Per-entity selection: skipped, computed-in-C++-only, or materialized. Field name is the entity; the value is the lifecycle level. Set per call

Configuration for the threaded PDF parser. Attributes: loglevel: Logging level ('fatal', 'error', 'warning', 'info'). threads: Nu

Threaded PDF parser that decodes pages from multiple documents in parallel.

How to compute page content (tuning, not selection). Fixed when the document is opened; the C++ page-decoder cache freezes these at first dec

Timing information from PDF page parsing. Provides detailed timing breakdown of the parsing process, useful for performance analysis and opti

Top-level timing breakdown for a threaded page decode task.

Outcome of one page processed by DoclingThreadedPdfParser.

Top-level timing breakdown for a threaded page render task.

PDF document annotations including form fields, language, metadata, and table of contents. Attributes: form: AcroForm data containing int

PDF table of contents entry (recursive structure). Attributes: title: The text of the TOC entry level: Nesting level in the hiera

One path-painting operator (S, s, f, F, f*, B, B*, b, b*): all subpaths of the painted path plus the graphics-state parameters needed to color it. Fil

One subpath of a painted path: an implicit move-to (x0, y0) followed by a run of segment ops. SEGMENT_LINE_TO consumes one coordinate pair from px/py,

How far a page entity travels. Ordered: SKIP < COMPUTE < COMPUTE_AND_MATERIALIZE.

Verify coordinates are bit-for-bit identical across locale switches. This catches subtle bugs where values are "close" but not exact, which w

End-to-end parsing of real PDFs under comma-decimal locale.

Verify that float parsing in the C++ layer is locale-independent.

Verify correct handling of edge-case numeric values in PDFs.

Verify the Python __init__.py locale guard works.

Loads BLFontFace objects from embedded PDF font programs and caches the result (success or failure) by the blob's content-hash cache key. The cache ke

Families of PDF colour spaces (8.6) that the SC/SCN/sc/scn operands are interpreted against once CS/cs has resolved a named /ColorSpace resource.

Render-facing view of an embedded font program (/FontFile, /FontFile2, /FontFile3). Deliberately free of qpdf and nlohmann types so that render instru

Distinguishes the /FontFile3 subtypes that embedded_font_file_kind folds into FONT_FILE_CFF. The renderer needs this: Blend2D loads only SFNT containe

Loads the embedded font program formats that Blend2D rejects — Type 1 (/FontFile) and bare CFF (/FontFile3 /Type1C, /CIDFontType0C) — through FreeType

--------------------------------------------------------------------------- Custom libjpeg error handler that longjmp's instead of calling exit() ----

Table 51 – Operator Categories [ p 119 ]