https://arxiv.org/abs/2606.28344. The end of web parsing. The beginning of scalable pixel-native search. link: https://pixelrag.ai/
fix(render): retain source tile indices for PDF subsets (#171)
OrdoAbChaoa0fb7cb8e68d
diff --git a/render/src/pixelrag_render/backends/pdf.py b/render/src/pixelrag_render/backends/pdf.pyindex 7241105..0db8594 100644--- a/render/src/pixelrag_render/backends/pdf.py+++ b/render/src/pixelrag_render/backends/pdf.py@@ -32,6 +32,7 @@ def render_pdf(output_dir: Directory to write the tile subdirectory into.dpi: Resolution for rendering (default 200 gives ~1650×2200px for A4).pages: 1-based list of page numbers to render. ``None`` renders all pages.+ Tile filenames and indices retain their zero-based source page index.quality: JPEG quality 1-100 (default 85).stem: Override for the tile directory name. Defaults to the PDF filenamestem. The pipeline passes the article_id here so directory names@@ -82,10 +83,11 @@ def render_pdf(saved_tiles: list[str] = []chunks_info: list[dict] = []- for idx, img in enumerate(images):+ first_index = convert_kwargs.get("first_page", 1) - 1+ for idx, img in enumerate(images, start=first_index):# If caller provided a sparse page list, skip pages not in the listif pages is not None:- page_num = min(pages) + idx+ page_num = idx + 1if page_num not in pages:continuediff --git a/tests/test_pdf_page_indices.py b/tests/test_pdf_page_indices.pynew file mode 100644index 0000000..0be77b0--- /dev/null+++ b/tests/test_pdf_page_indices.py@@ -0,0 +1,43 @@+"""Selected PDF pages retain the same zero-based IDs as a full render."""++import json+import sys+from types import ModuleType++import pytest+from PIL import Image+from pixelrag_render import render_pdf++[email protected]("pages", [None, [1], [3], [4], [2, 3], [1, 3], [2, 4], [4, 2]])+def test_pdf_page_indices_match_full_render(tmp_path, monkeypatch, pages):+ source = tmp_path / "document.pdf"+ source.touch()+ colors = ["red", "green", "blue", "yellow"]+ converter = ModuleType("pdf2image")++ def convert_from_path(**kwargs):+ first = kwargs.get("first_page", 1)+ last = kwargs.get("last_page", 4)+ return [+ Image.new("RGB", (16, 16), colors[i - 1]) for i in range(first, last + 1)+ ]++ converter.convert_from_path = convert_from_path+ monkeypatch.setitem(sys.modules, "pdf2image", converter)+ full_dir = render_pdf(source, tmp_path / "full")[0]+ selected_dir = render_pdf(source, tmp_path / "selected", pages=pages)[0]+ expected_ids = [0, 1, 2, 3] if pages is None else sorted(p - 1 for p in pages)+ expected_files = [f"tile_{i:04d}.jpg" for i in expected_ids]+ manifest = json.loads((selected_dir / "tiles.json").read_text())+ chunks = json.loads((selected_dir / "chunks.json").read_text())["chunks"]++ assert manifest["tiles"] == expected_files+ assert [c["tile_index"] for c in chunks] == expected_ids+ assert [c["file"] for c in chunks] == expected_files+ assert [c["tile"] for c in chunks] == expected_files+ assert sorted(p.name for p in selected_dir.glob("tile_*.jpg")) == expected_files+ for filename in expected_files:+ assert (selected_dir / filename).read_bytes() == (+ full_dir / filename+ ).read_bytes()