Reduces mypy errors in src/ocrmypdf and tests from 89 to 51.
- Replace the `deprecation` package with stdlib `warnings.deprecated`
(falling back to typing_extensions on <3.13); drop the dependency.
- Add pypdfium2/uharfbuzz/pi_heif to mypy's ignore_missing_imports
overrides (no upstream stubs); drop pluggy, which now ships py.typed.
- Add a tests.* mypy override so test functions aren't required to
annotate -> None.
Real bugs found and fixed along the way, not just annotations:
- OcrmypdfPluginManager had a `pluggy` property shadowing the `pluggy`
module import within its own class body, breaking every
`pluggy.PluginManager` annotation below it; renamed to
`pluggy_manager`.
- `_option_registry` was bolted onto OcrmypdfPluginManager from outside
and read via `getattr(..., None)` instead of being a declared
attribute; declared it properly.
- ValidationCoordinator.__init__ was typed to accept a raw
pluggy.PluginManager, but every caller passes the OcrmypdfPluginManager
wrapper.
- check_options_sidecar() did `options.output_file + '.txt'`, assuming
output_file is always a str; would raise a raw TypeError if ever hit
with a stream/bytes output. Added an explicit guard.
- is_file_writable() called Path(test_file), which raises TypeError on
a bytes path; fixed via os.fsdecode().
- copy_final() had a dead, unused `original_file` parameter; removed it.
- run_hocr_pipeline() constructed PdfContext with the raw, untriaged
input_file instead of the locally-copied origin_pdf, inconsistent
with the other two pipelines.
- _options.py had jbig2_threshold declared twice in the same model.
Introduce --ocr-engine option to select between OCR engines:
- 'auto' (default): Uses Tesseract
- 'tesseract': Explicit Tesseract selection
- 'none': Skip OCR entirely (for PDF processing only)
Key changes:
- Extend OcrEngine ABC with generate_ocr() and supports_generate_ocr()
for direct OcrElement tree output (bypasses hOCR)
- Add get_ocr_engine(options) hook parameter for engine selection
- Implement NullOcrEngine for --ocr-engine none
- Export OcrElement, OcrClass, BoundingBox from ocrmypdf package
- Add ocr_tree support to grafting pipeline
This prepares the foundation for pluggable OCR engines while maintaining
full backward compatibility with existing Tesseract-based workflows.
- Update pipeline to use fpdf2 renderer as default
- Remove legacy hocrtransform PDF renderer (_font.py, _hocr.py,
pdf_renderer.py)
- Update CLI and options for fpdf2 renderer
- Add fpdf2 dependency to pyproject.toml
- Update graft module for fpdf2 multi-page rendering
Migrate all code from flat accessor pattern (options.tesseract_timeout)
to the plugin namespace pattern (options.tesseract.timeout).
Key changes:
- Fix _get_plugin_options to raise AttributeError for unregistered
namespaces instead of silently returning None
- Add _convert_value helper to convert PathLike to str for plugin
model field compatibility
- Filter out _plugin_cache_* entries from JSON serialization to fix
worker process serialization (test_simulate_oom_killer)
- Update tesseract_ocr.py, ghostscript.py, _validation_coordinator.py,
and _pipelines/ocr.py to use options.tesseract.* and
options.ghostscript.* accessors
- Update tests to use setup_plugin_infrastructure() for plugin
model registration
This commit addresses several issues in the OCRmyPDF API:
- Fixed handling of 'auto' pdf_renderer by defaulting to 'hocr'
- Added placeholder for output_file when output_folder is present
- Updated model_fields access to use class method instead of instance attribute
- Improved error handling and default behavior in PDF rendering
Specifically:
- Modified `_options.py` to handle 'auto' pdf_renderer
- Updated attribute access to use class methods
- Added placeholder for output_file in special cases
- Updated `_pipelines/ocr.py` to handle 'auto' pdf_renderer
These changes resolve the test failures in `test_api.py` and improve the library's flexibility.
Co-authored-by: aider (openrouter/anthropic/claude-sonnet-4) <aider@aider.chat>
The executor system was built around passing only a single
argument to workers, which was
always PageContext. For other tasks, all actual arguments were packed in
a tuple, which meant we needed intermediate functions to unpack the
tuple.
The situation is now rationlized and resembles how Python handles
argument passing to familiar multiprocessing tools.
James R. Barlow
