v0.2.0 — positioned PDF text preview

Inspect positioned PDF text and check selected cells

Some text-based PDFs place table content by coordinates. Their extracted character columns can shift across pages. This preview adds an optional local helper that lets you inspect word coordinates, provide a page region and column edges, and extract simple single-line rows into one JSON table per page. The existing checker can then compare selected cells with values you verified in the original.

• Includes the original two-page synthetic example and English/Chinese instructions.
• Uses Python 3.10+ and the optional pinned pdfplumber==0.11.10 dependency. The helper does not detect tables, perform OCR, merge wrapped rows, or verify source truth; inspect the evidence against your PDF.
• A region or column boundary crossing a word, an empty selected region, invalid page bounds, and existing output paths stop the operation. All pages share one user-supplied region and edges; limit 100 pages.
• Tested the actual example and 15 Python tests on macOS; the PR's Linux CI and pinned PDF inspector smoke test also passed. These tests are not an overall extraction accuracy benchmark.

Download pdf-table-traps-v0.2.0.zip and verify it with SHA256SUMS.txt. See docs/POSITIONAL_COLUMNS.md or docs/POSITIONAL_COLUMNS.zh-CN.md after extraction. The older v0.1.0 package remains unchanged.

The ZIP's README links to GitHub's latest non-prerelease, which currently points to v0.1.0. For this preview, use the v0.2.0 release page; the repository README links directly here.