X-Git-Url: https://git.creatis.insa-lyon.fr/pubgit/?a=blobdiff_plain;f=src%2FgdcmDocument.cxx;h=11b88a2e6fefc1d8b1aebba0dd568c1c0cdeda05;hb=6278320cc85da00d2d56ffbf07806e84966892c3;hp=d9767fd616de8ec00c5588d209ede837870273fb;hpb=6c1b9eb78c07847c3dbfeccc55e031df5cf47a5f;p=gdcm.git diff --git a/src/gdcmDocument.cxx b/src/gdcmDocument.cxx index d9767fd6..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: 2005/10/28 12:44:26 $ - Version: $Revision: 1.319 $ + 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 @@ -33,7 +33,11 @@ #include // for isdigit #include // for atoi -namespace gdcm +#if defined(__BORLANDC__) + #include // for memset +#endif + +namespace GDCM_NAME_SPACE { //----------------------------------------------------------------------------- @@ -46,7 +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() @@ -58,12 +62,17 @@ Document::Document() Initialize(); SwapCode = 1234; Filetype = ExplicitVR; + CurrentOffsetPosition = 0; + OffsetOfPreviousParseDES =0; // Load will set it to true if sucessfull Group0002Parsed = false; IsDocumentAlreadyLoaded = false; IsDocumentModified = true; LoadMode = LD_ALL; // default : load everything, later + SetFileName(""); + changeFromUN=false; + UnexpectedEOF=false; } /** @@ -76,7 +85,6 @@ Document::~Document () //----------------------------------------------------------------------------- // Public - /** * \brief Loader. use SetLoadMode(), SetFileName() before ! * @return false if file cannot be open or no swap info was found, @@ -90,18 +98,24 @@ bool Document::Load( ) return false; } return DoTheLoadingDocumentJob( ); -} +} + + +//#ifndef GDCM_LEGACY_REMOVE /** * \brief Loader. (DEPRECATED : not to break the API) * @param fileName 'Document' (File or DicomDir) to be open for parsing * @return false if file cannot be open or no swap info was found, * or no tag was found. */ + /* bool Document::Load( std::string const &fileName ) { Filename = fileName; return DoTheLoadingDocumentJob( ); } +*/ +//#endif /** * \brief Performs the Loading Job (internal use only) @@ -127,10 +141,10 @@ bool Document::DoTheLoadingDocumentJob( ) gdcmDebugMacro( "Starting parsing of file: " << Filename.c_str()); - Fp->seekg(0, std::ios::end); - long lgt = Fp->tellg(); // total length of the file - - Fp->seekg(0, std::ios::beg); + // Computes the total length of the file + Fp->seekg(0, std::ios::end); // Once per Document ! + long lgt = Fp->tellg(); // Once per Document ! + Fp->seekg(0, std::ios::beg); // Once per Document ! // CheckSwap returns a boolean // (false if no swap info of any kind was found) @@ -148,6 +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() ) @@ -159,7 +174,7 @@ bool Document::DoTheLoadingDocumentJob( ) } IsDocumentAlreadyLoaded = true; - Fp->seekg( 0, std::ios::beg); + //Fp->seekg(0, std::ios::beg); // Once per Document! // Load 'non string' values @@ -212,13 +227,13 @@ bool Document::DoTheLoadingDocumentJob( ) } //FIXME later : how to use it? - SeqEntry *modLutSeq = GetSeqEntry(0x0028,0x3000); + SeqEntry *modLutSeq = GetSeqEntry(0x0028,0x3000); // Modality LUT Sequence if ( modLutSeq !=0 ) { SQItem *sqi= modLutSeq->GetFirstSQItem(); if ( sqi != 0 ) { - DataEntry *dataEntry = sqi->GetDataEntry(0x0028,0x3006); + DataEntry *dataEntry = sqi->GetDataEntry(0x0028,0x3006); // LUT Data if ( dataEntry != 0 ) { if ( dataEntry->GetLength() != 0 ) @@ -233,7 +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) @@ -245,7 +260,7 @@ bool Document::DoTheLoadingDocumentJob( ) if ( d == NULL) { - gdcmWarningMacro( "You asked toForce Load " << std::hex + gdcmWarningMacro( "You asked to ForceLoad " << std::hex << (*it).Group <<"|"<< (*it).Elem << " that doesn't exist" ); continue; @@ -254,7 +269,7 @@ bool Document::DoTheLoadingDocumentJob( ) LoadDocEntry(d, true); } - CloseFile(); + CloseFile(); // ---------------------------- // Specific code to allow gdcm to read ACR-LibIDO formated images @@ -268,10 +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); @@ -338,22 +356,33 @@ bool Document::SetShaDict(DictKey const &dictName) * @return false when we're 150 % sure it's NOT a Dicom/Acr file, * true otherwise. */ -bool Document::IsReadable() +bool Document::IsParsable() { if ( Filetype == Unknown ) { - gdcmErrorMacro( "Wrong filetype"); + gdcmWarningMacro( "Wrong filetype for " << GetFileName()); return false; } if ( IsEmpty() ) { - gdcmErrorMacro( "No tag in internal hash table."); + gdcmWarningMacro( "No tag in internal hash table."); return false; } return true; } +/** + * \brief This predicate tells us whether or not the current Document + * was properly parsed and contains at least *one* Dicom Element + * (and nothing more, sorry). + * @return false when we're 150 % sure it's NOT a Dicom/Acr file, + * true otherwise. + */ +bool Document::IsReadable() +{ + return IsParsable(); +} /** * \brief Predicate for dicom version 3 file. @@ -412,7 +441,10 @@ std::string Document::GetTransferSyntax() // The entry might be present but not loaded (parsing and loading // happen at different stages): try loading and proceed with check... - LoadDocEntrySafe(entry); + + // Well ... + // (parsing and loading happen at the very same stage!) + //LoadDocEntrySafe(entry); //JPRx if (DataEntry *dataEntry = dynamic_cast(entry) ) { std::string transfer = dataEntry->GetString(); @@ -555,9 +587,10 @@ double Document::SwapDouble(double a) // // -----------------File I/O --------------- /** - * \brief Tries to open the file \ref Document::Filename and - * checks the preamble when existing. - * @return The FILE pointer on success. + * \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. */ std::ifstream *Document::OpenFile() { @@ -600,22 +633,8 @@ std::ifstream *Document::OpenFile() return 0; } - //-- ACR or DICOM with no Preamble; may start with a Shadow Group -- - if ( - zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 || - zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 || - zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 || - zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 ) - { - std::string msg = Util::Format( - "ACR/DICOM starting by 0x(%04x) at the beginning of the file\n", zero); - // FIXME : is it a Warning message, or a Debug message? - gdcmWarningMacro( msg.c_str() ); - return Fp; - } - //-- DICOM -- - Fp->seekg(126L, std::ios::cur); + Fp->seekg(126L, std::ios::cur); // Once per Document char dicm[4]; // = {' ',' ',' ',' '}; Fp->read(dicm, (size_t)4); if ( Fp->eof() ) @@ -623,15 +642,35 @@ std::ifstream *Document::OpenFile() CloseFile(); return 0; } + if ( memcmp(dicm, "DICM", 4) == 0 ) { HasDCMPreamble = true; return Fp; } + //-- Broken ACR or DICOM (?) with no Preamble; may start with a Shadow Group -- + // FIXME : We cannot be sure the preable is only zeroes.. + // (see ACUSON-24-YBR_FULL-RLE.dcm ) + if ( + zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 || + zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 || + zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 || + zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 || + zero == 0x0028 || 0x2800 // worse : some ACR-NEMA like files + // start 00028 group ?!? + ) + { + std::string msg = Util::Format( + "ACR/DICOM starting by 0x(%04x) at the beginning of the file\n", zero); + // FIXME : is it a Warning message, or a Debug message? + gdcmWarningMacro( msg.c_str() ); + return Fp; + } + // -- Neither ACR/No Preamble Dicom nor DICOMV3 file CloseFile(); - // Don't user Warning nor Error, not to polute the output + // Don't user Warning nor Error, not to pollute the output // while directory recursive parsing ... gdcmDebugMacro( "Neither ACR/No Preamble Dicom nor DICOMV3 file: " << Filename.c_str()); @@ -659,12 +698,12 @@ 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 if ( filetype == ImplicitVR || filetype == ExplicitVR || - filetype == JPEG ) + filetype == JPEG || filetype == JPEG2000 ) { // writing Dicom File Preamble char filePreamble[128]; @@ -672,7 +711,6 @@ void Document::WriteContent(std::ofstream *fp, FileType filetype) fp->write(filePreamble, 128); fp->write("DICM", 4); } - /* * \todo rewrite later, if really usefull * - 'Group Length' element is optional in DICOM @@ -688,8 +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); } // ----------------------------------------- @@ -706,15 +746,15 @@ void Document::LoadEntryBinArea(uint16_t group, uint16_t elem) DocEntry *docEntry = GetDocEntry(group, elem); if ( !docEntry ) { - gdcmWarningMacro(std::hex << group << "|" << elem - << "doesn't exist" ); + gdcmDebugMacro(std::hex << group << "|" << elem + << " doesn't exist" ); return; } DataEntry *dataEntry = dynamic_cast(docEntry); if ( !dataEntry ) { gdcmWarningMacro(std::hex << group << "|" << elem - << "is NOT a DataEntry"); + << " is NOT a DataEntry"); return; } LoadEntryBinArea(dataEntry); @@ -726,7 +766,7 @@ void Document::LoadEntryBinArea(uint16_t group, uint16_t elem) * @param entry Entry whose binArea is going to be loaded */ void Document::LoadEntryBinArea(DataEntry *entry) -{ +{ if( entry->GetBinArea() ) return; @@ -734,8 +774,8 @@ void Document::LoadEntryBinArea(DataEntry *entry) if ( openFile ) OpenFile(); - size_t o =(size_t)entry->GetOffset(); - Fp->seekg(o, std::ios::beg); + //size_t o =(size_t)entry->GetOffset(); + Fp->seekg((size_t)entry->GetOffset(), std::ios::beg); // FIXME : for each DataEntry ! size_t l = entry->GetLength(); uint8_t *data = new uint8_t[l]; @@ -769,7 +809,6 @@ void Document::LoadEntryBinArea(DataEntry *entry) { case 1: { - gdcmDebugMacro( "Atomic lgt = 1 ; NO swap at reading"); break; } case 2: @@ -797,28 +836,29 @@ void Document::LoadEntryBinArea(DataEntry *entry) entry->SetBinArea(data); - if ( openFile ) + if ( openFile ) // The file is left in the state (open/close) it was at entrance CloseFile(); } /** * \brief Loads the element while preserving the current * underlying file position indicator as opposed to - * LoadDocEntry that modifies it. + * LoadDocEntry that modifies it + * \note seems to be unused!. * @param entry DocEntry whose value will be loaded. */ -void Document::LoadDocEntrySafe(DocEntry *entry) -{ - if ( Fp ) - { - long PositionOnEntry = Fp->tellg(); - LoadDocEntry(entry); - Fp->seekg(PositionOnEntry, std::ios::beg); - } -} +//void Document::LoadDocEntrySafe(DocEntry *entry) +//{ +// if ( Fp ) +// { +// long PositionOnEntry = Fp->tellg(); // LoadDocEntrySafe is not used +// LoadDocEntry(entry); +// Fp->seekg(PositionOnEntry, std::ios::beg); // LoadDocEntrySafe is not used +// } +//} /** - * \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 @@ -884,6 +924,27 @@ bool Document::operator<(Document &document) //----------------------------------------------------------------------------- // Protected + +/** + * \brief Reads a given length of bytes + * (in order to avoid to many CPU time-consuming fread-s) + * @param l length to read + */ +void Document::ReadBegBuffer(size_t l) + throw( FormatError ) +{ + Fp->read (BegBuffer, (size_t)l); + if ( Fp->fail() ) + { + throw FormatError( "Document::ReadBegBuffer()", " file error." ); + } + if ( Fp->eof() ) + { + throw FormatError( "Document::ReadBegBuffer()", "EOF." ); + } + PtrBegBuffer = BegBuffer; + CurrentOffsetPosition+=l; +} /** * \brief Reads a supposed to be 16 Bits integer * (swaps it depending on processor endianness) @@ -906,6 +967,18 @@ uint16_t Document::ReadInt16() return g; } +/** + * \brief Gets from BegBuffer a supposed to be 16 Bits integer + * (swaps it depending on processor endianness) + * @return read value + */ +uint16_t Document::GetInt16() +{ + uint16_t g = *((uint16_t*)PtrBegBuffer); + g = SwapShort(g); + PtrBegBuffer+=2; + return g; +} /** * \brief Reads a supposed to be 32 Bits integer * (swaps it depending on processor endianness) @@ -929,18 +1002,20 @@ uint32_t Document::ReadInt32() } /** - * \brief skips bytes inside the source file - * \warning NOT end user intended method ! - * @return + * \brief Gets from BegBuffer a supposed to be 32 Bits integer + * (swaps it depending on processor endianness) + * @return read value */ -void Document::SkipBytes(uint32_t nBytes) +uint32_t Document::GetInt32() { - //FIXME don't dump the returned value - Fp->seekg((long)nBytes, std::ios::cur); + uint32_t g = *((uint32_t*)PtrBegBuffer); + g = SwapLong(g); + PtrBegBuffer+=4; + return g; } /** - * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader + * \brief Re-computes the length of the Dicom group 0002. */ int Document::ComputeGroup0002Length( ) { @@ -964,19 +1039,13 @@ int Document::ComputeGroup0002Length( ) { vr = entry->GetVR(); - // FIXME : group 0x0002 is *always* Explicit VR! - // --> Except for Implicit VR Transfer Syntax (GE Private) !! - - //if ( filetype == ExplicitVR ) - //{ //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(); } } @@ -988,11 +1057,38 @@ int Document::ComputeGroup0002Length( ) return groupLength; } +/** + * \brief CallStartMethod + */ +void Document::CallStartMethod() +{ + Progress = 0.0f; + Abort = false; + CommandManager::ExecuteCommand(this,CMD_STARTPROGRESS); +} + +/** + * \brief CallProgressMethod + */ +void Document::CallProgressMethod() +{ + CommandManager::ExecuteCommand(this,CMD_PROGRESS); +} + +/** + * \brief CallEndMethod + */ +void Document::CallEndMethod() +{ + Progress = 1.0f; + CommandManager::ExecuteCommand(this,CMD_ENDPROGRESS); +} + //----------------------------------------------------------------------------- // Private /** * \brief Loads all the needed Dictionaries - * \warning NOT end user intended method ! + * \warning NOT end user intended method ! */ void Document::Initialize() { @@ -1014,40 +1110,53 @@ void Document::ParseDES(DocEntrySet *set, long offset, DocEntry *newDocEntry; DataEntry *newDataEntry; SeqEntry *newSeqEntry; - VRKey vr; + //VRKey vr; bool used; // will be set to false when something wrong happens to an Entry. // (Entry will then be deleted) bool delim_mode_intern = delim_mode; bool first = true; gdcmDebugMacro( "Enter in ParseDES, delim-mode " << delim_mode - << " at offset " << std::hex << offset ); + << " at offset " << std::hex << "0x(" << offset << ")" ); while (true) { - if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) - { - break; - } + + ///\todo FIXME : On 64 bits processors, tellg gives unexpected results after a while ? + /// Probabely a bug in gdcm code somewhere (some memory erased ?) - newDocEntry = ReadNextDocEntry( ); +// 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++ :-( + if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per DocEntry, when no delim mode + { + break; + } - // Uncoment this cerr line to be able to 'follow' the DocEntries - // when something *very* strange happens - if( Debug::GetDebugFlag() ) - std::cerr<GetKey()<<" "<GetVR()<GetKey()<<" "<GetVR()<IsItemStarter() ) - { - // Debug message within the method ! - newDocEntry = Backtrack(newDocEntry); + { + // Debug message within the method ! + newDocEntry = Backtrack(newDocEntry, set); } else { @@ -1061,14 +1170,14 @@ void Document::ParseDES(DocEntrySet *set, long offset, { //////////////////////////// DataEntry - vr = newDocEntry->GetVR(); + //vr = newDocEntry->GetVR(); // useless ? if ( !set->AddEntry( newDataEntry ) ) { gdcmDebugMacro( "in ParseDES : cannot add a DataEntry " - << newDataEntry->GetKey() - << " (at offset : " - << newDataEntry->GetOffset() << " )" ); + << newDataEntry->GetKey() + << " (at offset : 0x(" + << newDataEntry->GetOffset() << ") )" ); used=false; } else @@ -1077,7 +1186,6 @@ void Document::ParseDES(DocEntrySet *set, long offset, // Load only if we can add (not a duplicate key) LoadDocEntry( newDataEntry ); } - if ( newDataEntry->GetElement() == 0x0000 ) // if on group length { if ( newDataEntry->GetGroup()%2 != 0 ) // if Shadow Group @@ -1085,25 +1193,30 @@ void Document::ParseDES(DocEntrySet *set, long offset, if ( LoadMode & LD_NOSHADOW ) // if user asked to skip shad.gr { std::string strLgrGroup = newDataEntry->GetString(); + int lgrGroup; - if ( newDataEntry->IsUnfound() ) + //if ( newDataEntry->IsUnfound() ) /?!? JPR { lgrGroup = atoi(strLgrGroup.c_str()); - Fp->seekg(lgrGroup, std::ios::cur); - //used = false; // never used + Fp->seekg(lgrGroup, std::ios::cur); // Once per Shadow group, when NOSHADOW RemoveEntry( newDocEntry ); // Remove and delete - // bcc 5.5 is right "assigned a value that's never used" - // newDocEntry = 0; continue; } } } } - bool delimitor = newDataEntry->IsItemDelimitor(); + bool delimitor = newDataEntry->IsItemDelimitor(); + bool outOfBounds = false; + if (!delim_mode ) + if ( ((long)(Fp->tellg())-offset) >= l_max ) //Once per DataEntry when no delim mode + outOfBounds = true; - if ( (delimitor) || - (!delim_mode && ((long)(Fp->tellg())-offset) >= l_max) ) + // 'and then', 'or else' don't exist in C++ :-( + // if ( (delimitor) || + // (!delim_mode && ((long)(Fp->tellg())-offset) >= l_max) ) // Once per DataEntry + + if ( delimitor || outOfBounds ) { if ( !used ) newDocEntry->Delete(); @@ -1111,13 +1224,13 @@ void Document::ParseDES(DocEntrySet *set, long offset, } // Just to make sure we are at the beginning of next entry. - SkipToNextDocEntry(newDocEntry); + SkipToNextDocEntry(newDocEntry); // FIXME : once per DocEntry, segfault if commented out } else { /////////////////////// 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 ) @@ -1135,7 +1248,7 @@ void Document::ParseDES(DocEntrySet *set, long offset, // User asked to skip SeQuences *only* if they belong to Shadow Group if ( newDocEntry->GetGroup()%2 != 0 ) { - Fp->seekg( l, std::ios::cur); + Fp->seekg( l, std::ios::cur); // once per SQITEM, when NOSHADOWSEQ newDocEntry->Delete(); // Delete, not in the set continue; } @@ -1143,7 +1256,7 @@ void Document::ParseDES(DocEntrySet *set, long offset, if ( (LoadMode & LD_NOSEQ) && ! delim_mode_intern ) { // User asked to skip *any* SeQuence - Fp->seekg( l, std::ios::cur); + Fp->seekg( l, std::ios::cur); // Once per SQ, when NOSEQ newDocEntry->Delete(); // Delete, not in the set continue; } @@ -1175,30 +1288,33 @@ void Document::ParseDES(DocEntrySet *set, long offset, { // Don't try to parse zero-length sequences gdcmDebugMacro( "Entry in ParseSQ, delim " << delim_mode_intern - << " at offset " << std::hex - << newDocEntry->GetOffset() ); + << " 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 ) ) { gdcmWarningMacro( "in ParseDES : cannot add a SeqEntry " << newSeqEntry->GetKey() - << " (at offset : " - << newSeqEntry->GetOffset() << " )" ); + << " (at offset : 0x(" + << newSeqEntry->GetOffset() << ") )" ); used = false; } else { newDocEntry->Delete(); } + + // if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) // Once per SeqEntry - if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) + if ( !delim_mode ) // 'and then' doesn't exist in C++ :-( + if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per SeqEntry when no delim mode + { if ( !used ) newDocEntry->Delete(); @@ -1211,15 +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; @@ -1232,9 +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 ) { @@ -1245,10 +1368,13 @@ void Document::ParseSQ( SeqEntry *seqEntry, break; } } - if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) + else // ! delim_mode { - newDocEntry->Delete(); - break; + if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per SQItem when no delim mode + { + newDocEntry->Delete(); + break; + } } // create the current SQItem SQItem *itemSQ = SQItem::New( seqEntry->GetDepthLevel() ); @@ -1262,33 +1388,48 @@ 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; + } - // remove fff0,e000, created out of the SQItem - Fp->seekg(offsetStartCurrentSQItem, std::ios::beg); // 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(); + offsetStartCurrentSQItem = Fp->tellg(); // Once per SQItem seqEntry->AddSQItem( itemSQ, SQItemNumber ); itemSQ->Delete(); newDocEntry->Delete(); SQItemNumber++; - if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max ) + //if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max ) //JPRx + if ( !delim_mode && (offsetStartCurrentSQItem-offset ) >= l_max ) { 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 @@ -1301,9 +1442,10 @@ DocEntry *Document::Backtrack(DocEntry *docEntry) long offset = PreviousDocEntry->GetOffset(); gdcmDebugMacro( "Backtrack :" << std::hex << group - << "|" << elem - << " at offset " << offset ); - RemoveEntry( PreviousDocEntry ); + << "|" << elem + << " at offset 0x(" <RemoveEntry( PreviousDocEntry ); // forge the Seq Entry DocEntry *newEntry = NewSeqEntry(group, elem); @@ -1311,10 +1453,9 @@ DocEntry *Document::Backtrack(DocEntry *docEntry) newEntry->SetOffset(offset); // Move back to the beginning of the Sequence - Fp->seekg( 0, std::ios::beg); - Fp->seekg(offset, std::ios::cur); - return newEntry; + Fp->seekg(offset, std::ios::beg); // Only for Shadow Implicit VR SQ + return newEntry; // It will added where it has to be! } /** @@ -1325,18 +1466,17 @@ DocEntry *Document::Backtrack(DocEntry *docEntry) */ void Document::LoadDocEntry(DocEntry *entry, bool forceLoad) { - uint16_t group = entry->GetGroup(); - uint16_t elem = entry->GetElement(); + uint16_t group = entry->GetGroup(); + uint16_t elem = entry->GetElement(); const VRKey &vr = entry->GetVR(); - uint32_t length = entry->GetLength(); + uint32_t length = entry->GetLength(); - Fp->seekg((long)entry->GetOffset(), std::ios::beg); + // Fp->seekg((long)entry->GetOffset(), std::ios::beg); // JPRx // A SeQuence "contains" a set of Elements. // (fffe e000) tells us an Element is beginning // (fffe e00d) tells us an Element just ended // (fffe e0dd) tells us the current SeQuence just ended - // // (fffe 0000) is an 'impossible' tag value, // found in MR-PHILIPS-16-Multi-Seq.dcm @@ -1362,8 +1502,6 @@ void Document::LoadDocEntry(DocEntry *entry, bool forceLoad) // The elements whose length is bigger than the specified upper bound // are not loaded. - std::ostringstream s; - if (!forceLoad) { if (length > MaxSizeLoadEntry) @@ -1371,13 +1509,20 @@ void Document::LoadDocEntry(DocEntry *entry, bool forceLoad) dataEntryPtr->SetBinArea(NULL,true); dataEntryPtr->SetState(DataEntry::STATE_NOTLOADED); - // to be sure we are at the end of the value ... - Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(), - std::ios::beg); + // to be sure we are at the end of the value ... + // Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(), + // std::ios::beg); //JPRx return; } } - + + /// \todo: a method that *doesn't* load anything (maybe with MaxSizeLoadEntry=0 ?) + /// + a ForceLoad call on the +/- 20 'usefull' fields + /// Allow user to tell the fields he wants to ForceLoad + /// during initial stage. + /// Later, a GetString or GetBinArea will load the value from disk, if not loaded + /// + a method that load *everything* that's not yet loaded + LoadEntryBinArea(dataEntryPtr); // last one, not to erase length ! } @@ -1390,24 +1535,35 @@ 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); - uint32_t length32 = ReadInt32(); - if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff ) + //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" || vr == "OL") && length32 == 0xffffffff ) { uint32_t lengthOB; try { - lengthOB = FindDocEntryLengthOBOrOW(); + lengthOB = FindDocEntryLengthOBOrOW();// for encapsulation of encoded pixel } catch ( FormatUnexpected ) { @@ -1419,11 +1575,11 @@ void Document::FindDocEntryLength( DocEntry *entry ) gdcmWarningMacro( " Computing the length failed for " << entry->GetKey() <<" in " <tellg(); - Fp->seekg(0L,std::ios::end); + long currentPosition = Fp->tellg(); // Only for gdcm-JPEG-LossLess3a.dcm-like + Fp->seekg(0L,std::ios::end); // Only for gdcm-JPEG-LossLess3a.dcm-like - long lengthUntilEOF = (long)(Fp->tellg())-currentPosition; - Fp->seekg(currentPosition, std::ios::beg); + long lengthUntilEOF = (long)(Fp->tellg())-currentPosition; // Only for gdcm-JPEG-LossLess3a.dcm-like + Fp->seekg(currentPosition, std::ios::beg); // Only for gdcm-JPEG-LossLess3a.dcm-like entry->SetReadLength(lengthUntilEOF); entry->SetLength(lengthUntilEOF); @@ -1436,10 +1592,9 @@ void Document::FindDocEntryLength( DocEntry *entry ) FixDocEntryFoundLength(entry, length32); return; } - // Length is encoded on 2 bytes. - length16 = ReadInt16(); - + //length16 = ReadInt16(); + length16 = GetInt16(); // 0xffff means that we deal with 'No Length' Sequence // or 'No Length' SQItem if ( length16 == 0xffff) @@ -1461,8 +1616,9 @@ void Document::FindDocEntryLength( DocEntry *entry ) // Well ... group 0002 is always coded in 'Explicit VR Litle Endian' // even if Transfer Syntax is 'Implicit VR ...' // --> Except for 'Implicit VR Big Endian Transfer Syntax GE Private' - - FixDocEntryFoundLength( entry, ReadInt32() ); + // where Group 0x0002 is *also* encoded in Implicit VR ! + + FixDocEntryFoundLength( entry, GetInt32() /*ReadInt32()*/ ); return; } } @@ -1476,7 +1632,9 @@ uint32_t Document::FindDocEntryLengthOBOrOW() throw( FormatUnexpected ) { // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data. - long positionOnEntry = Fp->tellg(); + + long positionOnEntry = Fp->tellg(); // Only for OB,OW DataElements + bool foundSequenceDelimiter = false; uint32_t totalLength = 0; @@ -1484,27 +1642,32 @@ uint32_t Document::FindDocEntryLengthOBOrOW() { uint16_t group; uint16_t elem; + try - { - group = ReadInt16(); - elem = ReadInt16(); + { ///\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 } catch ( FormatError ) { throw FormatError("Unexpected end of file encountered during ", "Document::FindDocEntryLengthOBOrOW()"); } + group = GetInt16(); + elem = GetInt16(); + // We have to decount the group and element we just read totalLength += 4; if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) ) { - long filePosition = Fp->tellg(); gdcmWarningMacro( "Neither an Item tag nor a Sequence delimiter tag on :" - << std::hex << group << " , " << elem - << ") -before- position x(" << filePosition << ")" ); + << std::hex << group << "|" << elem << ") Pos. on entry was 0x(" <seekg(positionOnEntry, std::ios::beg); + Fp->seekg(positionOnEntry, std::ios::beg); // Once per fragment (if any) of OB,OW DataElements throw FormatUnexpected( "Neither an Item tag nor a Sequence delimiter tag."); } @@ -1512,7 +1675,7 @@ uint32_t Document::FindDocEntryLengthOBOrOW() { foundSequenceDelimiter = true; } - uint32_t itemLength = ReadInt32(); + uint32_t itemLength = ReadInt32(); // Once per fragment (if any) of OB,OW DataElements // We add 4 bytes since we just read the ItemLength with ReadInt32 totalLength += itemLength + 4; SkipBytes(itemLength); @@ -1522,7 +1685,7 @@ uint32_t Document::FindDocEntryLengthOBOrOW() break; } } - Fp->seekg( positionOnEntry, std::ios::beg); + Fp->seekg( positionOnEntry, std::ios::beg); // Only once for OB,OW DataElements return totalLength; } @@ -1533,26 +1696,46 @@ uint32_t Document::FindDocEntryLengthOBOrOW() VRKey Document::FindDocEntryVR() { if ( Filetype != ExplicitVR ) + { return GDCM_VRUNKNOWN; + } - long positionOnEntry = Fp->tellg(); + // Delimiters (0xfffe), are not explicit VR ... + if ( CurrentGroup == 0xfffe ) + return GDCM_VRUNKNOWN; + + //long positionOnEntry; + //if( Debug::GetWarningFlag() ) + // positionOnEntry = Fp->tellg(); // Only in Warning Mode + // Warning: we believe this is explicit VR (Value Representation) because // we used a heuristic that found "UL" in the first tag and/or // 'Transfer Syntax' told us it is. // Alas this doesn't guarantee that all the tags will be in explicit VR. - // In some cases one finds implicit VR tags mixed within an explicit VR file. + // In some cases one finds implicit VR tags mixed within an explicit VR file + // Well... + // 'Normaly' the only case is : group 0002 Explicit, and other groups Implicit + // // Hence we make sure the present tag is in explicit VR and try to fix things // if it happens not to be the case. VRKey vr; - Fp->read(&(vr[0]),(size_t)2); + //Fp->read(&(vr[0]),(size_t)2); + vr[0] = *PtrBegBuffer++; + vr[1] = *PtrBegBuffer++; + + //if ( !CheckDocEntryVR(vr) ) // avoid useless function call + if ( !Global::GetVR()->IsValidVR(vr) ) + { - if ( !CheckDocEntryVR(vr) ) - { - gdcmWarningMacro( "Unknown VR " << std::hex << "0x(" - << (unsigned int)vr[0] << "|" << (unsigned int)vr[1] - << ") at offset :" << positionOnEntry ); - Fp->seekg(positionOnEntry, std::ios::beg); + 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 + PtrBegBuffer-=2; return GDCM_VRUNKNOWN; } return vr; @@ -1573,7 +1756,6 @@ bool Document::CheckDocEntryVR(const VRKey &vr) /** * \brief Skip a given Header Entry - * \warning NOT end user intended method ! * @param entry entry to skip */ void Document::SkipDocEntry(DocEntry *entry) @@ -1583,19 +1765,18 @@ void Document::SkipDocEntry(DocEntry *entry) /** * \brief Skips to the beginning of the next Header Entry - * \warning NOT end user intended method ! * @param currentDocEntry entry to skip */ void Document::SkipToNextDocEntry(DocEntry *currentDocEntry) { - int l = currentDocEntry->GetReadLength(); - if ( l == -1 ) // length = 0xffff shouldn't appear here ... + long l = currentDocEntry->GetReadLength(); + if ( (uint32_t) l == (uint32_t)-1 ) // length = 0xffff shouldn't appear here ... // ... but PMS imagers happen ! return; - Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg); + Fp->seekg((size_t)(currentDocEntry->GetOffset()), std::ios::beg); //FIXME :each DocEntry if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb { - Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur); + Fp->seekg( l,std::ios::cur); //FIXME :each DocEntry } } @@ -1604,26 +1785,32 @@ void Document::SkipToNextDocEntry(DocEntry *currentDocEntry) * the parser went Jabberwocky) one can hope improving things by * applying some heuristics. * @param entry entry to check - * @param foundLength first assumption about length + * @param foundLength first assumption about length (before bug fix, or set to zero if =0xffffffff) */ void Document::FixDocEntryFoundLength(DocEntry *entry, uint32_t foundLength) { entry->SetReadLength( foundLength );// will be updated only if a bug is found + if ( foundLength == 0xffffffff) { - foundLength = 0; + //foundLength = 0; + //entry->SetLength(foundLength); + entry->SetLength(0); + return; // return ASAP; don't waist time on useless tests } - + uint16_t gr = entry->GetGroup(); uint16_t elem = entry->GetElement(); 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 @@ -1644,26 +1831,37 @@ void Document::FixDocEntryFoundLength(DocEntry *entry, // Occurence of such images is quite low (unless one leaves close to a // 'Leonardo' source. Hence, one might consider commenting out the // following fix on efficiency reasons. - else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) ) - { - foundLength = 4; - entry->SetReadLength(4); // a bug is to be fixed ! - } - + else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) ) + { + // Ideally we should check we are in Explicit and double check + // that VR=UL... this is done properly in gdcm2 + if( foundLength == 6 ) + { + gdcmWarningMacro( "Replacing Length from 6 into 4" ); + foundLength = 4; + entry->SetReadLength(4); // a bug is to be fixed ! + } + else if ( foundLength%4 ) + { + gdcmErrorMacro( "This looks like to a buggy Siemens DICOM file." + "The length of this tag seems to be wrong" ); + } + } + else if ( entry->GetVR() == "SQ" ) { - foundLength = 0; // ReadLength is unchanged - } - - //////// We encountered a 'delimiter' element i.e. a tag of the form + foundLength = 0; // ReadLength is unchanged + } + + //////// We encountered a 'delimiter' element i.e. a tag of the form // "fffe|xxxx" which is just a marker. Delimiters length should not be // taken into account. else if ( gr == 0xfffe ) - { + { // According to the norm, fffe|0000 shouldn't exist. BUT the Philips // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to // causes extra troubles... - if ( entry->GetElement() != 0x0000 ) + if ( elem != 0x0000 ) { foundLength = 0; } @@ -1671,7 +1869,7 @@ void Document::FixDocEntryFoundLength(DocEntry *entry, { foundLength=12; // to skip the mess that follows this bugged Tag ! } - } + } entry->SetLength(foundLength); } @@ -1690,7 +1888,7 @@ bool Document::IsDocEntryAnInteger(DocEntry *entry) // When we have some semantics on the element we just read, and if we // a priori know we are dealing with an integer, then we shall be - // able to swap it's element value properly. + // able to swap its element value properly. if ( elem == 0 ) // This is the group length of the group { if ( length == 4 ) @@ -1706,7 +1904,8 @@ bool Document::IsDocEntryAnInteger(DocEntry *entry) // test is useless (and might even look a bit paranoid), when we // encounter such an ill-formed image, we simply display a warning // message and proceed on parsing (while crossing fingers). - long filePosition = Fp->tellg(); + long filePosition = Fp->tellg(); // Only when elem 0x0000 length is not 4 (?!?) + (void)filePosition; gdcmWarningMacro( "Erroneous Group Length element length on : (" << std::hex << group << " , " << elem << ") -before- position x(" << filePosition << ")" @@ -1765,7 +1964,7 @@ bool Document::CheckSwap() // i.e. a total of 136 bytes. entCur = deb + 136; - // group 0x0002 *is always* Explicit VR Sometimes , + // group 0x0002 *is always* Explicit VR Sometimes, // even if elem 0002,0010 (Transfer Syntax) tells us the file is // *Implicit* VR (see former 'gdcmData/icone.dcm') @@ -1788,6 +1987,8 @@ bool Document::CheckSwap() << "Looks like a bugged Header!"); } + // Here, we assume that the file IS kosher Dicom ! + // (The meta elements - group 0x0002 - ARE little endian !) if ( net2host ) { SwapCode = 4321; @@ -1802,9 +2003,8 @@ bool Document::CheckSwap() // Position the file position indicator at first tag // (i.e. after the file preamble and the "DICM" string). - Fp->seekg(0, std::ios::beg); // FIXME : Is it usefull? - - Fp->seekg ( 132L, std::ios::beg); + Fp->seekg ( 132L, std::ios::beg); // Once per Document + CurrentOffsetPosition = 132; return true; } // ------------------------------- End of DicomV3 ---------------- @@ -1814,8 +2014,8 @@ bool Document::CheckSwap() gdcmWarningMacro( "Not a Kosher DICOM Version3 file (no preamble)"); - Fp->seekg(0, std::ios::beg); - + Fp->seekg(0, std::ios::beg); // Once per ACR-NEMA Document + CurrentOffsetPosition = 0; // Let's check 'No Preamble Dicom File' : // Should start with group 0x0002 // and be Explicit Value Representation @@ -1842,7 +2042,8 @@ bool Document::CheckSwap() memcmp(entCur, "AE", (size_t)2) == 0 || memcmp(entCur, "OB", (size_t)2) == 0 ) { - Filetype = ExplicitVR; + Filetype = ExplicitVR; // FIXME : not enough to say it's Explicit + // Wait untill reading Transfer Syntax gdcmDebugMacro( "Group 0002 : Explicit Value Representation"); return true; } @@ -1894,7 +2095,9 @@ bool Document::CheckSwap() // Find a trick to tell it the caller... s16 = *((uint16_t *)(deb)); - + + gdcmDebugMacro("not a DicomV3 nor a 'clean' ACR/NEMA;" + << " (->despaired wild guesses !)"); switch ( s16 ) { case 0x0001 : @@ -1905,6 +2108,7 @@ bool Document::CheckSwap() case 0x0006 : case 0x0007 : case 0x0008 : + case 0x0028 : SwapCode = 1234; Filetype = ACR; return true; @@ -1916,13 +2120,53 @@ bool Document::CheckSwap() case 0x0600 : case 0x0700 : case 0x0800 : + case 0x2800 : SwapCode = 4321; 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;itellg() ); + << " at: 0x" << std::hex << Fp->tellg() ); // Only when DEBUG if ( SwapCode == 1234 ) { SwapCode = 4321; @@ -1957,7 +2201,7 @@ void Document::SwitchByteSwapCode() * \brief during parsing, Header Elements too long are not loaded in memory * @param newSize new size */ -void Document::SetMaxSizeLoadEntry(long newSize) +void Document::SetMaxSizeLoadEntry(long newSize) { if ( newSize < 0 ) { @@ -1976,66 +2220,126 @@ void Document::SetMaxSizeLoadEntry(long newSize) * (read the 'Group Number', the 'Element Number', * gets the Dict Entry * gets the VR, gets the length, gets the offset value) - * @return On succes : the newly created DocEntry, NULL on failure. + * @return On succes : the newly created DocEntry, NULL on failure. */ DocEntry *Document::ReadNextDocEntry() { - uint16_t group; - uint16_t elem; - try { - group = ReadInt16(); - elem = ReadInt16(); + ReadBegBuffer(8); // Avoid to many time consuming freads + //CurrentGroup = ReadInt16(); + //CurrentElem = ReadInt16(); } catch ( FormatError ) { - // We reached the EOF (or an error occured) therefore + // We reached the EOF (or an error occured) therefore // header parsing has to be considered as finished. return 0; } - - // Sometimes file contains groups of tags with reversed endianess. - HandleBrokenEndian(group, elem); - + + changeFromUN = false; + CurrentGroup = GetInt16(); + CurrentElem = GetInt16(); + // In 'true DICOM' files Group 0002 is always little endian if ( HasDCMPreamble ) - HandleOutOfGroup0002(group, elem); - + { + if ( !Group0002Parsed && CurrentGroup != 0x0002) // avoid calling a function when useless + HandleOutOfGroup0002(CurrentGroup, CurrentElem); + else + // Sometimes file contains groups of tags with reversed endianess. + HandleBrokenEndian(CurrentGroup, CurrentElem); + } + VRKey vr = FindDocEntryVR(); - VRKey realVR = vr; if ( vr == GDCM_VRUNKNOWN ) { - if ( elem == 0x0000 ) // Group Length + if ( CurrentElem == 0x0000 ) // Group Length { realVR = "UL"; // must be UL } - else if (group%2 == 1 && (elem >= 0x0010 && elem <=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"; + else if (CurrentGroup == 0xfffe) // Don't get DictEntry for Delimitors + { + realVR = "UL"; + } + + // Was commented out in order not to generate 'Shadow Groups' where some + // Data Elements are Explicit VR and some other ones Implicit VR + // -> 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(group,elem); + DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem);//only when ImplicitVR if ( dictEntry ) { realVR = dictEntry->GetVR(); - dictEntry->Unregister(); + dictEntry->Unregister(); // GetDictEntry registered it } } } - // gdcmDebugMacro( "Found VR: " << vr << " / Real VR: " << realVR ); + + // 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) ) - newEntry = NewSeqEntry(group, elem); - else + //if ( Global::GetVR()->IsVROfSequence(realVR) ) + if (realVR == "SQ") + { + newEntry = NewSeqEntry(CurrentGroup, CurrentElem); + } + else { - newEntry = NewDataEntry(group, elem, realVR); + newEntry = NewDataEntry(CurrentGroup, CurrentElem, realVR); static_cast(newEntry)->SetState(DataEntry::STATE_NOTLOADED); } @@ -2045,14 +2349,14 @@ DocEntry *Document::ReadNextDocEntry() { // We thought this was explicit VR, but we end up with an // implicit VR tag. Let's backtrack. - if ( newEntry->GetGroup() != 0xfffe ) + + //if ( newEntry->GetGroup() != 0xfffe ) + if (CurrentGroup != 0xfffe) { - std::string msg; - int offset = Fp->tellg(); - msg = Util::Format( - "Entry (%04x,%04x) at x(%x) should be Explicit VR\n", - newEntry->GetGroup(), newEntry->GetElement(), offset ); - gdcmWarningMacro( msg.c_str() ); + int offset = Fp->tellg();//Only when heuristic for Explicit/Implicit was wrong + + gdcmWarningMacro("Entry (" << newEntry->GetKey() << ") at x(" + << std::hex << offset << ") should be Explicit VR"); } } newEntry->SetImplicitVR(); @@ -2069,8 +2373,7 @@ DocEntry *Document::ReadNextDocEntry() return 0; } - newEntry->SetOffset(Fp->tellg()); - + newEntry->SetOffset(Fp->tellg()); // for each DocEntry return newEntry; } @@ -2082,6 +2385,23 @@ DocEntry *Document::ReadNextDocEntry() */ void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem) { + // for strange PMS Gyroscan Intera images + // Item 'starter' has a tag : 0x3f3f,0x3f00, for no apparent reason + + // --- Feel free to remove this test *on your own coy of gdcm* + // if you are sure you'll never face this problem. + + if ((group == 0x3f3f) && (elem == 0x3f00)) + { + // start endian swap mark for group found + gdcmDebugMacro( " delimiter 0x3f3f found." ); + // fix the tag + group = 0xfffe; + elem = 0xe000; + return; + } + // --- End of removable code + // Endian reversion. // Some files contain groups of tags with reversed endianess. static int reversedEndian = 0; @@ -2089,6 +2409,7 @@ void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem) if ((group == 0xfeff) && (elem == 0x00e0)) { // start endian swap mark for group found + gdcmDebugMacro( "Start endian swap mark found." ); reversedEndian++; SwitchByteSwapCode(); // fix the tag @@ -2098,6 +2419,7 @@ void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem) else if (group == 0xfffe && elem == 0xe00d && reversedEndian) { // end of reversed endian group + gdcmDebugMacro( "End of reversed endian." ); reversedEndian--; SwitchByteSwapCode(); } @@ -2128,43 +2450,38 @@ void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem) { // Endian reversion. // Some files contain groups of tags with reversed endianess. - if ( !Group0002Parsed && group != 0x0002) - { + Group0002Parsed = true; // we just came out of group 0002 // if Transfer Syntax is Big Endian we have to change CheckSwap std::string ts = GetTransferSyntax(); - if ( ts == GDCM_UNKNOWN ) - { - gdcmDebugMacro("True DICOM File, with NO Transfer Syntax (?!) " ); - return; - } - if ( !Global::GetTS()->IsTransferSyntax(ts) ) - { - gdcmWarningMacro("True DICOM File, with illegal Transfer Syntax: [" - << ts << "]"); - return; - } + TS::SpecialType s = Global::GetTS()->GetSpecialTransferSyntax(ts); // Group 0002 is always 'Explicit ...' // even when Transfer Syntax says 'Implicit ..." - if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == - TS::ImplicitVRLittleEndian ) + if ( s == TS::ImplicitVRLittleEndian + || + s == TS::ImplicitVRBigEndianPrivateGE + ) { Filetype = ImplicitVR; } // FIXME Strangely, this works with - //'Implicit VR BigEndian Transfer Syntax (GE Private) + //'Implicit VR BigEndian Transfer Syntax' (GE Private) // // --> Probabely normal, since we considered we never have // to trust manufacturers. - // (we find very often 'Implicit VR' tag, + // (we often find 'Implicit VR' tag, // even when Transfer Syntax tells us it's Explicit ... - if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == - TS::ExplicitVRBigEndian ) + + // NEVER trust the meta elements! + // (see what ezDICOM does ...) + + /* + if ( s == TS::ExplicitVRBigEndian ) { gdcmDebugMacro("Transfer Syntax Name = [" << GetTransferSyntaxName() << "]" ); @@ -2172,7 +2489,45 @@ void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem) group = SwapShort(group); elem = SwapShort(elem); } - } + */ + //-- Broken ACR may start with a Shadow Group -- + // worse : some ACR-NEMA like files start 00028 group ?!? + if ( !( (group >= 0x0001 && group <= 0x0008) || group == 0x0028 ) ) + { + // We trust what we see. + SwitchByteSwapCode(); + group = SwapShort(group); + elem = SwapShort(elem); + // not what we where told (by meta elements) ! + gdcmDebugMacro("Transfer Syntax Name = [" + << GetTransferSyntaxName() << "]" ); + } + + /// \todo find a trick to warn user and stop processing + + if ( s == TS::DeflatedExplicitVRLittleEndian) + { + gdcmWarningMacro("Transfer Syntax [" + << GetTransferSyntaxName() << "] :" + << " not yet dealt with "); + return; + } + + // The following shouldn't occur very often + // Let's check at the very end. + + if ( ts == GDCM_UNKNOWN ) + { + gdcmDebugMacro("True DICOM File, with NO Transfer Syntax (?!) " ); + return; + } + + if ( !Global::GetTS()->IsTransferSyntax(ts) ) + { + gdcmWarningMacro("True DICOM File, with illegal Transfer Syntax: [" + << ts << "]"); + return; + } } //-----------------------------------------------------------------------------