jSrSSKJr SSKJrJr SSKJrJrJ r SSK J r J r \(aSSK Jr SSKJrJr SSKJr \ S r"S S \ 5r\"S S 55rg)uWhat the overlap matcher decides about each table on a page. Fusion matches the Set-of-Mark/Camelot tables against docling's tables and pictures one page at a time, by bounding-box overlap. A `RegionMatch` records one outcome: which SoM tables and which docling items overlapped, and what that cardinality means. The graft and annotation passes act on these records; they make no geometric decisions of their own. The seven kinds correspond one-to-one to the fusion matrix: - `replace` — one SoM table over one docling table, both agree. The Camelot body replaces the docling body; the SoM table is unchanged. - `som_merged` — more docling tables than SoM tables in the region: SoM fused stacked tables. `split_table` splits the SoM table; the splits graft in. - `docling_undercount` — more SoM tables than docling tables: docling dropped or merged a complex table. The docling table(s) in the region are deleted and the Camelot tables are inserted in their place, in vertical order. - `image_table` — a docling table with no text layer under it (read by OCR, Camelot empty): a real table rendered as an image. Docling's OCR'd table is kept; the SoM table is annotated as an image table. - `chart` — a SoM table over a docling picture and no docling table: SoM mistook an image for a table. Nothing grafts; the matched picture records that Set-of-Mark read it, and the SoM table is annotated as a chart. - `docling_miss` — a SoM table over nothing in docling: docling missed it. The Camelot table is inserted into the document. - `som_miss` — a docling table no SoM table matched: SoM missed a table docling found. Reported as an error; nothing is synthesized from docling. ) annotations) dataclassfield) TYPE_CHECKINGListLiteral) BaseModelField)DoclingDocument)CellFlagExtractedTable) ParseResult)replace som_mergeddocling_undercount image_tablechart docling_misssom_missc\rSrSr%SrS\S'S\S'\"\S9rS\S '\"\S9r S \S '\"\S9r S \S '\"\S S9r S \S'\"SSS9r S\S'\"SSS9r S\S'Srg) RegionMatch6aOne reconciliation outcome for a region on a page. `som_indices` index into the Set-of-Mark `ExtractedTable` list handed to the matcher. `docling_table_refs` / `docling_picture_refs` are docling `self_ref` strings. The combination of which lists are populated and `kind` drives the graft and annotation passes. After grafting, `docling_table_refs` reference the *unified* document: the graft deletes and inserts tables, which renumbers every later table's `self_ref`, so it rewrites each match's refs to the grafted tables' final positions. A consumer resolving these refs against the unified document (the report always ships beside it) lands on the table the match is about. intpage MatchKindkinddefault_factoryz List[int] som_indices List[str]docling_table_refsdocling_picture_refszBPredicted classes of the matched docling pictures (for chart kind))r descriptionpicture_classesgz9Representative box-overlap coverage for the region (0..1))defaultr#floatoverlapz Human-readable note on the matchstrdetailN)__name__ __module__ __qualname____firstlineno____doc____annotations__r listrr!r"r$r'r*__static_attributes__r+6/home/mande/repo/quber/src/quber/core/fusion/models.pyrr6s  I O"48K8$)$$? ?&+D&A)A!&X"OY!\GU0RSFCSr4rc~\rSrSr%SrS\S'S\S'\"\S9rS\S '\"\S9r S \S '\"\S9r S \S 'Sr S\S'Sr g) FusionResultTabThe two corrected outputs of one fusion. `document` is the unified `DoclingDocument`: docling's spine with the Camelot table bodies grafted in. `tables` is the corrected Set-of-Mark/Camelot output: merged regions split, charts and image tables annotated via `ExtractedTable.kind`. `matches` is the final per-region classification (after any split), and `errors` surfaces the regions that could not be fused: a table docling found that SoM missed, a merge the split pass declined, or a SoM table in a `docling_miss` or `docling_undercount` region left out of the unified document because Camelot read no body. z'DoclingDocument'documentzList['ExtractedTable']tablesrzList[RegionMatch]matchesr errorszList['CellFlag'] heading_flagsNz'ParseResult | None'parser+) r,r-r.r/r0r1rr2r;r<r=r>r3r+r4r5r7r7TsS   ""!&t!rFs]:#(//%?C.   T)T< '' 'r4