Import project sources

compose.py

@@ -0,0 +1,90 @@
+from pathlib import Path
+import json, base64, html, textwrap
+from io import BytesIO
+from PIL import Image, ImageDraw, ImageFont
+from reportlab.pdfgen import canvas
+from reportlab.lib.utils import ImageReader
+
+ROOT=Path(__file__).resolve().parent
+BOOK=json.loads((ROOT/'book.json').read_text())
+FONT='/System/Library/AssetsV2/com_apple_MobileAsset_Font7/8dc7805506cc9f233dcc19aabf593196842a47ae.asset/AssetData/Hannotate.ttc'
+SANS='/System/Library/Fonts/Hiragino Sans GB.ttc'
+W,H=1024,1696
+TOP=112
+SCALE=2
+def f(size,hand=True): return ImageFont.truetype(FONT if hand else SANS,size*SCALE)
+def wrap(t,font,maxw,draw):
+    out=[]
+    for raw in t.split('\n'):
+        line=''
+        for c in raw:
+            if line and draw.textlength(line+c,font=font)>maxw:
+                out.append(line); line=c
+            else: line+=c
+        out.append(line)
+    return out
+
+def compose(p):
+    num=p['number']
+    artpath=ROOT/'art'/f'page-{num:02d}-final.png'
+    art=Image.open(artpath).convert('RGB').resize((1024*SCALE,1536*SCALE),Image.Resampling.LANCZOS)
+    page=Image.new('RGB',(W*SCALE,H*SCALE),'white'); page.paste(art,(0,TOP*SCALE))
+    d=ImageDraw.Draw(page)
+    svg=[f'<svg xmlns="http://www.w3.org/2000/svg" width="1024" height="1696" viewBox="0 0 1024 1696"><rect width="1024" height="1696" fill="white"/>',
+         f'<image x="0" y="{TOP}" width="1024" height="1536" href="data:image/png;base64,{base64.b64encode(artpath.read_bytes()).decode()}"/>']
+    def txt(x,y,t,size=32,fill='#171719',hand=True):
+        d.text((x*SCALE,y*SCALE),t,font=f(size,hand),fill=fill,anchor='lt')
+        family='Hannotate SC' if hand else 'Hiragino Sans GB'
+        svg.append(f'<text x="{x}" y="{y}" dominant-baseline="text-before-edge" font-family="{family},sans-serif" font-size="{size}" fill="{fill}">{html.escape(t)}</text>')
+    def line(coords,fill='#171719',width=2):
+        d.line(tuple(int(v*SCALE) for v in coords),fill=fill,width=width*SCALE)
+        svg.append(f'<path d="M{coords[0]},{coords[1]} L{coords[2]},{coords[3]}" stroke="{fill}" stroke-width="{width}" fill="none" stroke-linecap="round"/>')
+    txt(24,14,BOOK['title'],22,fill='#727078')
+    txt(24,46,p['title'],44)
+    txt(896,54,f'{num:02d} / 07',22,fill='#727078',hand=False)
+    for j,utterances in enumerate(p['panels']):
+        col,row=j%2,j//2
+        x=col*512+38
+        compact=(num in (2,6) and j==2)
+        y=TOP+row*768+(16 if compact else 32)
+        textsize=28 if compact else 31
+        textfont=f(textsize)
+        for k,(speaker,words) in enumerate(utterances):
+            color=BOOK['characters'][speaker]['color']
+            txt(x,y,speaker,17 if compact else 19,fill=color)
+            y+=22 if compact else 27
+            lines=wrap(words,textfont,438*SCALE,d)
+            for t in lines:
+                txt(x,y,t,textsize); y+=33 if compact else 39
+            y+=8
+        # Speaker names keep freely positioned dialogue unambiguous.
+        if y>TOP+row*768+238: raise ValueError(f'Text overflows reserved art region: page {num}, panel {j+1}, y={y}')
+    footer='根据两帖评论改编 · 虚构合成人物 · 中文对白后期排版'
+    if num==5: footer='额度数字为帖内个别用户自述，不代表所有账号或统一规则。'
+    txt(24,1665,footer,17,fill='#6F6B75',hand=False)
+    svg.append('</svg>')
+    out=ROOT/'pages'/f'page-{num:02d}.png'; page.save(out)
+    (ROOT/'editable'/f'page-{num:02d}.svg').write_text('\n'.join(svg))
+    return out
+
+paths=[compose(p) for p in BOOK['pages']]
+pdfpath=ROOT/'output/pdf/最强模型茶话会.pdf'
+c=canvas.Canvas(str(pdfpath),pagesize=(512,848))
+c.setTitle(BOOK['title']); c.setAuthor('Codex · Reddit讨论改编')
+for p in paths:
+    encoded=BytesIO()
+    Image.open(p).convert('RGB').save(encoded,format='JPEG',quality=96,subsampling=0)
+    encoded.seek(0)
+    c.drawImage(ImageReader(encoded),0,0,width=512,height=848); c.showPage()
+c.save()
+# A contact sheet for visual inspection; page images remain full resolution.
+sheet=Image.new('RGB',(1024,1696),'#E9E7ED')
+for i,p in enumerate(paths):
+    im=Image.open(p); im.thumbnail((326,540))
+    x=(i%3)*342+(326-im.width)//2; y=(i//3)*564
+    sheet.paste(im,(x,y))
+sheet.save(ROOT/'overview.jpg',quality=92)
+index='''<!doctype html><html lang="zh-CN"><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1"><title>最强模型茶话会</title><style>body{margin:0;background:#eae7ee;font-family:system-ui;color:#25212b}header{max-width:900px;margin:30px auto;padding:0 16px}h1{font-size:28px}main{max-width:920px;margin:auto}img{display:block;width:100%;height:auto;margin:24px 0;box-shadow:0 5px 30px #211c3220}a{color:inherit}p{line-height:1.7}@media print{header{display:none}img{margin:0;box-shadow:none;break-after:page}}</style><header><h1>最强模型茶话会</h1><p>7页 · 6位虚构朋友 · 根据两帖公开讨论重新改编。评论为个人体验，非产品实测。</p><p><a href="output/pdf/最强模型茶话会.pdf">PDF阅读版</a> · <a href="README.md">改编说明</a> · <a href="book.json">对白源文件</a></p></header><main>'''
+index+=''.join(f'<img src="pages/page-{i:02d}.png" alt="第{i}页：{BOOK["pages"][i-1]["title"]}">' for i in range(1,8))+'</main></html>'
+(ROOT/'index.html').write_text(index)
+print(json.dumps({'pages':len(paths),'pdf':str(pdfpath),'size':pdfpath.stat().st_size,'text_overflow':False},ensure_ascii=False))