jSrSSKJr SSKrSSKJr SSKJr SSKJ r SSK J r \(a SSK r SSKJrJr \R$"\5r"S S \ 5r\SS j5rSS jrSS jrSSjrg)aExtract a PDF's outline (bookmarks / table-of-contents). The outline, when present, is the most authoritative heading-hierarchy signal in a PDF. Two extractors are provided so each backend uses its own native capability: * :func:`extract_outline_from_pdfium` -- for the pypdfium2 backend. Returns the richest data: title, depth, target page and vertical position. * :func:`extract_outline_from_docling_parse` -- for the docling-parse backends, using their native ``get_table_of_contents()`` (no pypdfium2 dependency). The native outline carries titles and hierarchy only, so page number and position are left unset. ``pypdfium2`` is imported lazily, inside the functions that use it, never at module level: ``datamodel.document`` imports this module for the ``_PdfOutlineItem`` model, which places it on the ``docling.service_client`` import chain. That chain must stay importable on any docling-slim install that does not enable the PDF pipeline (and therefore ships no ``pypdfium2``). ) annotationsN)cache) TYPE_CHECKING) BaseModel)pypdfium2_lock) PdfDocumentPdfTableOfContentscJ\rSrSr%SrS\S'S\S'SrS\S 'SrS \S 'S rg) _PdfOutlineItem)akA single PDF bookmark / table-of-contents entry (internal). Internal data-passing structure between a PDF backend's ``get_document_outline()`` and the heading-hierarchy stage; not part of the public datamodel or the serialized output. The list is kept flat and in document order; each entry carries its own ``level`` so no tree structure is needed for matching. strtitleintlevelNz int | Nonepage_noz float | Noney_top) __name__ __module__ __qualname____firstlineno____doc____annotations__rr__static_attributes__rV/home/mande/repo/quber/.venv/lib/python3.13/site-packages/docling/utils/pdf_outline.pyr r )s' J JGZE<rr crSSKJn URSURSURSUR S0$)alDestination view modes whose coordinates carry a usable vertical (top) position. Coordinates are in PDF space (bottom-left origin). Modes not listed (FIT, FITV, FITB, FITBV, unknown) provide no top. Returns: A mapping of each supported ``PDFDEST_VIEW_*`` mode to the index of the vertical (top) coordinate within that mode's position tuple: * ``PDFDEST_VIEW_XYZ`` -> ``1`` (position is ``[x, y, zoom]``) * ``PDFDEST_VIEW_FITH`` -> ``0`` (position is ``[y]``) * ``PDFDEST_VIEW_FITBH`` -> ``0`` (position is ``[y]``) * ``PDFDEST_VIEW_FITR`` -> ``3`` (position is ``[left, bottom, right, top]``) rN) pypdfium2.rawrawPDFDEST_VIEW_XYZPDFDEST_VIEW_FITHPDFDEST_VIEW_FITBHPDFDEST_VIEW_FITR)pdfium_cs r_view_top_indexr';s?"% !!1""A##Q""A  rcUR5nUR5up#[5RU5nUbU[ U5:aX4OSnX4$)zReturn ``(0-based page index, vertical top in PDF bottom-left coords)`` for a dest. Either element may be ``None`` when the destination does not encode it. N) get_indexget_viewr'getlen)dest page_indexmodeposidxy_pdfs r _dest_top_pdfr3VsP !J ID     %CC#c(NCHE  rc SSKJn /n0n[ [UR 55nUHnUR5=(d SR5nU(dM3SnSn UR5n U bJ[U 5upU b:U S-nU b2X;a&X n U R5X;'U R5 X;U - n UR[U[UR 5XS95 M SSS5 U$!Ua+n[ R SU5 /sSnAsSSS5 $SnAff=f!Ua Sn Nf=f!,(df  U$=f)aExtract the outline as a flat, document-ordered list of :class:`_PdfOutlineItem`. Vertical positions are converted to top-left origin (matching ``DocItem`` provenance) using the target page height. Returns an empty list when the document has no outline or it cannot be read. r) PdfiumErrorzCould not read PDF outline: %sNr)rrrr)pypdfium2._helpers.miscr5rlistget_toc_logdebug get_titlestripget_destr3 get_heightcloseappendr rr)pdocr5items page_heightstocexcbmrrrr-r.r2pages rextract_outline_from_pdfiumrIbsN4#%E%'L  t||~&C B\\^)r002E"&G"&E {{}$1$$7! )(1nG(%9#'#3D7;7HL4 JJL , 85 @ LLs288}g - F LA  JJ7 =I    ! F Ls^EC>9E&D26A>E>D/D*D/E*D//E2D=:E<D==E Ec^UR5nUc/$/n[UR=(d /5Vs/sHo3S4PM nnU(aUR5unmUR=(d UR =(d SR 5nU(aUR[UTS95 URU4Sj[UR=(d /555 U(aMU$s snf)aFlatten docling-parse's native table-of-contents into ordered ``_PdfOutlineItem``\ s. Walks the ``PdfTableOfContents`` tree returned by ``PdfDocument.get_table_of_contents()``, depth-first, assigning each node a 0-based ``level`` from its depth (top-level entries at level 0, matching the pypdfium2 extractor). The native outline exposes only the title and the tree structure -- no target page or vertical position -- so ``page_no`` and ``y_top`` are left ``None`` and the heading matcher falls back to title-only matching. ``get_table_of_contents()`` returns ``None`` for PDFs without an embedded outline, in which case an empty list is returned. rr6)rrc30># UH oTS-4v M g7f)rNr).0childrs r 5extract_outline_from_docling_parse..sS5REUQY'5Rs) get_table_of_contentsreversedchildrenpoptextorigr=rAr extend)dp_docrErCrMstacknoderrs @r"extract_outline_from_docling_parserZs  & & (C { #%E!)); <3 Qr5RSS % L3s C:)returnzdict[int, int])r-zpdfium.PdfDestr[ztuple[int | None, float | None])rBzpdfium.PdfDocumentr[list[_PdfOutlineItem])rWDoclingParsePdfDocumentr[r\)r __future__rlogging functoolsrtypingrpydanticrdocling.utils.locksr pypdfium2pdfiumdocling_parse.pdf_parserrr]r getLoggerrr:r r'r3rIrZrrrrhs{"# . "i$4 0f" #""r