Module inmem_raster 

In-memory Image-XObject → table lattice (WASM-safe; no pdfimages). In-memory raster table recovery (WASM-safe).

First principles: PDF Image XObjects are the pixel source (ISO 32000). Ruled tables are recoverable from projection of dark ink — no Poppler / Tesseract process for page render. Cell text uses crate::pdf::ocr.

Structs

RasterCandidate

Prepared OCR candidate for the two-phase (plan → OCR → finish) path.

RasterCandidateMeta

Serializable candidate metadata (no pixel buffer).

Enums

RasterCandidateKind

How OCR words from a raster candidate should be assembled.

Functions

bordered_table_is_plausible

True when a recovered bordered table looks contentful enough to keep.

build_table_from_candidate

Build a TableBorder from a prepared candidate + optional OCR words.

build_text_chunks_from_candidate

Convert OCR words from a text-blocks candidate into page-space TextChunks.

collect_raster_candidates

Collect OCR candidates from Image XObjects (no OCR yet).

ocr_words_to_text_chunks

Map OCR words into page-space text chunks (shared by TextBlocks assembly and Table→text fallback when a lattice cannot be formed).

prepare_candidate

Decode + pre-filter a single Image XObject into an OCR candidate.

recover_embedded_raster_tables

Recover bordered tables from embedded Image XObjects on a page.