X-Git-Url: https://git.creatis.insa-lyon.fr/pubgit/?a=blobdiff_plain;f=src%2FgdcmDocument.cxx;h=11b88a2e6fefc1d8b1aebba0dd568c1c0cdeda05;hb=6278320cc85da00d2d56ffbf07806e84966892c3;hp=1c50ab555b50df2015b30e3f8a369d96717e7408;hpb=87f5ff9abfff1e2cb824083b999638ee2d9f415d;p=gdcm.git diff --git a/src/gdcmDocument.cxx b/src/gdcmDocument.cxx index 1c50ab55..11b88a2e 100644 --- a/src/gdcmDocument.cxx +++ b/src/gdcmDocument.cxx @@ -3,8 +3,8 @@ Program: gdcm Module: $RCSfile: gdcmDocument.cxx,v $ Language: C++ - Date: $Date: 2007/04/12 13:22:49 $ - Version: $Revision: 1.357 $ + Date: $Date: 2008/09/15 15:49:22 $ + Version: $Revision: 1.381 $ Copyright (c) CREATIS (Centre de Recherche et d'Applications en Traitement de l'Image). All rights reserved. See Doc/License.txt or @@ -37,7 +37,7 @@ #include // for memset #endif -namespace gdcm +namespace GDCM_NAME_SPACE { //----------------------------------------------------------------------------- @@ -50,7 +50,7 @@ const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096 /** * \brief This default constructor neither loads nor parses the file. - * You should then invoke \ref Document::Load. + * You should then invoke Document::Load. * */ Document::Document() @@ -63,6 +63,7 @@ Document::Document() SwapCode = 1234; Filetype = ExplicitVR; CurrentOffsetPosition = 0; + OffsetOfPreviousParseDES =0; // Load will set it to true if sucessfull Group0002Parsed = false; IsDocumentAlreadyLoaded = false; @@ -70,6 +71,8 @@ Document::Document() LoadMode = LD_ALL; // default : load everything, later SetFileName(""); + changeFromUN=false; + UnexpectedEOF=false; } /** @@ -159,6 +162,7 @@ bool Document::DoTheLoadingDocumentJob( ) // Recursive call. // Loading is done during parsing + OffsetOfPreviousParseDES = beg; ParseDES( this, beg, lgt, false); // delim_mode is first defaulted to false if ( IsEmpty() ) @@ -244,7 +248,7 @@ bool Document::DoTheLoadingDocumentJob( ) // Force Loading some more elements if user asked to. - gdcm::DocEntry *d; + GDCM_NAME_SPACE::DocEntry *d; for (ListElements::iterator it = UserForceLoadList.begin(); it != UserForceLoadList.end(); ++it) @@ -279,10 +283,13 @@ bool Document::DoTheLoadingDocumentJob( ) // std::string RecCode; RecCode = GetEntryString(0x0008, 0x0010); // recognition code (RET) - if (RecCode == "ACRNEMA_LIBIDO_1.1" || - RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares - // with "little-endian strings" + + + if(RecCode.find("ACRNEMA_LIBIDO") == 0 || // any version + RecCode.find("CANRME_AILIBOD") == 0) // for brain-damaged softwares + // with "little-endian strings" { + Filetype = ACR_LIBIDO; std::string rows = GetEntryString(0x0028, 0x0010); std::string columns = GetEntryString(0x0028, 0x0011); @@ -580,7 +587,7 @@ double Document::SwapDouble(double a) // // -----------------File I/O --------------- /** - * \brief Tries to open the file \ref Document::Filename and + * \brief Tries to open the file Document::Filename and * checks the preamble when existing, * or if the file starts with an ACR-NEMA look-like element. * @return The FILE pointer on success, 0 on failure. @@ -691,7 +698,7 @@ bool Document::CloseFile() * @param filetype Type of the File to be written * (ACR-NEMA, ExplicitVR, ImplicitVR) */ -void Document::WriteContent(std::ofstream *fp, FileType filetype) +void Document::WriteContent(std::ofstream *fp, FileType filetype, bool, bool) { // Skip if user wants to write an ACR-NEMA file @@ -719,8 +726,10 @@ void Document::WriteContent(std::ofstream *fp, FileType filetype) * --> was too much tricky / we were [in a hurry / too lazy] * --> We don't write the element 0x0000 (group length) */ - - ElementSet::WriteContent(fp, filetype); // This one is recursive + // This one is recursive + // false : outside MetaElements + // false : outside Sequence + ElementSet::WriteContent(fp, filetype, false, false); } // ----------------------------------------- @@ -849,7 +858,7 @@ void Document::LoadEntryBinArea(DataEntry *entry) //} /** - * \brief Compares two documents, according to \ref DicomDir rules + * \brief Compares two documents, according to DicomDir rules * \warning Does NOT work with ACR-NEMA files * \todo Find a trick to solve the pb (use RET fields ?) * @param document to compare with current one @@ -918,7 +927,7 @@ bool Document::operator<(Document &document) /** * \brief Reads a given length of bytes - * (in order to avoid to many CPU time consuming fread-s) + * (in order to avoid to many CPU time-consuming fread-s) * @param l length to read */ void Document::ReadBegBuffer(size_t l) @@ -1031,10 +1040,10 @@ int Document::ComputeGroup0002Length( ) vr = entry->GetVR(); //if ( (vr == "OB")||(vr == "OW")||(vr == "UT")||(vr == "SQ")) - // (no SQ, OW, UT in group 0x0002;) + // (no SQ, OW, OL, UT in group 0x0002;) if ( vr == "OB" ) { - // explicit VR AND (OB, OW, SQ, UT) : 4 more bytes + // explicit VR AND (OB, OW, OL, SQ, UT, UN) : 4 more bytes groupLength += 4; } groupLength += 2 + 2 + 4 + entry->GetLength(); @@ -1112,14 +1121,14 @@ void Document::ParseDES(DocEntrySet *set, long offset, { ///\todo FIXME : On 64 bits processors, tellg gives unexpected results after a while ? - /// Probabely a bug in gdcm code somwhere (some memory erased ?) + /// Probabely a bug in gdcm code somewhere (some memory erased ?) // Uncomment to track the bug -/* + if( Debug::GetDebugFlag() ) std::cout << std::dec <<"(long)(Fp->tellg()) " << (long)(Fp->tellg()) // in Debug mode << std::hex << " 0x(" <<(long)(Fp->tellg()) << ")" << std::endl; - */ + // if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) // Once per DocEntry if ( !delim_mode ) // 'and then' doesn't exist in C++ :-( @@ -1147,7 +1156,7 @@ void Document::ParseDES(DocEntrySet *set, long offset, if ( !first && newDocEntry->IsItemStarter() ) { // Debug message within the method ! - newDocEntry = Backtrack(newDocEntry); + newDocEntry = Backtrack(newDocEntry, set); } else { @@ -1221,7 +1230,7 @@ void Document::ParseDES(DocEntrySet *set, long offset, { /////////////////////// SeqEntry : VR = "SQ" - unsigned long l = newDocEntry->GetReadLength(); + unsigned long l = newDocEntry->GetReadLength(); if ( l != 0 ) // don't mess the delim_mode for 'zero-length sequence' { if ( l == 0xffffffff ) @@ -1282,11 +1291,11 @@ void Document::ParseDES(DocEntrySet *set, long offset, << " at offset 0x(" << std::hex << newDocEntry->GetOffset() << ")"); - ParseSQ( newSeqEntry, - newDocEntry->GetOffset(), - l, delim_mode_intern); + bool res = ParseSQ( newSeqEntry, + newDocEntry->GetOffset(), + l, delim_mode_intern); - gdcmDebugMacro( "Exit from ParseSQ, delim " << delim_mode_intern); + gdcmDebugMacro( "Exit from ParseSQ, delim " << delim_mode_intern << " -->return : " << res); } if ( !set->AddEntry( newSeqEntry ) ) { @@ -1318,15 +1327,20 @@ void Document::ParseDES(DocEntrySet *set, long offset, newDocEntry->Delete(); } first = false; + + if (UnexpectedEOF) // some terminator was missing + { + break; + } } // end While gdcmDebugMacro( "Exit from ParseDES, delim-mode " << delim_mode ); } /** * \brief Parses a Sequence ( SeqEntry after SeqEntry) - * @return parsed length for this level + * @return false if expected fff0,e000 not found */ -void Document::ParseSQ( SeqEntry *seqEntry, +bool Document::ParseSQ( SeqEntry *seqEntry, long offset, long l_max, bool delim_mode) { int SQItemNumber = 0; @@ -1339,9 +1353,11 @@ void Document::ParseSQ( SeqEntry *seqEntry, DocEntry *newDocEntry = ReadNextDocEntry(); if ( !newDocEntry ) - { + { + // The most frequent is when a SQ terminator is missing (?!?) gdcmWarningMacro("in ParseSQ : should never get here!"); - break; + UnexpectedEOF = true; + return false; } if ( delim_mode ) { @@ -1373,8 +1389,22 @@ void Document::ParseSQ( SeqEntry *seqEntry, dlm_mod = false; } + // avoid infinite loop when Bad assumption was made on illegal 'unknown length' UN //JPRx + + if (offsetStartCurrentSQItem <= OffsetOfPreviousParseDES) + { + gdcmWarningMacro("Bad assumption was made on illegal 'unknown length' UN!" << std::endl << + "OffsetOfPreviousParseDES " << std::hex << OffsetOfPreviousParseDES + << " offsetStartCurrentSQItem " << offsetStartCurrentSQItem); + /// \todo when "Bad assumption (SQ) on illegal 'unknown length' UN", Backtrack again + try OB + return false; + } + else + { + OffsetOfPreviousParseDES = offsetStartCurrentSQItem; + } + // fill up the current SQItem, starting at the beginning of fff0,e000 - Fp->seekg(offsetStartCurrentSQItem, std::ios::beg); // Once per SQItem ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod); offsetStartCurrentSQItem = Fp->tellg(); // Once per SQItem @@ -1389,15 +1419,17 @@ void Document::ParseSQ( SeqEntry *seqEntry, break; } } + return true; } /** * \brief When a private Sequence + Implicit VR is encountered * we cannot guess it's a Sequence till we find the first * Item Starter. We then backtrack to do the job. - * @param docEntry Item Starter that warned us + * @param docEntry Item Starter that warned us + * @param set DocEntrySet (ElementSet/SQItem) the DocEntry will belong */ -DocEntry *Document::Backtrack(DocEntry *docEntry) +DocEntry *Document::Backtrack(DocEntry *docEntry, DocEntrySet *set) { // delete the Item Starter, built erroneously out of any Sequence // it's not yet in the HTable/chained list @@ -1412,7 +1444,8 @@ DocEntry *Document::Backtrack(DocEntry *docEntry) gdcmDebugMacro( "Backtrack :" << std::hex << group << "|" << elem << " at offset 0x(" <RemoveEntry( PreviousDocEntry ); // forge the Seq Entry DocEntry *newEntry = NewSeqEntry(group, elem); @@ -1422,7 +1455,7 @@ DocEntry *Document::Backtrack(DocEntry *docEntry) // Move back to the beginning of the Sequence Fp->seekg(offset, std::ios::beg); // Only for Shadow Implicit VR SQ - return newEntry; + return newEntry; // It will added where it has to be! } /** @@ -1502,20 +1535,30 @@ void Document::FindDocEntryLength( DocEntry *entry ) { const VRKey &vr = entry->GetVR(); uint16_t length16; - if ( Filetype == ExplicitVR && !entry->IsImplicitVR() ) { - if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UT" - || vr == "UN" ) + + // WARNING : + // + // For some images, length of UN elements is coded on 2 bytes (instead of 4) + // There are *not* readable ! + // You can make a quick and dirty patch, commenting out + //| vr == "UN" + // in the following line. + // (the 'straight' images will no longer be readable ...) + + if ( vr == "OB" || vr == "OW" || vr == "OL" || vr == "SQ" || vr == "UT" + || vr == "UN" || changeFromUN == true) { + changeFromUN = false; // The following reserved two bytes (see PS 3.5-2003, section // "7.1.2 Data element structure with explicit vr", p 27) must be // skipped before proceeding on reading the length on 4 bytes. - //Fp->seekg( 2L, std::ios::cur); // Once per OW,OB,SQ DocEntry - uint32_t length32 = ReadInt32(); // Once per OW,OB,SQ DocEntry + //Fp->seekg( 2L, std::ios::cur); // Once per OB,OW,OL,UT,UN,SQ DocEntry + uint32_t length32 = ReadInt32(); // Once per OB,OW,OL,UT,UN,SQ DocEntry CurrentOffsetPosition+=4; - if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff ) + if ( (vr == "OB" || vr == "OW" || vr == "OL") && length32 == 0xffffffff ) { uint32_t lengthOB; try @@ -1549,11 +1592,9 @@ void Document::FindDocEntryLength( DocEntry *entry ) FixDocEntryFoundLength(entry, length32); return; } - // Length is encoded on 2 bytes. //length16 = ReadInt16(); length16 = GetInt16(); - // 0xffff means that we deal with 'No Length' Sequence // or 'No Length' SQItem if ( length16 == 0xffff) @@ -1603,7 +1644,8 @@ uint32_t Document::FindDocEntryLengthOBOrOW() uint16_t elem; try - { + { ///\todo make sure there is never OL encoded pixel data! + //group = ReadInt16(); // Once per fragment (if any) of OB,OW DataElements //elem = ReadInt16(); // Once per fragment (if any) of OB,OW DataElements ReadBegBuffer(4); // Once per fragment (if any) of OB,OW DataElements @@ -1622,8 +1664,8 @@ uint32_t Document::FindDocEntryLengthOBOrOW() { gdcmWarningMacro( "Neither an Item tag nor a Sequence delimiter tag on :" - << std::hex << group << " , " << elem - << ")" ); + << std::hex << group << "|" << elem << ") Pos. on entry was 0x(" <seekg(positionOnEntry, std::ios::beg); // Once per fragment (if any) of OB,OW DataElements throw FormatUnexpected( @@ -1684,21 +1726,12 @@ VRKey Document::FindDocEntryVR() //if ( !CheckDocEntryVR(vr) ) // avoid useless function call if ( !Global::GetVR()->IsValidVR(vr) ) - { -/* -// std::cout << "================================================================Unknown VR" - << std::hex << "0x(" - << (unsigned int)vr[0] << "|" << (unsigned int)vr[1] - << ")" << "for : " << CurrentGroup - << " at offset : 0x(" << positionOnEntry << ")" - << std::endl; -*/ - gdcmWarningMacro( "Unknown VR " << std::hex << "0x(" - << (unsigned int)vr[0] << "|" << (unsigned int)vr[1] - << ")" - << " at offset : 0x(" << CurrentOffsetPosition-4<< ") for group " << CurrentGroup - ); + { + gdcmWarningMacro( "Unknown VR " << vr.GetHexaRepresentation() << std::hex + << " at offset : 0x(" << CurrentOffsetPosition-4 + << ") for group " << std::hex << CurrentGroup ); + //Fp->seekg(positionOnEntry, std::ios::beg); //JPRx //Fp->seekg((long)-2, std::ios::cur);// only for unrecognized VR (?!?) //see :MR_Philips_Intera_PrivateSequenceExplicitVR.dcm @@ -1737,7 +1770,7 @@ void Document::SkipDocEntry(DocEntry *entry) void Document::SkipToNextDocEntry(DocEntry *currentDocEntry) { long l = currentDocEntry->GetReadLength(); - if ( l == -1 ) // length = 0xffff shouldn't appear here ... + if ( (uint32_t) l == (uint32_t)-1 ) // length = 0xffff shouldn't appear here ... // ... but PMS imagers happen ! return; Fp->seekg((size_t)(currentDocEntry->GetOffset()), std::ios::beg); //FIXME :each DocEntry @@ -1772,10 +1805,12 @@ void Document::FixDocEntryFoundLength(DocEntry *entry, if ( foundLength % 2) { - gdcmWarningMacro( "Warning : Tag with uneven length " << foundLength - << " in x(" << std::hex << gr << "," << elem <<")"); + gdcmWarningMacro( "Warning : Tag (" << std::hex << gr << "|" << elem << ") with uneven length " + << std::dec << foundLength << " 0x(" << std::hex << foundLength << ") " + //<< " at offset x(" << offset << ")" + ); } - + //////// Fix for some naughty General Electric images. // Allthough not recent many such GE corrupted images are still present // on Creatis hard disks. Hence this fix shall remain when such images @@ -2062,8 +2097,7 @@ bool Document::CheckSwap() s16 = *((uint16_t *)(deb)); gdcmDebugMacro("not a DicomV3 nor a 'clean' ACR/NEMA;" - << " (->despaired wild guesses !)"); - + << " (->despaired wild guesses !)"); switch ( s16 ) { case 0x0001 : @@ -2091,9 +2125,48 @@ bool Document::CheckSwap() Filetype = ACR; return true; default : - gdcmWarningMacro("ACR/NEMA unfound swap info (Hopeless !)"); - Filetype = Unknown; - return false; + + s16 = *((uint16_t *)(deb)); + if (s16 != 0x0000) + return false; + s16 = *((uint16_t *)(deb+2)); + + Fp->seekg ( 0L, std::ios::beg); // Once per Document + CurrentOffsetPosition = 0; + switch(s16) // try an other trick! + // -> to be able to decode 0029|1010 DataElement + // -> and be not less cleaver than dcmdump ;-) + { + case 0x0004 : + SwapCode = 1234; + break; + case 0x0400 : + SwapCode = 3412; + break; + default: + gdcmWarningMacro("ACR/NEMA unfound swap info (Hopeless !)"); + Filetype = Unknown; + return false; + } + // Check if next 2 bytes are a VR + // Probabely something more time-consuming exists with std::string + const char VRvalues[] = "AEASATCSDADTFLFDISLOLTPNSHSLSSSTTMUIULUSUTOBOWOLOFATUNSQRT"; + int nbVal = 29; + const char *pt = VRvalues; + for (int i=0;i= 0x0010 && CurrentElem <=0x00ff )) - { - // DICOM PS 3-5 7.8.1 a) states that those - // (gggg-0010->00FF where gggg is odd) attributes have to be LO - realVR = "LO"; + // -> Better we fix the problem at Write time + + else if (CurrentGroup%2 == 1 ) + { + if (CurrentElem >= 0x0010 && CurrentElem <=0x00ff ) + { + // DICOM PS 3-5 7.8.1 a) states that : + // Private Creator Data Elements numbered (gggg,0010-00FF) (gggg is odd) + // attributes have to be LO (Long String) and the VM shall be equal to 1 + realVR = "LO"; + + // Seems not to be true + // Still in gdcmtk, David Clunnie disagrees, Marco Eichelberg says it's OK ... + // We let it for a while? + //(We should check length==4, for more security, but we don't have it yet !) + } + else if ( CurrentElem == 0x0001) + { + realVR = "UL"; // Private Group Length To End + } + else // check the private dictionary for shadow elements when Implicit VR! + { + DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem); + if ( dictEntry ) + { + realVR = dictEntry->GetVR(); + dictEntry->Unregister(); // GetDictEntry registered it + } + } } - */ + else { DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem);//only when ImplicitVR @@ -2214,6 +2309,28 @@ DocEntry *Document::ReadNextDocEntry() } } + // if UN found, let's check the dictionary, and trust it! + // (maybe a private dictionary exists?) + else if (vr == "UN") + { + DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem); + if ( dictEntry ) + { + realVR = dictEntry->GetVR(); + dictEntry->Unregister(); // GetDictEntry registered it + + // for VR = "UN", length is always stored on 4 bytes. + // remember this info, in order not to crash later + changeFromUN=true; + /// \todo : fixme If inside a vr = "UN" DataElement (but SQ according to a private dictionnary) + /// there is some more vr = "UN" DataElements, it will probabely fail. + /// --> find a -non time consuming- trick to store changeFromUN info at DataElement level, + /// not at the Document level. + /// --> ?!? JPR + + } + } + DocEntry *newEntry; //if ( Global::GetVR()->IsVROfSequence(realVR) ) if (realVR == "SQ") @@ -2239,7 +2356,7 @@ DocEntry *Document::ReadNextDocEntry() int offset = Fp->tellg();//Only when heuristic for Explicit/Implicit was wrong gdcmWarningMacro("Entry (" << newEntry->GetKey() << ") at x(" - << offset << ") should be Explicit VR"); + << std::hex << offset << ") should be Explicit VR"); } } newEntry->SetImplicitVR();