FROM source_records r JOIN sources s ON s.id=r.source_id WHERE r.paper_id=?',(row['id'],))] + submissions = [] + for sub in conn.execute('SELECT s.id,s.source_submission_id,s.submitted_at,v.acronym,e.year,e.track FROM submissions s JOIN editions e ON e.id=s.edition_id JOIN venues v ON v.id=e.venue_id WHERE s.paper_id=? ORDER BY e.year,s.id',(row['id'],)): + submission = dict(sub) + submission['decisions'] = [dict(d) for d in conn.execute('SELECT status,raw_value,evidence_url,observed_at,decision_date FROM decisions WHERE submission_id=? ORDER BY observed_at,id',(sub['id'],))] + submissions.append(submission) + paper['submissions'] = submissions result.append(paper) print(json.dumps(result,ensure_ascii=False,indent=2)) + conn.close() if __name__ == '__main__': diff --git a/schema.sql b/schema.sql index 8a93f24..58c947d 100644 --- a/schema.sql +++ b/schema.sql @@ -53,6 +53,7 @@ CREATE TABLE IF NOT EXISTS decisions ( raw_value TEXT NOT NULL, evidence_url TEXT NOT NULL, observed_at TEXT NOT NULL, decision_date TEXT, UNIQUE(submission_id,status,raw_value,evidence_url,observed_at) ); +CREATE INDEX IF NOT EXISTS decisions_submission_order ON decisions(submission_id,observed_at DESC,id DESC); CREATE TABLE IF NOT EXISTS tags ( id INTEGER PRIMARY KEY, namespace TEXT NOT NULL, value TEXT NOT NULL, description TEXT, UNIQUE(namespace,value) diff --git a/test_pmlr.py b/test_pmlr.py new file mode 100644 index 0000000..35122f5 --- /dev/null +++ b/test_pmlr.py @@ -0,0 +1,25 @@ +import unittest + +from collect_pmlr import load_metadata,venue_key + + +class MetadataTests(unittest.TestCase): + def test_legacy_binary_scalar_is_decoded_as_text(self): + self.assertEqual(load_metadata('title: !binary |\n SGVsbG8=\n')['title'],'Hello') + + def test_trailing_empty_document_is_allowed(self): + self.assertEqual(load_metadata('---\nyear: 2024\n---\n'),{'year':2024}) + + def test_conflicting_documents_are_not_silently_combined(self): + with self.assertRaisesRegex(ValueError,'exactly one'): + load_metadata('year: 2024\n---\nyear: 2025\n') + + def test_historical_conference_name_aliases(self): + self.assertEqual(venue_key('AISTATS 2010'),'AISTATS') + self.assertEqual(venue_key('AISTATS 2012'),'AISTATS') + self.assertEqual(venue_key('COLT 2012'),'COLT') + self.assertEqual(venue_key('ECML-PKDD'),'ECML-PKDD') + + +if __name__=='__main__': + unittest.main() diff --git a/test_query.py b/test_query.py new file mode 100644 index 0000000..0b83b50 --- /dev/null +++ b/test_query.py @@ -0,0 +1,39 @@ +import json +import subprocess +import sys +import tempfile +import unittest +from pathlib import Path + +from index_core import Store + + +class QueryTests(unittest.TestCase): + def test_venue_filter_is_case_insensitive_and_uses_latest_decision(self): + with tempfile.TemporaryDirectory() as tmp: + db=Path(tmp)/'papers.sqlite' + store=Store(db) + self.addCleanup(store.conn.close) + paper,record=store.record('peerread','paper1','https://example.org/paper1','A real submission','Abstract',['Jane Doe']) + store.decision(paper,record,'ICLR',2017,'main','paper1','https://example.org/paper1','rejected','explicit rejection') + store.record('arxiv','2601.00001','https://example.org/other','Another paper','Abstract',['John Smith']) + store.conn.commit() + command=[sys.executable,'-B',str(Path(__file__).with_name('query.py')),'--db',str(db),'--venue','iclr','--year','2017','--status','rejected'] + result=json.loads(subprocess.run(command,check=True,capture_output=True,text=True).stdout) + self.assertEqual([r['id'] for r in result],[paper]) + details_command=[sys.executable,'-B',str(Path(__file__).with_name('query.py')),'--db',str(db),'--paper',str(paper)] + details=json.loads(subprocess.run(details_command,check=True,capture_output=True,text=True).stdout)[0] + self.assertEqual(details['authors'],['Jane Doe']) + self.assertEqual(details['submissions'][0]['decisions'][0]['status'],'rejected') + self.assertEqual(details['sources'][0]['external_id'],'paper1') + stats_command=[sys.executable,'-B',str(Path(__file__).with_name('query.py')),'--db',str(db),'--stats'] + stats=json.loads(subprocess.run(stats_command,check=True,capture_output=True,text=True).stdout) + self.assertEqual(stats['papers'],2) + self.assertEqual({r['name']:r['imported_records'] for r in stats['records_by_source']},{'arxiv':1,'peerread':1}) + store.decision(paper,record,'ICLR',2017,'main','paper1','https://example.org/paper1','accepted','changed decision') + store.conn.commit() + self.assertEqual(json.loads(subprocess.run(command,check=True,capture_output=True,text=True).stdout),[]) + + +if __name__=='__main__': + unittest.main() -- 2.52.0 From a3539be4b18220e7eaaac64029dd69a1041bb9e6 Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 11:39:16 +0800 Subject: [PATCH 05/20] Index per-source tag refreshes for scalable bulk ingestion --- README.md | 2 +- merge_conferences.py | 19 +++++++++++++++---- schema.sql | 1 + 3 files changed, 17 insertions(+), 5 deletions(-) diff --git a/README.md b/README.md index 3362078..3c44c31 100644 --- a/README.md +++ b/README.md @@ -121,7 +121,7 @@ python3 -B collect_peerread.py --cache /data/paper-index/raw/http-cache \ --out /data/paper-index/collected ``` -数据任务以一次性 systemd 服务运行：`paper-atlas-arxiv-import`、`paper-atlas-pmlr-collect`、`paper-atlas-peerread-collect`；`paper-atlas-pmlr-verify` 用于全索引复核，`paper-atlas-conference-merge` 用于单写入合并。它们不是定时任务。主库只允许一个写入进程；会议采集先写独立 JSONL，待 arXiv 主批次结束后再合并；也可以使用 `merge_conferences.py` 在保存断点处暂停本项目的 arXiv unit、顺序合并会议清单并恢复。该操作应作为独立的一次性 systemd 任务运行，避免 SSH 断开影响恢复；不能同时启动多个合并任务。运行日志位于 `/data/paper-index/*-import.log`、`/data/paper-index/*-collect.log`。任何服务启动成功都不等于批次已完成，需检查批次状态、清单与完整性结果。 +数据任务以一次性 systemd 服务运行：`paper-atlas-arxiv-import`、`paper-atlas-pmlr-collect`、`paper-atlas-peerread-collect`；`paper-atlas-pmlr-verify` 用于全索引复核，`paper-atlas-conference-merge` 用于单写入合并。它们不是定时任务。主库只允许一个写入进程；会议采集先写独立 JSONL，待 arXiv 主批次结束后再合并；也可以使用 `merge_conferences.py` 在保存断点处暂停本项目的 arXiv unit、顺序合并会议清单并恢复。该操作应作为独立的一次性 systemd 任务运行，避免 SSH 断开影响恢复；不能同时启动多个合并任务。运行日志位于 `/data/paper-index/*-import.log`、`/data/paper-index/*-collect.log`。任何服务启动成功都不等于批次已完成，需检查批次状态、清单与完整性结果。标签按来源刷新使用 `paper_tags_source_evidence` 索引，避免每条论文都扫描整张标签表。运行中的 arXiv 可以通过同一协调脚本的 `--schema-only` 模式在断点更新索引后恢复；此模式要求已有 arXiv unit 正在运行。 ## 长期维护与更新 diff --git a/merge_conferences.py b/merge_conferences.py index a70098a..c869b14 100644 --- a/merge_conferences.py +++ b/merge_conferences.py @@ -10,6 +10,7 @@ import subprocess from pathlib import Path from import_records import ingest_file +from index_core import Store UNIT='paper-atlas-arxiv-import.service' @@ -27,6 +28,8 @@ def merge(db,archive,files): raise ValueError('Missing collected file or manifest: '+filename) if not Path(archive).is_file(): raise ValueError('Missing arXiv archive') + if not Path(db).is_file(): + raise ValueError('Missing existing database') if property_value('ActiveState')!='active': raise RuntimeError('Expected the existing arXiv unit to be active; inspect it before proceeding') subprocess.run(['systemctl','stop',UNIT],check=True) @@ -34,12 +37,17 @@ def merge(db,archive,files): with sqlite3.connect(db) as c: checkpoint=c.execute("SELECT records_written FROM ingest_runs WHERE source_id=(SELECT id FROM sources WHERE name='arxiv') ORDER BY id DESC LIMIT 1").fetchone() print('Saved arxiv checkpoint:',checkpoint,flush=True) + schema_store=Store(db) + schema_store.conn.close() for filename in files: ingest_file(db,filename) with sqlite3.connect(db) as c: - errors=c.execute('PRAGMA foreign_key_check').fetchall() - if errors: - raise RuntimeError('Foreign key violations after merge') + if files: + errors=c.execute('PRAGMA foreign_key_check').fetchall() + if errors: + raise RuntimeError('Foreign key violations after merge') + else: + print('Tag refresh plan:',c.execute('EXPLAIN QUERY PLAN DELETE FROM paper_tags WHERE source_record_id=? AND evidence_key=?',(1,'arxiv:categories')).fetchall(),flush=True) print('Decision totals:',c.execute('SELECT status,count(*) FROM decisions GROUP BY status').fetchall(),flush=True) finally: if property_value('LoadState')=='loaded': @@ -55,6 +63,9 @@ if __name__=='__main__': parser=argparse.ArgumentParser() parser.add_argument('--db',required=True) parser.add_argument('--archive',required=True) - parser.add_argument('files',nargs='+') + parser.add_argument('--schema-only',action='store_true',help='Apply idempotent schema/index updates at a saved checkpoint') + parser.add_argument('files',nargs='*') args=parser.parse_args() + if not args.files and not args.schema_only: + parser.error('Provide collected files or --schema-only') merge(args.db,args.archive,args.files) diff --git a/schema.sql b/schema.sql index 58c947d..09a73ec 100644 --- a/schema.sql +++ b/schema.sql @@ -63,6 +63,7 @@ CREATE TABLE IF NOT EXISTS paper_tags ( source_record_id INTEGER REFERENCES source_records(id), evidence_key TEXT NOT NULL, PRIMARY KEY(paper_id,tag_id,evidence_key) ); +CREATE INDEX IF NOT EXISTS paper_tags_source_evidence ON paper_tags(source_record_id,evidence_key); CREATE TABLE IF NOT EXISTS match_candidates ( left_paper_id INTEGER NOT NULL REFERENCES papers(id), right_paper_id INTEGER NOT NULL REFERENCES papers(id), method TEXT NOT NULL, score REAL, evidence_json TEXT NOT NULL, -- 2.52.0 From ad9967c1b21e26fee3565617cd26e3aae2c61b82 Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 12:04:18 +0800 Subject: [PATCH 06/20] Collect official COLM and NeurIPS abstracts with track-aware coverage --- README.md | 19 +++- collect_pmlr.py | 13 ++- collect_proceedings.py | 211 +++++++++++++++++++++++++++++++++++++++++ import_records.py | 3 +- merge_conferences.py | 5 +- test_import_arxiv.py | 9 +- test_index_core.py | 7 ++ test_pmlr.py | 13 ++- test_proceedings.py | 56 +++++++++++ 9 files changed, 323 insertions(+), 13 deletions(-) create mode 100644 collect_proceedings.py create mode 100644 test_proceedings.py diff --git a/README.md b/README.md index 3c44c31..5b35682 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 目标包括：完整历史元数据导入、逐年会议目录、跨平台论文去重、公开拒稿/撤稿信息、平台标签与自定义标签隔离、可恢复更新和覆盖率报告。 -**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR 官方元数据采集、带清单计数核验的会议导入和只读查询脚本。服务器上持续运行合成回归检查；完整 ICML 2025 元数据已采集 3,330 条，尚未作为全量交付。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），全量导入正在服务器运行。新增 PeerRead 公开 ICLR 2017 明确决定采集，其他会议适配器和全量历史导入尚未完成。其他会议适配器、跨平台疑似重复复核流程、完整覆盖率核验仍待实现。已配置 ces-control:host 对应的远程 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 +**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR、COLM、NeurIPS 官方元数据采集、带清单计数核验的会议导入和只读查询脚本。服务器上持续运行合成回归检查；完整 ICML 2025 元数据已采集 3,330 条，尚未作为全量交付。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），全量导入正在服务器运行。新增 PeerRead 公开 ICLR 2017 明确决定采集，其他会议适配器和全量历史导入尚未完成。其他会议适配器、跨平台疑似重复复核流程、完整覆盖率核验仍待实现。已配置 ces-control:host 对应的远程 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 ## 数据范围与来源 @@ -123,6 +123,23 @@ python3 -B collect_peerread.py --cache /data/paper-index/raw/http-cache \ 数据任务以一次性 systemd 服务运行：`paper-atlas-arxiv-import`、`paper-atlas-pmlr-collect`、`paper-atlas-peerread-collect`；`paper-atlas-pmlr-verify` 用于全索引复核，`paper-atlas-conference-merge` 用于单写入合并。它们不是定时任务。主库只允许一个写入进程；会议采集先写独立 JSONL，待 arXiv 主批次结束后再合并；也可以使用 `merge_conferences.py` 在保存断点处暂停本项目的 arXiv unit、顺序合并会议清单并恢复。该操作应作为独立的一次性 systemd 任务运行，避免 SSH 断开影响恢复；不能同时启动多个合并任务。运行日志位于 `/data/paper-index/*-import.log`、`/data/paper-index/*-collect.log`。任何服务启动成功都不等于批次已完成，需检查批次状态、清单与完整性结果。标签按来源刷新使用 `paper_tags_source_evidence` 索引，避免每条论文都扫描整张标签表。运行中的 arXiv 可以通过同一协调脚本的 `--schema-only` 模式在断点更新索引后恢复；此模式要求已有 arXiv unit 正在运行。 +### COLM 与 NeurIPS 官方摘要 + +```sh +python3 -B collect_proceedings.py --source colm \ + --cache /data/paper-index/raw/http-cache --out /data/paper-index/collected +python3 -B collect_proceedings.py --source neurips \ + --cache /data/paper-index/raw/http-cache --out /data/paper-index/collected +``` + +采集器从官方根目录发现实际发表届次，仅访问索引和 Abstract HTML 页面，不请求 PDF、补充材料或源码。可用重复 `--year` 限定实际存在的届次；不存在的年份会报错，局部重试保留其他届次的已有状态。每个来源串行请求，实际网络响应后等待 1 秒；缓存命中不额外等待。 + +COLM 当前目录提供 2024–2026 三届 pre-proceedings，按会议年份保存，保留平台实际出版日期和出版状态，不把晚补刊日期当作会议年份。NeurIPS 当前目录提供 1987–2025，并包含 2025 独立卷及 2021 独立 Datasets and Benchmarks 索引。主会、datasets_benchmarks、position、creative_ai 分开出文件，不混为同一个 track。首次真实样本核对：COLM 2024 官方目录 299 个摘要入口；NeurIPS 1987 为 90 个，作者与摘要均已验证解析。样本成功不代表所有页面采集完成。 + +`colm-index-status.json`、`neurips-index-status.json` 保存逐索引状态；对应 `*-last-run.json` 保存本次实际选择范围。每个 JSONL 的 manifest 记录索引入口数、track 记录数和缺失摘要数。论文数量与官方页面声明不一致会失败，缺失摘要会标记 partial；不会把页脚、姓名政策或其他正文当作摘要。 + +这两个采集器只写采集文件，合并仍通过单写入协调流程。全量历史覆盖和所有未实现会议仍待最终核验。 + ## 长期维护与更新 1. 每次采集先核对来源、许可、可访问性、空闲空间和现有运行批次。保持低并发，遵守来源限流、重试与人机验证要求。 diff --git a/collect_pmlr.py b/collect_pmlr.py index f43cfe3..2ba97c7 100644 --- a/collect_pmlr.py +++ b/collect_pmlr.py @@ -38,7 +38,7 @@ def venue_key(shortname): return re.sub(r'[\s_-]*(?:19|20)\d{2}$','',str(shortname).strip().upper()) -def fetch(url,cache,limit=64*1024**2): +def fetch(url,cache,limit=64*1024**2,delay=0.5): cache = Path(cache) cache.mkdir(parents=True,exist_ok=True) path = cache/hashlib.sha256(url.encode()).hexdigest() @@ -59,7 +59,7 @@ def fetch(url,cache,limit=64*1024**2): part.write_bytes(data) part.replace(path) path.with_suffix('.json').write_text(json.dumps({'url':url,'sha256':hashlib.sha256(data).hexdigest(),'bytes':len(data),'retrieved_at':time.strftime('%Y-%m-%dT%H:%M:%SZ',time.gmtime())})) - time.sleep(0.5) + time.sleep(delay) return data except urllib.error.HTTPError as exc: if exc.code not in (429,500,502,503,504) or attempt==2: @@ -128,14 +128,19 @@ def collect_volume(volume,cache,out,allowed=None): path.parent.mkdir(parents=True,exist_ok=True) part = path.with_suffix('.part') count = 0 + missing_abstracts = 0 with part.open('w') as output: for member_name,content in metadata_texts(volume,cache): parts = re.split(r'(?m)^---\s*$',content,maxsplit=2) if len(parts)<3 or parts[0].strip(): raise RuntimeError('Missing YAML frontmatter: '+member_name) row = load_metadata(parts[1]) - if row.get('layout') not in ('inproceedings','article') or not row.get('title'): + if row.get('layout') not in ('inproceedings','article'): continue + if not row.get('title'): + raise ValueError('Paper metadata has no title: '+member_name) + if not (row.get('abstract') or '').strip(): + missing_abstracts+=1 external = f"v{volume}/{row.get('id') or Path(member_name).stem}" ids = {} for key in ('doi','openreview','arxiv'): @@ -154,7 +159,7 @@ def collect_volume(volume,cache,out,allowed=None): part.unlink(missing_ok=True) raise RuntimeError('No papers parsed; refusing an empty successful collection') part.replace(path) - result = {'source':'pmlr','venue':venue,'year':year,'track':track,'volume':volume,'count':count,'list_url':f'https://proceedings.mlr.press/v{volume}/','file':str(path),'status':'collected'} + result = {'source':'pmlr','venue':venue,'year':year,'track':track,'volume':volume,'count':count,'missing_abstract_count':missing_abstracts,'list_url':f'https://proceedings.mlr.press/v{volume}/','file':str(path),'status':'partial' if missing_abstracts else 'collected'} path.with_suffix('.manifest.json').write_text(json.dumps(result,indent=2)) print(json.dumps(result),flush=True) return result diff --git a/collect_proceedings.py b/collect_proceedings.py new file mode 100644 index 0000000..19917c4 --- /dev/null +++ b/collect_proceedings.py @@ -0,0 +1,211 @@ +"""Collect COLM/NeurIPS official abstract pages without downloading PDFs.""" +import argparse +import json +import re +import time +from collections import defaultdict +from html.parser import HTMLParser +from pathlib import Path +from urllib.parse import urljoin,urlsplit + +from collect_pmlr import fetch + +ROOTS={'colm':'https://proceedings.colm.cc/','neurips':'https://proceedings.neurips.cc/'} + + +class Node: + def __init__(self,tag='',attrs=None): + self.tag=tag + self.attrs=dict(attrs or []) + self.children=[] + + def text(self): + if self.tag in ('script','style'): + return '' + content=''.join(child.text() if isinstance(child,Node) else child for child in self.children) + return ' '+content+' ' if self.tag in ('p','div','section','li','br') else content + + def walk(self): + yield self + for child in self.children: + if isinstance(child,Node): + yield from child.walk() + + +class Document(HTMLParser): + VOID={'area','base','br','col','embed','hr','img','input','link','meta','param','source','track','wbr'} + + def __init__(self,html): + super().__init__(convert_charrefs=True) + self.root=Node() + self.stack=[self.root] + self.feed(html) + + def handle_starttag(self,tag,attrs): + node=Node(tag,attrs) + self.stack[-1].children.append(node) + if tag not in self.VOID: + self.stack.append(node) + + def handle_endtag(self,tag): + for index in range(len(self.stack)-1,0,-1): + if self.stack[index].tag==tag: + del self.stack[index:] + break + + def handle_data(self,data): + self.stack[-1].children.append(data) + + +def clean(text): + return ' '.join(text.split()) + + +def track_key(text): + text=text.lower() + if 'dataset' in text or 'benchmark' in text: + return 'datasets_benchmarks' + if 'creative' in text: + return 'creative_ai' + if 'position' in text: + return 'position' + return 'main' + + +def paper_links(html,index_url): + doc=Document(html) + links=sorted({urljoin(index_url,n.attrs['href']) for n in doc.root.walk() if n.tag=='a' and '-Abstract' in n.attrs.get('href','') and urlsplit(urljoin(index_url,n.attrs['href'])).path.endswith('.html') and urlsplit(urljoin(index_url,n.attrs['href'])).netloc==urlsplit(index_url).netloc}) + declared=next((match for n in doc.root.walk() if n.tag in ('p','span','div') and not any(child.tag=='a' for child in n.walk()) and (match:=re.fullmatch(r'([\d,]+)\s+papers',clean(n.text()),re.I))),None) + if not links: + raise ValueError('Official index contains no abstract-page links') + if declared and int(declared[1].replace(',',''))!=len(links): + raise ValueError('Declared paper count differs from unique abstract-page links') + return links + + +def parse_paper(html,url,source,year,index_url): + nodes=list(Document(html).root.walk()) + meta=defaultdict(list) + for node in nodes: + if node.tag=='meta': + meta[node.attrs.get('name','').lower()].append(node.attrs.get('content','')) + title=next(iter(meta['citation_title']),None) + if not title: + title=next((clean(n.text()) for n in nodes if n.tag=='h1'),None) + abstract=next((clean(n.text()) for n in nodes if 'paper-abstract' in n.attrs.get('class','').split()),None) + if not abstract: + for index,node in enumerate(nodes): + if node.tag in ('h2','h3','h4','h5') and clean(node.text()).lower()=='abstract': + following=[] + covered=set() + for candidate in nodes[index+1:]: + if candidate.tag in ('h1','h2','h3','h4','h5'): + break + if candidate.tag=='p' and candidate not in covered: + following.append(clean(candidate.text())) + covered.update(candidate.walk()) + abstract=' '.join(following).strip() or None + break + if not title: + raise ValueError('Missing title on '+url) + # Never fall back to footer/policy text when the abstract is absent. + authors=[] + for author in meta['citation_author']: + if ',' in author: + family,given=author.split(',',1) + author=given.strip()+' '+family.strip() + authors.append(clean(author)) + track_text=' '.join(clean(n.text()) for n in nodes if 'paper-track' in n.attrs.get('class','').split()) + track=track_key(track_text or url) + identifiers={} + if meta['citation_doi']: + identifiers['doi']=meta['citation_doi'][0] + for node in nodes: + if node.tag!='a': + continue + link=urljoin(url,node.attrs.get('href','')) + parsed=urlsplit(link) + if parsed.netloc in ('arxiv.org','www.arxiv.org') and re.match(r'^/(abs|pdf)/',parsed.path): + identifiers['arxiv']=re.sub(r'^/(abs|pdf)/','',parsed.path).removesuffix('.pdf') + if parsed.netloc=='openreview.net' and parsed.path=='/forum': + from urllib.parse import parse_qs + forum=parse_qs(parsed.query).get('id') + if forum: identifiers['openreview']=forum[0] + venue='COLM' if source=='colm' else 'NeurIPS' + return {'source':source,'external_id':urlsplit(url).path,'url':url,'title':clean(title),'abstract':abstract,'authors':authors,'identifiers':identifiers,'venue':venue,'year':year,'track':track,'status':'accepted','raw_decision':'listed in official proceedings','list_url':index_url,'metadata':{'citation':dict(meta),'publication_state':'pre-proceedings' if 'pre-proceedings' in html.lower() else 'proceedings'}} + + +def discover(source,cache): + root=ROOTS[source] + html=fetch(root,cache,8*1024**2,delay=1.0).decode() + result={} + for node in Document(html).root.walk(): + if node.tag!='a':continue + url=urljoin(root,node.attrs.get('href','')) + parsed=urlsplit(url) + match=re.search(r'/paper(?:_files/paper)?/((?:19|20)\d{2})(?:/|$)',parsed.path) + if parsed.netloc==urlsplit(root).netloc and match and not any(term in parsed.path for term in ('/hash/','/file/','.pdf','-Abstract')): + result[url]=int(match[1]) + if source=='neurips' and parsed.netloc=='datasets-benchmarks-proceedings.neurips.cc' and match: + result[url]=int(match[1]) + if not result: + raise ValueError('No official year indexes found') + return sorted(result.items(),key=lambda entry:(entry[1],entry[0])) + + +def collect_index(source,index_url,year,cache,out): + html=fetch(index_url,cache,16*1024**2,delay=1.0).decode() + links=paper_links(html,index_url) + grouped=defaultdict(list) + missing=[] + for number,url in enumerate(links,1): + record=parse_paper(fetch(url,cache,2*1024**2,delay=1.0).decode(),url,source,year,index_url) + if not record['abstract']:missing.append(url) + grouped[record['track']].append(record) + if number%25==0: + print(json.dumps({'source':source,'year':year,'collected':number,'expected':len(links)}),flush=True) + results=[] + for track,records in grouped.items(): + # Distinct official index URLs may publish multiple volumes in one year. + import hashlib + suffix=hashlib.sha256(index_url.encode()).hexdigest()[:10] + path=Path(out)/f'{source}-{year}-{track}-{suffix}.jsonl' + path.parent.mkdir(parents=True,exist_ok=True) + part=path.with_suffix('.part') + with part.open('w') as output: + for record in records:output.write(json.dumps(record,ensure_ascii=False)+'\n') + part.replace(path) + missing_count=sum(not r['abstract'] for r in records) + result={'source':source,'venue':records[0]['venue'],'year':year,'track':track,'count':len(records),'index_link_count':len(links),'index_url':index_url,'list_url':index_url,'missing_abstract_count':missing_count,'file':str(path),'status':'collected' if not missing_count else 'partial'} + path.with_suffix('.manifest.json').write_text(json.dumps(result,indent=2)) + results.append(result) + print(json.dumps(result),flush=True) + return results + + +if __name__=='__main__': + parser=argparse.ArgumentParser() + parser.add_argument('--source',choices=ROOTS,required=True) + parser.add_argument('--cache',required=True) + parser.add_argument('--out',required=True) + parser.add_argument('--year',type=int,action='append') + args=parser.parse_args() + out=Path(args.out);out.mkdir(parents=True,exist_ok=True) + status_path=out/f'{args.source}-index-status.json' + catalogue=json.loads(status_path.read_text()) if status_path.exists() else {} + indexes=discover(args.source,args.cache) + if args.year and set(args.year)-{year for _,year in indexes}: + parser.error('Requested year is absent from the current official proceedings index') + requested=[] + for url,year in indexes: + if args.year and year not in args.year:continue + requested.append(url) + try: + files=collect_index(args.source,url,year,args.cache,args.out) + catalogue[url]={'status':'partial' if any(f['status']=='partial' for f in files) else 'collected','files':files} + except Exception as exc: + catalogue[url]={'status':'blocked','error':str(exc),'year':year} + print(json.dumps(catalogue[url]),flush=True) + part=status_path.with_suffix('.part');part.write_text(json.dumps(catalogue,indent=2));part.replace(status_path) + (out/f'{args.source}-last-run.json').write_text(json.dumps({'requested_indexes':requested,'year_filter':args.year},indent=2)) + if any(catalogue[url]['status']=='blocked' for url in requested):raise SystemExit(1) diff --git a/import_records.py b/import_records.py index 02ed87b..2b45792 100644 --- a/import_records.py +++ b/import_records.py @@ -47,7 +47,8 @@ def ingest_file(db,path): expected = manifest.get('count') if expected is not None and expected != count: raise ValueError(f'Manifest count {expected} differs from records {count}') - c.execute('INSERT INTO edition_sources(edition_id,source_id,list_url,status,expected_count,imported_count,last_success_at) VALUES(?,?,?,?,?,?,?) ON CONFLICT(edition_id,source_id,list_url) DO UPDATE SET status=excluded.status,expected_count=excluded.expected_count,imported_count=excluded.imported_count,last_success_at=excluded.last_success_at,last_error=NULL',(edition,source_id,manifest['list_url'],'complete' if expected==count else 'partial',expected,count,time.strftime('%Y-%m-%dT%H:%M:%SZ',time.gmtime()))) + complete=expected==count and manifest.get('status') not in ('partial','blocked') and not manifest.get('missing_abstract_count',0) + c.execute('INSERT INTO edition_sources(edition_id,source_id,list_url,status,expected_count,imported_count,last_success_at) VALUES(?,?,?,?,?,?,?) ON CONFLICT(edition_id,source_id,list_url) DO UPDATE SET status=excluded.status,expected_count=excluded.expected_count,imported_count=excluded.imported_count,last_success_at=excluded.last_success_at,last_error=NULL',(edition,source_id,manifest['list_url'],'complete' if complete else 'partial',expected,count,time.strftime('%Y-%m-%dT%H:%M:%SZ',time.gmtime()))) c.execute("UPDATE ingest_runs SET status='complete',finished_at=CURRENT_TIMESTAMP,records_seen=?,records_written=? WHERE id=?",(count,count,run)) c.commit() print(json.dumps({'file':str(path),'status':'imported','records':count}),flush=True) diff --git a/merge_conferences.py b/merge_conferences.py index c869b14..d971af9 100644 --- a/merge_conferences.py +++ b/merge_conferences.py @@ -4,6 +4,7 @@ Run as a transient systemd job so an SSH disconnect cannot strand the importer. Only the Paper Atlas arXiv unit is controlled; business services are untouched. """ import argparse +from contextlib import closing import json import sqlite3 import subprocess @@ -34,14 +35,14 @@ def merge(db,archive,files): raise RuntimeError('Expected the existing arXiv unit to be active; inspect it before proceeding') subprocess.run(['systemctl','stop',UNIT],check=True) try: - with sqlite3.connect(db) as c: + with closing(sqlite3.connect(db)) as c, c: checkpoint=c.execute("SELECT records_written FROM ingest_runs WHERE source_id=(SELECT id FROM sources WHERE name='arxiv') ORDER BY id DESC LIMIT 1").fetchone() print('Saved arxiv checkpoint:',checkpoint,flush=True) schema_store=Store(db) schema_store.conn.close() for filename in files: ingest_file(db,filename) - with sqlite3.connect(db) as c: + with closing(sqlite3.connect(db)) as c, c: if files: errors=c.execute('PRAGMA foreign_key_check').fetchall() if errors: diff --git a/test_import_arxiv.py b/test_import_arxiv.py index 4c2c6e8..d973cd4 100644 --- a/test_import_arxiv.py +++ b/test_import_arxiv.py @@ -1,4 +1,5 @@ import json +from contextlib import closing import sqlite3 import tempfile import unittest @@ -27,7 +28,7 @@ class ImportTests(unittest.TestCase): self.write([self.row('2601.00001','10.0000/example'),self.row('2601.00002','10.0000/example')]) ingest(self.db,self.archive) ingest(self.db,self.archive) - with sqlite3.connect(self.db) as c: + with closing(sqlite3.connect(self.db)) as c, c: self.assertEqual(c.execute('SELECT count(*) FROM papers').fetchone()[0],1) self.assertEqual(c.execute('SELECT count(*) FROM source_records').fetchone()[0],2) self.assertEqual(c.execute("SELECT count(*) FROM papers_fts WHERE papers_fts MATCH 'Searchable'").fetchone()[0],1) @@ -38,18 +39,18 @@ class ImportTests(unittest.TestCase): def test_title_alone_does_not_merge(self): self.write([self.row('2601.00001'),self.row('2601.00002')]) ingest(self.db,self.archive) - with sqlite3.connect(self.db) as c: + with closing(sqlite3.connect(self.db)) as c, c: self.assertEqual(c.execute('SELECT count(*) FROM papers').fetchone()[0],2) def test_platform_and_custom_tags_remain_distinct(self): self.write([self.row('2601.00001')]) ingest(self.db,self.archive) - with sqlite3.connect(self.db) as c: + with closing(sqlite3.connect(self.db)) as c, c: paper = c.execute('SELECT id FROM papers').fetchone()[0] tag = c.execute("INSERT INTO tags(namespace,value) VALUES('custom:topic','cs.AI')").lastrowid c.execute('INSERT INTO paper_tags VALUES(?,?,NULL,?)',(paper,tag,'manual:1')) ingest(self.db,self.archive) - with sqlite3.connect(self.db) as c: + with closing(sqlite3.connect(self.db)) as c, c: self.assertEqual(c.execute("SELECT count(*) FROM paper_tags pt JOIN tags t ON pt.tag_id=t.id WHERE t.value='cs.AI'").fetchone()[0],2) diff --git a/test_index_core.py b/test_index_core.py index 91a5979..29cb122 100644 --- a/test_index_core.py +++ b/test_index_core.py @@ -58,6 +58,13 @@ class StoreTests(unittest.TestCase): self.assertEqual(self.store.conn.execute('SELECT title FROM papers').fetchone()[0],'Updated research title') self.assertEqual(self.store.conn.execute("SELECT count(*) FROM papers_fts WHERE papers_fts MATCH 'searchable'").fetchone()[0],1) + def test_missing_abstract_does_not_become_complete_coverage(self): + path=Path(self.tmp.name)/'partial.jsonl' + path.write_text(json.dumps({'source':'colm','external_id':'paper1','url':'https://example.org/paper1','title':'Paper','abstract':None,'authors':[],'venue':'COLM','year':2024,'track':'main','status':'accepted'})+'\n') + path.with_suffix('.manifest.json').write_text(json.dumps({'source':'colm','list_url':'https://example.org/list','count':1,'status':'partial','missing_abstract_count':1})) + ingest_file(self.db,path) + self.assertEqual(self.store.conn.execute('SELECT status FROM edition_sources').fetchone()[0],'partial') + if __name__=='__main__': unittest.main() diff --git a/test_pmlr.py b/test_pmlr.py index 35122f5..5713a93 100644 --- a/test_pmlr.py +++ b/test_pmlr.py @@ -1,6 +1,8 @@ import unittest +import tempfile +from unittest.mock import patch -from collect_pmlr import load_metadata,venue_key +from collect_pmlr import load_metadata,venue_key,collect_volume class MetadataTests(unittest.TestCase): @@ -20,6 +22,15 @@ class MetadataTests(unittest.TestCase): self.assertEqual(venue_key('COLT 2012'),'COLT') self.assertEqual(venue_key('ECML-PKDD'),'ECML-PKDD') + def test_missing_abstract_is_reported_as_partial_collection(self): + config=b'shortname: ICML\nyear: 2025\nbooktitle: ICML\n' + paper='---\nid: paper1\nlayout: inproceedings\ntitle: Paper\nabstract:\n---\n' + with tempfile.TemporaryDirectory() as tmp,patch('collect_pmlr.fetch',return_value=config),patch('collect_pmlr.metadata_texts',return_value=iter([('paper.md',paper)])): + result=collect_volume(999,tmp,tmp) + self.assertEqual(result['count'],1) + self.assertEqual(result['missing_abstract_count'],1) + self.assertEqual(result['status'],'partial') + if __name__=='__main__': unittest.main() diff --git a/test_proceedings.py b/test_proceedings.py new file mode 100644 index 0000000..44c1153 --- /dev/null +++ b/test_proceedings.py @@ -0,0 +1,56 @@ +import unittest +import json +import runpy +import tempfile +from pathlib import Path +from unittest.mock import patch + +from collect_proceedings import paper_links,parse_paper + + +class ProceedingsTests(unittest.TestCase): + def test_colm_nested_abstract_and_citation_authors(self): + html='''

Abstract

Real abstract text.

Name Change Policy

Do not ingest footer.

''' + row=parse_paper(html,'https://proceedings.colm.cc/paper_files/paper/2024/hash/a-Abstract-Conference.html','colm',2024,'https://proceedings.colm.cc/paper_files/paper/2024') + self.assertEqual(row['abstract'],'Real abstract text.') + self.assertEqual(row['authors'],['Jane Doe']) + self.assertEqual(row['track'],'main') + self.assertEqual(row['year'],2024) + + def test_declared_index_count_detects_truncation(self): + html='

2 papers

Paper' + with self.assertRaisesRegex(ValueError,'count differs'): + paper_links(html,'https://proceedings.colm.cc/paper/2024') + + def test_missing_abstract_does_not_capture_footer(self): + row=parse_paper('

Paper

Name Change Policy

Footer

','https://example.org/a','colm',2024,'https://example.org/') + self.assertIsNone(row['abstract']) + + def test_paper_title_is_not_mistaken_for_index_paper_count(self): + html='
  • Lessons from 1000 papers
    • ' + self.assertEqual(len(paper_links(html,'https://datasets-benchmarks-proceedings.neurips.cc/paper/2021')),1) + + def test_neurips_track_and_legacy_abstract(self): + html='

    Paper

    Datasets and Benchmarks Track

    Abstract

    Actual abstract.

    Name Change Policy

    Footer

    ' + row=parse_paper(html,'https://proceedings.neurips.cc/a','neurips',2025,'https://proceedings.neurips.cc/') + self.assertEqual(row['abstract'],'Actual abstract.') + self.assertEqual(row['track'],'datasets_benchmarks') + + def test_focused_retry_preserves_other_index_status(self): + root='https://proceedings.colm.cc/' + url=root+'paper_files/paper/2024' + other=root+'paper_files/paper/2025' + paper=root+'paper_files/paper/2024/hash/a-Abstract-Conference.html' + documents={root:f'20242025',url:f'

    1 papers

    Paper',paper:'

    Paper

    Abstract

    '} + with tempfile.TemporaryDirectory() as tmp: + status=Path(tmp)/'colm-index-status.json' + status.write_text(json.dumps({other:{'status':'collected','files':[]}})) + argv=['collect_proceedings.py','--source','colm','--cache',tmp,'--out',tmp,'--year','2024'] + with patch('sys.argv',argv),patch('collect_pmlr.fetch',side_effect=lambda u,*a,**kw:documents[u].encode()): + runpy.run_path(str(Path(__file__).with_name('collect_proceedings.py')),run_name='__main__') + result=json.loads(status.read_text()) + self.assertEqual(result[other]['status'],'collected') + self.assertEqual(result[url]['files'][0]['count'],1) + + +if __name__=='__main__':unittest.main() -- 2.52.0 From adb4bf8f278951314da4d7dbbd93c2326cc5916a Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 12:35:15 +0800 Subject: [PATCH 07/20] Add authoritative ACL metadata and protect publication identities --- README.md | 22 ++++++- collect_acl.py | 131 +++++++++++++++++++++++++++++++++++++++++ collect_proceedings.py | 9 ++- import_records.py | 4 +- index_core.py | 4 +- metadata_checkout.py | 29 +++++++++ query.py | 10 +++- schema.sql | 4 ++ test_acl.py | 40 +++++++++++++ test_index_core.py | 10 +++- test_proceedings.py | 5 ++ test_query.py | 12 ++++ 12 files changed, 270 insertions(+), 10 deletions(-) create mode 100644 collect_acl.py create mode 100644 metadata_checkout.py create mode 100644 test_acl.py diff --git a/README.md b/README.md index 5b35682..861582c 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 目标包括：完整历史元数据导入、逐年会议目录、跨平台论文去重、公开拒稿/撤稿信息、平台标签与自定义标签隔离、可恢复更新和覆盖率报告。 -**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR、COLM、NeurIPS 官方元数据采集、带清单计数核验的会议导入和只读查询脚本。服务器上持续运行合成回归检查；完整 ICML 2025 元数据已采集 3,330 条，尚未作为全量交付。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），全量导入正在服务器运行。新增 PeerRead 公开 ICLR 2017 明确决定采集，其他会议适配器和全量历史导入尚未完成。其他会议适配器、跨平台疑似重复复核流程、完整覆盖率核验仍待实现。已配置 ces-control:host 对应的远程 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 +**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR、COLM、NeurIPS、ACL Anthology 官方元数据采集、带清单计数核验的会议导入和只读查询脚本。服务器上持续运行合成回归检查；完整 ICML 2025 元数据已采集 3,330 条，尚未作为全量交付。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），全量导入正在服务器运行。新增 PeerRead 公开 ICLR 2017 明确决定采集，其他会议适配器和全量历史导入尚未完成。其他会议适配器、跨平台疑似重复复核流程、完整覆盖率核验仍待实现。已配置 ces-control:host 对应的远程 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 ## 数据范围与来源 @@ -53,7 +53,7 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 | `papers` / `papers_fts` | 稳定论文实体、代表性标题摘要、FTS5 全文检索 | | `identifiers` | arXiv ID、DOI 等唯一外部标识 | | `source_records` / `sources` | 平台记录、原平台标题摘要与索引元数据、来源 URL 和抓取时间 | -| `venues` / `editions` | 会议或期刊系列、年份与 track，区分主会、Findings 和 workshop | +| `venues` / `venue_aliases` / `editions` | 会议或期刊系列、年份与 track，区分主会、Findings 和 workshop，联合会议用别名查询，不重复生成事件 | | `edition_sources` | 每届每个来源的覆盖状态、预计/已导入数量、错误与断点 | | `submissions` / `decisions` | 多次投稿及带证据的决定历史，支持改投、撤稿和决定变更 | | `tags` / `paper_tags` | 平台标签与人工标签，使用命名空间隔离 | @@ -134,12 +134,28 @@ python3 -B collect_proceedings.py --source neurips \ 采集器从官方根目录发现实际发表届次，仅访问索引和 Abstract HTML 页面，不请求 PDF、补充材料或源码。可用重复 `--year` 限定实际存在的届次；不存在的年份会报错，局部重试保留其他届次的已有状态。每个来源串行请求，实际网络响应后等待 1 秒；缓存命中不额外等待。 -COLM 当前目录提供 2024–2026 三届 pre-proceedings，按会议年份保存，保留平台实际出版日期和出版状态，不把晚补刊日期当作会议年份。NeurIPS 当前目录提供 1987–2025，并包含 2025 独立卷及 2021 独立 Datasets and Benchmarks 索引。主会、datasets_benchmarks、position、creative_ai 分开出文件，不混为同一个 track。首次真实样本核对：COLM 2024 官方目录 299 个摘要入口；NeurIPS 1987 为 90 个，作者与摘要均已验证解析。样本成功不代表所有页面采集完成。 +COLM 当前目录提供 2024–2026 三届 pre-proceedings，按会议年份保存，保留平台实际出版日期和出版状态，不把晚补刊日期当作会议年份。NeurIPS 当前目录提供 1987–2025，并包含 2025 独立卷及 2021 独立 Datasets and Benchmarks 索引。主会、datasets_benchmarks、position、creative_ai 分开出文件，不混为同一个 track。跨平台自标识仅从出版方 paper-actions 区域链接与 citation 元数据提取，不把摘要里的引用链接当作当前论文 ID。manifest 的 `parser_version` 至少为 2 才允许入库；旧文件须利用缓存重新采集，避免错误身份合并。首次真实样本核对：COLM 2024 官方目录 299 个摘要入口；NeurIPS 1987 为 90 个，作者与摘要均已验证解析。样本成功不代表所有页面采集完成。 `colm-index-status.json`、`neurips-index-status.json` 保存逐索引状态；对应 `*-last-run.json` 保存本次实际选择范围。每个 JSONL 的 manifest 记录索引入口数、track 记录数和缺失摘要数。论文数量与官方页面声明不一致会失败，缺失摘要会标记 partial；不会把页脚、姓名政策或其他正文当作摘要。 这两个采集器只写采集文件，合并仍通过单写入协调流程。全量历史覆盖和所有未实现会议仍待最终核验。 +### ACL Anthology 权威 XML + +```sh +python3 -B collect_acl.py \ + --checkout /data/paper-index/raw/metadata-checkouts/acl-anthology \ + --out /data/paper-index/collected +``` + +`metadata_checkout.py` 对官方仓库做只包含 `data/xml` 的 partial clone/sparse checkout，不获取 PDF，不执行仓库 hooks 或构建。每批固定到实际 Git commit。遍历 XML 的 volume venue 字段选择 ACL、EMNLP、NAACL、EACL、COLING、CoNLL；Findings 按所属会议独立成 track，不按文件名把其他同场 workshop 冒充主会。 + +新旧 Anthology ID 按官方规则构造，并核对源 DOI。保留 C69、D19-50 至 D19-66、W 系列的历史编号例外，排除编号 0 的前言。XML 混合标记通过 itertext 提取；无摘要保留为空并标记 partial。主会、Findings、workshop、demo、tutorial、industry、shared_task 分开。 + +联合举办会议保存一个会议事件及多个名称别名。例如 LREC-COLING 的论文可用 COLING 或 LREC 查询，仍只有一次该届投稿记录。可用 `--track main` 限定主会，避免把 Findings 或演示算入主会；别名数据在正常单写入合并时应用 schema。 + +`acl-collection-status.json` 记录每个 XML 及其 volume 的处理结果和排除原因；`acl-last-run.json` 记录源 commit 与本批范围。`--collection P17` 可限定一个确实存在的 collection。已验证真实 ACL 2017 长文 195 条、短文 107 条，二者摘要完整；其他 track 按源实际缺失情况记录，不能凭样本宣称整个历史完成。 + ## 长期维护与更新 1. 每次采集先核对来源、许可、可访问性、空闲空间和现有运行批次。保持低并发，遵守来源限流、重试与人机验证要求。 diff --git a/collect_acl.py b/collect_acl.py new file mode 100644 index 0000000..cfd9976 --- /dev/null +++ b/collect_acl.py @@ -0,0 +1,131 @@ +"""Collect authoritative ACL Anthology XML metadata, never PDFs.""" +import argparse +import json +import re +import xml.etree.ElementTree as ET +from pathlib import Path + +from metadata_checkout import checkout_metadata + +VENUES={'acl':'ACL','emnlp':'EMNLP','naacl':'NAACL','eacl':'EACL','coling':'COLING','conll':'CoNLL'} + + +def text(element): + return ' '.join(''.join(element.itertext()).split()) if element is not None else '' + + +def paper_id(collection,volume,number): + if not str(number).isdigit() or not re.fullmatch(r'[A-Za-z0-9_-]+',str(volume)): + raise ValueError('Invalid Anthology paper numbering') + if re.fullmatch(r'\d{4}\.[a-z0-9]+',collection): + return f'{collection}-{volume}.{int(number)}' + if not re.fullmatch(r'[A-Z]\d{2}',collection) or not str(volume).isdigit(): + raise ValueError('Unsupported historical identifier') + small=collection.startswith('W') or collection=='C69' or (collection=='D19' and 50<=int(volume)<=66) + width=2 if small else 1 + digits=2 if small else 3 + if len(str(int(volume)))>width or len(str(int(number)))>digits: + raise ValueError('Historical identifier exceeds documented width') + return f'{collection}-{int(volume):0{width}d}{int(number):0{digits}d}' + + +def volume_scope(meta,volume): + codes=[text(n).lower() for n in meta.findall('venue')] + book=text(meta.find('booktitle')) + lower=book.lower() + selected=sorted({VENUES[v] for v in codes if v in VENUES}) + if 'findings' in codes: + associated=volume.split('-')[0].lower() + if associated in VENUES:selected=[VENUES[associated]] + if not selected: + return None + joint_names=sorted({(VENUES|{'lrec':'LREC'})[v] for v in codes if v in VENUES or v=='lrec'}) + if len(joint_names)>1:selected=joint_names + track='main' + if 'findings' in codes or 'findings of' in lower:track='findings' + elif 'shared task' in lower:track='shared_task' + elif 'workshop' in lower or 'student research' in lower:track='workshop' + elif 'demonstration' in lower or 'system demonstrations' in lower:track='demo' + elif 'tutorial' in lower:track='tutorial' + elif 'industry' in lower:track='industry' + return '+'.join(selected),selected,track + + +def collect_xml(path,out,commit): + root=ET.parse(path).getroot() + collection=root.attrib['id'] + results=[] + for volume in root.findall('volume'): + vid=volume.attrib['id'] + meta=volume.find('meta') + if meta is None:raise ValueError('Volume metadata missing') + scope=volume_scope(meta,vid) + key=f'{collection}-{vid}' + if scope is None: + results.append({'volume':key,'status':'excluded_by_venue','venue_codes':[text(n) for n in meta.findall('venue')],'booktitle':text(meta.find('booktitle')),'year':text(meta.find('year'))}) + continue + venue,aliases,track=scope + year_text=text(meta.find('year')) + if not re.fullmatch(r'\d{4}',year_text):raise ValueError('No explicit official volume year: '+key) + year=int(year_text) + records=[] + frontmatter=0 + for paper in volume.findall('paper'): + pid=paper.attrib['id'] + if int(pid)==0: + frontmatter+=1 + continue + identifier=paper_id(collection,vid,pid) + doi=text(paper.find('doi')) + if doi.startswith('10.18653/v1/') and doi.split('10.18653/v1/',1)[1].casefold()!=identifier.casefold(): + raise ValueError('Official DOI disagrees with constructed Anthology ID: '+identifier) + title=text(paper.find('title')) + if not title:raise ValueError('Missing paper title: '+identifier) + authors=[{'given':text(a.find('first')),'family':text(a.find('last'))} for a in paper.findall('author')] + records.append({'source':'acl_anthology','external_id':identifier,'url':f'https://aclanthology.org/{identifier}/','title':title,'abstract':text(paper.find('abstract')) or None,'authors':authors,'identifiers':{'acl':identifier,'doi':doi},'venue':venue,'venue_aliases':aliases if len(aliases)>1 else [],'year':year,'track':track,'status':'accepted','raw_decision':'listed in authoritative ACL Anthology proceedings XML','metadata':{'collection':collection,'volume':vid,'booktitle':text(meta.find('booktitle')),'venue_codes':[text(n) for n in meta.findall('venue')],'pages':text(paper.find('pages')),'publisher':text(meta.find('publisher')),'metadata_commit':commit,'source_xml':f'https://github.com/acl-org/acl-anthology/blob/{commit}/data/xml/{collection}.xml'}}) + if not records: + results.append({'volume':key,'status':'unavailable','reason':'no research paper elements after excluding front matter'}) + continue + target=Path(out)/f'acl-{key}.jsonl' + target.parent.mkdir(parents=True,exist_ok=True) + part=target.with_suffix('.part') + with part.open('w') as output: + for record in records:output.write(json.dumps(record,ensure_ascii=False)+'\n') + part.replace(target) + missing=sum(not row['abstract'] for row in records) + manifest={'source':'acl_anthology','venue':venue,'year':year,'track':track,'count':len(records),'xml_paper_elements':len(volume.findall('paper')),'excluded_frontmatter':frontmatter,'missing_abstract_count':missing,'list_url':f'https://aclanthology.org/{key}/','metadata_commit':commit,'file':str(target),'status':'partial' if missing else 'collected'} + target.with_suffix('.manifest.json').write_text(json.dumps(manifest,indent=2)) + results.append(manifest) + print(json.dumps(manifest),flush=True) + return results + + +if __name__=='__main__': + parser=argparse.ArgumentParser() + parser.add_argument('--checkout',required=True) + parser.add_argument('--out',required=True) + parser.add_argument('--collection',action='append') + args=parser.parse_args() + commit=checkout_metadata('https://github.com/acl-org/acl-anthology.git',args.checkout,['/data/xml/']) + out=Path(args.out);out.mkdir(parents=True,exist_ok=True) + status_file=out/'acl-collection-status.json' + catalogue=json.loads(status_file.read_text()) if status_file.exists() else {} + paths=sorted((Path(args.checkout)/'data/xml').glob('*.xml')) + if not paths:raise SystemExit('No official XML files found') + if args.collection and set(args.collection)-{p.stem for p in paths}:parser.error('Requested collection is absent') + errors=[] + processed=0 + for path in paths: + if args.collection and path.stem not in args.collection:continue + try:catalogue[path.stem]={'status':'processed','metadata_commit':commit,'volumes':collect_xml(path,args.out,commit)} + except Exception as exc: + catalogue[path.stem]={'status':'blocked','metadata_commit':commit,'error':str(exc)} + errors.append(path.stem) + print(json.dumps({'collection':path.stem,'error':str(exc)}),flush=True) + processed+=1 + if processed%25==0: + part=status_file.with_suffix('.part');part.write_text(json.dumps(catalogue,indent=2));part.replace(status_file) + print(json.dumps({'xml_collections_processed':processed}),flush=True) + part=status_file.with_suffix('.part');part.write_text(json.dumps(catalogue,indent=2));part.replace(status_file) + (out/'acl-last-run.json').write_text(json.dumps({'metadata_commit':commit,'collection_filter':args.collection,'blocked_collections':errors},indent=2)) + if errors:raise SystemExit(1) diff --git a/collect_proceedings.py b/collect_proceedings.py index 19917c4..30e45fa 100644 --- a/collect_proceedings.py +++ b/collect_proceedings.py @@ -120,7 +120,12 @@ def parse_paper(html,url,source,year,index_url): identifiers={} if meta['citation_doi']: identifiers['doi']=meta['citation_doi'][0] - for node in nodes: + action_nodes=[] + for container in nodes: + if 'paper-actions' in container.attrs.get('class','').split(): + action_nodes.extend(container.walk()) + # Abstract links can cite other papers; only labelled self links identify this one. + for node in action_nodes: if node.tag!='a': continue link=urljoin(url,node.attrs.get('href','')) @@ -176,7 +181,7 @@ def collect_index(source,index_url,year,cache,out): for record in records:output.write(json.dumps(record,ensure_ascii=False)+'\n') part.replace(path) missing_count=sum(not r['abstract'] for r in records) - result={'source':source,'venue':records[0]['venue'],'year':year,'track':track,'count':len(records),'index_link_count':len(links),'index_url':index_url,'list_url':index_url,'missing_abstract_count':missing_count,'file':str(path),'status':'collected' if not missing_count else 'partial'} + result={'source':source,'parser_version':2,'venue':records[0]['venue'],'year':year,'track':track,'count':len(records),'index_link_count':len(links),'index_url':index_url,'list_url':index_url,'missing_abstract_count':missing_count,'file':str(path),'status':'collected' if not missing_count else 'partial'} path.with_suffix('.manifest.json').write_text(json.dumps(result,indent=2)) results.append(result) print(json.dumps(result),flush=True) diff --git a/import_records.py b/import_records.py index 2b45792..665ea13 100644 --- a/import_records.py +++ b/import_records.py @@ -11,6 +11,8 @@ def ingest_file(db,path): path = Path(path) manifest_path = path.with_suffix('.manifest.json') manifest = json.loads(manifest_path.read_text()) if manifest_path.exists() else {} + if manifest.get('source') in ('colm','neurips') and manifest.get('parser_version',0)<2: + raise ValueError('Recollect proceedings with parser_version>=2 before identity matching') store = Store(db) c = store.conn source_name = manifest.get('source') @@ -34,7 +36,7 @@ def ingest_file(db,path): seen.add(external) paper,record = store.record(source_name,external,row['url'],row['title'],row.get('abstract'),row.get('authors',[]),row.get('identifiers'),row.get('metadata')) if row.get('status') in ('accepted','rejected','desk_rejected','withdrawn','under_review','unknown'): - edition = store.decision(paper,record,row['venue'],row['year'],row['track'],external,row.get('evidence_url',row['url']),row['status'],row.get('raw_decision','listed in official proceedings' if source_name=='pmlr' else row['status'])) + edition = store.decision(paper,record,row['venue'],row['year'],row['track'],external,row.get('evidence_url',row['url']),row['status'],row.get('raw_decision','listed in official proceedings' if source_name=='pmlr' else row['status']),row.get('venue_aliases')) editions.add(edition) else: raise ValueError('This importer requires an explicit supported decision status') diff --git a/index_core.py b/index_core.py index 2c012e1..5c052d8 100644 --- a/index_core.py +++ b/index_core.py @@ -109,10 +109,12 @@ class Store: c.execute('INSERT OR IGNORE INTO match_candidates VALUES(?,?,?,?,?,?)',(min(paper,other),max(paper,other),'exact_title_unconfirmed',None,json.dumps({'title':title}), 'pending')) return paper,record - def decision(self,paper,record,venue,year,track,external_id,url,status,raw_value): + def decision(self,paper,record,venue,year,track,external_id,url,status,raw_value,aliases=None): c = self.conn c.execute("INSERT OR IGNORE INTO venues(acronym,name,field,venue_type) VALUES(?,?,?,'conference')",(venue,venue,'unclassified')) venue_id = c.execute('SELECT id FROM venues WHERE acronym=?',(venue,)).fetchone()[0] + for alias in aliases or []: + c.execute('INSERT OR IGNORE INTO venue_aliases VALUES(?,?)',(venue_id,alias)) c.execute('INSERT OR IGNORE INTO editions(venue_id,year,track) VALUES(?,?,?)',(venue_id,int(year),track)) edition = c.execute('SELECT id FROM editions WHERE venue_id=? AND year=? AND track=?',(venue_id,int(year),track)).fetchone()[0] c.execute('INSERT OR IGNORE INTO submissions(paper_id,edition_id,source_record_id,source_submission_id) VALUES(?,?,?,?)',(paper,edition,record,external_id)) diff --git a/metadata_checkout.py b/metadata_checkout.py new file mode 100644 index 0000000..4d19294 --- /dev/null +++ b/metadata_checkout.py @@ -0,0 +1,29 @@ +"""Public metadata-only Git checkouts, without credentials or project hooks.""" +import os +import subprocess +import tempfile +from pathlib import Path + + +def checkout_metadata(url,destination,patterns,branch='master'): + destination=Path(destination) + destination.parent.mkdir(parents=True,exist_ok=True) + env=dict(os.environ,GIT_TERMINAL_PROMPT='0',GIT_LFS_SKIP_SMUDGE='1') + def git(*args): + result=subprocess.run(['git','-c','credential.helper=','-c','http.extraHeader=','-c','core.hooksPath=/dev/null','-c','core.deltaBaseCacheLimit=32m',*args],env=env,capture_output=True,text=True,timeout=900) + if result.returncode: + raise RuntimeError('Metadata checkout failed: '+result.stderr[-500:]) + return result.stdout.strip() + if not destination.exists(): + with tempfile.TemporaryDirectory(prefix='metadata-clone-',dir=destination.parent) as tmp: + stage=Path(tmp)/'repo' + git('clone','--filter=blob:none','--depth','1','--single-branch','--branch',branch,'--no-checkout',url,str(stage)) + git('-C',str(stage),'sparse-checkout','set','--no-cone',*patterns) + git('-C',str(stage),'checkout',branch) + stage.rename(destination) + else: + if git('-C',str(destination),'status','--porcelain'): + raise RuntimeError('Metadata checkout has local changes; refusing to overwrite') + git('-C',str(destination),'fetch','--filter=blob:none','--depth','1','origin',branch) + git('-C',str(destination),'checkout','--detach','FETCH_HEAD') + return git('-C',str(destination),'rev-parse','HEAD') diff --git a/query.py b/query.py index c322f43..f1ba580 100644 --- a/query.py +++ b/query.py @@ -13,6 +13,7 @@ def main(): parser.add_argument('--tag', help='Filter as namespace=value, e.g. arxiv:category=cs.AI') parser.add_argument('--venue', help='Conference acronym') parser.add_argument('--year', type=int) + parser.add_argument('--track',help='Exact track, e.g. main, findings, workshop') parser.add_argument('--status', choices=['accepted','rejected','desk_rejected','withdrawn','under_review','unknown']) parser.add_argument('--limit', type=int, default=20) parser.add_argument('--stats', action='store_true') @@ -35,10 +36,15 @@ def main(): if args.tag: namespace,value = args.tag.split('=',1) where.append('EXISTS(SELECT 1 FROM paper_tags pt JOIN tags t ON t.id=pt.tag_id WHERE pt.paper_id=p.id AND t.namespace=? AND t.value=?)'); params.extend([namespace,value]) - if args.venue or args.year or args.status: + if args.venue or args.year or args.status or args.track: clause = ['1=1'] if args.venue: - clause.append('v.acronym=? COLLATE NOCASE'); params.append(args.venue) + if conn.execute("SELECT 1 FROM sqlite_master WHERE type='table' AND name='venue_aliases'").fetchone(): + clause.append('(v.acronym=? COLLATE NOCASE OR EXISTS(SELECT 1 FROM venue_aliases va WHERE va.venue_id=v.id AND va.alias=? COLLATE NOCASE))'); params.extend([args.venue,args.venue]) + else: + clause.append('v.acronym=? COLLATE NOCASE');params.append(args.venue) + if args.track: + clause.append('e.track=?');params.append(args.track) if args.year: clause.append('e.year=?'); params.append(args.year) if args.status: diff --git a/schema.sql b/schema.sql index 09a73ec..2e5cfb1 100644 --- a/schema.sql +++ b/schema.sql @@ -31,6 +31,10 @@ CREATE TABLE IF NOT EXISTS editions ( year INTEGER NOT NULL, track TEXT NOT NULL DEFAULT 'main', UNIQUE(venue_id,year,track) ); +CREATE TABLE IF NOT EXISTS venue_aliases ( + venue_id INTEGER NOT NULL REFERENCES venues(id), alias TEXT NOT NULL, + PRIMARY KEY(venue_id,alias) +); CREATE TABLE IF NOT EXISTS edition_sources ( edition_id INTEGER NOT NULL REFERENCES editions(id), source_id INTEGER NOT NULL REFERENCES sources(id), list_url TEXT NOT NULL, status TEXT NOT NULL DEFAULT 'pending' diff --git a/test_acl.py b/test_acl.py new file mode 100644 index 0000000..99775ba --- /dev/null +++ b/test_acl.py @@ -0,0 +1,40 @@ +import json +import tempfile +import unittest +import xml.etree.ElementTree as ET +from pathlib import Path + +from collect_acl import paper_id,volume_scope,collect_xml + + +class AnthologyTests(unittest.TestCase): + def test_documented_identifier_formats_and_exceptions(self): + for args,expected in [(('P17','1','1'),'P17-1001'),(('2025.acl','long','1'),'2025.acl-long.1'),(('C69','1','1'),'C69-0101'),(('D19','50','1'),'D19-5001'),(('W17','1','1'),'W17-0101')]: + self.assertEqual(paper_id(*args),expected) + + def test_findings_uses_associated_conference_and_separate_track(self): + meta=ET.fromstring('findingsFindings of ACL: EMNLP 2025') + self.assertEqual(volume_scope(meta,'emnlp'),('EMNLP',['EMNLP'],'findings')) + + def test_joint_event_has_one_identity_and_two_aliases(self): + meta=ET.fromstring('lreccolingLREC-COLING 2024') + self.assertEqual(volume_scope(meta,'main'),('COLING+LREC',['COLING','LREC'],'main')) + + def test_xml_mixed_text_frontmatter_and_missing_abstract(self): + xml='2017aclACLFront matterA <fixed-case>Mixed</fixed-case> TitleJaneDoe10.18653/v1/P17-1001' + with tempfile.TemporaryDirectory() as tmp: + path=Path(tmp)/'P17.xml';path.write_text(xml) + result=collect_xml(path,tmp,'a'*40)[0] + row=json.loads(Path(result['file']).read_text()) + self.assertEqual(result['count'],1) + self.assertEqual(result['excluded_frontmatter'],1) + self.assertEqual(result['status'],'partial') + self.assertEqual(row['external_id'],'P17-1001') + self.assertEqual(row['title'],'A Mixed Title') + + def test_non_target_workshop_is_not_main_conference(self): + meta=ET.fromstring('some-workshopwsWorkshop at ACL') + self.assertIsNone(volume_scope(meta,'1')) + + +if __name__=='__main__':unittest.main() diff --git a/test_index_core.py b/test_index_core.py index 29cb122..87e1fea 100644 --- a/test_index_core.py +++ b/test_index_core.py @@ -61,10 +61,18 @@ class StoreTests(unittest.TestCase): def test_missing_abstract_does_not_become_complete_coverage(self): path=Path(self.tmp.name)/'partial.jsonl' path.write_text(json.dumps({'source':'colm','external_id':'paper1','url':'https://example.org/paper1','title':'Paper','abstract':None,'authors':[],'venue':'COLM','year':2024,'track':'main','status':'accepted'})+'\n') - path.with_suffix('.manifest.json').write_text(json.dumps({'source':'colm','list_url':'https://example.org/list','count':1,'status':'partial','missing_abstract_count':1})) + path.with_suffix('.manifest.json').write_text(json.dumps({'source':'colm','parser_version':2,'list_url':'https://example.org/list','count':1,'status':'partial','missing_abstract_count':1})) ingest_file(self.db,path) self.assertEqual(self.store.conn.execute('SELECT status FROM edition_sources').fetchone()[0],'partial') + def test_legacy_proceedings_manifest_is_rejected_before_import(self): + path=Path(self.tmp.name)/'old.jsonl' + path.write_text('') + path.with_suffix('.manifest.json').write_text(json.dumps({'source':'colm','count':1})) + with self.assertRaisesRegex(ValueError,'parser_version'): + ingest_file(self.db,path) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM source_records').fetchone()[0],0) + if __name__=='__main__': unittest.main() diff --git a/test_proceedings.py b/test_proceedings.py index 44c1153..931b34d 100644 --- a/test_proceedings.py +++ b/test_proceedings.py @@ -52,5 +52,10 @@ class ProceedingsTests(unittest.TestCase): self.assertEqual(result[other]['status'],'collected') self.assertEqual(result[url]['files'][0]['count'],1) + def test_abstract_citations_are_not_used_as_self_identifiers(self): + html='
    OpenReview

    Related work and another paper.

    ' + row=parse_paper(html,'https://proceedings.colm.cc/paper','colm',2024,'https://proceedings.colm.cc/') + self.assertEqual(row['identifiers'],{'openreview':'selfForum'}) + if __name__=='__main__':unittest.main() diff --git a/test_query.py b/test_query.py index 0b83b50..068d4b9 100644 --- a/test_query.py +++ b/test_query.py @@ -9,6 +9,18 @@ from index_core import Store class QueryTests(unittest.TestCase): + def test_joint_conference_alias_does_not_duplicate_submission(self): + with tempfile.TemporaryDirectory() as tmp: + db=Path(tmp)/'papers.sqlite';store=Store(db);self.addCleanup(store.conn.close) + paper,record=store.record('acl_anthology','paper1','https://example.org/1','Joint event paper','Abstract',[]) + store.decision(paper,record,'COLING+LREC',2024,'main','paper1','https://example.org/1','accepted','official XML',['COLING','LREC']) + store.conn.commit() + for venue in ('COLING','LREC'): + command=[sys.executable,'-B',str(Path(__file__).with_name('query.py')),'--db',str(db),'--venue',venue,'--track','main'] + rows=json.loads(subprocess.run(command,check=True,capture_output=True,text=True).stdout) + self.assertEqual([r['id'] for r in rows],[paper]) + self.assertEqual(store.conn.execute('SELECT count(*) FROM submissions').fetchone()[0],1) + def test_venue_filter_is_case_insensitive_and_uses_latest_decision(self): with tempfile.TemporaryDirectory() as tmp: db=Path(tmp)/'papers.sqlite' -- 2.52.0 From dba943f03e41e9fbc5bc2764f58869c9768b7abe Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 12:58:02 +0800 Subject: [PATCH 08/20] Collect official ICLR programs with presentation identity checks --- README.md | 22 +++- collect_iclr.py | 246 +++++++++++++++++++++++++++++++++++++++++++++ import_records.py | 2 + index_core.py | 16 +++ test_iclr.py | 98 ++++++++++++++++++ test_index_core.py | 13 +++ venues.json | 13 +-- 7 files changed, 401 insertions(+), 9 deletions(-) create mode 100644 collect_iclr.py create mode 100644 test_iclr.py diff --git a/README.md b/README.md index 861582c..ab57d1c 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 目标包括：完整历史元数据导入、逐年会议目录、跨平台论文去重、公开拒稿/撤稿信息、平台标签与自定义标签隔离、可恢复更新和覆盖率报告。 -**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR、COLM、NeurIPS、ACL Anthology 官方元数据采集、带清单计数核验的会议导入和只读查询脚本。服务器上持续运行合成回归检查；完整 ICML 2025 元数据已采集 3,330 条，尚未作为全量交付。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），全量导入正在服务器运行。新增 PeerRead 公开 ICLR 2017 明确决定采集，其他会议适配器和全量历史导入尚未完成。其他会议适配器、跨平台疑似重复复核流程、完整覆盖率核验仍待实现。已配置 ces-control:host 对应的远程 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 +**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR、COLM、NeurIPS、ACL Anthology、ICLR 官方公开程序元数据采集、带清单核验的会议导入和只读查询。PeerRead 提供公开 ICLR 2017 明确决定的补充来源。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），全量导入正在服务器运行。其他会议适配器、历史缺失届次、完整投稿历史、跨平台疑似重复复核和覆盖率核验仍待完成。回归检查在远端执行，已配置 ces-control:host 对应的 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 ## 数据范围与来源 @@ -156,6 +156,26 @@ python3 -B collect_acl.py \ `acl-collection-status.json` 记录每个 XML 及其 volume 的处理结果和排除原因；`acl-last-run.json` 记录源 commit 与本批范围。`--collection P17` 可限定一个确实存在的 collection。已验证真实 ACL 2017 长文 195 条、短文 107 条，二者摘要完整；其他 track 按源实际缺失情况记录，不能凭样本宣称整个历史完成。 +### ICLR 官方公开程序元数据 + +```sh +python3 -B collect_iclr.py --cache /data/paper-index/raw/http-cache \ + --out /data/paper-index/collected +# 入库后按官方主题检索；主题值以实际来源为准 +python3 -B query.py --db /data/paper-index/papers.sqlite \ + --venue ICLR --tag 'iclr_program:topic=Reinforcement Learning->Deep RL' +``` + +从 ICLR 官方年份导航登记届次，读取对应 Papers 页面明确引用的 orals-posters/abstracts JSON。核对声明数量、分页状态及唯一 event ID；只下载公开程序元数据，不访问用户资料、PDF、海报、幻灯片或视频。`--year` 可限制年份；清单中的已采集数与数据库已入库数分别统计。 + +官方程序中的 event 数不是论文数：同一篇文章可以有 poster 和 oral 两场展示。程序还会给负 sourceid 的 oral 展示生成占位 OpenReview 链接，这些链接不能作为真实论文标识。采集器核对原投稿编号、标题与作者后合并展示，优先使用正编号记录的真实论文链接；原展示编号和类型保留在源元数据中。 + +同一投稿编号的展示标题可能有小幅修订。只有负编号 oral 与正编号原稿对应、至少两位作者完全一致且标题相似度满足严格阈值时，才保留为同稿的标题变体；该规则仅在一届官方程序内部生效，不用于跨平台模糊合并。 + +只将 Conference 来源的论文视为 ICLR 主会录用，排除 Blog 和期刊转会展示，并在审计清单保存排除计数。2020/2021 旧数据允许依据官方主会 paper program 形成录用证据；无 forum ID 时保留 program 身份和匹配候选，不伪造标识。历史年度返回 404 会登记 unavailable，未来未发表届次登记 not_yet_published，其他访问或解析失败登记 blocked；这些状态都不能宣称该年已覆盖。 + +主题、关键词使用 `iclr_program:topic`、`iclr_program:keyword` 命名空间。通用导入器按源记录刷新标签证据，保留同一论文其他源记录及 `custom:*` 标签。每年的 JSONL、manifest、`iclr-program-year-status.json` 和 `iclr-program-last-run.json` 保存实际范围、独立展示数、主会展示数、论文数、缺失摘要与缺失 forum 数。录用程序不能提供拒稿全集，公开投稿及拒稿补充仍需独立来源。 + ## 长期维护与更新 1. 每次采集先核对来源、许可、可访问性、空闲空间和现有运行批次。保持低并发，遵守来源限流、重试与人机验证要求。 diff --git a/collect_iclr.py b/collect_iclr.py new file mode 100644 index 0000000..ef0978f --- /dev/null +++ b/collect_iclr.py @@ -0,0 +1,246 @@ +"""Collect public ICLR program metadata, excluding blog/journal presentations.""" +import argparse +from collections import Counter, defaultdict +from datetime import datetime, timezone +from difflib import SequenceMatcher +import json +from pathlib import Path +import re +from urllib.error import HTTPError +from urllib.parse import parse_qs, urljoin, urlsplit + +from collect_pmlr import fetch +from collect_proceedings import Document, clean +from index_core import author_keys, normalized + +ROOT = 'https://iclr.cc/' +SOURCE = 'iclr_program' + + +class NotYetPublished(ValueError): + pass + + +def discover(html): + years = {} + for node in Document(html).root.walk(): + text = clean(node.text()) + if node.tag == 'a' and re.fullmatch(r'20\d{2}', text): + url = urljoin(ROOT, node.attrs.get('href', '')) + if urlsplit(url).netloc == 'iclr.cc': + years[int(text)] = url + if not years: + raise ValueError('No ICLR years in the official navigation') + return sorted(years.items()) + + +def asset_urls(html, page_url, year): + refs = re.findall(r'''["']([^"'\s]+\.json)["']''', html) + result = {} + for kind in ('orals-posters', 'abstracts'): + matches = {urljoin(page_url, ref) for ref in refs + if urlsplit(urljoin(page_url, ref)).netloc == 'iclr.cc' + and urlsplit(urljoin(page_url, ref)).path == f'/static/virtual/data/iclr-{year}-{kind}.json'} + if len(matches) > 1: + raise ValueError('Ambiguous official program asset') + if matches: + result[kind] = matches.pop() + if 'orals-posters' not in result: + raise ValueError('Official page has no supported program JSON reference') + return result + + +def main_event(row, year): + source = row.get('sourceurl') + if source: + parsed = urlsplit(source) + return (parsed.netloc == 'openreview.net' and parsed.path == '/group' + and parse_qs(parsed.query).get('id') == [f'ICLR.cc/{year}/Conference']) + # These historical official paper lists predate sourceurl/decision fields. + return year in (2020, 2021) and row.get('eventtype') in ('Poster', 'Oral', 'Spotlight') + + +def source_number(row): + value = row.get('sourceid') + return int(value) if re.fullmatch(r'-?[1-9]\d*', str(value)) else None + + +def identifiers(row): + # The program creates negative-ID oral events with fabricated forum URLs. + if (source_number(row) or 0) < 0: + return {} + links = [row.get('paper_url'), row.get('paper_pdf_url')] + links += [media.get('uri') for media in row.get('eventmedia', []) + if str(media.get('name') or '').casefold() in ('openreview', 'paper')] + forums = set() + for link in links: + if not isinstance(link, str): + continue + parsed = urlsplit(link) + if parsed.netloc == 'openreview.net' and parsed.path in ('/forum', '/pdf'): + values = parse_qs(parsed.query).get('id', []) + if len(values) == 1 and re.fullmatch(r'[\w-]+', values[0]) and not re.match(r'20\d{2}-Oral-', values[0]): + forums.add(values[0]) + if len(forums) > 1: + raise ValueError('Conflicting paper self-links in one program event') + return {'openreview': forums.pop()} if forums else {} + + +def parse_program(data, abstracts, year, index_url): + if not isinstance(data, dict) or not isinstance(data.get('results'), list): + raise ValueError('Expected a program results array with a declared count') + events = data['results'] + if data.get('next') or data.get('previous') or data.get('count') != len(events): + raise ValueError('Program JSON is paginated or its count differs from events') + if not isinstance(abstracts, dict): + raise ValueError('Expected an abstract mapping keyed by event ID') + event_ids = [str(row['id']) for row in events] + if len(event_ids) != len(set(event_ids)): + raise ValueError('Duplicate event IDs in program JSON') + grouped = defaultdict(list) + excluded = Counter() + for row in events: + if not main_event(row, year): + excluded[str(row.get('sourceurl') or 'unclassified')] += 1 + continue + decision = row.get('decision') + if decision and not (str(decision).casefold().startswith('accept') or decision in ('Poster', 'Oral', 'Spotlight')): + raise ValueError('Unexpected decision in an accepted conference program') + number = source_number(row) + identity = ('submission', abs(number)) if number else ('forum', identifiers(row).get('openreview')) + if identity == ('forum', None): + identity = ('event', row['id']) + grouped[identity].append(row) + records = [] + missing_identity = 0 + for identity, group in grouped.items(): + # Verify the negative oral event really describes its positive counterpart. + group.sort(key=lambda r: ((source_number(r) or 0) < 0, str(r['id']))) + primary = group[0] + title = normalized(primary.get('name', '')) + authors = [author_keys([a['fullname'] for a in row.get('authors', [])]) for row in group] + title_match = all(normalized(row.get('name', '')) == title or ( + identity[0] == 'submission' and (source_number(row) or 0) < 0 + and (source_number(primary) or 0) > 0 and len(authors[0]) >= 2 + and SequenceMatcher(None, title, normalized(row.get('name', ''))).ratio() >= 0.85) + for row in group) + if not title or not title_match or any(a != authors[0] for a in authors): + raise ValueError('Program submission ID collides across different papers') + own_ids = {} + for row in group: + for scheme, value in identifiers(row).items(): + if scheme in own_ids and own_ids[scheme] != value: + raise ValueError('Conflicting identifiers across presentation events') + own_ids[scheme] = value + abstract = next((text for row in group + if (text := row.get('abstract') or abstracts.get(str(row['id'])))), None) + if abstract is not None and not isinstance(abstract, str): + raise ValueError('Nontext abstract') + presentation_ids = [row['id'] for row in group] + forum = own_ids.get('openreview') + external = f'{year}/{forum}' if forum else f'{year}/program-{identity[1]}' + missing_identity += not bool(forum) + page = urljoin(ROOT, primary.get('virtualsite_url') or f'/virtual/{year}/{str(primary.get("eventtype", "Poster")).lower()}/{primary["id"]}') + topics = sorted({str(row['topic']).strip() for row in group if row.get('topic')}) + keywords = sorted({str(keyword).strip() for row in group for keyword in row.get('keywords', []) if keyword}) + records.append({'source': SOURCE, 'external_id': external, 'url': page, + 'title': clean(primary['name']), 'abstract': abstract.strip() if abstract else None, + 'authors': [a['fullname'] for a in primary.get('authors', [])], + 'identifiers': own_ids, 'venue': 'ICLR', 'year': year, 'track': 'main', + 'status': 'accepted', 'raw_decision': str(primary.get('decision') or 'listed in official conference paper program'), + 'evidence_url': index_url, 'list_url': index_url, + 'tags': [{'namespace': SOURCE+':'+kind, 'value': value} + for kind, values in [('topic', topics), ('keyword', keywords)] for value in values], + 'metadata': {'program_event_ids': presentation_ids, + 'presentation_titles': sorted({clean(r['name']) for r in group}), + 'program_source_id': identity[1], 'presentation_types': sorted({str(r.get('event_type') or r.get('eventtype')) for r in group}), + 'raw_decisions': sorted({str(r.get('decision')) for r in group if r.get('decision')}), + 'identity_state': 'forum' if forum else 'program_only'}}) + # Different source numbers can still identify the same forum. Keep one record. + by_id = {} + for row in records: + previous = by_id.get(row['external_id']) + if previous: + if normalized(previous['title']) != normalized(row['title']) or author_keys(previous['authors']) != author_keys(row['authors']): + raise ValueError('One forum identifies different program papers') + previous['metadata']['program_event_ids'] += row['metadata']['program_event_ids'] + previous['metadata']['presentation_types'] = sorted(set(previous['metadata']['presentation_types'] + row['metadata']['presentation_types'])) + previous['abstract'] = previous['abstract'] or row['abstract'] + else: + by_id[row['external_id']] = row + result = list(by_id.values()) + if not result: + raise ValueError('No verified main-conference papers in program') + return result, {'program_event_count': len(events), 'main_event_count': sum(map(len, grouped.values())), + 'excluded_events_by_source': dict(excluded), 'missing_forum_count': missing_identity, + 'missing_abstract_count': sum(not r['abstract'] for r in result)} + + +def collect_year(year, cache, out): + page = f'https://iclr.cc/virtual/{year}/papers.html' + assets = asset_urls(fetch(page, cache, 16*1024**2, delay=1).decode(), page, year) + data = json.loads(fetch(assets['orals-posters'], cache, 32*1024**2, delay=1)) + if data.get('count') == 0 and data.get('results') == [] and year > datetime.now(timezone.utc).year: + raise NotYetPublished('Official program is empty for a future conference') + abstracts = {} + abstract_asset_status = 'not_referenced' + if 'abstracts' in assets: + try: + abstracts = json.loads(fetch(assets['abstracts'], cache, 32*1024**2, delay=1)) + abstract_asset_status = 'available' + except HTTPError as exc: + if exc.code != 404: + raise + # Historical programs embed abstracts but retain a stale split-file URL. + exc.close() + abstract_asset_status = 'unavailable_404' + records, audit = parse_program(data, abstracts, year, page) + path = Path(out)/f'iclr-program-{year}-main.jsonl' + part = path.with_suffix('.part') + part.write_text(''.join(json.dumps(row, ensure_ascii=False)+'\n' for row in records)) + part.replace(path) + manifest = {'source': SOURCE, 'parser_version': 1, 'venue': 'ICLR', 'year': year, 'track': 'main', + 'count': len(records), 'list_url': page, 'asset_urls': assets, + 'abstract_asset_status': abstract_asset_status, 'file': str(path), + 'status': 'partial' if audit['missing_abstract_count'] else 'collected', **audit} + path.with_suffix('.manifest.json').write_text(json.dumps(manifest, indent=2)) + return manifest + + +def main(): + parser = argparse.ArgumentParser() + parser.add_argument('--cache', required=True) + parser.add_argument('--out', required=True) + parser.add_argument('--year', type=int, action='append') + args = parser.parse_args() + out = Path(args.out); out.mkdir(parents=True, exist_ok=True) + years = discover(fetch(ROOT, args.cache, 4*1024**2, delay=1).decode()) + if args.year and set(args.year)-{year for year, _ in years}: + parser.error('Requested year is absent from the official navigation') + status_path = out/'iclr-program-year-status.json' + catalogue = json.loads(status_path.read_text()) if status_path.exists() else {} + requested = [] + for year, conference_url in years: + if args.year and year not in args.year: + continue + requested.append(year) + try: + result = collect_year(year, args.cache, out) + except NotYetPublished as exc: + result = {'year': year, 'status': 'not_yet_published', 'error': str(exc)} + except HTTPError as exc: + result = {'year': year, 'status': 'not_yet_published' if year > datetime.now(timezone.utc).year and exc.code == 404 else 'unavailable' if exc.code == 404 else 'blocked', 'error': str(exc)} + exc.close() + except Exception as exc: + result = {'year': year, 'status': 'blocked', 'error': str(exc)} + result['conference_url'] = conference_url + catalogue[str(year)] = result + part = status_path.with_suffix('.part'); part.write_text(json.dumps(catalogue, indent=2)); part.replace(status_path) + print(json.dumps(result), flush=True) + (out/'iclr-program-last-run.json').write_text(json.dumps({'requested_years': requested, 'year_filter': args.year}, indent=2)) + if any(catalogue[str(y)]['status'] == 'blocked' for y in requested): + raise SystemExit(1) + + +if __name__ == '__main__': + main() diff --git a/import_records.py b/import_records.py index 665ea13..d287557 100644 --- a/import_records.py +++ b/import_records.py @@ -35,6 +35,8 @@ def ingest_file(db,path): raise ValueError('Duplicate external ID in one collected file: '+external) seen.add(external) paper,record = store.record(source_name,external,row['url'],row['title'],row.get('abstract'),row.get('authors',[]),row.get('identifiers'),row.get('metadata')) + if 'tags' in row: + store.record_tags(paper,record,source_name,row['tags']) if row.get('status') in ('accepted','rejected','desk_rejected','withdrawn','under_review','unknown'): edition = store.decision(paper,record,row['venue'],row['year'],row['track'],external,row.get('evidence_url',row['url']),row['status'],row.get('raw_decision','listed in official proceedings' if source_name=='pmlr' else row['status']),row.get('venue_aliases')) editions.add(edition) diff --git a/index_core.py b/index_core.py index 5c052d8..7227a30 100644 --- a/index_core.py +++ b/index_core.py @@ -109,6 +109,22 @@ class Store: c.execute('INSERT OR IGNORE INTO match_candidates VALUES(?,?,?,?,?,?)',(min(paper,other),max(paper,other),'exact_title_unconfirmed',None,json.dumps({'title':title}), 'pending')) return paper,record + def record_tags(self,paper,record,source,tags): + """Refresh this record's platform tags without removing other evidence.""" + values=set() + for tag in tags: + namespace,value=tag['namespace'],tag['value'] + if not isinstance(namespace,str) or not namespace.startswith(source+':') or not isinstance(value,str) or not value.strip(): + raise ValueError('Platform tag must have its own source namespace and a nonempty text value') + values.add((namespace,value.strip())) + evidence=f'{source}:record:{record}:tags' + c=self.conn + c.execute('DELETE FROM paper_tags WHERE source_record_id=? AND evidence_key=?',(record,evidence)) + for namespace,value in sorted(values): + c.execute('INSERT OR IGNORE INTO tags(namespace,value) VALUES(?,?)',(namespace,value)) + tag=c.execute('SELECT id FROM tags WHERE namespace=? AND value=?',(namespace,value)).fetchone()[0] + c.execute('INSERT OR IGNORE INTO paper_tags VALUES(?,?,?,?)',(paper,tag,record,evidence)) + def decision(self,paper,record,venue,year,track,external_id,url,status,raw_value,aliases=None): c = self.conn c.execute("INSERT OR IGNORE INTO venues(acronym,name,field,venue_type) VALUES(?,?,?,'conference')",(venue,venue,'unclassified')) diff --git a/test_iclr.py b/test_iclr.py new file mode 100644 index 0000000..d9f577e --- /dev/null +++ b/test_iclr.py @@ -0,0 +1,98 @@ +import json +from pathlib import Path +import tempfile +import unittest +from unittest.mock import patch +from urllib.error import HTTPError + +from collect_iclr import asset_urls, collect_year, discover, parse_program +from import_records import ingest_file +from index_core import Store + + +class ICLRTests(unittest.TestCase): + def event(self,identifier=1,sourceid=10,**updates): + row={'id':identifier,'sourceid':sourceid,'sourceurl':'https://openreview.net/group?id=ICLR.cc/2026/Conference', + 'name':'A reliable paper on abstract metadata indexing','authors':[{'fullname':'Jane Doe'},{'fullname':'John Smith'}], + 'eventtype':'Poster','event_type':'Poster','decision':'Accept (Oral)', + 'paper_url':'https://openreview.net/forum?id=realForum','topic':'Learning->Theory','keywords':[]} + return dict(row,**updates) + + def parse(self,events,abstracts=None,year=2026): + return parse_program({'count':len(events),'next':None,'previous':None,'results':events},abstracts or {},year,f'https://iclr.cc/virtual/{year}/papers.html') + + def test_oral_placeholder_is_not_an_identifier_or_another_submission(self): + poster=self.event() + oral=self.event(2,-10,eventtype='Oral',event_type='Oral',paper_url='https://openreview.net/forum?id=2026-Oral--10-placeholder') + rows,audit=self.parse([oral,poster],{'1':'The real abstract.'}) + self.assertEqual(len(rows),1) + self.assertEqual(rows[0]['identifiers'],{'openreview':'realForum'}) + self.assertEqual(rows[0]['abstract'],'The real abstract.') + self.assertEqual(rows[0]['metadata']['presentation_types'],['Oral','Poster']) + self.assertEqual(audit['main_event_count'],2) + + def test_blog_and_journal_presentations_are_not_main_acceptances(self): + rows,audit=self.parse([self.event(),self.event(2,20,sourceurl='https://openreview.net/group?id=ICLR.cc/2026/BlogPosts'),self.event(3,30,sourceurl='2026-TMLR-J2C-Spreadsheet')]) + self.assertEqual(len(rows),1) + self.assertEqual(sum(audit['excluded_events_by_source'].values()),2) + self.assertEqual(audit['missing_abstract_count'],1) + self.assertIsNone(rows[0]['abstract']) + + def test_title_revision_is_scoped_to_matching_signed_submission_and_authors(self): + poster=self.event(name='AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models') + oral=self.event(2,-10,eventtype='Oral',name='AlphaEdit: Null-Space Constrained Model Editing for Language Models') + rows,_=self.parse([oral,poster]) + self.assertEqual(len(rows),1) + self.assertEqual(rows[0]['title'],poster['name']) + self.assertEqual(len(rows[0]['metadata']['presentation_titles']),2) + with self.assertRaisesRegex(ValueError,'collides'): + self.parse([poster,dict(oral,authors=[{'fullname':'Unrelated Author'}])]) + + def test_truncation_and_identity_collision_fail(self): + with self.assertRaisesRegex(ValueError,'count differs'): + parse_program({'count':2,'results':[self.event()]},{},2026,'https://iclr.cc/') + with self.assertRaisesRegex(ValueError,'collides'): + self.parse([self.event(),self.event(2,-10,name='A different paper')]) + with self.assertRaisesRegex(ValueError,'Conflicting identifiers'): + self.parse([self.event(),self.event(2,10,paper_url='https://openreview.net/forum?id=otherForum')]) + + def test_legacy_embedded_abstracts_and_paper_links(self): + row=self.event(sourceurl=None,abstract='Historical abstract',paper_url=None,paper_pdf_url='https://openreview.net/forum?id=historicalForum',decision='Poster') + rows,_=self.parse([row],year=2021) + self.assertEqual(rows[0]['abstract'],'Historical abstract') + self.assertEqual(rows[0]['identifiers'],{'openreview':'historicalForum'}) + rows,_=self.parse([dict(row,paper_pdf_url=None)],year=2020) + self.assertEqual(rows[0]['metadata']['identity_state'],'program_only') + self.assertEqual(rows[0]['identifiers'],{}) + + def test_only_referenced_official_assets_and_real_navigation_years(self): + html='201320262025' + self.assertEqual([y for y,_ in discover(html)],[2013,2026]) + html='start("/static/virtual/data/iclr-2026-orals-posters.json",true,"https://other.example/abstracts.json")' + self.assertEqual(asset_urls(html,'https://iclr.cc/virtual/2026/papers.html',2026),{'orals-posters':'https://iclr.cc/static/virtual/data/iclr-2026-orals-posters.json'}) + + def test_missing_split_abstract_asset_does_not_discard_embedded_abstract(self): + html=b'start("/static/virtual/data/iclr-2021-orals-posters.json",true,"/static/virtual/data/iclr-2021-abstracts.json")' + event=self.event(sourceurl=None,decision='Poster',abstract='Embedded historical abstract') + responses=[html,json.dumps({'count':1,'results':[event]}).encode(),HTTPError('https://iclr.cc/abstracts',404,'Not Found',None,None)] + with tempfile.TemporaryDirectory() as tmp,patch('collect_iclr.fetch',side_effect=responses): + manifest=collect_year(2021,tmp,tmp) + self.assertEqual(manifest['status'],'collected') + self.assertEqual(manifest['missing_abstract_count'],0) + self.assertEqual(manifest['abstract_asset_status'],'unavailable_404') + + def test_import_repeat_preserves_one_submission_and_queryable_platform_tags(self): + rows,_=self.parse([self.event(),self.event(2,-10,eventtype='Oral')],{'1':'A searchable abstract.'}) + with tempfile.TemporaryDirectory() as tmp: + path=Path(tmp)/'iclr.jsonl';db=Path(tmp)/'papers.sqlite' + path.write_text(''.join(json.dumps(r)+'\n' for r in rows)) + path.with_suffix('.manifest.json').write_text(json.dumps({'source':'iclr_program','list_url':'https://iclr.cc/virtual/2026/papers.html','count':1,'status':'collected'})) + ingest_file(db,path);ingest_file(db,path) + store=Store(db);self.addCleanup(store.conn.close) + self.assertEqual(store.conn.execute('SELECT count(*) FROM submissions').fetchone()[0],1) + self.assertEqual(store.conn.execute('SELECT count(*) FROM decisions').fetchone()[0],1) + self.assertEqual(store.conn.execute('SELECT namespace,value FROM tags').fetchall(),[('iclr_program:topic','Learning->Theory')]) + self.assertEqual(store.conn.execute('PRAGMA foreign_key_check').fetchall(),[]) + + +if __name__=='__main__':unittest.main() diff --git a/test_index_core.py b/test_index_core.py index 87e1fea..aa08bf9 100644 --- a/test_index_core.py +++ b/test_index_core.py @@ -58,6 +58,19 @@ class StoreTests(unittest.TestCase): self.assertEqual(self.store.conn.execute('SELECT title FROM papers').fetchone()[0],'Updated research title') self.assertEqual(self.store.conn.execute("SELECT count(*) FROM papers_fts WHERE papers_fts MATCH 'searchable'").fetchone()[0],1) + def test_tag_refresh_keeps_other_record_evidence_and_custom_tags(self): + paper,first=self.record('iclr_program','2025/a',{'openreview':'sharedForum'}) + same,second=self.record('iclr_program','2025/b',{'openreview':'sharedForum'}) + self.store.record_tags(paper,first,'iclr_program',[{'namespace':'iclr_program:topic','value':'Shared'}]) + self.store.record_tags(same,second,'iclr_program',[{'namespace':'iclr_program:topic','value':'Shared'}]) + tag=self.store.conn.execute("INSERT INTO tags(namespace,value) VALUES('custom:topic','Manual')").lastrowid + self.store.conn.execute('INSERT INTO paper_tags VALUES(?,?,?,?)',(paper,tag,None,'manual')) + self.store.record_tags(paper,first,'iclr_program',[]) + self.assertEqual(self.store.conn.execute('SELECT source_record_id FROM paper_tags ORDER BY source_record_id').fetchall(),[(None,),(second,)]) + with self.assertRaisesRegex(ValueError,'namespace'): + self.store.record_tags(paper,second,'iclr_program',[{'namespace':'custom:topic','value':'No overwrite'}]) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM paper_tags').fetchone()[0],2) + def test_missing_abstract_does_not_become_complete_coverage(self): path=Path(self.tmp.name)/'partial.jsonl' path.write_text(json.dumps({'source':'colm','external_id':'paper1','url':'https://example.org/paper1','title':'Paper','abstract':None,'authors':[],'venue':'COLM','year':2024,'track':'main','status':'accepted'})+'\n') diff --git a/venues.json b/venues.json index 5f7adc6..ca25bc6 100644 --- a/venues.json +++ b/venues.json @@ -11,7 +11,8 @@ "aaai": "https://ojs.aaai.org/index.php/AAAI/issue/archive", "ijcai": "https://www.ijcai.org/all_proceedings/", "dblp": "https://dblp.org/db/conf/", - "colm": "https://proceedings.colm.cc/" + "colm": "https://proceedings.colm.cc/", + "iclr_program": "https://iclr.cc/virtual/2026/papers.html" }, "venues": [ { @@ -20,7 +21,6 @@ "venue_type": "conference", "discovery_sources": [ "openreview", - "pmlr", "neurips" ], "scope": "candidate_pending_scope_review", @@ -36,8 +36,7 @@ "venue_type": "conference", "discovery_sources": [ "openreview", - "pmlr", - "neurips" + "pmlr" ], "scope": "candidate_pending_scope_review", "years": "all_publicly_available", @@ -52,8 +51,7 @@ "venue_type": "conference", "discovery_sources": [ "openreview", - "pmlr", - "neurips" + "iclr_program" ], "scope": "candidate_pending_scope_review", "years": "all_publicly_available", @@ -68,8 +66,7 @@ "venue_type": "conference", "discovery_sources": [ "openreview", - "pmlr", - "neurips" + "colm" ], "scope": "candidate_pending_scope_review", "years": "all_publicly_available", -- 2.52.0 From 1e6d64f54148c54c554d1b21cbdb5f2ae44a4502 Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 13:01:05 +0800 Subject: [PATCH 09/20] Validate publication identities and completeness before bulk imports --- README.md | 4 ++++ collect_pmlr.py | 29 ++++++++++++++++++++++------- import_records.py | 8 ++++++-- test_index_core.py | 18 ++++++++++++++++++ test_pmlr.py | 24 ++++++++++++++++++++++++ 5 files changed, 74 insertions(+), 9 deletions(-) diff --git a/README.md b/README.md index ab57d1c..4d3796a 100644 --- a/README.md +++ b/README.md @@ -114,6 +114,10 @@ python3 -B import_records.py --db /data/paper-index/papers.sqlite \ 采集输出 JSONL 和对应 manifest，只有明确来源标签才产生相应决定：官方论文集记录为 accepted；PeerRead 使用其明确的 accepted 布尔标签，缺失标签保留 unknown。清单数目不一致时整个文件导入回滚，记录失败批次，不能标为覆盖完成。多个历史卷的全量覆盖仍需单独核验。旧卷可能把 PDF 放入归档；采集器对前 50 卷使用 Git partial clone 和 sparse checkout，仅取 `_posts` 和卷配置，不执行出版方项目代码。较新归档若超出 64 MiB 限额，也转为此方式。已对原本超限的 AISTATS 2009（第 5 卷）验证采集 84 条且工作区无 PDF。旧 YAML 的 `!binary` 文本、空文档尾标记会安全解析；带年份的 AISTATS/COLT 等历史 shortname 会规范化，避免名称别名漏届；缺失 year 时仅从官方会议 start 日期提取年份。`pmlr-volume-status.json` 逐卷保存结果和候选范围归属，`pmlr-last-run.json` 记录本批实际请求范围，局部重试不能证明全历史完成。缓存默认有效 24 小时；同一次批次可复用缓存，后续更新会重新读取动态入口。 +PMLR 的同一出版 ID 可能对应多份 Git 元数据，例如 CoRL 2023 的两篇论文分别存在不同日期的副本。只有论文内容与标识完全一致、差异限于元数据日期时才合为一条，并保留源文件路径和日期版本；冲突副本会使该卷采集失败，保留上一份有效文件。manifest 分别记录 metadata_post_count、duplicate_post_count 和论文 count。 + +会议导入器还会独立检查每条记录的会议、年份、track 与 manifest 是否一致，并从实际记录计算缺失摘要数；旧 manifest 即使声称摘要完整，也不能让缺失摘要的来源被标为 complete。缺失的源摘要与合并后论文实体是否从其他来源补得摘要分别统计。 + PeerRead 采集仅选择作者公开发布的 ICLR 2017 投稿元数据，固定到源仓库 commit。只依据明确的布尔 `accepted` 标签形成 accepted/rejected；空标签为 unknown，不使用其他数据分区中的 probably-rejected 推断。不会保存 reviews 或 parsed_pdfs 正文。 ```sh diff --git a/collect_pmlr.py b/collect_pmlr.py index 2ba97c7..4ac8303 100644 --- a/collect_pmlr.py +++ b/collect_pmlr.py @@ -127,8 +127,8 @@ def collect_volume(volume,cache,out,allowed=None): path = Path(out)/f'pmlr-v{volume}.jsonl' path.parent.mkdir(parents=True,exist_ok=True) part = path.with_suffix('.part') - count = 0 - missing_abstracts = 0 + records = {} + duplicate_posts = 0 with part.open('w') as output: for member_name,content in metadata_texts(volume,cache): parts = re.split(r'(?m)^---\s*$',content,maxsplit=2) @@ -139,9 +139,9 @@ def collect_volume(volume,cache,out,allowed=None): continue if not row.get('title'): raise ValueError('Paper metadata has no title: '+member_name) - if not (row.get('abstract') or '').strip(): - missing_abstracts+=1 - external = f"v{volume}/{row.get('id') or Path(member_name).stem}" + if not row.get('id'): + raise ValueError('Paper metadata has no publication ID: '+member_name) + external = f"v{volume}/{row['id']}" ids = {} for key in ('doi','openreview','arxiv'): value = row.get(key) @@ -153,13 +153,28 @@ def collect_volume(volume,cache,out,allowed=None): value = value.split('id=')[-1].split('&')[0] ids[key] = value record = {'source':'pmlr','external_id':external,'url':f"https://proceedings.mlr.press/v{volume}/{row.get('id')}.html",'title':row['title'],'abstract':row.get('abstract'),'authors':row.get('author',[]),'identifiers':ids,'venue':venue,'year':year,'track':track,'status':'accepted','list_url':f'https://proceedings.mlr.press/v{volume}/','metadata':{key:row[key] for key in ('id','volume','date','page','doi','openreview','arxiv','publisher') if key in row}} + version={'path':member_name,'date':row.get('date')} + if external in records: + previous=records[external] + if ({k:v for k,v in previous.items() if k!='metadata'} != {k:v for k,v in record.items() if k!='metadata'} + or {k:v for k,v in previous['metadata'].items() if k not in ('date','source_versions')} != {k:v for k,v in record['metadata'].items() if k!='date'}): + raise ValueError('Conflicting metadata posts share a publication ID: '+external) + previous['metadata']['source_versions'].append(version) + if str(row.get('date') or '')>str(previous['metadata'].get('date') or ''): + previous['metadata']['date']=row['date'] + duplicate_posts+=1 + continue + record['metadata']['source_versions']=[version] + records[external]=record + for record in records.values(): output.write(json.dumps(record,ensure_ascii=False,default=str)+'\n') - count += 1 + count=len(records) + missing_abstracts=sum(not (record.get('abstract') or '').strip() for record in records.values()) if count==0: part.unlink(missing_ok=True) raise RuntimeError('No papers parsed; refusing an empty successful collection') part.replace(path) - result = {'source':'pmlr','venue':venue,'year':year,'track':track,'volume':volume,'count':count,'missing_abstract_count':missing_abstracts,'list_url':f'https://proceedings.mlr.press/v{volume}/','file':str(path),'status':'partial' if missing_abstracts else 'collected'} + result = {'source':'pmlr','parser_version':2,'venue':venue,'year':year,'track':track,'volume':volume,'count':count,'metadata_post_count':count+duplicate_posts,'duplicate_post_count':duplicate_posts,'missing_abstract_count':missing_abstracts,'list_url':f'https://proceedings.mlr.press/v{volume}/','file':str(path),'status':'partial' if missing_abstracts else 'collected'} path.with_suffix('.manifest.json').write_text(json.dumps(result,indent=2)) print(json.dumps(result),flush=True) return result diff --git a/import_records.py b/import_records.py index d287557..eaed872 100644 --- a/import_records.py +++ b/import_records.py @@ -22,6 +22,7 @@ def ingest_file(db,path): run = c.execute('INSERT INTO ingest_runs(source_id,checkpoint_json) VALUES(?,?)',(source_id,json.dumps({'file':str(path)}))).lastrowid c.commit() count = 0 + missing_abstracts = 0 seen = set() editions = set() try: @@ -30,6 +31,8 @@ def ingest_file(db,path): row = json.loads(line) if row['source'] != source_name: raise ValueError('Source differs from manifest') + if any(key in manifest and manifest[key]!=row[key] for key in ('venue','year','track')): + raise ValueError('Conference edition differs from manifest') external = row['external_id'] if external in seen: raise ValueError('Duplicate external ID in one collected file: '+external) @@ -43,6 +46,7 @@ def ingest_file(db,path): else: raise ValueError('This importer requires an explicit supported decision status') count += 1 + missing_abstracts += not bool((row.get('abstract') or '').strip()) if not count: raise ValueError('Empty metadata file') if len(editions) != 1: @@ -51,11 +55,11 @@ def ingest_file(db,path): expected = manifest.get('count') if expected is not None and expected != count: raise ValueError(f'Manifest count {expected} differs from records {count}') - complete=expected==count and manifest.get('status') not in ('partial','blocked') and not manifest.get('missing_abstract_count',0) + complete=expected==count and manifest.get('status') not in ('partial','blocked') and not manifest.get('missing_abstract_count',0) and not missing_abstracts c.execute('INSERT INTO edition_sources(edition_id,source_id,list_url,status,expected_count,imported_count,last_success_at) VALUES(?,?,?,?,?,?,?) ON CONFLICT(edition_id,source_id,list_url) DO UPDATE SET status=excluded.status,expected_count=excluded.expected_count,imported_count=excluded.imported_count,last_success_at=excluded.last_success_at,last_error=NULL',(edition,source_id,manifest['list_url'],'complete' if complete else 'partial',expected,count,time.strftime('%Y-%m-%dT%H:%M:%SZ',time.gmtime()))) c.execute("UPDATE ingest_runs SET status='complete',finished_at=CURRENT_TIMESTAMP,records_seen=?,records_written=? WHERE id=?",(count,count,run)) c.commit() - print(json.dumps({'file':str(path),'status':'imported','records':count}),flush=True) + print(json.dumps({'file':str(path),'status':'imported','records':count,'missing_abstract_count':missing_abstracts}),flush=True) except BaseException as exc: c.rollback() c.execute("UPDATE ingest_runs SET status='failed',finished_at=CURRENT_TIMESTAMP,error=? WHERE id=?",(str(exc),run)) diff --git a/test_index_core.py b/test_index_core.py index aa08bf9..36f5c18 100644 --- a/test_index_core.py +++ b/test_index_core.py @@ -86,6 +86,24 @@ class StoreTests(unittest.TestCase): ingest_file(self.db,path) self.assertEqual(self.store.conn.execute('SELECT count(*) FROM source_records').fetchone()[0],0) + def test_missing_abstract_is_checked_even_when_manifest_claims_complete(self): + path=Path(self.tmp.name)/'incomplete.jsonl' + row={'source':'pmlr','external_id':'v1/paper','url':'https://example.org/paper','title':'Paper','abstract':None,'authors':[],'venue':'ICML','year':2025,'track':'main','status':'accepted'} + path.write_text(json.dumps(row)+'\n') + path.with_suffix('.manifest.json').write_text(json.dumps({'source':'pmlr','list_url':'https://example.org/list','count':1,'status':'collected','missing_abstract_count':0})) + ingest_file(self.db,path) + self.assertEqual(self.store.conn.execute('SELECT status FROM edition_sources').fetchone()[0],'partial') + + def test_manifest_edition_mismatch_rolls_back_instead_of_marking_coverage(self): + path=Path(self.tmp.name)/'wrong-edition.jsonl' + row={'source':'pmlr','external_id':'v1/paper','url':'https://example.org/paper','title':'Paper','abstract':'Abstract','authors':[],'venue':'ICML','year':2025,'track':'main','status':'accepted'} + path.write_text(json.dumps(row)+'\n') + path.with_suffix('.manifest.json').write_text(json.dumps({'source':'pmlr','list_url':'https://example.org/list','count':1,'year':2024})) + with self.assertRaisesRegex(ValueError,'edition differs'): + ingest_file(self.db,path) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM source_records').fetchone()[0],0) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM edition_sources').fetchone()[0],0) + if __name__=='__main__': unittest.main() diff --git a/test_pmlr.py b/test_pmlr.py index 5713a93..0af571d 100644 --- a/test_pmlr.py +++ b/test_pmlr.py @@ -1,5 +1,7 @@ import unittest import tempfile +import json +from pathlib import Path from unittest.mock import patch from collect_pmlr import load_metadata,venue_key,collect_volume @@ -31,6 +33,28 @@ class MetadataTests(unittest.TestCase): self.assertEqual(result['missing_abstract_count'],1) self.assertEqual(result['status'],'partial') + def test_duplicate_publication_posts_preserve_date_evidence_once(self): + config=b'shortname: CoRL\nyear: 2023\nbooktitle: CoRL\n' + paper='---\nid: paper1\nlayout: inproceedings\ntitle: Paper\nabstract: Real abstract\ndate: {date}\n---\n' + posts=[('2023-11-13-paper1.md',paper.format(date='2023-11-13')),('2023-12-02-paper1.md',paper.format(date='2023-12-02'))] + with tempfile.TemporaryDirectory() as tmp,patch('collect_pmlr.fetch',return_value=config),patch('collect_pmlr.metadata_texts',return_value=iter(posts)): + result=collect_volume(229,tmp,tmp) + rows=[json.loads(line) for line in Path(result['file']).read_text().splitlines()] + self.assertEqual(result['count'],1) + self.assertEqual(result['metadata_post_count'],2) + self.assertEqual(result['duplicate_post_count'],1) + self.assertEqual(len(rows[0]['metadata']['source_versions']),2) + self.assertEqual(rows[0]['metadata']['date'],'2023-12-02') + + def test_conflicting_duplicate_posts_do_not_replace_previous_collection(self): + config=b'shortname: CoRL\nyear: 2023\nbooktitle: CoRL\n' + paper='---\nid: paper1\nlayout: inproceedings\ntitle: {title}\nabstract: Abstract\n---\n' + with tempfile.TemporaryDirectory() as tmp,patch('collect_pmlr.fetch',return_value=config),patch('collect_pmlr.metadata_texts',return_value=iter([('a.md',paper.format(title='First paper')),('b.md',paper.format(title='Other paper'))])): + previous=Path(tmp)/'pmlr-v229.jsonl';previous.write_text('previous valid collection') + with self.assertRaisesRegex(ValueError,'Conflicting metadata posts'): + collect_volume(229,tmp,tmp) + self.assertEqual(previous.read_text(),'previous valid collection') + if __name__=='__main__': unittest.main() -- 2.52.0 From 396b7dbaf2e7a7e0a4c7b378a17014a31b65da44 Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 13:15:20 +0800 Subject: [PATCH 10/20] Collect CVPR ICCV and ECCV official abstract metadata --- README.md | 21 +++++- collect_vision.py | 189 ++++++++++++++++++++++++++++++++++++++++++++++ test_vision.py | 72 ++++++++++++++++++ venues.json | 9 +-- 4 files changed, 282 insertions(+), 9 deletions(-) create mode 100644 collect_vision.py create mode 100644 test_vision.py diff --git a/README.md b/README.md index 4d3796a..af7cb53 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 目标包括：完整历史元数据导入、逐年会议目录、跨平台论文去重、公开拒稿/撤稿信息、平台标签与自定义标签隔离、可恢复更新和覆盖率报告。 -**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR、COLM、NeurIPS、ACL Anthology、ICLR 官方公开程序元数据采集、带清单核验的会议导入和只读查询。PeerRead 提供公开 ICLR 2017 明确决定的补充来源。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），全量导入正在服务器运行。其他会议适配器、历史缺失届次、完整投稿历史、跨平台疑似重复复核和覆盖率核验仍待完成。回归检查在远端执行，已配置 ces-control:host 对应的 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 +**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR、COLM、NeurIPS、ACL Anthology、ICLR 官方公开程序、CVF/ECVA 视觉会议元数据采集、带清单核验的会议导入和只读查询。PeerRead 提供公开 ICLR 2017 明确决定的补充来源。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），全量导入正在服务器运行。其他会议适配器、历史缺失届次、完整投稿历史、跨平台疑似重复复核和覆盖率核验仍待完成。回归检查在远端执行，已配置 ces-control:host 对应的 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 ## 数据范围与来源 @@ -180,6 +180,21 @@ python3 -B query.py --db /data/paper-index/papers.sqlite \ 主题、关键词使用 `iclr_program:topic`、`iclr_program:keyword` 命名空间。通用导入器按源记录刷新标签证据，保留同一论文其他源记录及 `custom:*` 标签。每年的 JSONL、manifest、`iclr-program-year-status.json` 和 `iclr-program-last-run.json` 保存实际范围、独立展示数、主会展示数、论文数、缺失摘要与缺失 forum 数。录用程序不能提供拒稿全集，公开投稿及拒稿补充仍需独立来源。 +### CVF / ECVA 视觉会议元数据 + +```sh +python3 -B collect_vision.py --source cvf \ + --cache /data/paper-index/raw/http-cache --out /data/paper-index/collected +python3 -B collect_vision.py --source ecva \ + --cache /data/paper-index/raw/http-cache --out /data/paper-index/collected +``` + +CVF 从官方根目录发现 CVPR、ICCV 主会与 Findings，分别保存 track；不把 workshop 或 WACV 算入这些会议。优先跟随页面明确提供的 All Papers 链接，避免只收录默认显示的第一天。ECVA 从官方总目录按论文 URL 的会议与年份分组，当前总目录列有 ECCV 2018、2020、2022、2024。不存在于来源的 `--venue`、`--year` 或二者组合会报错，不补造双年会议届次，也不能因此推断旧届次未举办。 + +只访问论文索引与摘要 HTML/PHP 页面。作者优先取 citation 元数据，旧页面尊重声明编码；无 citation 作者时取索引中的作者行，排除论文集出版信息。DOI 只取该论文条目所属的 DOI 链接或 citation 字段；冲突会失败，不从摘要引用中抽取当前论文 ID。不请求 PDF、补充材料、视频或代码。 + +每份清单记录索引入口数、论文数、缺失摘要与作者数，缺失内容保留为空并标记 partial。存在未被标题条目解析的论文链接时会失败，防止静默漏项。逐来源的 discovery、index-status、last-run 文件分别保存发现目录、逐届结果和实际请求范围。可重复使用 `--year` 或 `--venue` 限定重试；其他届次状态会保留。官方开放目录之外的早期届次、尚未刊出的新届次和独立出版方历史数据仍需后续补齐，采集器完成不代表全历史覆盖已完成。 + ## 长期维护与更新 1. 每次采集先核对来源、许可、可访问性、空闲空间和现有运行批次。保持低并发，遵守来源限流、重试与人机验证要求。 @@ -194,6 +209,6 @@ python3 -B query.py --db /data/paper-index/papers.sqlite \ 2026-11-06 从目标服务器测试：Kaggle 元数据 GET 可下载；OpenReview 会议 group 接口可读，但 notes 接口返回 `ChallengeRequiredError`。拒稿记录导入尚受此限制，不能用录用名单替代全部投稿，也不能声称已覆盖全部拒稿。需通过平台支持的访问方式或有合法来源的公开数据导出继续补全。 -已验证的阶段性结果：ICML 2025 的 3,330 条记录和 PeerRead ICLR 2017 的 427 条记录已合并主库，合计 3,502 条 accepted 决定、255 条 rejected 决定；arXiv 曾在 529,000 条断点暂停合并，已验证恢复。这些是阶段性值，不是全量交付数据量。 +已核验的采集文件计数：ICLR 2020–2026 主会 15,534 条、COLM 2024–2026 共 1,573 条，摘要均已取得；候选范围内的 PMLR 文件为 31,428 条，已剔除两份重复出版元数据。ACL Anthology 已选文件为 63,297 条，其中 15,147 条源记录缺摘要。这些是采集文件的来源记录数，不能相加当作去重论文数，也不能代替主库实时统计；实际入库量使用 query.py --stats 查看。 -代码在 `codex/metadata-index` 工作分支交付，[Draft PR #1](https://gitea.entroia.com/Zayden-Jung/paper-atlas/pulls/1) 尚未合并，等待全量就绪后用户手动合并。已推送的前三个开发提交均有成功的远程 CI；后续提交仍需检查其各自 CI。全量导入、其他会议适配器、去重复核和覆盖率核验继续进行，应随实际结果更新此说明。 +代码在 `codex/metadata-index` 工作分支交付，[Draft PR #1](https://gitea.entroia.com/Zayden-Jung/paper-atlas/pulls/1) 尚未合并，等待全量就绪后用户手动合并。各提交分别由远程 CI 验证，结果以对应提交的 Gitea Actions 检查为准。全量导入、其他会议适配器、去重复核和覆盖率核验继续进行，应随实际结果更新此说明。 diff --git a/collect_vision.py b/collect_vision.py new file mode 100644 index 0000000..9709222 --- /dev/null +++ b/collect_vision.py @@ -0,0 +1,189 @@ +"""Collect CVF/ECVA public titles, abstracts and bibliographic identifiers.""" +import argparse +from collections import defaultdict +import json +from pathlib import Path +import re +import time +from urllib.parse import parse_qs, unquote, urljoin, urlsplit + +from collect_pmlr import fetch +from collect_proceedings import Document, Node, clean + +ROOTS={'cvf':'https://openaccess.thecvf.com/','ecva':'https://www.ecva.net/papers.php'} +HOSTS={'openaccess.thecvf.com','www.ecva.net','ecva.net'} + + +def decode_html(data): + try: + return data.decode('utf-8-sig') + except UnicodeDecodeError: + charset=re.search(br'charset\s*=\s*["\x27]?([\w-]+)',data[:4096],re.I) + if not charset: + raise + return data.decode(charset[1].decode('ascii')) + + +def publication(url): + parsed=urlsplit(url) + match=re.search(r'(CVPR|ICCV|ECCV)[_/]?(20\d{2})',parsed.path,re.I) + if parsed.netloc in HOSTS and '/html/' in parsed.path and parsed.path.endswith(('.html','.php')) and match: + return match[1].upper(),int(match[2]) + + +def doi_link(url): + parsed=urlsplit(url) + value=None + if parsed.netloc in ('doi.org','dx.doi.org'): + value=unquote(parsed.path.lstrip('/')) + elif parsed.netloc=='link.springer.com' and parsed.path.startswith('/chapter/'): + value=unquote(parsed.path.removeprefix('/chapter/')) + return value.lower() if value and re.fullmatch(r'10\.\d{4,9}/\S+',value) else None + + +def index_entries(html,index_url): + records={} + raw_count=0 + for dl in (n for n in Document(html).root.walk() if n.tag=='dl'): + current=None + for node in (n for n in dl.children if isinstance(n,Node)): + if node.tag=='dt': + current=None + links=[urljoin(index_url,n.attrs['href']) for n in node.walk() if n.tag=='a' and n.attrs.get('href') and publication(urljoin(index_url,n.attrs['href']))] + if not links: + continue + if len(set(links))!=1: + raise ValueError('Ambiguous publication links in one index title') + url=links[0];raw_count+=1 + if url in records and records[url]['index_title']!=clean(node.text()): + raise ValueError('Different index titles share a publication URL') + current=records.setdefault(url,{'url':url,'index_title':clean(node.text()),'index_authors':None,'doi':None}) + elif node.tag=='dd' and current is not None: + text=clean(node.text()) + # Author rows precede download/bibliography rows. + if current['index_authors'] is None and text and not text.startswith('['): + current['index_authors']=text + for link in (n for n in node.walk() if n.tag=='a' and clean(n.text()).casefold()=='doi'): + value=doi_link(urljoin(index_url,link.attrs.get('href',''))) + if value: + if current['doi'] and current['doi']!=value: + raise ValueError('Conflicting DOI links in one index entry') + current['doi']=value + if not records: + raise ValueError('No official paper entries found') + all_links={urljoin(index_url,n.attrs['href']) for n in Document(html).root.walk() if n.tag=='a' and n.attrs.get('href') and publication(urljoin(index_url,n.attrs['href']))} + if all_links!=set(records):raise ValueError('Paper links exist outside parsed title entries') + return list(records.values()),raw_count + + +def discover(source,cache): + root=ROOTS[source] + html=decode_html(fetch(root,cache,32*1024**2,delay=1)) + if source=='ecva': + entries,_=index_entries(html,root) + groups=defaultdict(list) + for entry in entries: + venue,year=publication(entry['url']) + if venue=='ECCV':groups[(venue,year)].append(entry) + return [{'venue':venue,'year':year,'track':'main','index_url':root,'entries':entries} + for (venue,year),entries in sorted(groups.items(),key=lambda item:item[0][1])] + indexes={} + for node in Document(html).root.walk(): + if node.tag!='a':continue + url=urljoin(root,node.attrs.get('href',''));parsed=urlsplit(url) + match=re.fullmatch(r'/(CVPR|ICCV)(20\d{2})(_findings)?(?:\.py)?/?',parsed.path) + if parsed.netloc=='openaccess.thecvf.com' and match: + indexes[url]={'venue':match[1],'year':int(match[2]),'track':'findings' if match[3] else 'main','index_url':url} + if not indexes:raise ValueError('No official CVPR/ICCV indexes found') + return sorted(indexes.values(),key=lambda row:(row['year'],row['venue'],row['track']!='main',row['track'])) + + +def parse_paper(html,entry,source,venue,year,track,index_url): + nodes=list(Document(html).root.walk());meta=defaultdict(list) + for node in nodes: + if node.tag=='meta':meta[node.attrs.get('name','').lower()].append(node.attrs.get('content','')) + title=next(iter(meta['citation_title']),None) or next((clean(n.text()) for n in nodes if n.attrs.get('id')=='papertitle'),None) + if not title:raise ValueError('Missing official paper title: '+entry['url']) + abstract=next((clean(n.text()) or None for n in nodes if n.attrs.get('id')=='abstract'),None) + authors=[] + for author in meta['citation_author']: + if ',' in author: + family,given=author.split(',',1);author=given.strip()+' '+family.strip() + if clean(author):authors.append(clean(author)) + if not authors: + raw=next((clean(n.text()) for n in nodes if n.attrs.get('id')=='authors'),'') or entry.get('index_authors') or '' + raw=raw.split(';',1)[0] + if ' and ' in raw and all(',' in name for name in raw.split(' and ')): + names=[' '.join(reversed([part.strip() for part in name.split(',',1)])) for name in raw.split(' and ')] + else:names=raw.split(',') + authors=[clean(name).removesuffix('*').strip() for name in names if clean(name).removesuffix('*').strip()] + dois={value for doi in meta['citation_doi'] if doi and (value:=doi_link(doi) or doi.strip().lower()) and re.fullmatch(r'10\.\d{4,9}/\S+',value)} + if entry.get('doi'):dois.add(entry['doi']) + if len(dois)>1:raise ValueError('Citation DOI differs from the official index') + identifiers={'doi':dois.pop()} if dois else {} + # Abstract/reference links are never interpreted as this paper's identifiers. + return {'source':source,'external_id':urlsplit(entry['url']).path,'url':entry['url'], + 'title':clean(title),'abstract':abstract,'authors':authors,'identifiers':identifiers, + 'venue':venue,'year':year,'track':track,'status':'accepted', + 'raw_decision':'listed in official proceedings','list_url':index_url, + 'metadata':{'citation':dict(meta),'index_title':entry['index_title'],'index_authors':entry.get('index_authors')}} + + +def collect_index(source,index,cache,out): + url=index['index_url'];venue=index['venue'];year=index['year'];track=index['track'] + entries=index.get('entries') + raw_count=len(entries) if entries is not None else None + if entries is None: + html=decode_html(fetch(url,cache,32*1024**2,delay=1)) + # Some defaults contain only day one; prefer the explicitly linked all-papers page. + all_pages={urljoin(url,n.attrs.get('href','')) for n in Document(html).root.walk() if n.tag=='a' and parse_qs(urlsplit(urljoin(url,n.attrs.get('href',''))).query).get('day')==['all'] and urlsplit(urljoin(url,n.attrs.get('href',''))).netloc==urlsplit(url).netloc and urlsplit(urljoin(url,n.attrs.get('href',''))).path==urlsplit(url).path} + if len(all_pages)>1:raise ValueError('Ambiguous all-papers index') + if all_pages: + url=all_pages.pop();html=decode_html(fetch(url,cache,32*1024**2,delay=1)) + entries,raw_count=index_entries(html,url) + records=[] + for number,entry in enumerate(entries,1): + if publication(entry['url'])!=(venue,year):raise ValueError('Publication path differs from requested edition') + record=parse_paper(decode_html(fetch(entry['url'],cache,2*1024**2,delay=1)),entry,source,venue,year,track,url) + records.append(record) + if number%25==0:print(json.dumps({'source':source,'venue':venue,'year':year,'track':track,'collected':number,'expected':len(entries)}),flush=True) + if not records:raise ValueError('Empty publication list') + path=Path(out)/f'{source}-{venue}{year}-{track}.jsonl';path.parent.mkdir(parents=True,exist_ok=True) + part=path.with_suffix('.part');part.write_text(''.join(json.dumps(r,ensure_ascii=False)+'\n' for r in records));part.replace(path) + missing=sum(not r['abstract'] for r in records);missing_authors=sum(not r['authors'] for r in records) + manifest={'source':source,'parser_version':1,'venue':venue,'year':year,'track':track, + 'count':len(records),'index_link_count':len(entries),'raw_index_entry_count':raw_count, + 'missing_abstract_count':missing,'missing_authors_count':missing_authors,'file':str(path), + 'list_url':url,'status':'partial' if missing or missing_authors else 'collected'} + path.with_suffix('.manifest.json').write_text(json.dumps(manifest,indent=2));print(json.dumps(manifest),flush=True) + return manifest + + +def main(): + parser=argparse.ArgumentParser();parser.add_argument('--source',choices=ROOTS,required=True) + parser.add_argument('--cache',required=True);parser.add_argument('--out',required=True) + parser.add_argument('--year',type=int,action='append');parser.add_argument('--venue',choices=['CVPR','ICCV','ECCV'],action='append') + args=parser.parse_args();out=Path(args.out);out.mkdir(parents=True,exist_ok=True) + indexes=discover(args.source,args.cache) + if args.year and set(args.year)-{r['year'] for r in indexes}:parser.error('Requested year is absent from the official source') + if args.venue and set(args.venue)-{r['venue'] for r in indexes}:parser.error('Requested venue is absent from this source') + if not any((not args.year or r['year'] in args.year) and (not args.venue or r['venue'] in args.venue) for r in indexes):parser.error('Requested venue/year combination is absent from the official source') + path=out/f'{args.source}-index-status.json';catalogue=json.loads(path.read_text()) if path.exists() else {};requested=[] + (out/f'{args.source}-discovery.json').write_text(json.dumps({'source':args.source,'root_url':ROOTS[args.source], + 'discovered_at':time.strftime('%Y-%m-%dT%H:%M:%SZ',time.gmtime()),'indexes':[ + {**{k:index[k] for k in ('venue','year','track','index_url')}, + 'index_link_count':len(index['entries']) if 'entries' in index else None} for index in indexes]},indent=2)) + for index in indexes: + if args.year and index['year'] not in args.year or args.venue and index['venue'] not in args.venue:continue + key=f'{index["venue"]}:{index["year"]}:{index["track"]}';requested.append(key) + try: + manifest=collect_index(args.source,index,args.cache,args.out);catalogue[key]={'status':manifest['status'],'files':[manifest]} + except Exception as exc: + catalogue[key]={'status':'blocked','year':index['year'],'venue':index['venue'],'track':index['track'],'index_url':index['index_url'],'error':str(exc)} + print(json.dumps(catalogue[key]),flush=True) + part=path.with_suffix('.part');part.write_text(json.dumps(catalogue,indent=2));part.replace(path) + (out/f'{args.source}-last-run.json').write_text(json.dumps({'requested_indexes':requested,'year_filter':args.year,'venue_filter':args.venue},indent=2)) + if any(catalogue[key]['status']=='blocked' for key in requested):raise SystemExit(1) + + +if __name__=='__main__':main() diff --git a/test_vision.py b/test_vision.py new file mode 100644 index 0000000..e236bc4 --- /dev/null +++ b/test_vision.py @@ -0,0 +1,72 @@ +import tempfile +import unittest +from unittest.mock import patch + +from collect_vision import collect_index, decode_html, discover, index_entries, parse_paper + + +class VisionTests(unittest.TestCase): + def entry(self): + return {'url':'https://www.ecva.net/papers/eccv_2024/papers_ECCV/html/1_ECCV_2024_paper.php','index_title':'Paper','index_authors':'Jane Doe*, John Smith','doi':'10.1007/example_1'} + + def test_ecva_doi_belongs_to_its_own_title_entry(self): + html='
    First paper
    Jane Doe*, John Smith
    DOI
    Second paper
    Other Author
    DOI
    ' + entries,count=index_entries(html,'https://www.ecva.net/papers.php') + self.assertEqual(count,2) + self.assertEqual([e['doi'] for e in entries],['10.1007/example_1','10.1007/example_2']) + html='
    Paper
    Abstract citing another paper and a reference.
    ' + row=parse_paper(html,entries[0],'ecva','ECCV',2024,'main','https://www.ecva.net/papers.php') + self.assertEqual(row['identifiers'],{'doi':'10.1007/example_1'}) + self.assertEqual(row['authors'],['Jane Doe','John Smith']) + + def test_cvf_meta_authors_and_missing_abstract_never_capture_footer(self): + html='
    Do not include the proceedings citation

    Copyright footer

    ' + entry=dict(self.entry(),doi=None) + row=parse_paper(html,entry,'cvf','CVPR',2013,'main','https://openaccess.thecvf.com/CVPR2013') + self.assertEqual(row['authors'],['Jane Doe']) + self.assertIsNone(row['abstract']) + self.assertEqual(row['title'],'A vision paper') + + def test_conflicting_self_doi_is_rejected(self): + html='' + with self.assertRaisesRegex(ValueError,'DOI differs'): + parse_paper(html,self.entry(),'ecva','ECCV',2024,'main','https://www.ecva.net/papers.php') + + def test_legacy_ecva_author_formats_do_not_become_extra_people(self): + entry=dict(self.entry(),index_authors='Doe, Jane and Smith, John') + page='
    Paper
    Jane Doe, John Smith; The European Conference on Computer Vision (ECCV), 2018, pp. 1-12
    ' + row=parse_paper(page,entry,'ecva','ECCV',2018,'main','https://www.ecva.net/papers.php') + self.assertEqual(row['authors'],['Jane Doe','John Smith']) + row=parse_paper('
    Paper
    ',entry,'ecva','ECCV',2018,'main','https://www.ecva.net/papers.php') + self.assertEqual(row['authors'],['Jane Doe','John Smith']) + + def test_cvpr_day_one_is_replaced_by_explicit_all_papers_index(self): + root='https://openaccess.thecvf.com/CVPR2026';all_url=root+'?day=all' + first='https://openaccess.thecvf.com/content/CVPR2026/html/First_CVPR_2026_paper.html' + second='https://openaccess.thecvf.com/content/CVPR2026/html/Second_CVPR_2026_paper.html' + def item(url):return f'
    Paper
    Jane Doe, John Smith
    [pdf]
    ' + pages={root:f'All papers
    {item(first)}
    ',all_url:f'
    {item(first)}{item(second)}
    ',first:'
    First paper
    First abstract
    ',second:'
    Second paper

    Footer without an abstract

    '} + with tempfile.TemporaryDirectory() as tmp,patch('collect_vision.fetch',side_effect=lambda url,*a,**kw:pages[url].encode()) as fetch: + manifest=collect_index('cvf',{'venue':'CVPR','year':2026,'track':'main','index_url':root},tmp,tmp) + self.assertEqual(manifest['count'],2) + self.assertEqual(manifest['list_url'],all_url) + self.assertEqual(manifest['status'],'partial') + self.assertEqual(manifest['missing_abstract_count'],1) + self.assertEqual({call.args[0] for call in fetch.call_args_list},{root,all_url,first,second}) + + def test_discovery_keeps_findings_separate_and_excludes_workshops(self): + html='MainFindingsMainLegacy mainWorkshopsWACV' + with patch('collect_vision.fetch',return_value=html.encode()):indexes=discover('cvf','unused') + self.assertEqual({(x['venue'],x['year'],x['track']) for x in indexes},{('CVPR',2013,'main'),('CVPR',2026,'main'),('CVPR',2026,'findings'),('ICCV',2025,'main')}) + + def test_unparsed_publication_link_does_not_silently_reduce_coverage(self): + html='Other paper
    Paper
    ' + with self.assertRaisesRegex(ValueError,'outside parsed'): + index_entries(html,'https://openaccess.thecvf.com/') + + def test_legacy_declared_encoding_is_respected(self): + html='
    André
    ' + self.assertIn('André',decode_html(html.encode('iso-8859-1'))) + + +if __name__=='__main__':unittest.main() diff --git a/venues.json b/venues.json index ca25bc6..1784f8a 100644 --- a/venues.json +++ b/venues.json @@ -235,8 +235,7 @@ "venue_type": "conference", "discovery_sources": [ "cvf", - "ecva", - "openreview" + "dblp" ], "scope": "candidate_pending_scope_review", "years": "all_publicly_available", @@ -251,8 +250,7 @@ "venue_type": "conference", "discovery_sources": [ "cvf", - "ecva", - "openreview" + "dblp" ], "scope": "candidate_pending_scope_review", "years": "all_publicly_available", @@ -266,9 +264,8 @@ "field": "vision", "venue_type": "conference", "discovery_sources": [ - "cvf", "ecva", - "openreview" + "dblp" ], "scope": "candidate_pending_scope_review", "years": "all_publicly_available", -- 2.52.0 From a7a0729df206a8008878269462554fd7e01df414 Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 13:29:30 +0800 Subject: [PATCH 11/20] Download fixed metadata snapshots with checksum and resume safeguards --- .gitignore | 6 +++ README.md | 29 +++++++++++ download_snapshot.py | 106 ++++++++++++++++++++++++++++++++++++++ test_download_snapshot.py | 71 +++++++++++++++++++++++++ 4 files changed, 212 insertions(+) create mode 100644 download_snapshot.py create mode 100644 test_download_snapshot.py diff --git a/.gitignore b/.gitignore index adfd07a..f0d97c8 100644 --- a/.gitignore +++ b/.gitignore @@ -3,8 +3,14 @@ __pycache__/ *.sqlite *.sqlite-* *.zip +*.gz +*.bz2 +*.jsonl *.part +*.part.json +*.download.lock .env .DS_Store raw/ data/ +collected/ diff --git a/README.md b/README.md index af7cb53..9d5f8f2 100644 --- a/README.md +++ b/README.md @@ -127,6 +127,35 @@ python3 -B collect_peerread.py --cache /data/paper-index/raw/http-cache \ 数据任务以一次性 systemd 服务运行：`paper-atlas-arxiv-import`、`paper-atlas-pmlr-collect`、`paper-atlas-peerread-collect`；`paper-atlas-pmlr-verify` 用于全索引复核，`paper-atlas-conference-merge` 用于单写入合并。它们不是定时任务。主库只允许一个写入进程；会议采集先写独立 JSONL，待 arXiv 主批次结束后再合并；也可以使用 `merge_conferences.py` 在保存断点处暂停本项目的 arXiv unit、顺序合并会议清单并恢复。该操作应作为独立的一次性 systemd 任务运行，避免 SSH 断开影响恢复；不能同时启动多个合并任务。运行日志位于 `/data/paper-index/*-import.log`、`/data/paper-index/*-collect.log`。任何服务启动成功都不等于批次已完成，需检查批次状态、清单与完整性结果。标签按来源刷新使用 `paper_tags_source_evidence` 索引，避免每条论文都扫描整张标签表。运行中的 arXiv 可以通过同一协调脚本的 `--schema-only` 模式在断点更新索引后恢复；此模式要求已有 arXiv unit 正在运行。 +### 固定版本的补充元数据快照 + +`download_snapshot.py` 流式下载到 `.part`，核对完整 HTTP 响应、预期大小和可信来源发布的校验和后，才发布正式文件及 manifest。已存在的正式文件会先重新核验；错误校验和不会覆盖它。默认保留 8 GiB 空闲空间，同一输出只允许一个下载进程。 + +DBLP 官方月度快照用于补充出版目录和历史届次，元数据按 CC0 发布。固定版本可复现，不把每日变动下载链接当作同一快照： + +```sh +python3 -B download_snapshot.py \ + --url https://drops.dagstuhl.de/storage/artifacts/dblp/xml/2026/dblp-2026-11-01.xml.gz \ + --md5 18ad5db59592406ec74289975c4b44cb \ + --output /data/paper-index/raw/dblp-2026-11-01.xml.gz +``` + +[官方版本与校验值](https://doi.org/10.4230/dblp.xml.2026-11-01)。下载器另外计算 SHA-256，记录来源 URL、大小和核验时间。DBLP 没有提供的摘要不能编造，出版目录也不能作为拒稿证据。 + +公开 ICLR 投稿快照的固定 Release 供后续筛选投稿和明确决定；它是第三方保存的 OpenReview 历史副本，不等于最新官方全量数据： + +```sh +python3 -B download_snapshot.py \ + --url https://github.com/qhjqhj00/iclr-openreview-reviews/releases/download/v1.0/iclr.tar.bz2 \ + --sha256 350071daf25ebf1aef854ee1d5dc3a81834e1c468917c82a4f78af5548de34fc \ + --expected-bytes 512940722 \ + --output /data/paper-index/raw/iclr-openreview-v1.0.tar.bz2 +``` + +[快照来源与复现脚本](https://github.com/qhjqhj00/iclr-openreview-reviews)。原包包含评审回复；后续解析只筛选公开投稿的索引、标题、摘要、平台标签及带证据的明确决定，不提取评审正文、作者邮件或论文全文。必须核对原始 invitation、readers、签名及决定字段，不能根据评分或录用名单缺席推断拒稿。该解析适配尚在开发，下载成功不代表已经入库或核验覆盖。 + +下载异常后先检查对应进程或 CI 任务确已终止，再使用同一参数增加 `--resume`。只有断点身份一致且服务器实际返回正确 Range 响应才续传；不支持时保留断点并报错，不自动重下。正式文件与下载清单均放在仓库外，不提交大型来源包。 + ### COLM 与 NeurIPS 官方摘要 ```sh diff --git a/download_snapshot.py b/download_snapshot.py new file mode 100644 index 0000000..4f40848 --- /dev/null +++ b/download_snapshot.py @@ -0,0 +1,106 @@ +"""Checksum-verified streaming downloads; existing valid snapshots stay intact.""" +import argparse +import fcntl +import hashlib +import json +from pathlib import Path +import re +import shutil +import time +from urllib.parse import urlsplit +import urllib.request + + +def digest_file(path): + sha=hashlib.sha256();md5=hashlib.md5(usedforsecurity=False);size=0 + with path.open('rb') as stream: + while block:=stream.read(1024**2): + sha.update(block);md5.update(block);size+=len(block) + return {'sha256':sha.hexdigest(),'md5':md5.hexdigest(),'bytes':size} + + +def verify(actual,sha256,md5,expected_bytes): + if expected_bytes is not None and actual['bytes']!=expected_bytes: + raise ValueError('Snapshot size differs from trusted expected size') + if sha256 and actual['sha256']!=sha256: + raise ValueError('Snapshot SHA256 mismatch; partial file retained') + if md5 and actual['md5']!=md5: + raise ValueError('Snapshot MD5 mismatch; partial file retained') + + +def write_manifest(output,url,actual): + manifest={'status':'verified','file':str(output),'url':url,**actual, + 'verified_at':time.strftime('%Y-%m-%dT%H:%M:%SZ',time.gmtime())} + metadata=Path(str(output)+'.manifest.json');stage=Path(str(metadata)+'.part') + stage.write_text(json.dumps(manifest,indent=2));stage.replace(metadata) + return manifest + + +def download(url,output,sha256=None,md5=None,expected_bytes=None,resume=False,min_free=8*1024**3): + parsed=urlsplit(url) + if parsed.scheme!='https' or not parsed.netloc or parsed.username or parsed.password: + raise ValueError('A public HTTPS URL without embedded credentials is required') + if not (sha256 or md5):raise ValueError('A trusted published checksum is required') + if expected_bytes is not None and expected_bytes<=0:raise ValueError('Expected byte count must be positive') + if min_free<0:raise ValueError('Disk reserve must not be negative') + for name,value,length in [('SHA256',sha256,64),('MD5',md5,32)]: + if value and not re.fullmatch('[0-9a-f]{'+str(length)+'}',value):raise ValueError('Invalid '+name) + output=Path(output);output.parent.mkdir(parents=True,exist_ok=True) + if output.is_symlink():raise ValueError('Refusing a symlink output') + intent={'url':url,'expected_sha256':sha256,'expected_md5':md5,'expected_bytes':expected_bytes} + with Path(str(output)+'.download.lock').open('a') as lock: + fcntl.flock(lock,fcntl.LOCK_EX|fcntl.LOCK_NB) + if output.exists(): + actual=digest_file(output);verify(actual,sha256,md5,expected_bytes) + if not Path(str(output)+'.manifest.json').exists():write_manifest(output,url,actual) + return {'status':'already_verified','file':str(output),**actual} + part=Path(str(output)+'.part');intent_path=Path(str(part)+'.json') + offset=0 + if part.exists(): + if not resume or not intent_path.is_file() or json.loads(intent_path.read_text())!=intent: + raise ValueError('Existing partial requires --resume and the same snapshot identity') + offset=part.stat().st_size + else: + intent_path.write_text(json.dumps(intent,indent=2)) + remaining=max(0,(expected_bytes or 0)-offset) + if shutil.disk_usage(output.parent).freetotal:raise ValueError('HTTP response exceeds advertised size') + if written-reported>=64*1024**2: + print(json.dumps({'status':'downloading','file':str(output),'bytes':written,'expected_bytes':total}),flush=True);reported=written + if shutil.disk_usage(output.parent).free Date: Fri, 6 Nov 2026 14:05:37 +0800 Subject: [PATCH 12/20] Preserve canonical submission identities and decision source evidence --- .gitignore | 1 + README.md | 22 ++++- backfill_submission_evidence.py | 63 ++++++++++++ collect_iclr.py | 3 +- import_records.py | 17 +++- index_core.py | 24 +++-- merge_conferences.py | 49 +++++++--- query.py | 19 +++- schema.sql | 27 ++++++ submission_evidence.py | 110 +++++++++++++++++++++ test_iclr.py | 2 +- test_index_core.py | 1 + test_maintenance.py | 41 ++++++++ test_submission_evidence.py | 163 ++++++++++++++++++++++++++++++++ 14 files changed, 515 insertions(+), 27 deletions(-) create mode 100644 backfill_submission_evidence.py create mode 100644 submission_evidence.py create mode 100644 test_maintenance.py create mode 100644 test_submission_evidence.py diff --git a/.gitignore b/.gitignore index f0d97c8..b0a58ff 100644 --- a/.gitignore +++ b/.gitignore @@ -9,6 +9,7 @@ __pycache__/ *.part *.part.json *.download.lock +*.maintenance.lock .env .DS_Store raw/ diff --git a/README.md b/README.md index 9d5f8f2..73b9b4e 100644 --- a/README.md +++ b/README.md @@ -55,7 +55,9 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 | `source_records` / `sources` | 平台记录、原平台标题摘要与索引元数据、来源 URL 和抓取时间 | | `venues` / `venue_aliases` / `editions` | 会议或期刊系列、年份与 track，区分主会、Findings 和 workshop，联合会议用别名查询，不重复生成事件 | | `edition_sources` | 每届每个来源的覆盖状态、预计/已导入数量、错误与断点 | -| `submissions` / `decisions` | 多次投稿及带证据的决定历史，支持改投、撤稿和决定变更 | +| `submissions` / `submission_identifiers` / `submission_sources` | 投稿事件、按届次限定的原平台强标识、多来源证据；同稿的不同投稿尝试分别保留 | +| `decisions` / `decision_sources` | 明确决定、原始决定日期、来源观察日期及证据，不把缺失决定编造成拒稿 | +| `submission_merges` / `decision_merges` | 投稿事件与重复决定观察的合并审计，保留来源关联并避免重用已合并的编号 | | `tags` / `paper_tags` | 平台标签与人工标签，使用命名空间隔离 | | `match_candidates` / `paper_merges` | 潜在重复复核状态、确定合并审计 | | `ingest_runs` | 导入批次、状态、错误和恢复位置 | @@ -127,6 +129,24 @@ python3 -B collect_peerread.py --cache /data/paper-index/raw/http-cache \ 数据任务以一次性 systemd 服务运行：`paper-atlas-arxiv-import`、`paper-atlas-pmlr-collect`、`paper-atlas-peerread-collect`；`paper-atlas-pmlr-verify` 用于全索引复核，`paper-atlas-conference-merge` 用于单写入合并。它们不是定时任务。主库只允许一个写入进程；会议采集先写独立 JSONL，待 arXiv 主批次结束后再合并；也可以使用 `merge_conferences.py` 在保存断点处暂停本项目的 arXiv unit、顺序合并会议清单并恢复。该操作应作为独立的一次性 systemd 任务运行，避免 SSH 断开影响恢复；不能同时启动多个合并任务。运行日志位于 `/data/paper-index/*-import.log`、`/data/paper-index/*-collect.log`。任何服务启动成功都不等于批次已完成，需检查批次状态、清单与完整性结果。标签按来源刷新使用 `paper_tags_source_evidence` 索引，避免每条论文都扫描整张标签表。运行中的 arXiv 可以通过同一协调脚本的 `--schema-only` 模式在断点更新索引后恢复；此模式要求已有 arXiv unit 正在运行。 +### 投稿事件身份与旧数据回填 + +论文实体与投稿事件分开去重。同一篇论文可以在同一年向同一会议投稿多次，不能仅凭 paper_id、会议、标题相同就把投稿历史压成一次。跨来源只有原平台 forum ID、已验证的投稿编号，或正式发表稿的同届同 track 出版 DOI 等强证据才能关联同一事件。程序展示的内部编号仅在经过独立核验并带有 program_submission_number_verified 标记时才能用作投稿编号。 + +新版 ICLR manifest 的 parser_version 为 2，明确区别投稿编号、forum 和展示事件编号。旧版文件应先用缓存重新采集；事件证据回填只接受已入库、与 manifest 的会议/年份/track 和数量一致的文件。队列 JSON 包含 files 及对应的 file_sha256，文件变化后必须重新核验，不能继续使用旧哈希。 + +```sh +python3 -B merge_conferences.py --db /data/paper-index/papers.sqlite \ + --archive /data/paper-index/raw/arxiv-metadata.zip \ + --backfill-queue /data/paper-index/collected/submission-evidence-queue.json +``` + +协调器只维护本项目的 arXiv unit，并对维护过程加互斥锁。若 arXiv 正在运行，会保存断点、应用 schema、回填来源身份、导入本次附加文件、检查外键后恢复；若批次已经读至 EOF 并标记 complete，则不再启动它。若导入停下但未完成，会报错要求先调查，不以“unit 不存在”推断成功或重新全量运行。应以一次性 systemd 任务运行该命令，保持单写入。 + +首次升级会给已有投稿与决定补上来源关联；身份回填不重新导入标题摘要，不更改原抓取时间。确认同一次投稿后，可以合并镜像事件并保留所有决定及来源。当前稿件的原始强标识也会写入 source_records 元数据，后续源 URL 更新会同步保存。 + +通用 JSONL 导入支持 decision_events 数组及 observed_at、decision_date、submitted_at。没有取得明确决定时，可用 unknown 状态和空数组仅保存投稿，不生成虚假的决定；`--status unknown` 也能查询这些投稿。历史快照使用其实际观察时间，不能因今天导入就覆盖较新来源的状态；同一观察时间内按原始决定日期排序。反复导入相同证据只补来源关联，不重复制造决定事件。全量结束后仍须单独核验重复候选、冲突决定及来源覆盖，迁移检查不能代替最终全库审计。 + ### 固定版本的补充元数据快照 `download_snapshot.py` 流式下载到 `.part`，核对完整 HTTP 响应、预期大小和可信来源发布的校验和后，才发布正式文件及 manifest。已存在的正式文件会先重新核验；错误校验和不会覆盖它。默认保留 8 GiB 空闲空间，同一输出只允许一个下载进程。 diff --git a/backfill_submission_evidence.py b/backfill_submission_evidence.py new file mode 100644 index 0000000..6f84430 --- /dev/null +++ b/backfill_submission_evidence.py @@ -0,0 +1,63 @@ +"""Attach authoritative identities to existing events without reimporting text.""" +import argparse +import hashlib +import json +from pathlib import Path + +from index_core import Store +from submission_evidence import attach_submission,identities_for_row + + +def queue_files(path): + queue=json.loads(Path(path).read_text()) + files=queue['files'] + if len(files)!=len(set(files)):raise ValueError('Duplicate paths in evidence queue') + for filename in files: + if hashlib.sha256(Path(filename).read_bytes()).hexdigest()!=queue['file_sha256'][filename]: + raise ValueError('Evidence file differs from audited queue: '+filename) + return files + + +def backfill(db,files): + for filename in files: + manifest=json.loads(Path(filename).with_suffix('.manifest.json').read_text()) + if manifest['source'] in ('colm','neurips','iclr_program') and manifest.get('parser_version',0)<2: + raise ValueError('Recollect old parser output before submission evidence backfill') + store=Store(db);c=store.conn;count=0 + before=c.execute('SELECT count(*) FROM submissions').fetchone()[0] + try: + for filename in files: + path=Path(filename);manifest=json.loads(path.with_suffix('.manifest.json').read_text());seen=set() + with path.open() as stream: + for line in stream: + row=json.loads(line) + if row['external_id'] in seen:raise ValueError('Repeated source ID in evidence file') + seen.add(row['external_id']) + if any(row.get(key)!=manifest.get(key) for key in ('source','venue','year','track')): + raise ValueError('Evidence row differs from edition manifest') + record=c.execute('SELECT r.id,r.paper_id,r.metadata_json FROM source_records r JOIN sources s ON s.id=r.source_id WHERE s.name=? AND r.external_id=?',(row['source'],row['external_id'])).fetchone() + edition=c.execute('SELECT e.id FROM editions e JOIN venues v ON v.id=e.venue_id WHERE v.acronym=? AND e.year=? AND e.track=?',(row['venue'],row['year'],row['track'])).fetchone() + if not record or not edition:raise ValueError('Evidence backfill requires an already imported record and edition') + if not c.execute('SELECT 1 FROM submission_sources ss JOIN submissions s ON s.id=ss.submission_id WHERE ss.source_record_id=? AND ss.source_submission_id=? AND s.edition_id=?',(record[0],row['external_id'],edition[0])).fetchone(): + raise ValueError('No existing submission to backfill') + attach_submission(store,edition[0],record[1],record[0],row['external_id'],identities_for_row(row)) + metadata=json.loads(record[2]);metadata['record_identifiers_raw']=row.get('identifiers') or {} + if row['source']=='iclr_program': + for key in ('program_identity_type','program_submission_number_verified','program_identity_verification'): + if key in row['metadata']:metadata[key]=row['metadata'][key] + c.execute('UPDATE source_records SET metadata_json=? WHERE id=?',(json.dumps(metadata,ensure_ascii=False),record[0])) + count+=1 + if len(seen)!=manifest['count']:raise ValueError('Evidence count differs from manifest') + conflict=c.execute('SELECT 1 FROM submission_sources ss JOIN submissions s ON s.id=ss.submission_id JOIN source_records r ON r.id=ss.source_record_id WHERE s.paper_id!=r.paper_id LIMIT 1').fetchone() + if conflict:raise ValueError('Submission evidence points to a different paper entity') + c.commit() + result={'status':'backfilled','records':count,'submissions_before':before,'submissions_after':c.execute('SELECT count(*) FROM submissions').fetchone()[0]} + print(json.dumps(result),flush=True);return result + except BaseException: + c.rollback();raise + finally:c.close() + + +if __name__=='__main__': + parser=argparse.ArgumentParser();parser.add_argument('--db',required=True);parser.add_argument('--queue',required=True) + args=parser.parse_args();backfill(args.db,queue_files(args.queue)) diff --git a/collect_iclr.py b/collect_iclr.py index ef0978f..6eb4eb0 100644 --- a/collect_iclr.py +++ b/collect_iclr.py @@ -152,6 +152,7 @@ def parse_program(data, abstracts, year, index_url): 'tags': [{'namespace': SOURCE+':'+kind, 'value': value} for kind, values in [('topic', topics), ('keyword', keywords)] for value in values], 'metadata': {'program_event_ids': presentation_ids, + 'program_identity_type': identity[0], 'presentation_titles': sorted({clean(r['name']) for r in group}), 'program_source_id': identity[1], 'presentation_types': sorted({str(r.get('event_type') or r.get('eventtype')) for r in group}), 'raw_decisions': sorted({str(r.get('decision')) for r in group if r.get('decision')}), @@ -199,7 +200,7 @@ def collect_year(year, cache, out): part = path.with_suffix('.part') part.write_text(''.join(json.dumps(row, ensure_ascii=False)+'\n' for row in records)) part.replace(path) - manifest = {'source': SOURCE, 'parser_version': 1, 'venue': 'ICLR', 'year': year, 'track': 'main', + manifest = {'source': SOURCE, 'parser_version': 2, 'venue': 'ICLR', 'year': year, 'track': 'main', 'count': len(records), 'list_url': page, 'asset_urls': assets, 'abstract_asset_status': abstract_asset_status, 'file': str(path), 'status': 'partial' if audit['missing_abstract_count'] else 'collected', **audit} diff --git a/import_records.py b/import_records.py index eaed872..fd5c61d 100644 --- a/import_records.py +++ b/import_records.py @@ -5,6 +5,7 @@ import time from pathlib import Path from index_core import Store +from submission_evidence import identities_for_row,timestamp def ingest_file(db,path): @@ -13,6 +14,8 @@ def ingest_file(db,path): manifest = json.loads(manifest_path.read_text()) if manifest_path.exists() else {} if manifest.get('source') in ('colm','neurips') and manifest.get('parser_version',0)<2: raise ValueError('Recollect proceedings with parser_version>=2 before identity matching') + if manifest.get('source')=='iclr_program' and manifest.get('parser_version',0)<2: + raise ValueError('Recollect ICLR with parser_version>=2 before submission identity matching') store = Store(db) c = store.conn source_name = manifest.get('source') @@ -41,7 +44,19 @@ def ingest_file(db,path): if 'tags' in row: store.record_tags(paper,record,source_name,row['tags']) if row.get('status') in ('accepted','rejected','desk_rejected','withdrawn','under_review','unknown'): - edition = store.decision(paper,record,row['venue'],row['year'],row['track'],external,row.get('evidence_url',row['url']),row['status'],row.get('raw_decision','listed in official proceedings' if source_name=='pmlr' else row['status']),row.get('venue_aliases')) + identities=identities_for_row(row) + events=row.get('decision_events') + if events is None: + events=[{'status':row['status'],'raw_value':row.get('raw_decision','listed in official proceedings' if source_name=='pmlr' else row['status']),'evidence_url':row.get('evidence_url',row['url'])}] + if not isinstance(events,list):raise ValueError('Decision events must be a list') + if not events and row['status']!='unknown':raise ValueError('Known status requires decision evidence') + if row['status']!='unknown' and not any(event.get('status')==row['status'] for event in events):raise ValueError('Summary status has no corresponding decision evidence') + row_observed=row.get('observed_at') or timestamp() + for event in events or [{'status':None}]: + status=event['status'] + if events and status is None:raise ValueError('Explicit decision event requires a status') + if status is not None and status not in ('accepted','rejected','desk_rejected','withdrawn','under_review','unknown'):raise ValueError('Unsupported decision event') + edition=store.decision(paper,record,row['venue'],row['year'],row['track'],external,event.get('evidence_url',row['url']),status,event.get('raw_value',''),row.get('venue_aliases'),identities=identities,observed_at=event.get('observed_at') or row_observed,decision_date=event.get('decision_date'),source_decision_id=event.get('source_decision_id',''),submitted_at=row.get('submitted_at')) editions.add(edition) else: raise ValueError('This importer requires an explicit supported decision status') diff --git a/index_core.py b/index_core.py index 7227a30..df7ce7f 100644 --- a/index_core.py +++ b/index_core.py @@ -4,6 +4,7 @@ import re import sqlite3 import unicodedata from pathlib import Path +from submission_evidence import attach_submission,migrate_sources,write_decision def normalized(value): @@ -31,6 +32,7 @@ class Store: self.conn.executescript(Path(__file__).with_name('schema.sql').read_text()) self.conn.execute('PRAGMA synchronous=NORMAL') self.conn.execute('PRAGMA cache_size=-32768') + migrate_sources(self.conn) def source(self,name,url): self.conn.execute('INSERT OR IGNORE INTO sources(name,homepage) VALUES(?,?)',(name,url)) @@ -40,6 +42,12 @@ class Store: c = self.conn if target == other: return + columns=('abstract','authors_json','first_published') + target_values=c.execute('SELECT abstract,authors_json,first_published FROM papers WHERE id=?',(target,)).fetchone() + other_values=c.execute('SELECT abstract,authors_json,first_published FROM papers WHERE id=?',(other,)).fetchone() + if not target_values or not other_values:raise ValueError('Missing paper entity during merge') + fill=[(column,value) for column,current,value in zip(columns,target_values,other_values) if current in (None,'','[]') and value not in (None,'','[]')] + if fill:c.execute('UPDATE papers SET '+','.join(column+'=?' for column,_ in fill)+' WHERE id=?',(*[value for _,value in fill],target)) c.execute('INSERT INTO paper_merges(former_paper_id,surviving_paper_id,reason) VALUES(?,?,?)',(other,target,reason)) c.execute('UPDATE identifiers SET paper_id=? WHERE paper_id=?',(target,other)) c.execute('UPDATE source_records SET paper_id=? WHERE paper_id=?',(target,other)) @@ -102,7 +110,8 @@ class Store: paper = c.execute('INSERT INTO papers(title,abstract,title_key,authors_json,first_published) VALUES(?,?,?,?,?)',(title,abstract,key,json.dumps(authors,ensure_ascii=False),published)).lastrowid for scheme,value in ids: c.execute('INSERT OR IGNORE INTO identifiers VALUES(?,?,?)',(scheme,value,paper)) - c.execute('INSERT INTO source_records(source_id,external_id,paper_id,url,title,abstract,authors_json,metadata_json) VALUES(?,?,?,?,?,?,?,?) ON CONFLICT(source_id,external_id) DO UPDATE SET paper_id=excluded.paper_id,title=excluded.title,abstract=excluded.abstract,authors_json=excluded.authors_json,metadata_json=excluded.metadata_json,retrieved_at=CURRENT_TIMESTAMP',(source_id,external_id,paper,url,title,abstract,json.dumps(authors,ensure_ascii=False),json.dumps(metadata or {},ensure_ascii=False,default=str))) + metadata=dict(metadata or {});metadata['record_identifiers']=dict(ids) + c.execute('INSERT INTO source_records(source_id,external_id,paper_id,url,title,abstract,authors_json,metadata_json) VALUES(?,?,?,?,?,?,?,?) ON CONFLICT(source_id,external_id) DO UPDATE SET paper_id=excluded.paper_id,url=excluded.url,title=excluded.title,abstract=excluded.abstract,authors_json=excluded.authors_json,metadata_json=excluded.metadata_json,retrieved_at=CURRENT_TIMESTAMP',(source_id,external_id,paper,url,title,abstract,json.dumps(authors,ensure_ascii=False),json.dumps(metadata,ensure_ascii=False,default=str))) record = c.execute('SELECT id FROM source_records WHERE source_id=? AND external_id=?',(source_id,external_id)).fetchone()[0] for other,_,_ in title_matches: if other != paper and c.execute('SELECT 1 FROM papers WHERE id=?',(other,)).fetchone(): @@ -125,7 +134,7 @@ class Store: tag=c.execute('SELECT id FROM tags WHERE namespace=? AND value=?',(namespace,value)).fetchone()[0] c.execute('INSERT OR IGNORE INTO paper_tags VALUES(?,?,?,?)',(paper,tag,record,evidence)) - def decision(self,paper,record,venue,year,track,external_id,url,status,raw_value,aliases=None): + def decision(self,paper,record,venue,year,track,external_id,url,status,raw_value,aliases=None,identities=None,observed_at=None,decision_date=None,source_decision_id='',submitted_at=None): c = self.conn c.execute("INSERT OR IGNORE INTO venues(acronym,name,field,venue_type) VALUES(?,?,?,'conference')",(venue,venue,'unclassified')) venue_id = c.execute('SELECT id FROM venues WHERE acronym=?',(venue,)).fetchone()[0] @@ -133,10 +142,13 @@ class Store: c.execute('INSERT OR IGNORE INTO venue_aliases VALUES(?,?)',(venue_id,alias)) c.execute('INSERT OR IGNORE INTO editions(venue_id,year,track) VALUES(?,?,?)',(venue_id,int(year),track)) edition = c.execute('SELECT id FROM editions WHERE venue_id=? AND year=? AND track=?',(venue_id,int(year),track)).fetchone()[0] - c.execute('INSERT OR IGNORE INTO submissions(paper_id,edition_id,source_record_id,source_submission_id) VALUES(?,?,?,?)',(paper,edition,record,external_id)) - submission = c.execute('SELECT id FROM submissions WHERE source_record_id=? AND edition_id=? AND source_submission_id=?',(record,edition,external_id)).fetchone()[0] - if not c.execute('SELECT 1 FROM decisions WHERE submission_id=? AND status=? AND raw_value=? AND evidence_url=?',(submission,status,raw_value,url)).fetchone(): - c.execute("INSERT INTO decisions(submission_id,status,raw_value,evidence_url,observed_at) VALUES(?,?,?,?,strftime('%Y-%m-%dT%H:%M:%fZ','now'))",(submission,status,raw_value,url)) + # A previous event in this row may already have merged the input paper. + paper=c.execute('SELECT paper_id FROM source_records WHERE id=?',(record,)).fetchone()[0] + submission=attach_submission(self,edition,paper,record,external_id,identities or []) + if submitted_at: + from submission_evidence import timestamp + c.execute('UPDATE submissions SET submitted_at=COALESCE(submitted_at,?) WHERE id=?',(timestamp(submitted_at),submission)) + if status is not None:write_decision(self,submission,record,status,raw_value,url,observed_at,decision_date,source_decision_id) return edition def accepted(self,paper,record,venue,year,track,external_id,url): diff --git a/merge_conferences.py b/merge_conferences.py index d971af9..0d07a41 100644 --- a/merge_conferences.py +++ b/merge_conferences.py @@ -6,12 +6,14 @@ Only the Paper Atlas arXiv unit is controlled; business services are untouched. import argparse from contextlib import closing import json +import fcntl import sqlite3 import subprocess from pathlib import Path from import_records import ingest_file from index_core import Store +from backfill_submission_evidence import backfill,queue_files UNIT='paper-atlas-arxiv-import.service' @@ -21,29 +23,37 @@ def property_value(name): return subprocess.check_output(['systemctl','show',UNIT,'-p',name,'--value'],text=True).strip() -def merge(db,archive,files): +def merge_locked(db,archive,files,backfill_queue=None): # Validate inputs before stopping the existing writer. for filename in files: path=Path(filename) if not path.is_file() or not path.with_suffix('.manifest.json').is_file(): raise ValueError('Missing collected file or manifest: '+filename) + evidence_files=queue_files(backfill_queue) if backfill_queue else [] if not Path(archive).is_file(): raise ValueError('Missing arXiv archive') if not Path(db).is_file(): raise ValueError('Missing existing database') - if property_value('ActiveState')!='active': - raise RuntimeError('Expected the existing arXiv unit to be active; inspect it before proceeding') - subprocess.run(['systemctl','stop',UNIT],check=True) + resume_arxiv=property_value('ActiveState')=='active' + if not resume_arxiv: + with closing(sqlite3.connect(Path(db).resolve().as_uri()+'?mode=ro',uri=True)) as c: + last=c.execute("SELECT status FROM ingest_runs WHERE source_id=(SELECT id FROM sources WHERE name='arxiv') ORDER BY id DESC LIMIT 1").fetchone() + if not last or last[0]!='complete': + raise RuntimeError('arXiv is neither active nor verified complete; inspect its state before maintenance') try: + if resume_arxiv:subprocess.run(['systemctl','stop',UNIT],check=True) with closing(sqlite3.connect(db)) as c, c: - checkpoint=c.execute("SELECT records_written FROM ingest_runs WHERE source_id=(SELECT id FROM sources WHERE name='arxiv') ORDER BY id DESC LIMIT 1").fetchone() + checkpoint=c.execute("SELECT records_written,status FROM ingest_runs WHERE source_id=(SELECT id FROM sources WHERE name='arxiv') ORDER BY id DESC LIMIT 1").fetchone() print('Saved arxiv checkpoint:',checkpoint,flush=True) + if checkpoint and checkpoint[1]=='complete':resume_arxiv=False schema_store=Store(db) schema_store.conn.close() + if evidence_files:backfill(db,evidence_files) for filename in files: ingest_file(db,filename) with closing(sqlite3.connect(db)) as c, c: - if files: + c.execute('PRAGMA cache_size=-32768') + if files or evidence_files: errors=c.execute('PRAGMA foreign_key_check').fetchall() if errors: raise RuntimeError('Foreign key violations after merge') @@ -51,13 +61,21 @@ def merge(db,archive,files): print('Tag refresh plan:',c.execute('EXPLAIN QUERY PLAN DELETE FROM paper_tags WHERE source_record_id=? AND evidence_key=?',(1,'arxiv:categories')).fetchall(),flush=True) print('Decision totals:',c.execute('SELECT status,count(*) FROM decisions GROUP BY status').fetchall(),flush=True) finally: - if property_value('LoadState')=='loaded': - subprocess.run(['systemctl','start',UNIT],check=True) - else: - subprocess.run(['systemd-run','--unit='+UNIT,'--property=WorkingDirectory='+str(Path(__file__).parent),'--property=CPUQuota=50%','--property=MemoryMax=512M','--property=Nice=15','--property=IOWeight=10','--property=StandardOutput=append:/data/paper-index/arxiv-import.log','--property=StandardError=append:/data/paper-index/arxiv-import.log','/usr/bin/env','PYTHONDONTWRITEBYTECODE=1','python3','-B',str(Path(__file__).with_name('import_arxiv.py')),'--db',db,'--archive',archive],check=True) - if property_value('ActiveState')!='active': - raise RuntimeError('arXiv unit did not resume; inspect the logs') - print('ARXIV_RESUMED',flush=True) + if resume_arxiv: + if property_value('LoadState')=='loaded': + subprocess.run(['systemctl','start',UNIT],check=True) + else: + subprocess.run(['systemd-run','--unit='+UNIT,'--property=WorkingDirectory='+str(Path(__file__).parent),'--property=CPUQuota=50%','--property=MemoryMax=512M','--property=Nice=15','--property=IOWeight=10','--property=StandardOutput=append:/data/paper-index/arxiv-import.log','--property=StandardError=append:/data/paper-index/arxiv-import.log','/usr/bin/env','PYTHONDONTWRITEBYTECODE=1','python3','-B',str(Path(__file__).with_name('import_arxiv.py')),'--db',db,'--archive',archive],check=True) + if property_value('ActiveState')!='active': + raise RuntimeError('arXiv unit did not resume; inspect the logs') + print('ARXIV_RESUMED',flush=True) + else:print('ARXIV_ALREADY_COMPLETE',flush=True) + + +def merge(db,archive,files,backfill_queue=None): + with Path(str(db)+'.maintenance.lock').open('a') as lock: + fcntl.flock(lock,fcntl.LOCK_EX|fcntl.LOCK_NB) + return merge_locked(db,archive,files,backfill_queue) if __name__=='__main__': @@ -65,8 +83,9 @@ if __name__=='__main__': parser.add_argument('--db',required=True) parser.add_argument('--archive',required=True) parser.add_argument('--schema-only',action='store_true',help='Apply idempotent schema/index updates at a saved checkpoint') + parser.add_argument('--backfill-queue',help='Audited existing-source queue for canonical submission evidence') parser.add_argument('files',nargs='*') args=parser.parse_args() - if not args.files and not args.schema_only: + if not args.files and not args.schema_only and not args.backfill_queue: parser.error('Provide collected files or --schema-only') - merge(args.db,args.archive,args.files) + merge(args.db,args.archive,args.files,args.backfill_queue) diff --git a/query.py b/query.py index f1ba580..04354fb 100644 --- a/query.py +++ b/query.py @@ -48,7 +48,15 @@ def main(): if args.year: clause.append('e.year=?'); params.append(args.year) if args.status: - clause.append('EXISTS(SELECT 1 FROM decisions d WHERE d.submission_id=s.id AND d.status=? AND NOT EXISTS(SELECT 1 FROM decisions newer WHERE newer.submission_id=d.submission_id AND (newer.observed_at>d.observed_at OR (newer.observed_at=d.observed_at AND newer.id>d.id))))'); params.append(args.status) + latest='''EXISTS ( + SELECT 1 FROM decisions d WHERE d.submission_id=s.id AND d.status=? + AND NOT EXISTS ( + SELECT 1 FROM decisions newer WHERE newer.submission_id=d.submission_id + AND (newer.observed_at,COALESCE(newer.decision_date,newer.observed_at),newer.id) + > (d.observed_at,COALESCE(d.decision_date,d.observed_at),d.id) + ) + )''' + clause.append('(NOT EXISTS(SELECT 1 FROM decisions d WHERE d.submission_id=s.id) OR '+latest+')' if args.status=='unknown' else latest); params.append(args.status) where.append('p.id IN (SELECT s.paper_id FROM submissions s JOIN editions e ON e.id=s.edition_id JOIN venues v ON v.id=e.venue_id WHERE '+' AND '.join(clause)+')') sql = 'SELECT p.* FROM papers p'+(' WHERE '+' AND '.join(where) if where else '')+' ORDER BY p.id LIMIT ?' params.append(max(1,min(args.limit,200))) @@ -63,7 +71,14 @@ def main(): submissions = [] for sub in conn.execute('SELECT s.id,s.source_submission_id,s.submitted_at,v.acronym,e.year,e.track FROM submissions s JOIN editions e ON e.id=s.edition_id JOIN venues v ON v.id=e.venue_id WHERE s.paper_id=? ORDER BY e.year,s.id',(row['id'],)): submission = dict(sub) - submission['decisions'] = [dict(d) for d in conn.execute('SELECT status,raw_value,evidence_url,observed_at,decision_date FROM decisions WHERE submission_id=? ORDER BY observed_at,id',(sub['id'],))] + submission['decisions'] = [] + for d in conn.execute('SELECT id,status,raw_value,evidence_url,observed_at,decision_date FROM decisions WHERE submission_id=? ORDER BY observed_at,COALESCE(decision_date,observed_at),id',(sub['id'],)): + decision=dict(d) + if conn.execute("SELECT 1 FROM sqlite_master WHERE type='table' AND name='decision_sources'").fetchone(): + decision['sources']=[dict(r) for r in conn.execute('SELECT s.name,r.external_id,ds.source_decision_id FROM decision_sources ds JOIN source_records r ON r.id=ds.source_record_id JOIN sources s ON s.id=r.source_id WHERE ds.decision_id=?',(d['id'],))] + submission['decisions'].append(decision) + if conn.execute("SELECT 1 FROM sqlite_master WHERE type='table' AND name='submission_sources'").fetchone(): + submission['sources']=[dict(r) for r in conn.execute('SELECT so.name,r.external_id,ss.source_submission_id FROM submission_sources ss JOIN source_records r ON r.id=ss.source_record_id JOIN sources so ON so.id=r.source_id WHERE ss.submission_id=?',(sub['id'],))] submissions.append(submission) paper['submissions'] = submissions result.append(paper) diff --git a/schema.sql b/schema.sql index 2e5cfb1..5a0b268 100644 --- a/schema.sql +++ b/schema.sql @@ -58,6 +58,33 @@ CREATE TABLE IF NOT EXISTS decisions ( decision_date TEXT, UNIQUE(submission_id,status,raw_value,evidence_url,observed_at) ); CREATE INDEX IF NOT EXISTS decisions_submission_order ON decisions(submission_id,observed_at DESC,id DESC); +CREATE TABLE IF NOT EXISTS submission_sources ( + submission_id INTEGER NOT NULL REFERENCES submissions(id), + source_record_id INTEGER NOT NULL REFERENCES source_records(id), source_submission_id TEXT NOT NULL, + PRIMARY KEY(submission_id,source_record_id,source_submission_id) +); +CREATE INDEX IF NOT EXISTS submission_sources_record ON submission_sources(source_record_id,source_submission_id); +CREATE TABLE IF NOT EXISTS submission_identifiers ( + edition_id INTEGER NOT NULL REFERENCES editions(id), scheme TEXT NOT NULL, value TEXT NOT NULL, + submission_id INTEGER NOT NULL REFERENCES submissions(id), PRIMARY KEY(edition_id,scheme,value) +); +CREATE INDEX IF NOT EXISTS submission_identifiers_submission ON submission_identifiers(submission_id); +CREATE TABLE IF NOT EXISTS decision_sources ( + decision_id INTEGER NOT NULL REFERENCES decisions(id), + source_record_id INTEGER NOT NULL REFERENCES source_records(id), source_decision_id TEXT NOT NULL DEFAULT '', + PRIMARY KEY(decision_id,source_record_id,source_decision_id) +); +CREATE TABLE IF NOT EXISTS submission_merges ( + former_submission_id INTEGER PRIMARY KEY, surviving_submission_id INTEGER NOT NULL REFERENCES submissions(id), + reason TEXT NOT NULL, merged_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP +); +CREATE TABLE IF NOT EXISTS decision_merges ( + former_decision_id INTEGER PRIMARY KEY, surviving_decision_id INTEGER NOT NULL REFERENCES decisions(id), + reason TEXT NOT NULL, merged_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP +); +CREATE TABLE IF NOT EXISTS schema_migrations ( + name TEXT PRIMARY KEY, applied_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP +); CREATE TABLE IF NOT EXISTS tags ( id INTEGER PRIMARY KEY, namespace TEXT NOT NULL, value TEXT NOT NULL, description TEXT, UNIQUE(namespace,value) diff --git a/submission_evidence.py b/submission_evidence.py new file mode 100644 index 0000000..43d3502 --- /dev/null +++ b/submission_evidence.py @@ -0,0 +1,110 @@ +"""Canonical submission events with independent source and decision evidence.""" +from datetime import datetime, timezone +import re + + +PUBLICATIONS={'pmlr','acl_anthology','colm','neurips','cvf','ecva','dblp'} + + +def timestamp(value=None): + date=datetime.now(timezone.utc) if value is None else datetime.fromisoformat(value.replace('Z','+00:00')) + if date.tzinfo is None:raise ValueError('Decision timestamp requires a timezone') + return date.astimezone(timezone.utc).isoformat(timespec='milliseconds').replace('+00:00','Z') + + +def identities_for_row(row): + identities=list(row.get('submission_identifiers') or []) + own=row.get('identifiers') or {} + if own.get('openreview'): + identities.append({'scheme':'openreview','value':own['openreview']}) + if row['source'] in PUBLICATIONS and row.get('status')=='accepted' and own.get('doi'): + doi=re.sub(r'^https?://(?:dx\.)?doi\.org/','',str(own['doi']),flags=re.I).strip().lower() + if re.fullmatch(r'10\.\d{4,9}/\S+',doi):identities.append({'scheme':'publication_doi','value':doi}) + meta=row.get('metadata') or {} + if row['source']=='iclr_program' and row['venue']=='ICLR' and meta.get('program_identity_type')=='submission' and meta.get('program_submission_number_verified') is True: + identities.append({'scheme':'iclr_number','value':str(meta['program_source_id'])}) + result=set() + for identity in identities: + scheme,value=identity['scheme'],str(identity['value']).strip() + if scheme=='openreview': + valid=re.fullmatch(r'[A-Za-z0-9_-]+',value) and not re.match(r'20\d{2}-Oral-',value) + elif scheme=='iclr_number':valid=re.fullmatch(r'[1-9]\d*',value) and row['venue']=='ICLR' + elif scheme=='publication_doi':valid=re.fullmatch(r'10\.\d{4,9}/\S+',value) + else:valid=False + if not valid:raise ValueError('Invalid authoritative submission identity') + result.add((scheme,value)) + return sorted(result) + + +def migrate_sources(c): + if c.execute("SELECT 1 FROM schema_migrations WHERE name='submission_source_baseline_v1'").fetchone():return + with c: + c.execute('INSERT OR IGNORE INTO submission_sources SELECT id,source_record_id,source_submission_id FROM submissions') + c.execute("INSERT OR IGNORE INTO decision_sources SELECT d.id,s.source_record_id,'' FROM decisions d JOIN submissions s ON s.id=d.submission_id") + c.execute("INSERT INTO schema_migrations(name) VALUES('submission_source_baseline_v1')") + + +def next_id(c,table,audit,former): + # Merged IDs are kept in audit; never allocate them to a later unrelated event. + return max(c.execute('SELECT COALESCE(max(id),0) FROM '+table).fetchone()[0],c.execute('SELECT COALESCE(max('+former+'),0) FROM '+audit).fetchone()[0])+1 + + +def merge_submissions(c,target,other): + if target==other:return + editions=c.execute('SELECT id,edition_id FROM submissions WHERE id IN (?,?)',(target,other)).fetchall() + if len(editions)!=2 or editions[0][1]!=editions[1][1]:raise ValueError('Cannot combine submissions from different editions') + c.execute('INSERT OR IGNORE INTO submission_sources SELECT ?,source_record_id,source_submission_id FROM submission_sources WHERE submission_id=?',(target,other)) + c.execute('DELETE FROM submission_sources WHERE submission_id=?',(other,)) + for decision,status,raw,url,observed,date in c.execute('SELECT id,status,raw_value,evidence_url,observed_at,decision_date FROM decisions WHERE submission_id=?',(other,)).fetchall(): + duplicate=c.execute('SELECT id,decision_date FROM decisions WHERE submission_id=? AND status=? AND raw_value=? AND evidence_url=? AND observed_at=?',(target,status,raw,url,observed)).fetchone() + if duplicate: + surviving=duplicate[0] + if date and duplicate[1] and date!=duplicate[1]:raise ValueError('Conflicting dates for the same decision observation') + if date and not duplicate[1]:c.execute('UPDATE decisions SET decision_date=? WHERE id=?',(date,surviving)) + c.execute('INSERT OR IGNORE INTO decision_sources SELECT ?,source_record_id,source_decision_id FROM decision_sources WHERE decision_id=?',(surviving,decision)) + c.execute('DELETE FROM decision_sources WHERE decision_id=?',(decision,)) + c.execute('UPDATE decision_merges SET surviving_decision_id=? WHERE surviving_decision_id=?',(surviving,decision)) + c.execute("INSERT INTO decision_merges(former_decision_id,surviving_decision_id,reason) VALUES(?,?,'same decision observation on merged submission')",(decision,surviving)) + c.execute('DELETE FROM decisions WHERE id=?',(decision,)) + else:c.execute('UPDATE decisions SET submission_id=? WHERE id=?',(target,decision)) + c.execute('UPDATE submission_identifiers SET submission_id=? WHERE submission_id=?',(target,other)) + c.execute('UPDATE submission_merges SET surviving_submission_id=? WHERE surviving_submission_id=?',(target,other)) + c.execute("INSERT INTO submission_merges(former_submission_id,surviving_submission_id,reason) VALUES(?,?,'shared authoritative submission identity')",(other,target)) + c.execute('DELETE FROM submissions WHERE id=?',(other,)) + + +def attach_submission(store,edition,paper,record,external_id,identities): + c=store.conn + matches={r[0] for r in c.execute('SELECT s.id FROM submission_sources p JOIN submissions s ON s.id=p.submission_id WHERE p.source_record_id=? AND p.source_submission_id=? AND s.edition_id=?',(record,external_id,edition))} + matches.update(r[0] for r in c.execute('SELECT id FROM submissions WHERE source_record_id=? AND source_submission_id=? AND edition_id=?',(record,external_id,edition))) + for scheme,value in identities: + found=c.execute('SELECT submission_id FROM submission_identifiers WHERE edition_id=? AND scheme=? AND value=?',(edition,scheme,value)).fetchone() + if found:matches.add(found[0]) + if matches: + papers={paper}|{c.execute('SELECT paper_id FROM submissions WHERE id=?',(sid,)).fetchone()[0] for sid in matches} + surviving=min(papers) + for other in sorted(papers-{surviving}):store.merge(surviving,other,'shared authoritative conference submission identity') + submission=min(matches) + for other in sorted(matches-{submission}):merge_submissions(c,submission,other) + c.execute('UPDATE submissions SET paper_id=? WHERE id=?',(surviving,submission)) + else: + submission=next_id(c,'submissions','submission_merges','former_submission_id') + c.execute('INSERT INTO submissions(id,paper_id,edition_id,source_record_id,source_submission_id) VALUES(?,?,?,?,?)',(submission,paper,edition,record,external_id)) + c.execute('INSERT OR IGNORE INTO submission_sources VALUES(?,?,?)',(submission,record,external_id)) + for scheme,value in identities: + c.execute('INSERT OR IGNORE INTO submission_identifiers VALUES(?,?,?,?)',(edition,scheme,value,submission)) + return submission + + +def write_decision(store,submission,record,status,raw,url,observed_at=None,decision_date=None,source_decision_id=''): + c=store.conn + observed=timestamp(observed_at) + date=timestamp(decision_date) if decision_date else None + # Re-observing unchanged evidence adds provenance, not a new decision event. + old=c.execute('SELECT id FROM decisions WHERE submission_id=? AND status=? AND raw_value=? AND evidence_url=? AND COALESCE(decision_date,\'\')=COALESCE(?,\'\')',(submission,status,raw,url,date)).fetchone() + if old:decision=old[0] + else: + decision=next_id(c,'decisions','decision_merges','former_decision_id') + c.execute('INSERT INTO decisions(id,submission_id,status,raw_value,evidence_url,observed_at,decision_date) VALUES(?,?,?,?,?,?,?)',(decision,submission,status,raw,url,observed,date)) + c.execute('INSERT OR IGNORE INTO decision_sources VALUES(?,?,?)',(decision,record,source_decision_id or '')) + return decision diff --git a/test_iclr.py b/test_iclr.py index d9f577e..fb9b983 100644 --- a/test_iclr.py +++ b/test_iclr.py @@ -86,7 +86,7 @@ class ICLRTests(unittest.TestCase): with tempfile.TemporaryDirectory() as tmp: path=Path(tmp)/'iclr.jsonl';db=Path(tmp)/'papers.sqlite' path.write_text(''.join(json.dumps(r)+'\n' for r in rows)) - path.with_suffix('.manifest.json').write_text(json.dumps({'source':'iclr_program','list_url':'https://iclr.cc/virtual/2026/papers.html','count':1,'status':'collected'})) + path.with_suffix('.manifest.json').write_text(json.dumps({'source':'iclr_program','parser_version':2,'list_url':'https://iclr.cc/virtual/2026/papers.html','count':1,'status':'collected'})) ingest_file(db,path);ingest_file(db,path) store=Store(db);self.addCleanup(store.conn.close) self.assertEqual(store.conn.execute('SELECT count(*) FROM submissions').fetchone()[0],1) diff --git a/test_index_core.py b/test_index_core.py index 36f5c18..c55c331 100644 --- a/test_index_core.py +++ b/test_index_core.py @@ -56,6 +56,7 @@ class StoreTests(unittest.TestCase): updated,_=self.store.record('arxiv','2601.00001','https://arxiv.org/abs/2601.00001','Updated research title','New searchable content',['Jane Doe']) self.assertEqual(updated,paper) self.assertEqual(self.store.conn.execute('SELECT title FROM papers').fetchone()[0],'Updated research title') + self.assertEqual(self.store.conn.execute('SELECT url FROM source_records').fetchone()[0],'https://arxiv.org/abs/2601.00001') self.assertEqual(self.store.conn.execute("SELECT count(*) FROM papers_fts WHERE papers_fts MATCH 'searchable'").fetchone()[0],1) def test_tag_refresh_keeps_other_record_evidence_and_custom_tags(self): diff --git a/test_maintenance.py b/test_maintenance.py new file mode 100644 index 0000000..057f9d2 --- /dev/null +++ b/test_maintenance.py @@ -0,0 +1,41 @@ +from pathlib import Path +import tempfile +import unittest +from unittest.mock import patch + +from index_core import Store +from merge_conferences import merge + + +class MaintenanceTests(unittest.TestCase): + def fixture(self,tmp,status): + db=Path(tmp)/'papers.sqlite';archive=Path(tmp)/'metadata.zip';archive.write_bytes(b'not read by schema-only maintenance') + store=Store(db);source=store.source('arxiv','https://arxiv.org') + store.conn.execute('INSERT INTO ingest_runs(source_id,status) VALUES(?,?)',(source,status));store.conn.commit();store.conn.close() + return db,archive + + def test_completed_archive_is_not_started_again_after_maintenance(self): + with tempfile.TemporaryDirectory() as tmp: + db,archive=self.fixture(tmp,'complete') + with patch('merge_conferences.property_value',return_value='inactive'),patch('merge_conferences.subprocess.run') as run: + merge(db,archive,[]) + run.assert_not_called() + + def test_inactive_incomplete_arxiv_requires_investigation(self): + with tempfile.TemporaryDirectory() as tmp: + db,archive=self.fixture(tmp,'failed') + with patch('merge_conferences.property_value',return_value='inactive'),patch('merge_conferences.subprocess.run') as run: + with self.assertRaisesRegex(RuntimeError,'neither active nor verified complete'): + merge(db,archive,[]) + run.assert_not_called() + + def test_arxiv_finishing_at_the_stop_boundary_is_not_restarted(self): + with tempfile.TemporaryDirectory() as tmp: + db,archive=self.fixture(tmp,'complete') + with patch('merge_conferences.property_value',return_value='active'),patch('merge_conferences.subprocess.run') as run: + merge(db,archive,[]) + self.assertEqual(run.call_args_list[0].args[0],['systemctl','stop','paper-atlas-arxiv-import.service']) + self.assertEqual(run.call_count,1) + + +if __name__=='__main__':unittest.main() diff --git a/test_submission_evidence.py b/test_submission_evidence.py new file mode 100644 index 0000000..b0ed694 --- /dev/null +++ b/test_submission_evidence.py @@ -0,0 +1,163 @@ +import json +from pathlib import Path +import sqlite3 +import subprocess +import sys +import tempfile +import unittest + +from import_records import ingest_file +from index_core import Store +from submission_evidence import identities_for_row +from backfill_submission_evidence import backfill,queue_files + + +class SubmissionTests(unittest.TestCase): + def setUp(self): + self.tmp=tempfile.TemporaryDirectory();self.addCleanup(self.tmp.cleanup) + self.db=Path(self.tmp.name)/'papers.sqlite';self.store=Store(self.db);self.addCleanup(self.store.conn.close) + + def record(self,source,external,authors): + return self.store.record(source,external,'https://example.org/'+external,'A long paper about reliable metadata and historical events','Abstract',authors) + + def decision(self,paper,record,external,identities,status='accepted',raw='Accept',url='https://example.org/decision',observed='2026-05-08T00:00:00Z',date=None): + return self.store.decision(paper,record,'ICLR',2026,'main',external,url,status,raw,identities=identities,observed_at=observed,decision_date=date) + + def query(self,*args): + self.store.conn.commit() + command=[sys.executable,'-B',str(Path(__file__).with_name('query.py')),'--db',str(self.db),*args] + result=subprocess.run(command,capture_output=True,text=True) + self.assertEqual(result.returncode,0,result.stderr) + return json.loads(result.stdout) + + def test_same_submission_across_sources_has_one_event_and_both_provenances(self): + first,one=self.store.record('iclr_program','2026/number-42','https://example.org/program','A long paper about reliable metadata and historical events',None,['Jane Doe','John Smith']) + second,two=self.record('openreview_snapshot','realForum',['Anonymous']) + self.assertNotEqual(first,second) + self.decision(first,one,'2026/number-42',[('iclr_number','42')]) + self.decision(second,two,'realForum',[('iclr_number','42'),('openreview','realForum')],url='https://openreview.net/forum?id=realForum¬eId=note1') + c=self.store.conn + self.assertEqual(c.execute('SELECT count(*) FROM papers').fetchone()[0],1) + self.assertEqual(c.execute('SELECT abstract FROM papers').fetchone()[0],'Abstract') + self.assertEqual(c.execute('SELECT count(*) FROM submissions').fetchone()[0],1) + self.assertEqual(c.execute('SELECT count(*) FROM submission_sources').fetchone()[0],2) + details=self.query('--paper',str(first))[0]['submissions'][0] + self.assertEqual({r['name'] for r in details['sources']},{'iclr_program','openreview_snapshot'}) + self.assertEqual(len(details['decisions']),2) + self.assertTrue(all(d['sources'] for d in details['decisions'])) + self.assertEqual(c.execute('PRAGMA foreign_key_check').fetchall(),[]) + + def test_same_paper_can_have_two_distinct_submission_attempts(self): + paper,one=self.store.record('openreview_snapshot','forum1','https://example.org/1','First version','Abstract',[],{'arxiv':'2601.00001','openreview':'forum1'}) + other,two=self.store.record('openreview_snapshot','forum2','https://example.org/2','Another version','Abstract',[],{'arxiv':'2601.00001','openreview':'forum2'}) + self.assertEqual(paper,other) + self.decision(paper,one,'forum1',[('openreview','forum1'),('iclr_number','1')],status='rejected',raw='Reject') + self.decision(paper,two,'forum2',[('openreview','forum2'),('iclr_number','2')]) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM submissions').fetchone()[0],2) + + def test_late_identity_bridge_preserves_decisions_and_does_not_reuse_merged_ids(self): + first,one=self.record('iclr_program','number-42',['Jane Doe','John Smith']) + second,two=self.record('openreview_snapshot','realForum',['Anonymous']) + self.decision(first,one,'number-42',[('iclr_number','42')]) + self.decision(second,two,'realForum',[('openreview','realForum')]) + self.decision(second,two,'realForum',[('openreview','realForum'),('iclr_number','42')]) + c=self.store.conn + self.assertEqual(c.execute('SELECT count(*) FROM submissions').fetchone()[0],1) + self.assertEqual(c.execute('SELECT count(*) FROM decisions').fetchone()[0],1) + self.assertEqual(c.execute('SELECT count(*) FROM decision_sources').fetchone()[0],2) + self.assertEqual(c.execute('SELECT count(*) FROM submission_merges').fetchone()[0],1) + self.assertEqual(c.execute('SELECT count(*) FROM decision_merges').fetchone()[0],1) + third,record=self.record('other','different',['Other Author']) + self.decision(third,record,'different',[('openreview','differentForum')],status='rejected',raw='Reject') + self.assertGreater(c.execute('SELECT max(id) FROM submissions').fetchone()[0],2) + self.assertGreater(c.execute('SELECT max(id) FROM decisions').fetchone()[0],2) + self.assertEqual(c.execute('PRAGMA foreign_key_check').fetchall(),[]) + + def test_old_snapshot_does_not_override_a_newer_official_observation(self): + paper,one=self.record('iclr_program','number-42',['Jane Doe','John Smith']) + _,two=self.record('openreview_snapshot','realForum',['Jane Doe','John Smith']) + self.decision(paper,one,'number-42',[('iclr_number','42')],observed='2026-11-06T00:00:00Z') + self.decision(paper,two,'realForum',[('iclr_number','42')],status='rejected',raw='Reject',observed='2026-05-08T00:00:00Z') + self.assertEqual(len(self.query('--venue','ICLR','--status','accepted')),1) + self.assertEqual(self.query('--venue','ICLR','--status','rejected'),[]) + + def test_decision_date_orders_events_with_the_same_snapshot_time(self): + paper,record=self.record('openreview_snapshot','realForum',['Jane Doe']) + self.decision(paper,record,'realForum',[('openreview','realForum')],date='2026-02-01T00:00:00Z') + self.decision(paper,record,'realForum',[('openreview','realForum')],status='rejected',raw='Reject',date='2026-01-01T00:00:00Z') + self.assertEqual(len(self.query('--venue','ICLR','--status','accepted')),1) + self.assertEqual(self.query('--venue','ICLR','--status','rejected'),[]) + + def test_unknown_without_a_decision_is_queryable_without_invented_verdict(self): + path=Path(self.tmp.name)/'unknown.jsonl' + row={'source':'openreview_snapshot','external_id':'realForum','url':'https://example.org/paper','title':'Unknown result','abstract':'Abstract','authors':[],'identifiers':{'openreview':'realForum'},'venue':'ICLR','year':2026,'track':'main','status':'unknown','decision_events':[]} + path.write_text(json.dumps(row)+'\n');path.with_suffix('.manifest.json').write_text(json.dumps({'source':row['source'],'list_url':'https://example.org/list','count':1})) + ingest_file(self.db,path) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM decisions').fetchone()[0],0) + self.assertEqual(len(self.query('--venue','ICLR','--status','unknown')),1) + + def test_known_summary_without_corresponding_decision_rolls_back(self): + path=Path(self.tmp.name)/'missing-evidence.jsonl' + row={'source':'openreview_snapshot','external_id':'realForum','url':'https://example.org/paper','title':'Unsupported result','abstract':'Abstract','authors':[],'venue':'ICLR','year':2026,'track':'main','status':'accepted','decision_events':[{'status':None}]} + path.write_text(json.dumps(row)+'\n');path.with_suffix('.manifest.json').write_text(json.dumps({'source':row['source'],'list_url':'https://example.org/list','count':1})) + with self.assertRaisesRegex(ValueError,'no corresponding decision evidence'): + ingest_file(self.db,path) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM submissions').fetchone()[0],0) + + def test_event_number_is_not_assumed_to_be_a_submission_number(self): + row={'source':'iclr_program','venue':'ICLR','metadata':{'program_source_id':42,'program_identity_type':'event'},'identifiers':{}} + self.assertEqual(identities_for_row(row),[]) + row['metadata']['program_identity_type']='submission' + self.assertEqual(identities_for_row(row),[]) + row['metadata']['program_submission_number_verified']=True + self.assertEqual(identities_for_row(row),[('iclr_number','42')]) + with self.assertRaisesRegex(ValueError,'Invalid authoritative'): + identities_for_row({'source':'snapshot','venue':'ICLR','submission_identifiers':[{'scheme':'openreview','value':'2026-Oral--42-placeholder'}]}) + + def test_legacy_database_migration_preserves_source_and_decision_links(self): + legacy=Path(self.tmp.name)/'legacy.sqlite';sql=Path(__file__).with_name('schema.sql').read_text() + sql=sql[:sql.index('CREATE TABLE IF NOT EXISTS submission_sources')]+sql[sql.index('CREATE TABLE IF NOT EXISTS tags'):] + c=sqlite3.connect(legacy);c.executescript(sql) + c.execute("INSERT INTO sources VALUES(1,'legacy','https://example.org')") + c.execute("INSERT INTO papers(id,title,title_key) VALUES(1,'Paper','paper')") + c.execute("INSERT INTO source_records VALUES(1,1,'paper',1,'https://example.org/p','Paper','Abstract','[]','{}',CURRENT_TIMESTAMP)") + c.execute("INSERT INTO venues VALUES(1,'ICLR','ICLR','machine_learning','conference')") + c.execute("INSERT INTO editions VALUES(1,1,2026,'main')") + c.execute("INSERT INTO submissions VALUES(1,1,1,1,'paper',NULL)") + c.execute("INSERT INTO decisions VALUES(1,1,'accepted','Accept','https://example.org/decision','2026-01-01T00:00:00Z',NULL)") + c.commit();c.close() + store=Store(legacy);self.addCleanup(store.conn.close) + self.assertEqual(store.conn.execute('SELECT * FROM submission_sources').fetchall(),[(1,1,'paper')]) + self.assertEqual(store.conn.execute('SELECT * FROM decision_sources').fetchall(),[(1,1,'')]) + self.assertEqual(store.conn.execute('PRAGMA foreign_key_check').fetchall(),[]) + + def test_existing_program_identity_backfill_is_idempotent_and_preserves_text(self): + paper,record=self.record('iclr_program','2026/program-42',['Jane Doe','John Smith']) + self.decision(paper,record,'2026/program-42',[]) + self.store.conn.commit() + original=self.store.conn.execute('SELECT title,abstract,retrieved_at FROM source_records WHERE id=?',(record,)).fetchone() + path=Path(self.tmp.name)/'program.jsonl' + row={'source':'iclr_program','external_id':'2026/program-42','venue':'ICLR','year':2026,'track':'main','status':'accepted','identifiers':{},'metadata':{'program_identity_type':'submission','program_source_id':42,'program_submission_number_verified':True}} + path.write_text(json.dumps(row)+'\n');path.with_suffix('.manifest.json').write_text(json.dumps({'source':'iclr_program','parser_version':2,'venue':'ICLR','year':2026,'track':'main','count':1})) + first=backfill(self.db,[path]);second=backfill(self.db,[path]) + self.assertEqual(first,second) + self.assertEqual(self.store.conn.execute('SELECT scheme,value FROM submission_identifiers').fetchall(),[('iclr_number','42')]) + self.assertEqual(self.store.conn.execute('SELECT title,abstract,retrieved_at FROM source_records WHERE id=?',(record,)).fetchone(),original) + other,snapshot=self.record('openreview_snapshot','realForum',['Anonymous']) + self.decision(other,snapshot,'realForum',[('iclr_number','42'),('openreview','realForum')]) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM submissions').fetchone()[0],1) + + def test_evidence_queue_rejects_changed_file_before_mutation(self): + path=Path(self.tmp.name)/'program.jsonl';path.write_text('changed bytes') + queue=Path(self.tmp.name)/'queue.json';queue.write_text(json.dumps({'files':[str(path)],'file_sha256':{str(path):'0'*64}})) + with self.assertRaisesRegex(ValueError,'differs from audited'): + queue_files(queue) + + def test_old_untyped_program_files_require_recollection(self): + path=Path(self.tmp.name)/'old.jsonl';path.write_text('') + path.with_suffix('.manifest.json').write_text(json.dumps({'source':'iclr_program','parser_version':1})) + with self.assertRaisesRegex(ValueError,'Recollect ICLR'): + ingest_file(self.db,path) + + +if __name__=='__main__':unittest.main() -- 2.52.0 From 121336982420e0376cc7a81c9bfe91db01d676b1 Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 15:01:15 +0800 Subject: [PATCH 13/20] Retry transient metadata requests and handle legacy empty HTML attributes --- README.md | 2 ++ collect_pmlr.py | 18 ++++++++++++++-- collect_proceedings.py | 2 +- test_fetch.py | 49 ++++++++++++++++++++++++++++++++++++++++++ test_proceedings.py | 5 +++++ 5 files changed, 73 insertions(+), 3 deletions(-) create mode 100644 test_fetch.py diff --git a/README.md b/README.md index 73b9b4e..bd33c39 100644 --- a/README.md +++ b/README.md @@ -185,6 +185,8 @@ python3 -B collect_proceedings.py --source neurips \ --cache /data/paper-index/raw/http-cache --out /data/paper-index/collected ``` +公共元数据请求对连接重置、超时及不完整响应最多尝试三次，保留原有有效缓存；HTTP 403 等访问拒绝会直接记录失败。声明 Content-Length 与实际响应不符时不发布缓存，已有缓存也受字节上限约束。旧页面没有赋值的 HTML 属性按空值处理，避免某一个空 class 属性阻断整届采集。 + 采集器从官方根目录发现实际发表届次，仅访问索引和 Abstract HTML 页面，不请求 PDF、补充材料或源码。可用重复 `--year` 限定实际存在的届次；不存在的年份会报错，局部重试保留其他届次的已有状态。每个来源串行请求，实际网络响应后等待 1 秒；缓存命中不额外等待。 COLM 当前目录提供 2024–2026 三届 pre-proceedings，按会议年份保存，保留平台实际出版日期和出版状态，不把晚补刊日期当作会议年份。NeurIPS 当前目录提供 1987–2025，并包含 2025 独立卷及 2021 独立 Datasets and Benchmarks 索引。主会、datasets_benchmarks、position、creative_ai 分开出文件，不混为同一个 track。跨平台自标识仅从出版方 paper-actions 区域链接与 citation 元数据提取，不把摘要里的引用链接当作当前论文 ID。manifest 的 `parser_version` 至少为 2 才允许入库；旧文件须利用缓存重新采集，避免错误身份合并。首次真实样本核对：COLM 2024 官方目录 299 个摘要入口；NeurIPS 1987 为 90 个，作者与摘要均已验证解析。样本成功不代表所有页面采集完成。 diff --git a/collect_pmlr.py b/collect_pmlr.py index 4ac8303..a92422f 100644 --- a/collect_pmlr.py +++ b/collect_pmlr.py @@ -1,11 +1,13 @@ """Collect official PMLR metadata, never the PDFs linked by the metadata.""" import argparse import hashlib +from http.client import IncompleteRead import io import json import re import os import subprocess +import ssl import tarfile import time import urllib.error @@ -43,6 +45,7 @@ def fetch(url,cache,limit=64*1024**2,delay=0.5): cache.mkdir(parents=True,exist_ok=True) path = cache/hashlib.sha256(url.encode()).hexdigest() if path.exists() and time.time()-path.stat().st_mtime < 24*3600: + if path.stat().st_size>limit:raise RuntimeError('Metadata response exceeds byte limit') return path.read_bytes() for attempt in range(3): try: @@ -50,11 +53,14 @@ def fetch(url,cache,limit=64*1024**2,delay=0.5): with urllib.request.urlopen(request,timeout=90) as response: if response.status != 200: raise RuntimeError('Expected complete HTTP 200 response') - if int(response.headers.get('Content-Length','0'))>limit: + expected=int(response.headers.get('Content-Length','0')) + if expected>limit: raise RuntimeError('Metadata response exceeds byte limit') data = response.read(limit+1) if len(data)>limit: raise RuntimeError('Metadata response exceeds byte limit') + if expected and len(data)!=expected: + raise IncompleteRead(data,max(0,expected-len(data))) part = path.with_suffix('.part') part.write_bytes(data) part.replace(path) @@ -63,8 +69,16 @@ def fetch(url,cache,limit=64*1024**2,delay=0.5): return data except urllib.error.HTTPError as exc: if exc.code not in (429,500,502,503,504) or attempt==2: + exc.close() raise - time.sleep(min(30,max(2**attempt,int(exc.headers.get('Retry-After','1'))))) + retry_after=exc.headers.get('Retry-After','1');exc.close() + try:retry_after=int(retry_after) + except ValueError:retry_after=5 + time.sleep(min(30,max(2**attempt,retry_after))) + except (urllib.error.URLError,TimeoutError,ConnectionError,IncompleteRead,ssl.SSLError) as exc: + reason=getattr(exc,'reason',exc) + if isinstance(reason,ssl.SSLCertVerificationError) or attempt==2:raise + time.sleep(min(30,2**(attempt+1))) def metadata_texts(volume,cache): diff --git a/collect_proceedings.py b/collect_proceedings.py index 30e45fa..6c19bf3 100644 --- a/collect_proceedings.py +++ b/collect_proceedings.py @@ -16,7 +16,7 @@ ROOTS={'colm':'https://proceedings.colm.cc/','neurips':'https://proceedings.neur class Node: def __init__(self,tag='',attrs=None): self.tag=tag - self.attrs=dict(attrs or []) + self.attrs={key:value or '' for key,value in attrs or []} self.children=[] def text(self): diff --git a/test_fetch.py b/test_fetch.py new file mode 100644 index 0000000..720bf4b --- /dev/null +++ b/test_fetch.py @@ -0,0 +1,49 @@ +import hashlib +import io +from pathlib import Path +import tempfile +import unittest +from unittest.mock import patch +from urllib.error import HTTPError,URLError + +from collect_pmlr import fetch + + +class Response(io.BytesIO): + def __init__(self,body,expected=None): + super().__init__(body);self.status=200;self.headers={'Content-Length':str(expected if expected is not None else len(body))} + + +class FetchTests(unittest.TestCase): + def test_transient_connection_failure_is_retried(self): + with tempfile.TemporaryDirectory() as tmp,patch('collect_pmlr.time.sleep'),patch('collect_pmlr.urllib.request.urlopen',side_effect=[URLError(ConnectionResetError('reset')),Response(b'complete')]) as network: + self.assertEqual(fetch('https://example.org/metadata',tmp),b'complete') + self.assertEqual(network.call_count,2) + + def test_truncated_response_is_not_cached_as_complete(self): + with tempfile.TemporaryDirectory() as tmp,patch('collect_pmlr.time.sleep'),patch('collect_pmlr.urllib.request.urlopen',side_effect=[Response(b'bad',expected=8),Response(b'complete')]): + self.assertEqual(fetch('https://example.org/metadata',tmp),b'complete') + path=Path(tmp)/hashlib.sha256(b'https://example.org/metadata').hexdigest() + self.assertEqual(path.read_bytes(),b'complete') + + def test_forbidden_response_is_not_retried_or_bypassed(self): + body=io.BytesIO(b'Forbidden');error=HTTPError('https://example.org/metadata',403,'Forbidden',{},body) + with tempfile.TemporaryDirectory() as tmp,patch('collect_pmlr.time.sleep'),patch('collect_pmlr.urllib.request.urlopen',side_effect=error) as network: + with self.assertRaises(HTTPError):fetch('https://example.org/metadata',tmp) + self.assertEqual(network.call_count,1) + self.assertTrue(body.closed) + + def test_transport_retries_are_bounded(self): + with tempfile.TemporaryDirectory() as tmp,patch('collect_pmlr.time.sleep'),patch('collect_pmlr.urllib.request.urlopen',side_effect=ConnectionResetError('reset')) as network: + with self.assertRaises(ConnectionResetError):fetch('https://example.org/metadata',tmp) + self.assertEqual(network.call_count,3) + + def test_cached_response_still_obeys_the_size_limit(self): + with tempfile.TemporaryDirectory() as tmp,patch('collect_pmlr.urllib.request.urlopen') as network: + path=Path(tmp)/hashlib.sha256(b'https://example.org/metadata').hexdigest();path.write_bytes(b'too large') + with self.assertRaisesRegex(RuntimeError,'exceeds byte limit'): + fetch('https://example.org/metadata',tmp,limit=2) + network.assert_not_called() + + +if __name__=='__main__':unittest.main() diff --git a/test_proceedings.py b/test_proceedings.py index 931b34d..4c9f607 100644 --- a/test_proceedings.py +++ b/test_proceedings.py @@ -9,6 +9,11 @@ from collect_proceedings import paper_links,parse_paper class ProceedingsTests(unittest.TestCase): + def test_legacy_empty_class_attribute_does_not_break_a_paper(self): + html='
    Legacy formatting

    Paper

    Real abstract.

    ' + row=parse_paper(html,'https://proceedings.neurips.cc/paper/2000/hash/a-Abstract.html','neurips',2000,'https://proceedings.neurips.cc/paper/2000') + self.assertEqual(row['abstract'],'Real abstract.') + def test_colm_nested_abstract_and_citation_authors(self): html='''

    Abstract

    Real abstract text.

    Name Change Policy

    Do not ingest footer.

    ''' row=parse_paper(html,'https://proceedings.colm.cc/paper_files/paper/2024/hash/a-Abstract-Conference.html','colm',2024,'https://proceedings.colm.cc/paper_files/paper/2024') -- 2.52.0 From 53a83a85a6622fd21e148c921da7812c2aa18f4b Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 15:50:26 +0800 Subject: [PATCH 14/20] Extract public ICLR snapshot submissions and signed decision evidence --- README.md | 16 ++- collect_openreview_snapshot.py | 200 +++++++++++++++++++++++++++++++++ test_openreview_snapshot.py | 107 ++++++++++++++++++ 3 files changed, 321 insertions(+), 2 deletions(-) create mode 100644 collect_openreview_snapshot.py create mode 100644 test_openreview_snapshot.py diff --git a/README.md b/README.md index bd33c39..15f5b9a 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 目标包括：完整历史元数据导入、逐年会议目录、跨平台论文去重、公开拒稿/撤稿信息、平台标签与自定义标签隔离、可恢复更新和覆盖率报告。 -**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR、COLM、NeurIPS、ACL Anthology、ICLR 官方公开程序、CVF/ECVA 视觉会议元数据采集、带清单核验的会议导入和只读查询。PeerRead 提供公开 ICLR 2017 明确决定的补充来源。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），全量导入正在服务器运行。其他会议适配器、历史缺失届次、完整投稿历史、跨平台疑似重复复核和覆盖率核验仍待完成。回归检查在远端执行，已配置 ces-control:host 对应的 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 +**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR、COLM、NeurIPS、ACL Anthology、ICLR 官方公开程序、CVF/ECVA 视觉会议元数据采集、带清单核验的会议导入和只读查询。PeerRead 提供公开 ICLR 2017 明确决定的历史补充；近期 ICLR 投稿与决定快照解析器正在验证。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），主批次已读至 EOF 并完成 3,233,620 条源记录的导入；这不是去重后的论文数，最终全库审计尚未完成。其他会议适配器、历史缺失届次、完整投稿历史、跨平台疑似重复复核和覆盖率核验仍待完成。回归检查在远端执行，已配置 ces-control:host 对应的 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 ## 数据范围与来源 @@ -172,7 +172,19 @@ python3 -B download_snapshot.py \ --output /data/paper-index/raw/iclr-openreview-v1.0.tar.bz2 ``` -[快照来源与复现脚本](https://github.com/qhjqhj00/iclr-openreview-reviews)。原包包含评审回复；后续解析只筛选公开投稿的索引、标题、摘要、平台标签及带证据的明确决定，不提取评审正文、作者邮件或论文全文。必须核对原始 invitation、readers、签名及决定字段，不能根据评分或录用名单缺席推断拒稿。该解析适配尚在开发，下载成功不代表已经入库或核验覆盖。 +[快照来源与复现脚本](https://github.com/qhjqhj00/iclr-openreview-reviews)。原包包含评审回复；解析只筛选公开投稿的索引、标题、摘要、平台标签及带证据的明确决定，不提取评审正文、作者邮件或论文全文。必须核对原始 invitation、readers、签名及决定字段，不能根据评分或录用名单缺席推断拒稿。 + +优先处理近期三届，再单独补全 2020–2023；不以 2017 年的 PeerRead 样本代表近期拒稿覆盖： + +```sh +python3 -B collect_openreview_snapshot.py \ + --archive /data/paper-index/raw/iclr-openreview-v1.0.tar.bz2 \ + --out /data/paper-index/collected --year 2024 --year 2025 --year 2026 +``` + +解析器先核验固定 SHA-256，再流式读取 tar，以临时 SQLite 按年份和 forum 关联投稿及决定，兼容回复先于投稿的归档顺序。只接受明确对 everyone 公开的投稿和具有会议主席签名的 Decision；保留原始决定值、note ID、决定日期、invitation 和签名证据。无明确决定的投稿保留 unknown，不产生拒稿推断。无法关联到公开投稿元数据的公开决定另存 unmatched-decisions 文件，供复核，不擅自创建论文。输出清单分别统计投稿状态、决定事件、缺失摘要和排除原因；重复副本去重，内容冲突会中止。 + +该版本是 2026-05-08 发布的第三方历史副本，不代表平台当前状态或全部投稿；缺失决定及公开性筛选均需核验。采集成功也不等于主库已更新，后续应通过单写入协调器导入，并核对同届官方程序与公开决定的差异。 下载异常后先检查对应进程或 CI 任务确已终止，再使用同一参数增加 `--resume`。只有断点身份一致且服务器实际返回正确 Range 响应才续传；不支持时保留断点并报错，不自动重下。正式文件与下载清单均放在仓库外，不提交大型来源包。 diff --git a/collect_openreview_snapshot.py b/collect_openreview_snapshot.py new file mode 100644 index 0000000..b2c3302 --- /dev/null +++ b/collect_openreview_snapshot.py @@ -0,0 +1,200 @@ +"""Extract public paper metadata and signed Decision notes from a pinned release.""" +import argparse +from collections import Counter,defaultdict +from contextlib import closing +from datetime import datetime,timezone +import json +from pathlib import Path +import re +import sqlite3 +import tarfile +import tempfile + +from download_snapshot import digest_file + + +SOURCE='openreview_snapshot' +SHA256='350071daf25ebf1aef854ee1d5dc3a81834e1c468917c82a4f78af5548de34fc' +RELEASE='https://github.com/qhjqhj00/iclr-openreview-reviews/releases/tag/v1.0' +AS_OF='2026-05-08T04:50:23.000Z' +EXPECTED={2020:2213,2021:2594,2022:2617,2023:3792,2024:7404,2025:11672,2026:19814} + + +def value(content,key,default=None): + result=content.get(key,default) + return result.get('value',default) if isinstance(result,dict) else result + + +def invitations(note): + result=note.get('invitations') or [] + if isinstance(result,str):result=[result] + if note.get('invitation'):result=list(result)+[note['invitation']] + return [i for i in result if isinstance(i,str)] + + +def public(note): + return isinstance(note.get('readers'),list) and 'everyone' in note['readers'] + + +def date_ms(raw): + if raw is None:return None + return datetime.fromtimestamp(int(raw)/1000,timezone.utc).isoformat(timespec='milliseconds').replace('+00:00','Z') + + +def decision_status(raw): + key=' '.join(raw.casefold().replace('_',' ').split()) + if key.startswith(('desk reject','desk-reject')):return 'desk_rejected' + if key in ('accept','accepted','poster','oral','spotlight') or key.startswith(('accept (','accept:','accept -')):return 'accepted' + if key in ('reject','rejected','reject and resubmit') or key.startswith(('reject (','reject:','reject -')):return 'rejected' + if key in ('withdraw','withdrawn','withdrawal'):return 'withdrawn' + return 'unknown' + + +def paper_record(note,year): + if not public(note):return None,'not_public' + base=f'ICLR.cc/{year}/Conference/-/' + if not any(i in (base+'Submission',base+'Blind_Submission') for i in invitations(note)):return None,'wrong_invitation' + forum=note.get('id') + if not isinstance(forum,str) or not re.fullmatch(r'[A-Za-z0-9_-]+',forum) or note.get('forum') not in (None,forum):return None,'not_submission_root' + content=note.get('content') or {};title=value(content,'title');abstract=value(content,'abstract') + if not isinstance(title,str) or not title.strip():return None,'missing_title' + if abstract is not None and not isinstance(abstract,str):return None,'nontext_abstract' + authors=value(content,'authors',[]) or [] + if isinstance(authors,str):authors=[a.strip() for a in authors.split(',')] + if not isinstance(authors,list):return None,'invalid_authors' + authors=[a.strip() for a in authors if isinstance(a,str) and a.strip() and not re.fullmatch(r'anonymous(?: authors?)?',a.strip(),re.I)] + number=note.get('number');number=int(number) if re.fullmatch(r'[1-9]\d*',str(number)) else None + identifiers={'openreview':forum} + for field in ('doi','arxiv','arxiv_id'): + own=value(content,field) + if not isinstance(own,str):continue + own=own.strip() + if field=='doi': + own=re.sub(r'^https?://(?:dx\.)?doi\.org/','',own,flags=re.I).lower() + if re.fullmatch(r'10\.\d{4,9}/\S+',own):identifiers['doi']=own + else: + own=re.sub(r'^https?://arxiv.org/(?:abs|pdf)/','',own).removesuffix('.pdf') + if re.fullmatch(r'(?:\d{4}\.\d{4,5}|[a-zA-Z.-]+/\d{7})(?:v\d+)?',own):identifiers['arxiv']=own + tags=[];index_fields={} + for field in ('keywords','primary_area','secondary_area','Please_choose_the_closest_area_that_your_submission_falls_into'): + raw=value(content,field) + if raw is None:continue + index_fields[field]=raw + terms=raw if isinstance(raw,list) else re.split(r'[,;\n]',raw) if field=='keywords' and isinstance(raw,str) else [raw] + namespace=SOURCE+':'+('keyword' if field=='keywords' else field) + tags.extend({'namespace':namespace,'value':term.strip()} for term in terms if isinstance(term,str) and term.strip()) + ids=[{'scheme':'openreview','value':forum}] + if number:ids.append({'scheme':'iclr_number','value':str(number)}) + return {'source':SOURCE,'external_id':f'{year}/{forum}','url':'https://openreview.net/forum?id='+forum, + 'title':' '.join(title.split()),'abstract':abstract.strip() if abstract else None,'authors':authors, + 'identifiers':identifiers,'submission_identifiers':ids,'venue':'ICLR','year':year,'track':'main', + 'submitted_at':date_ms(note.get('cdate') or note.get('tcdate')),'observed_at':AS_OF,'tags':tags, + 'metadata':{'forum':forum,'source_submission_number':number,'original_invitations':invitations(note), + 'public_readers_verified':True,'index_fields':index_fields,'venue':value(content,'venue'), + 'venueid':value(content,'venueid'),'snapshot_release':RELEASE, + 'snapshot_observation_upper_bound':AS_OF}},None + + +def decision_record(note,year): + if not public(note):return None,'not_public' + base=f'ICLR.cc/{year}/Conference/' + if not any(i.startswith(base) and i.endswith('/Decision') for i in invitations(note)):return None,'not_decision' + if not any(isinstance(s,str) and s.startswith(base) and s.rsplit('/',1)[-1] in ('Program_Chairs','Area_Chairs','Senior_Area_Chairs') for s in note.get('signatures',[])): + return None,'unverified_signature' + if note.get('ddate'):return None,'deleted_decision' + forum=note.get('forum');identifier=note.get('id') + if not all(isinstance(v,str) and re.fullmatch(r'[A-Za-z0-9_-]+',v) for v in (forum,identifier)):return None,'invalid_identity' + if note.get('_paper_forum',forum)!=forum:return None,'forum_mismatch' + raw=value(note.get('content') or {},'decision') + if not isinstance(raw,str) or not raw.strip():return None,'missing_decision_value' + return {'id':identifier,'forum':forum,'status':decision_status(raw),'raw_value':raw, + 'evidence_url':f'https://openreview.net/forum?id={forum}¬eId={identifier}', + 'observed_at':AS_OF,'decision_date':date_ms(note.get('cdate') or note.get('tcdate') or note.get('mdate')), + 'source_decision_id':identifier,'authority':{'invitations':invitations(note), + 'signatures':note.get('signatures',[]),'public_readers_verified':True}},None + + +def collect(archive,out,years,expected_sha=SHA256): + actual=digest_file(Path(archive)) + if actual['sha256']!=expected_sha:raise ValueError('Snapshot checksum mismatch before parsing') + if expected_sha==SHA256 and actual['bytes']!=512940722:raise ValueError('Unexpected pinned snapshot size') + out=Path(out);out.mkdir(parents=True,exist_ok=True);years=set(years) + counts=defaultdict(Counter);members=set() + with tempfile.TemporaryDirectory(prefix='openreview-stage-',dir=out) as tmp,closing(sqlite3.connect(Path(tmp)/'metadata.sqlite')) as c: + c.executescript('PRAGMA synchronous=OFF; CREATE TABLE papers(year INTEGER,forum TEXT,payload TEXT,PRIMARY KEY(year,forum)); CREATE TABLE decisions(year INTEGER,id TEXT,forum TEXT,payload TEXT,PRIMARY KEY(year,id)); CREATE INDEX decisions_forum ON decisions(year,forum);') + with tarfile.open(archive,mode='r|bz2') as tar: + for member in tar: + match=re.fullmatch(r'(?:\./)?data/iclr_(20\d{2})/(papers|reviews)\.jsonl',member.name) + if not match or int(match[1]) not in years:continue + year=int(match[1]);kind=match[2] + if not member.isfile() or (year,kind) in members:raise ValueError('Invalid or repeated metadata member') + members.add((year,kind));lines=0 + with tar.extractfile(member) as stream: + while line:=stream.readline(16*1024**2+1): + if len(line)>16*1024**2:raise ValueError('Unexpectedly large metadata line') + if not line.strip():continue + lines+=1;counts[year][kind+'_lines']+=1 + if kind=='reviews' and b'/Decision' not in line:continue + note=json.loads(line) + row,reason=paper_record(note,year) if kind=='papers' else decision_record(note,year) + if row is None: + counts[year][kind+'_excluded_'+reason]+=1;continue + if kind=='papers': + forum=row['metadata']['forum'];payload=json.dumps(row,ensure_ascii=False) + old=c.execute('SELECT payload FROM papers WHERE year=? AND forum=?',(year,forum)).fetchone() + if old and old[0]!=payload:raise ValueError('Conflicting copies of one submission') + if old:counts[year]['duplicate_papers']+=1 + else:c.execute('INSERT INTO papers VALUES(?,?,?)',(year,forum,payload)) + else: + payload=json.dumps(row,ensure_ascii=False);old=c.execute('SELECT payload FROM decisions WHERE year=? AND id=?',(year,row['id'])).fetchone() + if old and old[0]!=payload:raise ValueError('Conflicting copies of one decision note') + if old:counts[year]['duplicate_decisions']+=1 + else:c.execute('INSERT INTO decisions VALUES(?,?,?,?)',(year,row['id'],row['forum'],payload)) + if lines%5000==0:print(json.dumps({'source':SOURCE,'year':year,'member':kind,'lines':lines}),flush=True) + c.commit();print(json.dumps({'source':SOURCE,'year':year,'member':kind,'lines':lines,'status':'member_processed'}),flush=True) + if members!={(year,kind) for year in years for kind in ('papers','reviews')}:raise ValueError('Selected year metadata members are missing') + if expected_sha==SHA256: + for year in years: + if counts[year]['papers_lines']!=EXPECTED.get(year):raise ValueError('Pinned release paper count differs from published inventory') + status_path=out/'openreview-snapshot-year-status.json';catalogue=json.loads(status_path.read_text()) if status_path.exists() else {} + for year in sorted(years,reverse=True): + target=out/f'openreview-snapshot-{year}-main.jsonl';part=target.with_suffix('.part') + statuses=Counter();event_counts=Counter();missing=0;written=0 + with part.open('w') as output: + for forum,payload in c.execute('SELECT forum,payload FROM papers WHERE year=? ORDER BY forum',(year,)): + row=json.loads(payload) + events=[json.loads(r[0]) for r in c.execute('SELECT payload FROM decisions WHERE year=? AND forum=?',(year,forum))] + events.sort(key=lambda event:(event['decision_date'] or '',event['id'])) + row['decision_events']=events;row['status']=events[-1]['status'] if events else 'unknown' + row['metadata']['snapshot_sha256']=actual['sha256'];row['metadata']['decision_note_ids']=[e['id'] for e in events] + row['metadata']['decision_authority']=[{'note_id':e['id'],**e['authority']} for e in events] + output.write(json.dumps(row,ensure_ascii=False)+'\n');written+=1;missing+=not bool(row['abstract']);statuses[row['status']]+=1 + event_counts.update(e['status'] for e in events) + if not written:raise ValueError('No verified public submissions for selected year') + unmatched=c.execute('SELECT count(*) FROM decisions d WHERE d.year=? AND NOT EXISTS(SELECT 1 FROM papers p WHERE p.year=d.year AND p.forum=d.forum)',(year,)).fetchone()[0] + unmatched_path=None + if unmatched: + unresolved=out/f'openreview-snapshot-{year}-unmatched-decisions.jsonl' + with unresolved.with_suffix('.part').open('w') as output: + for payload, in c.execute('SELECT payload FROM decisions d WHERE d.year=? AND NOT EXISTS(SELECT 1 FROM papers p WHERE p.year=d.year AND p.forum=d.forum)',(year,)): + event=json.loads(payload);event.update(year=year,reason='public decision lacks public submission metadata in snapshot',snapshot_sha256=actual['sha256']) + output.write(json.dumps(event,ensure_ascii=False)+'\n') + unresolved.with_suffix('.part').replace(unresolved);unmatched_path=str(unresolved) + part.replace(target) + manifest={'source':SOURCE,'parser_version':1,'venue':'ICLR','year':year,'track':'main','count':written, + 'list_url':RELEASE,'snapshot_sha256':actual['sha256'],'snapshot_observation_upper_bound':AS_OF, + 'missing_abstract_count':missing,'submission_status_counts':dict(statuses),'decision_event_counts':dict(event_counts), + 'unmatched_public_decisions':unmatched,'archive_counts':dict(counts[year]),'file':str(target), + 'unmatched_evidence_file':unmatched_path, + 'status':'partial' if missing or statuses['unknown'] or unmatched or any('_excluded_' in k for k in counts[year]) else 'collected'} + target.with_suffix('.manifest.json').write_text(json.dumps(manifest,indent=2));catalogue[str(year)]=manifest + print(json.dumps(manifest),flush=True) + stage=status_path.with_suffix('.part');stage.write_text(json.dumps(catalogue,indent=2));stage.replace(status_path) + (out/'openreview-snapshot-last-run.json').write_text(json.dumps({'requested_years':sorted(years),'sha256':actual['sha256'],'release':RELEASE,'observed_at_upper_bound':AS_OF},indent=2)) + return {year:catalogue[str(year)] for year in years} + + +if __name__=='__main__': + parser=argparse.ArgumentParser();parser.add_argument('--archive',required=True);parser.add_argument('--out',required=True) + parser.add_argument('--year',type=int,action='append');args=parser.parse_args() + collect(args.archive,args.out,args.year or sorted(EXPECTED)) diff --git a/test_openreview_snapshot.py b/test_openreview_snapshot.py new file mode 100644 index 0000000..ddb4dc8 --- /dev/null +++ b/test_openreview_snapshot.py @@ -0,0 +1,107 @@ +import hashlib +import io +import json +from pathlib import Path +import sqlite3 +import tarfile +import tempfile +import unittest + +from collect_openreview_snapshot import collect,decision_record,paper_record +from import_records import ingest_file + + +class SnapshotTests(unittest.TestCase): + def setUp(self): + self.tmp=tempfile.TemporaryDirectory();self.addCleanup(self.tmp.cleanup) + self.root=Path(self.tmp.name) + + def paper(self,forum='paper1',v2=True): + content={'title':'A public paper','abstract':'A public abstract','authors':['Jane Doe'], + 'authorids':['private@example.org'],'keywords':['learning']} + if v2:content={key:{'value':value} for key,value in content.items()} + return {'id':forum,'forum':forum,'number':42,'readers':['everyone'],'cdate':1760000000000, + 'invitations':['ICLR.cc/2026/Conference/-/Submission'],'content':content} + + def decision(self,forum='paper1',identifier='decision1',value='Reject'): + return {'id':identifier,'forum':forum,'_paper_forum':forum,'readers':['everyone'], + 'invitations':['ICLR.cc/2026/Conference/Submission42/-/Decision'], + 'signatures':['ICLR.cc/2026/Conference/Program_Chairs'],'cdate':1770000000000, + 'content':{'decision':{'value':value},'comment':{'value':'PRIVATE REVIEW BODY'}}} + + def archive(self,papers,reviews): + path=self.root/'snapshot.tar.bz2' + with tarfile.open(path,'w:bz2') as archive: + # Real release order is not guaranteed; decisions precede submissions. + for kind,rows in [('reviews',reviews),('papers',papers)]: + data=(''.join(json.dumps(row)+'\n' for row in rows)).encode() + entry=tarfile.TarInfo('data/iclr_2026/'+kind+'.jsonl');entry.size=len(data) + archive.addfile(entry,io.BytesIO(data)) + return path,hashlib.sha256(path.read_bytes()).hexdigest() + + def test_public_v1_and_v2_metadata_omit_authorids_and_review_bodies(self): + for v2 in (False,True): + row,reason=paper_record(self.paper(v2=v2),2026) + self.assertIsNone(reason);self.assertEqual(row['authors'],['Jane Doe']) + self.assertNotIn('private@example.org',json.dumps(row)) + self.assertEqual(row['identifiers']['openreview'],'paper1') + note=self.paper();note['readers']=['chairs'] + self.assertEqual(paper_record(note,2026)[1],'not_public') + + def test_only_public_signed_formal_decisions_count(self): + note=self.decision();row,reason=decision_record(note,2026) + self.assertIsNone(reason);self.assertEqual(row['status'],'rejected') + self.assertNotIn('PRIVATE REVIEW BODY',json.dumps(row)) + note['invitations']=['ICLR.cc/2026/Conference/Submission42/-/Official_Review'] + self.assertEqual(decision_record(note,2026)[1],'not_decision') + note=self.decision();note['signatures']=['ICLR.cc/2026/Conference/Submission42/Reviewer_1'] + self.assertEqual(decision_record(note,2026)[1],'unverified_signature') + note=self.decision();note['readers']=['chairs'] + self.assertEqual(decision_record(note,2026)[1],'not_public') + note=self.decision(value='Borderline Accept') + self.assertEqual(decision_record(note,2026)[0]['status'],'unknown') + + def test_unordered_snapshot_import_is_idempotent_and_preserves_unknown_and_orphans(self): + path,sha=self.archive([self.paper(),self.paper('paper2'),self.paper('paper3')], + [self.decision(),self.decision('paper2','decision2','Accept (poster)'),self.decision('absent','orphan')]) + out=self.root/'out';manifest=collect(path,out,[2026],sha)[2026] + self.assertEqual(manifest['submission_status_counts'],{'rejected':1,'accepted':1,'unknown':1}) + self.assertEqual(manifest['unmatched_public_decisions'],1) + self.assertEqual(len(Path(manifest['unmatched_evidence_file']).read_text().splitlines()),1) + rows=[json.loads(line) for line in Path(manifest['file']).read_text().splitlines()] + self.assertEqual(rows[2]['decision_events'],[]) + self.assertNotIn('PRIVATE REVIEW BODY',json.dumps(rows)) + db=self.root/'db.sqlite' + ingest_file(db,manifest['file']);ingest_file(db,manifest['file']) + with sqlite3.connect(db) as connection: + self.assertEqual(connection.execute('SELECT count(*) FROM submissions').fetchone()[0],3) + self.assertEqual(connection.execute('SELECT count(*) FROM decisions').fetchone()[0],2) + self.assertEqual(connection.execute('PRAGMA foreign_key_check').fetchall(),[]) + + def test_identical_copies_coalesce_and_conflicting_copies_fail(self): + paper=self.paper();decision=self.decision() + path,sha=self.archive([paper,paper],[decision,decision]) + manifest=collect(path,self.root/'out',[2026],sha)[2026] + self.assertEqual(manifest['count'],1) + self.assertEqual(manifest['archive_counts']['duplicate_decisions'],1) + different=self.paper();different['content']['title']={'value':'Conflicting title'} + path,sha=self.archive([paper,different],[decision]) + with self.assertRaisesRegex(ValueError,'Conflicting copies'): + collect(path,self.root/'bad',[2026],sha) + self.assertFalse((self.root/'bad'/'openreview-snapshot-2026-main.jsonl').exists()) + + def test_checksum_failure_does_not_publish_output(self): + path,sha=self.archive([self.paper()],[self.decision()]) + with self.assertRaisesRegex(ValueError,'checksum mismatch'): + collect(path,self.root/'out',[2026],'0'*64) + self.assertFalse((self.root/'out').exists()) + + def test_legacy_invitation_and_plain_decision_value(self): + note=self.decision();note['invitation']=note.pop('invitations')[0] + note['content']['decision']='Desk Reject' + self.assertEqual(decision_record(note,2026)[0]['status'],'desk_rejected') + note['_paper_forum']='wrong' + self.assertEqual(decision_record(note,2026)[1],'forum_mismatch') + + +if __name__=='__main__':unittest.main() -- 2.52.0 From 8bfe52df68155709ab3eb5fff931a47226b2de7d Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 16:32:09 +0800 Subject: [PATCH 15/20] Reject contradictory submission identities and correct snapshot fixtures --- README.md | 2 +- submission_evidence.py | 5 +++++ test_openreview_snapshot.py | 3 ++- test_submission_evidence.py | 10 ++++++++++ 4 files changed, 18 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 15f5b9a..d2b7f95 100644 --- a/README.md +++ b/README.md @@ -182,7 +182,7 @@ python3 -B collect_openreview_snapshot.py \ --out /data/paper-index/collected --year 2024 --year 2025 --year 2026 ``` -解析器先核验固定 SHA-256，再流式读取 tar，以临时 SQLite 按年份和 forum 关联投稿及决定，兼容回复先于投稿的归档顺序。只接受明确对 everyone 公开的投稿和具有会议主席签名的 Decision；保留原始决定值、note ID、决定日期、invitation 和签名证据。无明确决定的投稿保留 unknown，不产生拒稿推断。无法关联到公开投稿元数据的公开决定另存 unmatched-decisions 文件，供复核，不擅自创建论文。输出清单分别统计投稿状态、决定事件、缺失摘要和排除原因；重复副本去重，内容冲突会中止。 +解析器先核验固定 SHA-256，再流式读取 tar，以临时 SQLite 按年份和 forum 关联投稿及决定，兼容回复先于投稿的归档顺序。只接受明确对 everyone 公开的投稿和具有会议主席签名的 Decision；保留原始决定值、note ID、决定日期、invitation 和签名证据。无明确决定的投稿保留 unknown，不产生拒稿推断。无法关联到公开投稿元数据的公开决定另存 unmatched-decisions 文件，供复核，不擅自创建论文。输出清单分别统计投稿状态、决定事件、缺失摘要和排除原因；重复副本去重，内容冲突会中止。导入时若同一编号指向不同 forum，或同一 forum 指向不同投稿编号，会报冲突并回滚，不强制合并。 该版本是 2026-05-08 发布的第三方历史副本，不代表平台当前状态或全部投稿；缺失决定及公开性筛选均需核验。采集成功也不等于主库已更新，后续应通过单写入协调器导入，并核对同届官方程序与公开决定的差异。 diff --git a/submission_evidence.py b/submission_evidence.py index 43d3502..c1d350b 100644 --- a/submission_evidence.py +++ b/submission_evidence.py @@ -80,6 +80,11 @@ def attach_submission(store,edition,paper,record,external_id,identities): for scheme,value in identities: found=c.execute('SELECT submission_id FROM submission_identifiers WHERE edition_id=? AND scheme=? AND value=?',(edition,scheme,value)).fetchone() if found:matches.add(found[0]) + for scheme in ('openreview','iclr_number'): + values={value for kind,value in identities if kind==scheme} + for sid in matches: + values.update(value for value, in c.execute('SELECT value FROM submission_identifiers WHERE submission_id=? AND scheme=?',(sid,scheme))) + if len(values)>1:raise ValueError('Conflicting authoritative submission identities: '+scheme) if matches: papers={paper}|{c.execute('SELECT paper_id FROM submissions WHERE id=?',(sid,)).fetchone()[0] for sid in matches} surviving=min(papers) diff --git a/test_openreview_snapshot.py b/test_openreview_snapshot.py index ddb4dc8..46e1640 100644 --- a/test_openreview_snapshot.py +++ b/test_openreview_snapshot.py @@ -20,7 +20,8 @@ class SnapshotTests(unittest.TestCase): content={'title':'A public paper','abstract':'A public abstract','authors':['Jane Doe'], 'authorids':['private@example.org'],'keywords':['learning']} if v2:content={key:{'value':value} for key,value in content.items()} - return {'id':forum,'forum':forum,'number':42,'readers':['everyone'],'cdate':1760000000000, + number=int(forum[-1]) if forum[-1].isdigit() else 42 + return {'id':forum,'forum':forum,'number':number,'readers':['everyone'],'cdate':1760000000000, 'invitations':['ICLR.cc/2026/Conference/-/Submission'],'content':content} def decision(self,forum='paper1',identifier='decision1',value='Reject'): diff --git a/test_submission_evidence.py b/test_submission_evidence.py index b0ed694..71e195c 100644 --- a/test_submission_evidence.py +++ b/test_submission_evidence.py @@ -55,6 +55,16 @@ class SubmissionTests(unittest.TestCase): self.decision(paper,two,'forum2',[('openreview','forum2'),('iclr_number','2')]) self.assertEqual(self.store.conn.execute('SELECT count(*) FROM submissions').fetchone()[0],2) + def test_conflicting_forums_cannot_be_bridged_by_one_submission_number(self): + paper,one=self.record('snapshot','realForum',['Jane Doe']) + other,two=self.record('snapshot','differentForum',['Other Author']) + self.decision(paper,one,'realForum',[('openreview','realForum'),('iclr_number','42')]) + with self.assertRaisesRegex(ValueError,'Conflicting authoritative submission identities'): + self.decision(other,two,'differentForum',[('openreview','differentForum'),('iclr_number','42')]) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM papers').fetchone()[0],2) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM submissions').fetchone()[0],1) + self.assertEqual(self.store.conn.execute('SELECT count(*) FROM submission_merges').fetchone()[0],0) + def test_late_identity_bridge_preserves_decisions_and_does_not_reuse_merged_ids(self): first,one=self.record('iclr_program','number-42',['Jane Doe','John Smith']) second,two=self.record('openreview_snapshot','realForum',['Anonymous']) -- 2.52.0 From a40efaa77ca2a15309844a6bf36ac8525864b2d5 Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 16:39:18 +0800 Subject: [PATCH 16/20] Audit public ICLR snapshots against official same-year programs --- README.md | 10 ++++ audit_iclr_snapshot.py | 91 +++++++++++++++++++++++++++++++++++++ test_iclr_snapshot_audit.py | 58 +++++++++++++++++++++++ 3 files changed, 159 insertions(+) create mode 100644 audit_iclr_snapshot.py create mode 100644 test_iclr_snapshot_audit.py diff --git a/README.md b/README.md index d2b7f95..b92072a 100644 --- a/README.md +++ b/README.md @@ -186,6 +186,16 @@ python3 -B collect_openreview_snapshot.py \ 该版本是 2026-05-08 发布的第三方历史副本，不代表平台当前状态或全部投稿；缺失决定及公开性筛选均需核验。采集成功也不等于主库已更新，后续应通过单写入协调器导入，并核对同届官方程序与公开决定的差异。 +入库前按真实 forum 对照同届官方程序，生成逐年状态、名单缺口、历史决定差异和文本版本差异报告，以及带 SHA-256 的现有程序身份回填队列： + +```sh +python3 -B audit_iclr_snapshot.py --root /data/paper-index/collected \ + --year 2024 --year 2025 --year 2026 \ + --report /data/paper-index/collected/iclr-recent-audit.json +``` + +该检查要求清单和记录的届次、数量、摘要缺失统计一致，拒绝重复 forum、编号冲突及缺失公开决定证据。报告中的名单差异和 unknown 不会转换成拒稿。不同时间来源的决定保留各自证据，强 forum 标识负责关联同一次投稿；报告通过不等于平台全量覆盖或最终全库去重已经完成。生成的 `*-backfill-queue.json` 仅用于已入库程序记录的身份回填，快照文件仍须由协调器单独导入。 + 下载异常后先检查对应进程或 CI 任务确已终止，再使用同一参数增加 `--resume`。只有断点身份一致且服务器实际返回正确 Range 响应才续传；不支持时保留断点并报错，不自动重下。正式文件与下载清单均放在仓库外，不提交大型来源包。 ### COLM 与 NeurIPS 官方摘要 diff --git a/audit_iclr_snapshot.py b/audit_iclr_snapshot.py new file mode 100644 index 0000000..8501d78 --- /dev/null +++ b/audit_iclr_snapshot.py @@ -0,0 +1,91 @@ +"""Cross-check collected public snapshots against same-edition official programs.""" +import argparse +from collections import Counter +import hashlib +import json +from pathlib import Path + +from index_core import author_keys,normalized +from submission_evidence import identities_for_row,timestamp + + +def load(path,source,year): + path=Path(path);manifest=json.loads(path.with_suffix('.manifest.json').read_text()) + minimum=2 if source=='iclr_program' else 1 + if manifest.get('parser_version',0) Date: Fri, 6 Nov 2026 16:43:18 +0800 Subject: [PATCH 17/20] Inventory pinned DBLP conference metadata before scope classification --- README.md | 17 ++++++ catalogue_dblp.py | 136 +++++++++++++++++++++++++++++++++++++++++ test_dblp_catalogue.py | 54 ++++++++++++++++ 3 files changed, 207 insertions(+) create mode 100644 catalogue_dblp.py create mode 100644 test_dblp_catalogue.py diff --git a/README.md b/README.md index b92072a..5a95742 100644 --- a/README.md +++ b/README.md @@ -162,6 +162,23 @@ python3 -B download_snapshot.py \ [官方版本与校验值](https://doi.org/10.4230/dblp.xml.2026-11-01)。下载器另外计算 SHA-256，记录来源 URL、大小和核验时间。DBLP 没有提供的摘要不能编造，出版目录也不能作为拒稿证据。 +DBLP 目录盘点先核验固定快照与[官方 DTD](https://doi.org/10.4230/dblp.xml.dtd.2023-06-28)，再流式读取 XML，直至 gzip EOF 检查通过： + +```sh +python3 -B download_snapshot.py \ + --url https://drops.dagstuhl.de/storage/artifacts/dblp/xml/2023/dblp-2023-06-28.dtd \ + --md5 96e4283e71886dec2173718957293648 \ + --output /data/paper-index/raw/dblp-2023-06-28.dtd +python3 -B catalogue_dblp.py \ + --archive /data/paper-index/raw/dblp-2026-11-01.xml.gz \ + --dtd /data/paper-index/raw/dblp-2023-06-28.dtd \ + --out /data/paper-index/collected/dblp +``` + +`dblp-catalogue.sqlite` 保存论文集目录、按 crossref/年份/booktitle 的记录数量、各目录族数量及少量样本；`dblp-ai-candidates.jsonl` 保存候选会议族的原始索引元数据。候选范围包含历史名称与 workshop，尚未赋予主会归属或录用状态。后续须依据具体论文集和届次核对目标范围、track 与出版标识，才能生成入库文件；不会把同一目录下的所有记录都视为主会。该盘点不下载 PDF，也不更改论文主库。 + +XML 只允许校验过的本地 `dblp.dtd`，拒绝其他外部实体和内部 DTD 子集，不访问 XML 声明的任意网络地址。临时目录库或 gzip 检查失败时不发布正式候选文件，需先调查残留的 `.part` 状态，再决定是否重新运行。 + 公开 ICLR 投稿快照的固定 Release 供后续筛选投稿和明确决定；它是第三方保存的 OpenReview 历史副本，不等于最新官方全量数据： ```sh diff --git a/catalogue_dblp.py b/catalogue_dblp.py new file mode 100644 index 0000000..2623981 --- /dev/null +++ b/catalogue_dblp.py @@ -0,0 +1,136 @@ +"""Inventory a pinned DBLP snapshot and stage candidate metadata for scope review. + +This does not label candidates accepted or import them into the paper database. +Only the checksum-verified local DTD is allowed; XML cannot fetch remote entities. +""" +import argparse +from collections import Counter,deque +import gzip +import json +from pathlib import Path +import sqlite3 +from xml.parsers import expat + +from download_snapshot import digest_file +from submission_evidence import timestamp + + +XML_SHA='0376ad019b216553faff0a14e1757b2d4c7a3b84c1beacf0511ef83695211b18' +DTD_MD5='96e4283e71886dec2173718957293648' +RELEASE='https://doi.org/10.4230/dblp.xml.2026-11-01' +# Broad discovery candidates, including historical names; not a venue mapping. +FAMILIES=set('nips icml iclr colm colt aistats uai clear acl emnlp naacl eacl coling conll cvpr iccv eccv aaai ijcai ecai kdd sigir www wsdm recsys cikm icdm sdm ecml pkdd ecmlpkdd icra iros rss corl aips icaps kr atal aamas icassp interspeech eurospeech icslp ismir miccai fat facct lrec'.split()) + + +def records(archive,dtd_bytes): + parser=expat.ParserCreate();parser.buffer_text=True + pending=deque();depth=0;row=None;field=None;text=[];doctype=False + + def document_type(name,system_id,public_id,internal): + nonlocal doctype + if name!='dblp' or system_id!='dblp.dtd' or public_id or internal: + raise ValueError('Only the pinned external DBLP DTD is allowed') + doctype=True + + def external(context,base,system_id,public_id): + if context is not None or system_id!='dblp.dtd' or public_id: + raise ValueError('Unexpected external XML entity') + child=parser.ExternalEntityParserCreate(context) + child.ExternalEntityRefHandler=lambda *args: (_ for _ in ()).throw(ValueError('Nested external entity')) + child.Parse(dtd_bytes,True);return 1 + + def start(name,attrs): + nonlocal depth,row,field,text + depth+=1 + if depth==1 and name!='dblp':raise ValueError('Unexpected XML root') + if depth==2 and name in ('inproceedings','proceedings'): + row={'record_type':name,'key':attrs.get('key'),'publtype':attrs.get('publtype'),'fields':{}} + elif depth==3 and row is not None: + field=name;text=[] + + def content(value): + if row is not None and field is not None:text.append(value) + + def end(name): + nonlocal depth,row,field,text + if depth==3 and row is not None and field is not None: + row['fields'].setdefault(field,[]).append(' '.join(''.join(text).split())) + field=None;text=[] + if depth==2 and row is not None: + pending.append(row);row=None + depth-=1 + + parser.StartDoctypeDeclHandler=document_type;parser.ExternalEntityRefHandler=external + parser.SetParamEntityParsing(expat.XML_PARAM_ENTITY_PARSING_ALWAYS) + parser.StartElementHandler=start;parser.EndElementHandler=end;parser.CharacterDataHandler=content + with gzip.open(archive,'rb') as stream: + while chunk:=stream.read(128*1024): + parser.Parse(chunk,False) + while pending:yield pending.popleft() + parser.Parse(b'',True) + while pending:yield pending.popleft() + if not doctype:raise ValueError('DBLP DTD declaration missing') + + +def family(key): + parts=(key or '').split('/') + return parts[1] if len(parts)>=3 and parts[0]=='conf' else '' + + +def catalogue(archive,dtd,out,expected_sha=XML_SHA,expected_dtd_md5=DTD_MD5): + archive=Path(archive);out=Path(out);dtd=Path(dtd) + identity=digest_file(archive);dtd_identity=digest_file(dtd) + if identity['sha256']!=expected_sha:raise ValueError('DBLP snapshot checksum mismatch') + if dtd_identity['md5']!=expected_dtd_md5:raise ValueError('DBLP DTD checksum mismatch') + if dtd.stat().st_size>128*1024:raise ValueError('Unexpectedly large DTD') + out.mkdir(parents=True,exist_ok=True) + target=out/'dblp-catalogue.sqlite';stage=out/'dblp-catalogue.sqlite.part' + candidate=out/'dblp-ai-candidates.jsonl';part=candidate.with_suffix('.part') + if stage.exists():raise ValueError('Unfinished catalogue exists; inspect it before another run') + c=sqlite3.connect(stage) + c.executescript('PRAGMA journal_mode=DELETE; PRAGMA synchronous=NORMAL; CREATE TABLE proceedings(key TEXT PRIMARY KEY,family TEXT,metadata_json TEXT); CREATE INDEX proceedings_family ON proceedings(family); CREATE TABLE volume_counts(crossref TEXT,year TEXT,booktitle TEXT,record_type TEXT,count INTEGER,PRIMARY KEY(crossref,year,booktitle,record_type)); CREATE TABLE family_counts(family TEXT,record_type TEXT,count INTEGER,PRIMARY KEY(family,record_type)); CREATE TABLE family_samples(family TEXT,seq INTEGER,metadata_json TEXT,PRIMARY KEY(family,seq));') + totals=Counter();families=Counter();volumes=Counter();samples=Counter();staged=Counter();seen=0 + + def flush(): + c.executemany('INSERT INTO volume_counts VALUES(?,?,?,?,?) ON CONFLICT(crossref,year,booktitle,record_type) DO UPDATE SET count=count+excluded.count',((*key,count) for key,count in volumes.items())) + volumes.clear();c.commit() + + try: + with part.open('w') as output: + for row in records(archive,dtd.read_bytes()): + seen+=1;kind=row['record_type'];key=row['key'];fields=row['fields'];prefix=family(key) + if not key:raise ValueError('Bibliographic record lacks a key') + totals[kind]+=1 + if kind=='proceedings': + c.execute('INSERT INTO proceedings VALUES(?,?,?)',(key,prefix,json.dumps(row,ensure_ascii=False))) + if prefix: + families[(prefix,kind)]+=1 + if kind=='inproceedings': + crossrefs=fields.get('crossref') or [''];years=fields.get('year') or [''];books=fields.get('booktitle') or [''] + if len(crossrefs)!=1 or len(years)!=1 or len(books)!=1:raise ValueError('Ambiguous publication volume metadata') + volumes[(crossrefs[0],years[0],books[0],kind)]+=1 + if samples[prefix]<2: + c.execute('INSERT INTO family_samples VALUES(?,?,?)',(prefix,samples[prefix],json.dumps(row,ensure_ascii=False)));samples[prefix]+=1 + if prefix in FAMILIES: + output.write(json.dumps(row,ensure_ascii=False)+'\n');staged[prefix]+=1 + if seen%50000==0: + flush();print(json.dumps({'source':'dblp','records_scanned':seen,'candidate_records':sum(staged.values())}),flush=True) + flush() + c.executemany('INSERT INTO family_counts VALUES(?,?,?)',((*key,count) for key,count in families.items()));c.commit() + if c.execute('PRAGMA quick_check').fetchall()!=[('ok',)]:raise ValueError('Catalogue integrity failed') + c.close();c=None + manifest={'source':'dblp','status':'catalogued_candidates_pending_scope_review','source_release':RELEASE,'snapshot_sha256':identity['sha256'], + 'dtd_md5':dtd_identity['md5'],'finished_at':timestamp(),'records_scanned':seen,'record_types':dict(totals), + 'candidate_records':sum(staged.values()),'candidate_family_counts':dict(staged),'candidate_sha256':digest_file(part)['sha256'], + 'file':str(candidate),'catalogue':str(target),'note':'Candidate families include workshops and historical names. No acceptance labels or normalized venue assignments have been inferred.'} + stage.replace(target);part.replace(candidate) + manifest_path=candidate.with_suffix('.manifest.json');manifest_part=manifest_path.with_suffix('.part') + manifest_part.write_text(json.dumps(manifest,indent=2));manifest_part.replace(manifest_path) + print(json.dumps(manifest),flush=True);return manifest + finally: + if c is not None:c.close() + + +if __name__=='__main__': + parser=argparse.ArgumentParser();parser.add_argument('--archive',required=True);parser.add_argument('--dtd',required=True);parser.add_argument('--out',required=True) + args=parser.parse_args();catalogue(args.archive,args.dtd,args.out) diff --git a/test_dblp_catalogue.py b/test_dblp_catalogue.py new file mode 100644 index 0000000..2995374 --- /dev/null +++ b/test_dblp_catalogue.py @@ -0,0 +1,54 @@ +import gzip +import hashlib +import json +from pathlib import Path +import sqlite3 +import tempfile +import unittest + +from catalogue_dblp import catalogue,records + + +class DblpCatalogueTests(unittest.TestCase): + def setUp(self): + self.tmp=tempfile.TemporaryDirectory();self.addCleanup(self.tmp.cleanup);self.root=Path(self.tmp.name) + self.dtd=self.root/'dblp.dtd';self.dtd.write_bytes(b'') + + def archive(self,body,doctype=''): + path=self.root/'dblp.xml.gz' + with gzip.open(path,'wb') as stream:stream.write((''+doctype+''+body+'').encode()) + return path + + def test_streamed_entities_nested_titles_and_candidate_scope_remain_explicit(self): + path=self.archive('AAAI Workshop proceedings2025RenéLearn <i>deep</i> models2025AAAI Workshopconf/aaai/2025https://doi.org/10.1234/exampleOther2020') + manifest=catalogue(path,self.dtd,self.root/'out',hashlib.sha256(path.read_bytes()).hexdigest(),hashlib.md5(self.dtd.read_bytes()).hexdigest()) + self.assertEqual(manifest['records_scanned'],3);self.assertEqual(manifest['candidate_records'],1) + row=json.loads(Path(manifest['file']).read_text()) + self.assertEqual(row['fields']['author'],['René']);self.assertEqual(row['fields']['title'],['Learn deep models']) + self.assertNotIn('status',row);self.assertNotIn('venue',row) + with sqlite3.connect(manifest['catalogue']) as connection: + self.assertEqual(connection.execute('SELECT sum(count) FROM volume_counts').fetchone()[0],2) + self.assertEqual(connection.execute('SELECT count(*) FROM proceedings').fetchone()[0],1) + + def test_remote_external_dtd_is_rejected_without_fetching(self): + path=self.archive('',doctype='') + with self.assertRaisesRegex(ValueError,'pinned external'):list(records(path,self.dtd.read_bytes())) + + def test_internal_entities_are_rejected(self): + path=self.archive('',doctype=']>') + with self.assertRaisesRegex(ValueError,'pinned external'):list(records(path,self.dtd.read_bytes())) + + def test_checksum_failure_preserves_existing_output(self): + path=self.archive('');out=self.root/'out';out.mkdir();target=out/'dblp-ai-candidates.jsonl';target.write_text('previous') + with self.assertRaisesRegex(ValueError,'snapshot checksum'):catalogue(path,self.dtd,out,'0'*64) + self.assertEqual(target.read_text(),'previous') + + def test_truncated_gzip_cannot_publish_a_catalogue(self): + path=self.archive('One') + path.write_bytes(path.read_bytes()[:-8]);out=self.root/'out' + with self.assertRaises((EOFError,OSError)): + catalogue(path,self.dtd,out,hashlib.sha256(path.read_bytes()).hexdigest(),hashlib.md5(self.dtd.read_bytes()).hexdigest()) + self.assertFalse((out/'dblp-ai-candidates.jsonl').exists()) + + +if __name__=='__main__':unittest.main() -- 2.52.0 From f9bdab49f2488dcf608aa6f69f7b43194198d95d Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 16:47:08 +0800 Subject: [PATCH 18/20] Resolve the official dated DBLP DTD locally and preserve source encoding --- README.md | 2 +- catalogue_dblp.py | 5 +++-- test_dblp_catalogue.py | 6 ++++++ 3 files changed, 10 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index 5a95742..5ad83b4 100644 --- a/README.md +++ b/README.md @@ -177,7 +177,7 @@ python3 -B catalogue_dblp.py \ `dblp-catalogue.sqlite` 保存论文集目录、按 crossref/年份/booktitle 的记录数量、各目录族数量及少量样本；`dblp-ai-candidates.jsonl` 保存候选会议族的原始索引元数据。候选范围包含历史名称与 workshop，尚未赋予主会归属或录用状态。后续须依据具体论文集和届次核对目标范围、track 与出版标识，才能生成入库文件；不会把同一目录下的所有记录都视为主会。该盘点不下载 PDF，也不更改论文主库。 -XML 只允许校验过的本地 `dblp.dtd`，拒绝其他外部实体和内部 DTD 子集，不访问 XML 声明的任意网络地址。临时目录库或 gzip 检查失败时不发布正式候选文件，需先调查残留的 `.part` 状态，再决定是否重新运行。 +XML 只允许校验过的本地 DTD；官方通用名 `dblp.dtd` 与月度快照使用的 `dblp-2023-06-28.dtd` 均解析到同一份已校验文件。拒绝其他外部实体和内部 DTD 子集，不访问 XML 声明的任意网络地址，并遵循源 XML 的字符编码。临时目录库或 gzip 检查失败时不发布正式候选文件，需先调查残留的 `.part` 状态，再决定是否重新运行。 公开 ICLR 投稿快照的固定 Release 供后续筛选投稿和明确决定；它是第三方保存的 OpenReview 历史副本，不等于最新官方全量数据： diff --git a/catalogue_dblp.py b/catalogue_dblp.py index 2623981..e1c947c 100644 --- a/catalogue_dblp.py +++ b/catalogue_dblp.py @@ -17,6 +17,7 @@ from submission_evidence import timestamp XML_SHA='0376ad019b216553faff0a14e1757b2d4c7a3b84c1beacf0511ef83695211b18' DTD_MD5='96e4283e71886dec2173718957293648' +DTD_NAMES={'dblp.dtd','dblp-2023-06-28.dtd'} RELEASE='https://doi.org/10.4230/dblp.xml.2026-11-01' # Broad discovery candidates, including historical names; not a venue mapping. FAMILIES=set('nips icml iclr colm colt aistats uai clear acl emnlp naacl eacl coling conll cvpr iccv eccv aaai ijcai ecai kdd sigir www wsdm recsys cikm icdm sdm ecml pkdd ecmlpkdd icra iros rss corl aips icaps kr atal aamas icassp interspeech eurospeech icslp ismir miccai fat facct lrec'.split()) @@ -28,12 +29,12 @@ def records(archive,dtd_bytes): def document_type(name,system_id,public_id,internal): nonlocal doctype - if name!='dblp' or system_id!='dblp.dtd' or public_id or internal: + if name!='dblp' or system_id not in DTD_NAMES or public_id or internal: raise ValueError('Only the pinned external DBLP DTD is allowed') doctype=True def external(context,base,system_id,public_id): - if context is not None or system_id!='dblp.dtd' or public_id: + if context is not None or system_id not in DTD_NAMES or public_id: raise ValueError('Unexpected external XML entity') child=parser.ExternalEntityParserCreate(context) child.ExternalEntityRefHandler=lambda *args: (_ for _ in ()).throw(ValueError('Nested external entity')) diff --git a/test_dblp_catalogue.py b/test_dblp_catalogue.py index 2995374..bfb5ee0 100644 --- a/test_dblp_catalogue.py +++ b/test_dblp_catalogue.py @@ -34,6 +34,12 @@ class DblpCatalogueTests(unittest.TestCase): path=self.archive('',doctype='') with self.assertRaisesRegex(ValueError,'pinned external'):list(records(path,self.dtd.read_bytes())) + def test_real_monthly_dtd_name_and_latin1_encoding(self): + path=self.root/'monthly.xml.gz' + xml='RenéMetadata' + with gzip.open(path,'wb') as stream:stream.write(xml.encode('latin1')) + self.assertEqual(list(records(path,self.dtd.read_bytes()))[0]['fields']['author'],['René']) + def test_internal_entities_are_rejected(self): path=self.archive('',doctype=']>') with self.assertRaisesRegex(ValueError,'pinned external'):list(records(path,self.dtd.read_bytes())) -- 2.52.0 From 827ba17a998a084df9176c2b948aabc5c5f46e94 Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 16:53:31 +0800 Subject: [PATCH 19/20] Preserve multi-volume DBLP metadata without inflating record counts --- README.md | 2 +- catalogue_dblp.py | 12 ++++++++---- test_dblp_catalogue.py | 8 ++++++++ 3 files changed, 17 insertions(+), 5 deletions(-) diff --git a/README.md b/README.md index 5ad83b4..498ed4f 100644 --- a/README.md +++ b/README.md @@ -175,7 +175,7 @@ python3 -B catalogue_dblp.py \ --out /data/paper-index/collected/dblp ``` -`dblp-catalogue.sqlite` 保存论文集目录、按 crossref/年份/booktitle 的记录数量、各目录族数量及少量样本；`dblp-ai-candidates.jsonl` 保存候选会议族的原始索引元数据。候选范围包含历史名称与 workshop，尚未赋予主会归属或录用状态。后续须依据具体论文集和届次核对目标范围、track 与出版标识，才能生成入库文件；不会把同一目录下的所有记录都视为主会。该盘点不下载 PDF，也不更改论文主库。 +`dblp-catalogue.sqlite` 保存论文集目录、按 crossref/年份/booktitle 的记录数量、各目录族数量及少量样本；`dblp-ai-candidates.jsonl` 保存候选会议族的原始索引元数据。存在多个出版卷字段的记录另存 `volume_anomalies`，保留全部值供归属核验；统计中仍只计一条，不任取第一个字段或展开成多篇。候选范围包含历史名称与 workshop，尚未赋予主会归属或录用状态。后续须依据具体论文集和届次核对目标范围、track 与出版标识，才能生成入库文件；不会把同一目录下的所有记录都视为主会。该盘点不下载 PDF，也不更改论文主库。 XML 只允许校验过的本地 DTD；官方通用名 `dblp.dtd` 与月度快照使用的 `dblp-2023-06-28.dtd` 均解析到同一份已校验文件。拒绝其他外部实体和内部 DTD 子集，不访问 XML 声明的任意网络地址，并遵循源 XML 的字符编码。临时目录库或 gzip 检查失败时不发布正式候选文件，需先调查残留的 `.part` 状态，再决定是否重新运行。 diff --git a/catalogue_dblp.py b/catalogue_dblp.py index e1c947c..6d7bb16 100644 --- a/catalogue_dblp.py +++ b/catalogue_dblp.py @@ -89,7 +89,7 @@ def catalogue(archive,dtd,out,expected_sha=XML_SHA,expected_dtd_md5=DTD_MD5): candidate=out/'dblp-ai-candidates.jsonl';part=candidate.with_suffix('.part') if stage.exists():raise ValueError('Unfinished catalogue exists; inspect it before another run') c=sqlite3.connect(stage) - c.executescript('PRAGMA journal_mode=DELETE; PRAGMA synchronous=NORMAL; CREATE TABLE proceedings(key TEXT PRIMARY KEY,family TEXT,metadata_json TEXT); CREATE INDEX proceedings_family ON proceedings(family); CREATE TABLE volume_counts(crossref TEXT,year TEXT,booktitle TEXT,record_type TEXT,count INTEGER,PRIMARY KEY(crossref,year,booktitle,record_type)); CREATE TABLE family_counts(family TEXT,record_type TEXT,count INTEGER,PRIMARY KEY(family,record_type)); CREATE TABLE family_samples(family TEXT,seq INTEGER,metadata_json TEXT,PRIMARY KEY(family,seq));') + c.executescript('PRAGMA journal_mode=DELETE; PRAGMA synchronous=NORMAL; CREATE TABLE proceedings(key TEXT PRIMARY KEY,family TEXT,metadata_json TEXT); CREATE INDEX proceedings_family ON proceedings(family); CREATE TABLE volume_counts(crossref TEXT,year TEXT,booktitle TEXT,record_type TEXT,count INTEGER,PRIMARY KEY(crossref,year,booktitle,record_type)); CREATE TABLE family_counts(family TEXT,record_type TEXT,count INTEGER,PRIMARY KEY(family,record_type)); CREATE TABLE family_samples(family TEXT,seq INTEGER,metadata_json TEXT,PRIMARY KEY(family,seq)); CREATE TABLE volume_anomalies(key TEXT PRIMARY KEY,metadata_json TEXT);') totals=Counter();families=Counter();volumes=Counter();samples=Counter();staged=Counter();seen=0 def flush(): @@ -108,8 +108,11 @@ def catalogue(archive,dtd,out,expected_sha=XML_SHA,expected_dtd_md5=DTD_MD5): families[(prefix,kind)]+=1 if kind=='inproceedings': crossrefs=fields.get('crossref') or [''];years=fields.get('year') or [''];books=fields.get('booktitle') or [''] - if len(crossrefs)!=1 or len(years)!=1 or len(books)!=1:raise ValueError('Ambiguous publication volume metadata') - volumes[(crossrefs[0],years[0],books[0],kind)]+=1 + if any(len(values)!=1 for values in (crossrefs,years,books)): + c.execute('INSERT INTO volume_anomalies VALUES(?,?)',(key,json.dumps(row,ensure_ascii=False))) + totals['multiple_volume_fields']+=1 + group=tuple(values[0] if len(values)==1 else '@multiple:'+json.dumps(values,ensure_ascii=False) for values in (crossrefs,years,books)) + volumes[(*group,kind)]+=1 if samples[prefix]<2: c.execute('INSERT INTO family_samples VALUES(?,?,?)',(prefix,samples[prefix],json.dumps(row,ensure_ascii=False)));samples[prefix]+=1 if prefix in FAMILIES: @@ -121,7 +124,8 @@ def catalogue(archive,dtd,out,expected_sha=XML_SHA,expected_dtd_md5=DTD_MD5): if c.execute('PRAGMA quick_check').fetchall()!=[('ok',)]:raise ValueError('Catalogue integrity failed') c.close();c=None manifest={'source':'dblp','status':'catalogued_candidates_pending_scope_review','source_release':RELEASE,'snapshot_sha256':identity['sha256'], - 'dtd_md5':dtd_identity['md5'],'finished_at':timestamp(),'records_scanned':seen,'record_types':dict(totals), + 'dtd_md5':dtd_identity['md5'],'finished_at':timestamp(),'records_scanned':seen,'record_types':{kind:totals[kind] for kind in ('proceedings','inproceedings')}, + 'multiple_volume_field_records':totals['multiple_volume_fields'], 'candidate_records':sum(staged.values()),'candidate_family_counts':dict(staged),'candidate_sha256':digest_file(part)['sha256'], 'file':str(candidate),'catalogue':str(target),'note':'Candidate families include workshops and historical names. No acceptance labels or normalized venue assignments have been inferred.'} stage.replace(target);part.replace(candidate) diff --git a/test_dblp_catalogue.py b/test_dblp_catalogue.py index bfb5ee0..67bc23a 100644 --- a/test_dblp_catalogue.py +++ b/test_dblp_catalogue.py @@ -40,6 +40,14 @@ class DblpCatalogueTests(unittest.TestCase): with gzip.open(path,'wb') as stream:stream.write(xml.encode('latin1')) self.assertEqual(list(records(path,self.dtd.read_bytes()))[0]['fields']['author'],['René']) + def test_multiple_volume_fields_remain_reviewable_without_double_counting(self): + path=self.archive('Joint publication2024First meetingSecond meetingconf/aaai/joint2024') + manifest=catalogue(path,self.dtd,self.root/'out',hashlib.sha256(path.read_bytes()).hexdigest(),hashlib.md5(self.dtd.read_bytes()).hexdigest()) + with sqlite3.connect(manifest['catalogue']) as connection: + raw=json.loads(connection.execute('SELECT metadata_json FROM volume_anomalies').fetchone()[0]) + self.assertEqual(raw['fields']['booktitle'],['First meeting','Second meeting']) + self.assertEqual(connection.execute('SELECT sum(count) FROM volume_counts').fetchone()[0],1) + def test_internal_entities_are_rejected(self): path=self.archive('',doctype=']>') with self.assertRaisesRegex(ValueError,'pinned external'):list(records(path,self.dtd.read_bytes())) -- 2.52.0 From 16f428cae9230729c05522bab07ba999ad6cf5db Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Fri, 6 Nov 2026 17:01:40 +0800 Subject: [PATCH 20/20] Report every target venue and expose missing historical coverage --- README.md | 11 ++++++++++- coverage_report.py | 41 +++++++++++++++++++++++++++++++++++++++++ test_coverage_report.py | 35 +++++++++++++++++++++++++++++++++++ 3 files changed, 86 insertions(+), 1 deletion(-) create mode 100644 coverage_report.py create mode 100644 test_coverage_report.py diff --git a/README.md b/README.md index 498ed4f..02d83d0 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ Paper Atlas 是一个**长期维护、按需或按批次运行的研究数据工 目标包括：完整历史元数据导入、逐年会议目录、跨平台论文去重、公开拒稿/撤稿信息、平台标签与自定义标签隔离、可恢复更新和覆盖率报告。 -**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR、COLM、NeurIPS、ACL Anthology、ICLR 官方公开程序、CVF/ECVA 视觉会议元数据采集、带清单核验的会议导入和只读查询。PeerRead 提供公开 ICLR 2017 明确决定的历史补充；近期 ICLR 投稿与决定快照解析器正在验证。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），主批次已读至 EOF 并完成 3,233,620 条源记录的导入；这不是去重后的论文数，最终全库审计尚未完成。其他会议适配器、历史缺失届次、完整投稿历史、跨平台疑似重复复核和覆盖率核验仍待完成。回归检查在远端执行，已配置 ces-control:host 对应的 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 +**目前仍在开发，不能视为完整数据集。** 已实现 SQLite 数据结构、arXiv 元数据 ZIP 流式导入、断点恢复、跨平台标识合并及审计、PMLR、COLM、NeurIPS、ACL Anthology、ICLR 官方公开程序、CVF/ECVA 视觉会议元数据采集、带清单核验的会议导入和只读查询。PeerRead 提供公开 ICLR 2017 明确决定的历史补充；近期 ICLR 公开决定快照已解析，实际主库进度以运行日志与统计为准。arXiv 元数据 ZIP 已完整下载（1,878,831,586 字节），主批次读取 3,233,620 行至 EOF；主库有 3,233,527 条 arXiv 来源记录，93 行差额仍需按原始 ID 独立核验。这些不是去重后的全库论文数，最终全库审计尚未完成。其他会议适配器、历史缺失届次、完整投稿历史、跨平台疑似重复复核和覆盖率核验仍待完成。回归检查在远端执行，已配置 ces-control:host 对应的 CI 工作流，实际 CI 结果以 Gitea Actions 为准。`venues.json` 的 41 个会议是候选覆盖范围，不代表已完成采集，也不是与 COLM 比较的客观排名。 ## 数据范围与来源 @@ -100,6 +100,15 @@ python3 -B query.py --db /data/paper-index/papers.sqlite --venue COLM --year 202 查询连接使用 SQLite 只读模式，结果输出 JSON。`--stats` 中 `papers` 是按当前规则去重后的论文实体数，`records_by_source` 是各平台已入库记录数，`decision_events_by_status` 是决定事件数。采集文件 manifest 的 `count` 是已采集数量，不能等同已入库数量，也不能把各平台记录数相加当成独立论文数。 +全部目标会议的覆盖报告会保留尚无入库记录的项目，避免只展示已经做完的会议： + +```sh +python3 -B coverage_report.py --db /data/paper-index/papers.sqlite \ + --out /data/paper-index/collected/venue-coverage-current.json +``` + +报告按会议、年份和 track 列出投稿数、独立论文数及来源清单状态，包含联合会议别名。没有数据的目标明确标为 no_imported_records；已有数据也保持“历史待独立核验”，不会因单个文件导入通过就宣称历年覆盖完成。同一论文可以出现于多个会议或年份，各会议论文数也不能直接相加作为全库数量。 + 用 `--paper 531782` 可查看单篇论文的所有已知来源、平台标签、投稿届次及带证据的决定历史。会议筛选不区分大小写，状态筛选使用该次投稿的最新已知决定。当前没有公共 Web 页面或网络查询服务。 ## 官方会议元数据采集 diff --git a/coverage_report.py b/coverage_report.py new file mode 100644 index 0000000..11cdbac --- /dev/null +++ b/coverage_report.py @@ -0,0 +1,41 @@ +"""Show every target venue, including those with no imported records.""" +import argparse +import json +from pathlib import Path +import sqlite3 + +from submission_evidence import timestamp + + +def report(db,registry): + registry=json.loads(Path(registry).read_text());targets=registry['venues'] + names=[row['acronym'].casefold() for row in targets] + if len(names)!=len(set(names)):raise ValueError('Duplicate target venue') + scope='(v.acronym=? COLLATE NOCASE OR EXISTS(SELECT 1 FROM venue_aliases a WHERE a.venue_id=v.id AND a.alias=? COLLATE NOCASE))' + c=sqlite3.connect(Path(db).resolve().as_uri()+'?mode=ro',uri=True);c.row_factory=sqlite3.Row + try: + c.execute('BEGIN');venues=[] + for target in targets: + acronym=target['acronym'];params=(acronym,acronym) + editions=[] + for row in c.execute('SELECT e.id,v.acronym AS canonical_venue,e.year,e.track,count(s.id) AS submissions,count(DISTINCT s.paper_id) AS papers FROM editions e JOIN venues v ON v.id=e.venue_id LEFT JOIN submissions s ON s.edition_id=e.id WHERE '+scope+' GROUP BY e.id ORDER BY e.year,e.track',params): + edition=dict(row) + edition['sources']=[dict(source) for source in c.execute('SELECT so.name,es.status,es.expected_count,es.imported_count,es.list_url,es.last_success_at,es.last_error FROM edition_sources es JOIN sources so ON so.id=es.source_id WHERE es.edition_id=? ORDER BY so.name',(row['id'],))] + editions.append(edition) + totals=c.execute('SELECT count(s.id),count(DISTINCT s.paper_id) FROM submissions s JOIN editions e ON e.id=s.edition_id JOIN venues v ON v.id=e.venue_id WHERE '+scope,params).fetchone() + venues.append({'acronym':acronym,'field':target['field'],'imported_submissions':totals[0],'distinct_papers':totals[1], + 'history_status':'partial_history_pending_independent_audit' if totals[0] else 'no_imported_records', + 'discovery_sources':target.get('discovery_sources',[]),'editions':editions}) + return {'checked_at':timestamp(),'target_venue_count':len(targets),'venues_with_imported_records':sum(bool(v['imported_submissions']) for v in venues), + 'scope_note':registry['scope_note'],'interpretation':'All registry targets are listed. Edition source complete means that source file passed import checks; it does not prove full historical or abstract coverage. Papers can appear in multiple venues/years, so do not sum venue paper counts as a global total.', + 'venues':venues} + finally:c.close() + + +if __name__=='__main__': + parser=argparse.ArgumentParser();parser.add_argument('--db',required=True) + parser.add_argument('--registry',default=str(Path(__file__).with_name('venues.json')));parser.add_argument('--out') + args=parser.parse_args();result=report(args.db,args.registry);text=json.dumps(result,ensure_ascii=False,indent=2) + if args.out: + path=Path(args.out);stage=path.with_suffix('.part');stage.write_text(text+'\n');stage.replace(path) + else:print(text) diff --git a/test_coverage_report.py b/test_coverage_report.py new file mode 100644 index 0000000..c2c4e94 --- /dev/null +++ b/test_coverage_report.py @@ -0,0 +1,35 @@ +import json +from pathlib import Path +import tempfile +import unittest + +from coverage_report import report +from index_core import Store + + +class CoverageTests(unittest.TestCase): + def setUp(self): + self.tmp=tempfile.TemporaryDirectory();self.addCleanup(self.tmp.cleanup) + self.root=Path(self.tmp.name);self.db=self.root/'papers.sqlite';self.store=Store(self.db);self.addCleanup(self.store.conn.close) + self.registry=self.root/'venues.json';self.registry.write_text(json.dumps({'scope_note':'Candidate scope','venues':[{'acronym':'COLING','field':'language'},{'acronym':'AAAI','field':'general_ai'}]})) + + def test_zero_coverage_targets_are_listed_and_joint_alias_is_counted_once(self): + paper,record=self.store.record('acl_anthology','joint-paper','https://example.org','Joint paper','Abstract',['Jane Doe']) + edition=self.store.decision(paper,record,'COLING+LREC',2024,'main','joint-paper','https://example.org','accepted','Accepted',aliases=['COLING','LREC']) + self.store.decision(paper,record,'COLING+LREC',2024,'main','joint-paper','https://example.org/another-observation','accepted','Accepted',aliases=['COLING']) + self.store.conn.commit();result=report(self.db,self.registry) + self.assertEqual(result['target_venue_count'],2);self.assertEqual(result['venues_with_imported_records'],1) + coling,aaai=result['venues'];self.assertEqual(coling['imported_submissions'],1) + self.assertEqual(coling['editions'][0]['canonical_venue'],'COLING+LREC') + self.assertEqual(aaai['history_status'],'no_imported_records');self.assertEqual(aaai['editions'],[]) + self.assertNotEqual(coling['history_status'],'complete') + + def test_same_paper_across_years_is_not_two_distinct_papers(self): + paper,record=self.store.record('source','one','https://example.org','A paper','Abstract',[]) + for year in (2024,2025): + self.store.decision(paper,record,'COLING',year,'main',str(year),'https://example.org','accepted','Accepted') + self.store.conn.commit();coling=report(self.db,self.registry)['venues'][0] + self.assertEqual(coling['imported_submissions'],2);self.assertEqual(coling['distinct_papers'],1) + + +if __name__=='__main__':unittest.main() -- 2.52.0