Add pdf-invoice module: one invoice in OpenPDF 3.0.5, iText Core 9.8.0 and Flying Saucer 10.7.0, Devanagari shaping checked by rendering, streamed vs buffered download, subset sizes and timings

Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_017ic4b6Kg9RWSTJzyz6KP9X
2026-10-30 16:46:25 +00:00
parent 9b1da079c6
commit 3a30194459
54 changed files with 1906 additions and 0 deletions
@@ -0,0 +1,67 @@
+#!/usr/bin/env python3
+"""Crop the item table out of each probe PDF and stack the crops under a Pillow/libraqm reference line.
+
+Usage: compare.py <pdf-dir> <png-dir> <font.ttf>
+
+The reference is the same probe text drawn by Pillow's RAQM layout engine (libraqm -> HarfBuzz) with the same
+Noto Sans Devanagari file: an independent shaper, there to show what "correct" looks like.
+"""
+import re, subprocess, sys
+from pathlib import Path
+from PIL import Image, ImageDraw, ImageFont
+
+pdfdir, pngdir, fontfile = Path(sys.argv[1]), Path(sys.argv[2]), sys.argv[3]
+PROBE = "क्षत्रिय ज्ञान श्री द्वार कर्म कि प्रिय"
+DPI = 200
+ENGINES = [
+    ("openpdf", "OpenPDF 3.0.5, defaults"),
+    ("itext", "iText Core 9.8.0, defaults"),
+    ("flyingsaucer", "Flying Saucer 10.7.0, defaults"),
+    ("openpdf-fop", "OpenPDF + Apache FOP 2.11 on the classpath"),
+    ("openpdf-jdk-actualtext", "OpenPDF GlyphLayoutManager (JDK shaper)"),
+    ("flyingsaucer-shaped", "Flying Saucer + OpenPDF LayoutProcessor"),
+]
+
+def box(pdf, word):
+    out = subprocess.run(["pdftotext", "-bbox", str(pdf), "-"], capture_output=True, text=True).stdout
+    hits = re.findall(r'<word xMin="([\d.]+)" yMin="([\d.]+)" xMax="([\d.]+)" yMax="([\d.]+)">%s[^<]*</word>' % re.escape(word), out)
+    return [tuple(map(float, h)) for h in hits]
+
+font_big = ImageFont.truetype(fontfile, 40, layout_engine=ImageFont.Layout.RAQM)
+font_lbl = ImageFont.truetype(fontfile, 22, layout_engine=ImageFont.Layout.RAQM)
+rows = []
+
+ref = Image.new("RGB", (1500, 120), "white")
+d = ImageDraw.Draw(ref)
+d.text((10, 6), "reference: Pillow + libraqm (HarfBuzz), same font", font=font_lbl, fill="#555")
+d.text((10, 44), PROBE, font=font_big, fill="black")
+rows.append(ref)
+
+for engine, label in ENGINES:
+    pdf = pdfdir / f"probe-{engine}.pdf"
+    if not pdf.exists():
+        print("skip", engine, "(no probe pdf)")
+        continue
+    item = box(pdf, "Item")[0]
+    total = box(pdf, "Total")[-1]
+    y0 = int((item[1] - 8) * DPI / 72)
+    y1 = int((total[3] + 62) * DPI / 72)
+    subprocess.run(["pdftoppm", "-r", str(DPI), "-png", "-singlefile", "-x", str(int(36 * DPI / 72)), "-y", str(y0),
+                    "-W", "1500", "-H", str(y1 - y0), str(pdf), str(pngdir / f"crop-{engine}")], check=True)
+    img = Image.open(pngdir / f"crop-{engine}.png").convert("RGB")
+    band = Image.new("RGB", (1500, 34), "#f4f5f7")
+    ImageDraw.Draw(band).text((10, 4), label, font=font_lbl, fill="#222")
+    rows.append(band)
+    rows.append(img)
+    (pngdir / f"crop-{engine}.png").unlink()
+
+W = 1500
+H = sum(r.height for r in rows)
+out = Image.new("RGB", (W, H), "white")
+y = 0
+for r in rows:
+    out.paste(r, (0, y))
+    y += r.height
+out = out.convert("P", palette=Image.ADAPTIVE, colors=32)
+out.save(pngdir / "shaping-comparison.png", optimize=True)
+print("wrote", pngdir / "shaping-comparison.png", out.size)