feat(format): B3.5 — PDF smart-fit rewrite + critical fontname=None fix
All checks were successful
Deploy to Production / Build and Deploy (push) Successful in 2m44s

ROOT CAUSE FIX: PyMuPDF silently raised AttributeError when fontname=None
was passed to insert_textbox. The try/except in _try_insert was swallowing
the error and returning None, causing every block to be skipped via the
graceful failure path. Setting fontname='helv' as the default unblocks
the entire PDF translation pipeline.

SMART-FIT: rewrite _write_translated_block with proper tier-fallback:
  - Tier 0: original bbox at original size
  - Tier 1: expanded horizontal
  - Tier 2: expanded vertical (3x original height)
  - Tier 3: shrink once (0.93x)
  - Tier 4: shrink twice (0.87x cumulative)
  - Tier 5: min size floor (90% for headings, 75% for body)
  - Tier 6: graceful skip with visible placeholder

REDACTION: single redaction per block (was per sub-bbox, creating 100+
redaction rectangles per page). Now only 1 redaction per text block.

FEATURE FLAG: PDF_SMART_FIT_ENABLED (default true, observation-first).

METRICS: text_overflow -> format_elements_lost_total.

RESULT ON REAL PDF:
  Before: fonts shrunk 22pt->5.6pt, hierarchy destroyed
  After:  fonts EXACT match: [8, 11, 12, 14, 16, 22] preserved
This commit is contained in:
2026-07-14 18:36:12 +02:00
parent 12cd0c6893
commit e706cef5d6
16 changed files with 1294 additions and 62 deletions

View File

@@ -0,0 +1,81 @@
"""Debug the smart-fit behavior on a real PDF."""
import sys
sys.path.insert(0, '.')
import importlib.util
spec = importlib.util.spec_from_file_location("pdf_mod", "translators/pdf_translator.py")
pdf_mod = importlib.util.module_from_spec(spec)
spec.loader.exec_module(pdf_mod)
import fitz
# Open the test PDF
doc = fitz.open("sample_files/test_corpus/test_pdf.pdf")
page = doc[0]
# Extract blocks
translator = pdf_mod.PDFTranslator()
raw_blocks = translator._extract_text_blocks(page)
merged = translator._merge_adjacent_blocks(raw_blocks, page.rect)
# Show block info
for i, block in enumerate(merged[:5]):
print(f"Block {i}:")
print(f" bbox: {block['bbox']}")
print(f" font_size: {block['font_size']}")
print(f" text: {block['text'][:50]!r}")
print()
# Manually try the smart-fit on the title block
title = merged[0]
print("=" * 60)
print("Smart-fit trace on title block:")
print("=" * 60)
import fitz
original_rect = fitz.Rect(title["bbox"])
page_rect = page.rect
print(f" original_rect: {original_rect}")
print(f" page_rect: {page_rect}")
print(f" page width: {page_rect.x1 - page_rect.x0}")
print(f" original width: {original_rect.width}, height: {original_rect.height}")
# Simulate the expansion
margin = 18
max_x1 = page_rect.x1 - margin
expanded_h = fitz.Rect(
max(original_rect.x0, page_rect.x0 + margin),
original_rect.y0,
max_x1,
original_rect.y1,
)
print(f" expanded_h: {expanded_h}, width: {expanded_h.width}")
next_block_y = page_rect.y1 - margin
max_expand_y = min(
next_block_y - original_rect.y1,
original_rect.height * pdf_mod.MAX_VERTICAL_EXPANSION,
)
expanded_v = fitz.Rect(
expanded_h.x0,
expanded_h.y0,
expanded_h.x1,
expanded_h.y1 + max_expand_y,
)
print(f" expanded_v: {expanded_v}, width: {expanded_v.width}, height: {expanded_v.height}")
# Try the actual insert
from fitz import TEXT_ALIGN_LEFT
french_text = "Spécification technique : Office Translator v3.0"
rc = page.insert_textbox(
expanded_v,
french_text,
fontsize=title["font_size"],
fontname="helv",
color=(0.1, 0.2, 0.5),
align=TEXT_ALIGN_LEFT,
overlay=True,
)
print(f" insert_textbox rc (with expanded_v at original size): {rc}")
print(f" >= 0 = fit, < 0 = overflow of {rc} points")
doc.close()