j>SSKrSSKrSSKrSSKJr SSKJr SSKJr SSK J r J r J r J r JrJrJrJr SSKJr SSKJr SSKJrJrJr SS KJr SS KJrJrJrJ r J!r! SS K"J#r# SS K$J%r% SS K&J'r'J(r( SSK)J*r*J+r+J,r, \RZ"\.5r/\"SS55r0\"SS55r1S\#S\2\2\3\44S4SS4Sjr5"SS\'5r6g)N)Iterator) dataclass)Path) BoundingBox CoordOrigin DocItemLabelDoclingDocumentDocumentOriginImageRefProvenanceItemSize)SegmentedPdfPage)AbstractDocumentBackend)PdfDocumentBackendPdfPageBackenditer_pdf_page_backends)"ThreadedDoclingParseBackendOptions)ConversionStatusDoclingComponentType ErrorItemFailureCategoryPage)ConversionResultNativePdfPipelineOptions)ConvertPipelineget_expected_page_nos)ProfilingScopeTimeIntervalRecorder TimeRecorderc@\rSrSr%Sr\\S'Sr\\S'Sr\\S'Sr g)_NativeBuildTimings4 parser_wait materializepage_image_fetchN) __name__ __module__ __qualname____firstlineno__r%float__annotations__r&r'__static_attributes__r(a/home/mande/repo/quber/.venv/lib/python3.13/site-packages/docling/pipeline/native_pdf_pipeline.pyr"r"4s!KK!e!r0r"cj\rSrSr%Sr\\S'Sr\\S'Sr\\S'Sr \ \S'Sr \ \S'Sr \ \S 'S r g ) _NativeAssemblyTimings;r$page_image_encode text_items picture_itemsr n_text_items n_picturesn_picture_imagesr(N)r)r*r+r,r5r-r.r6r7r8intr9r:r/r(r0r1r3r3;s@"u"JM5L#Jcr0r3conv_restimings.returncUH>up#[X[RS9nURU5 UR 5 M@ g)z8Publish native phases so `--profiling` reports them too.scopeN)rrDOCUMENTaddclose)r<r=keysecondsrecorders r1_record_timingsrHFs8 '^=T=TU W r0c ^\rSrSrSrS\SS4U4SjjrS\S\4SjrS \ S \ \ S\ \ 4S jr S \ SS4S jrS \ S\S\ S\S-4SjrS\S\ S\SS4SjrS\S\4SjrS \S\S\S\S\SS4 SjrS \S\S\S\S\SS4 Sjr\S\S\S\4Sj5rS\S\4SjrS\SS4Sjr \!S\4Sj5r"\!S \#S\$4Sj5r%Sr&U=r'$) NativePdfPipelinePaModel-free PDF pipeline built on the native content of the PDF. Every text cell reported by the PDF backend becomes a plain `TextItem` with a provenance box, every embedded bitmap becomes a `PictureItem`, and - when `generate_page_images` is set - the rendered page is attached to the page. No layout, OCR or table model runs, so conversion is fast, but the document has no reading order, headings or tables: items appear in the order the parser reports them. pipeline_optionsr>NcR>[TU]U5 XlURUlgN)super__init__rLgenerate_page_images keep_images)selfrL __class__s r1rPNativePdfPipeline.__init__[s% )*:J+@@r0r<c URRn[U[5(d9[ S[ U5R SURRS35e[U5nU(d[RUl U$[5nURU5 [R"5n[!US["R$S9 UR'U[)X#55GH)nUR+XAU5nUbUR,R/U5 UR15 UR2R4n[R"5U- n UcMwX:dM~SUSSU SS [7UR,5S [7U5S 3 n [8R;U 5 UR<R/[?[@RBURDR U [FRHS 95 [RJUl O UR,RMS S9 SSS5 [8ROSURRRP[7UR,5[R"5U- UR2RRURTURVURX5 UR,(d[RUl [[USURT4SURV4SURX445 U$!,(df  N=f)NzThe selected backend z for z is not a PDF backend. Can not convert this with the native PDF pipeline. Please check your format configuration on DocumentConverter. doc_buildr@z&Document processing timeout: exceeded z.3fzs limit after z s. Processed /z pages.)component_type module_name error_messagecategorycUR$rNpage_no)pages r13NativePdfPipeline._build_document..sr0)rEzNative build of %s: %d page(s) in %.2fs with %d parser thread(s) [waiting for the parser %.2fs, materializing cells %.2fs, page images %.2fs]native_parser_waitnative_materializenative_page_image_fetch).input_backend isinstancer RuntimeErrortyper)filerrFAILUREstatusr"_warn_on_rerendertime monotonicr rrB_timed_page_backendsr _parse_pagepagesappendunloadrLdocument_timeoutlen_logwarningerrorsrrPIPELINErTrTIMEOUTPARTIAL_SUCCESSsortinfonameparser_threadsr%r&r'rH) rSr<backendexpected_page_nosr= start_time page_backendr`timeoutelapsed timeout_msgs r1_build_document!NativePdfPipeline._build_document`s..))'#566'W (>(>'?uX^^EXEXDYZOP  2(; .66HOO%' w'^^% (K~7N7N O $ 9 9/K! ''<H#NN))$/##%//@@..*Z7&7+<@ ^"3-}S5H4I012'; LL-OO**!+?+H+H(,(?(?*5%4%<%< '7&F&FHO5!@ NN  $=  >CPF  ! NN   $ $   NN z )  ! ! 0 0        $ $ ~~.66HO %w':':;%w':':;*G,D,DE  wP OsBL3*L31C L33 Mr= page_backendsc## [R"5n[U5nU=R[R"5U- - slUv MS![a. U=R[R"5U- - slgf=f!U=R[R"5U- - slf=f7f)aYield the page backends, timing how long each one is waited for. That wait is the pipeline's view of the parser's threaded decode: with enough parser threads the next page is already finished and the wait is near zero, while a single thread makes it the bulk of the build. N)rorpnext StopIterationr%)rSr=rstartedrs r1rq&NativePdfPipeline._timed_page_backendssnn&G B#M2 ##t~~'7''AA# ! ##t~~'7''AA# ##t~~'7''AA#s9C A1C B!B",CBB-B??CrcURR(dgURn[U[5(dgUR (dg[ R"URURR5(ag[RSURURRURR5 g)a6Warn when every page image will be rasterized a second time. The threaded parser rasterizes each page on its worker threads while decoding it, but only at its own `render_scale`. Asking for a page image at any other scale re-rasterizes it here, on the calling thread, which is both duplicate work and serial. NzThe backend rasterizes pages at scale %.2f but the pipeline asks for page images at scale %.2f, so every page is rasterized twice, the second time serially. Set the backend's render_scale to %.2f.) rLrQoptionsrhr render_pagesmathisclose render_scale images_scalerxry)rSrrs r1rn#NativePdfPipeline._warn_on_rerenders$$99 //'#EFF ##  <<,,d.C.C.P.P Q Q   L   ! ! . .  ! ! . .  r0rc0UR5(d0UR5nURUUU(aSU3OS5 g[R"5n[ UR S9nUR5UlUR5Ul U=R[R"5U- - sl URR(a[R"5nX6lURRUlUR#URRS9 SUlU=R$[R"5U- - slU$![&anURX#SU35 SnAgSnAff=f)zFCollect the native content of one page, or None if it failed to parse.zPage failed to parse: zPage failed to parse.Nr^scale)is_validget_error_message_record_page_failurerorprr_get_sizesizeget_segmented_page parsed_pager&rLrQrgr_default_image_scale get_imager' Exception)rSr=r<rdetailrr`excs r1rrNativePdfPipeline._parse_pagesX$$&&!335F  % %)1,   nn&G 4 45D$--/DI+>>@D    4>>#3g#= = $$99..* , ,0,A,A,N,N)T%:%:%G%GH $ ((DNN, NN   $ $      $ $  % %     ! !   +W-F-FG$g&8&89')>)>?   $&!'AS Rs IM  Mrr_rrcURRnSnURU5HynURn U (dMUR [ R U U [UURURR5U5S[U 54S9S9 US- nM{ U=RU- sl US:Xa4UR(a"[RSUUR 5 ggg)Nrr_bboxcharspan)labeltextorigprovz\Page %d has text, but no %s cells: the PDF backend does not materialize that text cell unit.)rLtext_cell_unit iterate_cellsradd_textrTEXTr _to_prov_bboxrectto_bounding_boxrwr8textline_cellsrxryvalue) rSr=rr_rrunitn_cellscellrs r1r!NativePdfPipeline._add_text_items_s$$33--d3D99D LL"''##++DII,E,E,GNT^   qLG4 ' a> " " NN # # * * , #r0c[5$rNr)clss r1get_default_options%NativePdfPipeline.get_default_optionss '))r0c"[U[5$rN)rhr)rrs r1is_backend_supported&NativePdfPipeline.is_backend_supporteds'#566r0)rRrL)(r)r*r+r,__doc__rrPrrr"rrrqrrnrrrstrrrr3r r;rr rr staticmethodrrrrr classmethodrrboolrr/ __classcell__)rTs@r1rJrJPsA)AAdA M(8M=MM^$ / . ! * );  6)$)#)% )  )V;(;8F;QT; ;"B+;B@PBH!'!! ! & !  ! !F:':: : & :  : :.CKCtC CC *:?O - 0-T-*$<**7+B7t77r0rJ)7loggingrrocollections.abcr dataclassesrpathlibrdocling_core.types.docrrrr r r r r docling_core.types.doc.pager docling.backend.abstract_backendrdocling.backend.pdf_backendrrr!docling.datamodel.backend_optionsrdocling.datamodel.base_modelsrrrrrdocling.datamodel.documentr"docling.datamodel.pipeline_optionsrdocling.pipeline.base_pipelinerrdocling.utils.profilingrrr getLoggerr)rxr"r3tuplerr-rHrJr(r0r1rs $!   9D Q8G " "" "   ).uS%Z/@#/E)F b7b7r0