97 lines
4.3 KiB
Python
97 lines
4.3 KiB
Python
"""Libpq service connections; credentials never belong in CLI arguments."""
|
|
from collections.abc import Mapping
|
|
import re
|
|
from paper_atlas.postgres import text
|
|
|
|
SEARCH_VECTOR = "to_tsvector('english'::regconfig,coalesce(title,'') || ' ' || coalesce(abstract,''))"
|
|
|
|
|
|
def driver():
|
|
try:
|
|
import psycopg
|
|
except ImportError as error:
|
|
raise RuntimeError('Install paper-atlas[postgres] on the approved remote runner') from error
|
|
return psycopg
|
|
|
|
|
|
def connect(service, *, readonly=False):
|
|
if not re.fullmatch(r'[A-Za-z0-9_-]+', service):
|
|
raise ValueError('Use a libpq service name, never a password or connection URL')
|
|
connection = driver().connect(service=service, connect_timeout=15, autocommit=True,
|
|
application_name='paper-atlas')
|
|
connection.execute("SET timezone='UTC'")
|
|
connection.execute("SET lock_timeout='5s'")
|
|
connection.execute("SET work_mem='2MB'")
|
|
connection.execute("SET maintenance_work_mem='16MB'")
|
|
connection.execute('SET max_parallel_workers_per_gather=0')
|
|
connection.execute('SET max_parallel_maintenance_workers=0')
|
|
if readonly:
|
|
connection.execute('BEGIN ISOLATION LEVEL REPEATABLE READ READ ONLY')
|
|
return connection
|
|
|
|
|
|
class Row(Mapping):
|
|
"""The small mapping/positional interface used by existing query output."""
|
|
def __init__(self, columns, values):
|
|
self.columns, self.values = columns, values
|
|
|
|
def __getitem__(self, key):
|
|
return self.values[key if isinstance(key, int) else self.columns.index(key)]
|
|
|
|
def __iter__(self):
|
|
return iter(self.columns)
|
|
|
|
def __len__(self):
|
|
return len(self.columns)
|
|
|
|
|
|
def row_factory(cursor, *, encoded=False):
|
|
columns = [column.name for column in cursor.description] if cursor.description else []
|
|
return lambda values: Row(columns, tuple(map(text.decode, values)) if encoded else values)
|
|
|
|
|
|
class ReadConnection:
|
|
postgres = True
|
|
|
|
def __init__(self, service, schema=None):
|
|
if schema is not None and not re.fullmatch(r'[a-z][a-z0-9_]{0,35}', schema):
|
|
raise ValueError('Invalid schema name')
|
|
self.connection = connect(service, readonly=True)
|
|
from paper_atlas.postgres.schema import quote
|
|
if schema is not None:
|
|
self.connection.execute('SET LOCAL search_path TO ' + quote(schema) + ',pg_catalog')
|
|
if not self.has_table('papers'):
|
|
self.close()
|
|
raise RuntimeError('PostgreSQL publication is not ready in the configured search_path')
|
|
current = self.connection.execute('SELECT current_schema()').fetchone()[0]
|
|
# Schema comments are readable by other projects granted table access;
|
|
# consumers need no privileges on the private migration controller.
|
|
annotation = self.connection.execute("SELECT obj_description(oid,'pg_namespace') FROM pg_namespace WHERE nspname=%s", (current,)).fetchone()[0]
|
|
self.encoded = False
|
|
if annotation and annotation.startswith('paper-atlas:text-codec='):
|
|
codec = annotation.removeprefix('paper-atlas:text-codec=')
|
|
if codec != text.CODEC:
|
|
self.close()
|
|
raise RuntimeError('Unsupported PostgreSQL text storage codec')
|
|
self.encoded = codec == text.CODEC
|
|
self.connection.row_factory = lambda cursor: row_factory(cursor, encoded=self.encoded)
|
|
|
|
def execute(self, query, params=()):
|
|
# All caller SQL is static; user values are always bound parameters.
|
|
if 'sqlite_master' in query:
|
|
match = re.search(r"name='([a-z_]+)'", query)
|
|
if not match or match[1] not in ('decision_sources', 'submission_sources', 'venue_aliases'):
|
|
raise ValueError('Unsupported legacy table lookup')
|
|
return self.connection.execute('''SELECT 1 FROM information_schema.tables
|
|
WHERE table_schema=current_schema() AND table_name=%s''',
|
|
(match[1],))
|
|
values = tuple(map(text.encode, params)) if self.encoded else params
|
|
return self.connection.execute(query.replace('?', '%s'), values)
|
|
|
|
def has_table(self, name):
|
|
return bool(self.connection.execute('''SELECT 1 FROM information_schema.tables
|
|
WHERE table_schema=current_schema() AND table_name=%s''', (name,)).fetchone())
|
|
|
|
def close(self):
|
|
self.connection.close()
|