_i2|SrSSKrSSKJr SSKJrJrJrJr SSKJ r SSK J r SSK J r Jr SS KJr "S S 5rg) z Enhanced Table Parser with Header Consolidation Parses tables from Docling documents and consolidates consecutive headers that immediately precede tables, following strict rules to avoid incorrect consolidation across content boundaries. N)datetime)AnyListOptionalTuple) InputFormat)PdfPipelineOptions)DocumentConverterPdfFormatOption) TableMetadatac \rSrSrSrSS\S\S\S\4SjjrS\S \ \ \ \ 44S jr S \ \ S \S \ \ \\ \44S jrS\ \S \4SjrSS \ \ S \S\S \\4SjjrS\ 4SjrS\ \ S\4SjrSrg)EnhancedTableParserz Enhanced table parser with intelligent header consolidation. Consolidates consecutive headers of the same level that immediately precede tables, with strict rules to prevent incorrect consolidation. separator max_lookbackenable_consolidationdebugc4XlX lX0lX@lg)a Initialize the enhanced table parser. Args: separator: String to join multiple headers (e.g., " ^ " or " | ") max_lookback: Maximum elements to look back for headers enable_consolidation: Whether to consolidate consecutive headers debug: Enable debug output N)rrrr)selfrrrrs 5/home/mande/repo/quber/src/quber/core/table_parser.py__init__EnhancedTableParser.__init__s #($8! pdf_pathreturnc[5nSUlSUlSURl[ [ R[US90S9nURU5nURn/nSn[UR55nUV V s/sHupU PM n n n [U 5V V s0sHup[U 5U _M nn n URGHn[!US5(aHUR"(a7[!UR"SS5(aUR"SR$OSnUR'[U5S 5n/n/nUS:aUR)U U5unnS nUR*(aU(aUR-U5nSnSnS n[!US 5(aUR.(aUR.R0nUR.R2nUS:aUS:a[!UR.S 5(aUR.R4(an[7UR.R45S:aUR.R4SO/nU(a)[7U5S:aUS(a[9US5OS nS n[!US5(avUR"(ae[!UR"SS 5(aGUR"SR:nUR<UR>UR@URB4nUREU USS9n[!US5(aURGU5OS n[IUUUUU(aUSOS UUUUUS9 nURKU5 US- nURL(dGMUROU5 GM XV4$s sn n fs sn n f)z Parse document and extract all tables with enhanced metadata. Args: pdf_path: Path to PDF file Returns: Tuple of (document object, list of TableMetadata) FT)pipeline_options)format_optionsrprovpage_nor Ndatagridbbox)limitexport_to_markdown) table_index page_numberheadersconsolidated_headerdescriptive_texttable_markdownpreceding_textrowscolsfirst_cell_content)(r do_ocrdo_table_structuretable_structure_optionsdo_cell_matchingr rPDFr convertdocumentlist iterate_items enumerateidtableshasattrr r!get_find_preceding_headersr_consolidate_headersr#num_rowsnum_colsr$lenstrr%ltrb_get_preceding_textr(r appendr_print_debug_info)rrr converterresultr9tables_metadatar)elements_with_depthitemdepth elements_listielemelement_to_indextablepage_num table_idxr+descriptive_textsr,r0r1 first_cell first_rowr%rJr/table_mdmetadatas rparse_document"EnhancedTableParser.parse_document1sY./"'.2+DH00A&'OO_N^-_` ""8,?? #8#9#9#;<1DE1D+$1D E7@7OP7OGABtHaK7OP__EE6**uzzgejjQRmU^>_>_ 1 %% ),,RY;IG " 1}-1-I-I-Yb-c**#' ((W&*&?&?&H#DDJuf%%%**zz**zz**!8quzz622uzz:=ejjoo:NQR:REJJOOA$6XZ $Y!);>GlYq\):PTJDuf%%%**APV9W9WJJqM&&QSS!##qss+"55mYVW5XN>EUL`=a=au//9gkH%'$$79J!21!5PT'-#- H  " "8 , 1 Kzzz&&x0M%P((WFPs N6(N<elementsrZc n/n/nSnSn[[URU55GHnX'- S- nXn [U 5Rn U S;a$UR (a[ SU S35 X44$U S:Xa'U(a UR (a [ S5 X44$[U S 5(aU R(dMU RR5n U (dM[U S 5=(a U RSLn U (aU Rn Uc@U nURS U 5 UR (a[ S U S U SSS35 GM6GM9X:XaEU(a>URS U 5 UR (a[ SU S U SSS35 GMGMUR (a[ SU SUS35 X44$U(dJ[U5S :Xa;URS U 5 UR (a[ SU SSS35 GMGMU(dGMSnGM X44$)a Find headers that immediately precede a table. Only consolidates headers that are: 1. Consecutive (no non-header content between them) 2. Of the same level 3. Immediately before the table 4. Not separated by boundary elements (tables, images, etc.) Args: elements: List of document elements table_idx: Index of the table in the elements list Returns: Tuple of (headers list, descriptive texts list) NTr ) TableItem PictureItemz Hit boundary element (z), stopping header searchListItemz/ Hit list item after finding headers, stoppingtextlevelrz Found first header at level z: 2...z$ Found consecutive header at level z Found different level header (z vs z ), stoppingz Found descriptive text: F) rangeminrtype__name__rprintr?rgstriprhinsertrE)rrbrZr+r[ current_level consecutiverUidxrV elem_typerg is_headerrhs rrA+EnhancedTableParser._find_preceding_headerss,"  s4,,i89A-!#C=DT ++I88::4YK?XYZn))iJ&7::KLb))]4(( 99??$D g.I4::T3II !($)MNN1d+zz >ugRSb {RUVW"+ NN1d+zz DUG2dSVTVi[X[\]" zz @tM?Zefg ))3'8#9Q#>%,,Q5zz :49+SIJ"W"'K}:D))rr+cpU(dg[U5S:XaUS$URRU5$)z Consolidate multiple headers into a single string. Args: headers: List of header texts to consolidate Returns: Consolidated header string Nr r)rErjoin)rr+s rrB(EnhancedTableParser._consolidate_headerss5 w<1 1: ~~""7++rr'c/nSn[[URU55HnX&- S- nXn[U5Rn U S:XaUR SS5 M=[ US5(dMPUR(dMcURR5n U (dMUR SU 5 US- nXS:dM O U(aSRU5$S$)a Get text content immediately preceding the table. Args: elements: List of document elements table_idx: Index of the table limit: Maximum number of text elements to include Returns: Preceding text context or None, with [Table] placeholders for intervening tables rr rdz[Table]rg N) rkrlrrmrnrqr?rgrpry) rrbrZr'textscountrUrtrVrurgs rrK'EnhancedTableParser._get_preceding_textss4,,i89A-!#C=DT ++IK' Q *tV$$yy(4LLD)QJE~!:$$)tyy2d2rr_c[SURS-SURS35 UR(a[SUR35 UR(a[SUR35 UR (a[SUR SS S 35 [S UR S URS 35 UR(a[SURSSS 35 gg)z$Print debug information for a table.z === Table r z (Page z) ===zConsolidated Header: zIndividual Headers: zDescriptive Text: NdrjzSize: u rows × z colsz First Cell: ri) ror)r*r,r+r-r0r1r2)rr_s rrM%EnhancedTableParser._print_debug_info+s X11A56gh>R>R=SSXYZ  ' ' )(*F*F)GH I    ()9)9(:; <  $ $ &x'@'@#'F&GsK L x}}oYx}}oUCD  & & L!)r0r1)r)r*r,r+r-r/ dimensionsr\r>wzutf-8)encodingF)indent ensure_asciiNzMetadata saved to: )rnow isoformatrEr)r*r,r+r-r/r0r1r2rLopenjsondumpro)rrrr#meta table_dictfs r save_metadata!EnhancedTableParser.save_metadata8s (||~779 .  "D#//#//'+'?'?<<$($9$9"&"5"5'+yy$))D"55 J N ! !* -"+sW 5 IIdae <6 #K=126 5s C// C=)rrrrN)z ^ TF)r&)rn __module__ __qualname____firstlineno____doc__rFintboolrrrrr r`rArBrrKrMr__static_attributes__rrrrs %) #   *j)sj)uS$}:M5M/Nj)XY*S Y*cY*eTXY\T]_cdg_hThNiY*v,DI,#, !3DI!3#!3c!3ZbcfZg!3F H- H 34 +> 3S 3rr)rrrtypingrrrrdocling.datamodel.base_modelsr"docling.datamodel.pipeline_optionsr docling.document_converterr r modelsr rrrrrs0 --5AI!D3D3r