X-Git-Url: https://git.creatis.insa-lyon.fr/pubgit/?a=blobdiff_plain;f=src%2FgdcmDocument.cxx;h=11b88a2e6fefc1d8b1aebba0dd568c1c0cdeda05;hb=6278320cc85da00d2d56ffbf07806e84966892c3;hp=c784c14d8b4d0d287cb3093c74e00f02a85060cc;hpb=4c54e5135106a491c069b82cbcd0e41ad0837a89;p=gdcm.git diff --git a/src/gdcmDocument.cxx b/src/gdcmDocument.cxx index c784c14d..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: 2005/10/11 14:38:05 $ - Version: $Revision: 1.287 $ + 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 @@ -17,8 +17,6 @@ =========================================================================*/ #include "gdcmDocument.h" -#include "gdcmValEntry.h" -#include "gdcmBinEntry.h" #include "gdcmSeqEntry.h" #include "gdcmGlobal.h" #include "gdcmUtil.h" @@ -27,6 +25,7 @@ #include "gdcmDictSet.h" #include "gdcmDocEntrySet.h" #include "gdcmSQItem.h" +#include "gdcmDataEntry.h" #include #include @@ -34,7 +33,11 @@ #include // for isdigit #include // for atoi -namespace gdcm +#if defined(__BORLANDC__) + #include // for memset +#endif + +namespace GDCM_NAME_SPACE { //----------------------------------------------------------------------------- @@ -47,11 +50,11 @@ 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() - :ElementSet(-1) + :ElementSet() { Fp = 0; @@ -59,37 +62,19 @@ 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; } -/** - * \brief Constructor (DEPRECATED : not to break the API) - * @param fileName 'Document' (File or DicomDir) to be open for parsing - */ -Document::Document( std::string const &fileName ) - :ElementSet(-1) -{ - Fp = 0; - - SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE); - Initialize(); - SwapCode = 1234; - Filetype = ExplicitVR; - Group0002Parsed = false; - LoadMode = LD_ALL; // Load everything, later - - // Load will set it to true if sucessfull - IsDocumentAlreadyLoaded = false; - IsDocumentModified = true; - - SetFileName(fileName); - Load( ); -} /** * \brief Canonical destructor. */ @@ -100,7 +85,6 @@ Document::~Document () //----------------------------------------------------------------------------- // Public - /** * \brief Loader. use SetLoadMode(), SetFileName() before ! * @return false if file cannot be open or no swap info was found, @@ -114,18 +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) @@ -137,41 +127,24 @@ bool Document::DoTheLoadingDocumentJob( ) if ( ! IsDocumentModified ) // Nothing to do ! return true; - // if ( Filename == fileName ) - // { - // gdcmWarningMacro( "The file was already parsed inside this " - // << "gdcm::Document (its name is: " - // << Filename.c_str() ); - // return true; - // } - - //gdcmWarningMacro( "A file was already parsed inside this " - // << "gdcm::Document (previous name was: " - // << Filename.c_str() << ". New name is :" - // << fileName ); - // clean out the Entries, if already parsed - // (probabely a mistake from the user) - ClearEntry(); Fp = 0; if ( !OpenFile() ) { // warning already performed in OpenFile() - //gdcmWarningMacro( "Unable to open as an ACR/DICOM file: " - // << Filename.c_str() ); Filetype = Unknown; return false; } Group0002Parsed = false; - gdcmWarningMacro( "Starting parsing of file: " << Filename.c_str()); + 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) @@ -189,22 +162,23 @@ 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() ) { - gdcmWarningMacro( "No tag in internal hash table for: " + gdcmErrorMacro( "No tag in internal hash table for: " << Filename.c_str()); CloseFile(); return false; } IsDocumentAlreadyLoaded = true; - Fp->seekg( 0, std::ios::beg); + //Fp->seekg(0, std::ios::beg); // Once per Document! // Load 'non string' values - std::string PhotometricInterpretation = GetEntryValue(0x0028,0x0004); + std::string PhotometricInterpretation = GetEntryString(0x0028,0x0004); if ( PhotometricInterpretation == "PALETTE COLOR " ) { // FIXME @@ -213,6 +187,13 @@ bool Document::DoTheLoadingDocumentJob( ) LoadEntryBinArea(0x0028,0x1200); // gray LUT /// FIXME + /// --> FIXME : The difference between BinEntry and DataEntry + /// --> no longer exists, but the alteration of Dicom Dictionary remains. + /// --> Old comment restored on purpose. + /// --> New one (replacing both BinEntry and ValEntry by DataEntry) + /// --> had absolutely no meaning. + /// --> The whole comment will be removed when the stuff is cleaned ! + /// --> /// The tags refered by the three following lines used to be CORRECTLY /// defined as having an US Value Representation in the public /// dictionary. BUT the semantics implied by the three following @@ -221,12 +202,18 @@ bool Document::DoTheLoadingDocumentJob( ) /// In order to fix things "Quick and Dirty" the dictionary was /// altered on PURPOSE but now contains a WRONG value. /// In order to fix things and restore the dictionary to its - /// correct value, one needs to decided of the semantics by deciding + /// correct value, one needs to decide of the semantics by deciding /// whether the following tags are either : /// - multivaluated US, and hence loaded as ValEntry, but afterwards /// also used as BinEntry, which requires the proper conversion, /// - OW, and hence loaded as BinEntry, but afterwards also used /// as ValEntry, which requires the proper conversion. + + // --> OB (byte aray) or OW (short int aray) + // The actual VR has to be deduced from other entries. + // Our way of loading them may fail in some cases : + // We must or not SwapByte depending on other field values. + LoadEntryBinArea(0x0028,0x1201); // R LUT LoadEntryBinArea(0x0028,0x1202); // G LUT LoadEntryBinArea(0x0028,0x1203); // B LUT @@ -240,18 +227,20 @@ 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 ) { - BinEntry *b = sqi->GetBinEntry(0x0028,0x3006); - if ( b != 0 ) + DataEntry *dataEntry = sqi->GetDataEntry(0x0028,0x3006); // LUT Data + if ( dataEntry != 0 ) { - if ( b->GetLength() != 0 ) + if ( dataEntry->GetLength() != 0 ) { - LoadEntryBinArea(b); //LUT Data (CTX dependent) + // FIXME : CTX dependent means : contexted dependant. + // see upper comment. + LoadEntryBinArea(dataEntry); //LUT Data (CTX dependent) } } } @@ -259,47 +248,28 @@ 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) { - gdcmWarningMacro( "Force Load " << std::hex + gdcmDebugMacro( "Force Load " << std::hex << (*it).Group << "|" <<(*it).Elem ); d = GetDocEntry( (*it).Group, (*it).Elem); 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; } - if ( dynamic_cast(d) ) - { - LoadDocEntry(d, true); - continue; - } - - BinEntry *b = dynamic_cast(d); - if ( b ) - { - LoadEntryBinArea(b); - b->SetValue(GDCM_BINLOADED); - continue; - } - - if ( dynamic_cast(d) ) - { - gdcmWarningMacro( "You cannot 'ForceLoad' a SeqEntry :" << std::hex - << (*it).Group <<"|"<< (*it).Elem ); - continue; - } + LoadDocEntry(d, true); } - CloseFile(); + CloseFile(); // ---------------------------- // Specific code to allow gdcm to read ACR-LibIDO formated images @@ -312,16 +282,19 @@ bool Document::DoTheLoadingDocumentJob( ) // we switch lineNumber and columnNumber // std::string RecCode; - RecCode = GetEntryValue(0x0008, 0x0010); // recognition code (RET) - if (RecCode == "ACRNEMA_LIBIDO_1.1" || - RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares - // with "little-endian strings" + RecCode = GetEntryString(0x0008, 0x0010); // recognition code (RET) + + + 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 = GetEntryValue(0x0028, 0x0010); - std::string columns = GetEntryValue(0x0028, 0x0011); - SetValEntry(columns, 0x0028, 0x0010); - SetValEntry(rows , 0x0028, 0x0011); + std::string rows = GetEntryString(0x0028, 0x0010); + std::string columns = GetEntryString(0x0028, 0x0011); + SetEntryString(columns, 0x0028, 0x0010); + SetEntryString(rows , 0x0028, 0x0011); } // --- End of ACR-LibIDO kludge --- return true; @@ -335,7 +308,7 @@ bool Document::DoTheLoadingDocumentJob( ) */ void Document::AddForceLoadElement (uint16_t group, uint16_t elem) { - Element el; + DicomElement el; el.Group = group; el.Elem = elem; UserForceLoadList.push_back(el); @@ -383,11 +356,11 @@ 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 ) { - gdcmWarningMacro( "Wrong filetype"); + gdcmWarningMacro( "Wrong filetype for " << GetFileName()); return false; } @@ -399,6 +372,17 @@ bool Document::IsReadable() 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. @@ -444,7 +428,8 @@ FileType Document::GetFileType() * \brief Accessor to the Transfer Syntax (when present) of the * current document (it internally handles reading the * value from disk when only parsing occured). - * @return The encountered Transfer Syntax of the current document. + * @return The encountered Transfer Syntax of the current document, if DICOM. + * GDCM_UNKNOWN for ACR-NEMA files (or broken headers ...) */ std::string Document::GetTransferSyntax() { @@ -456,21 +441,31 @@ 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); - if (ValEntry *valEntry = dynamic_cast< ValEntry* >(entry) ) + + // Well ... + // (parsing and loading happen at the very same stage!) + //LoadDocEntrySafe(entry); //JPRx + if (DataEntry *dataEntry = dynamic_cast(entry) ) { - std::string transfer = valEntry->GetValue(); + std::string transfer = dataEntry->GetString(); // The actual transfer (as read from disk) might be padded. We // first need to remove the potential padding. We can make the // weak assumption that padding was not executed with digits... if ( transfer.length() == 0 ) { // for brain damaged headers + gdcmWarningMacro( "Transfer Syntax has length = 0."); return GDCM_UNKNOWN; } while ( !isdigit((unsigned char)transfer[transfer.length()-1]) ) { transfer.erase(transfer.length()-1, 1); + if ( transfer.length() == 0 ) + { + // for brain damaged headers + gdcmWarningMacro( "Transfer Syntax contains no valid character."); + return GDCM_UNKNOWN; + } } return transfer; } @@ -484,7 +479,7 @@ std::string Document::GetTransferSyntax() std::string Document::GetTransferSyntaxName() { // use the TS (TS : Transfer Syntax) - std::string transferSyntax = GetEntryValue(0x0002,0x0010); + std::string transferSyntax = GetEntryString(0x0002,0x0010); if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) ) { @@ -494,7 +489,7 @@ std::string Document::GetTransferSyntaxName() } if ( transferSyntax == GDCM_UNFOUND ) { - gdcmWarningMacro( "Unfound Transfer Syntax (0002,0010)"); + gdcmDebugMacro( "Unfound Transfer Syntax (0002,0010)"); return "Uncompressed ACR-NEMA"; } @@ -547,18 +542,55 @@ uint32_t Document::SwapLong(uint32_t a) a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) ); break; default : - gdcmErrorMacro( "Unset swap code:" << SwapCode ); + gdcmErrorMacro( "Unexpected swap code:" << SwapCode ); a = 0; } return a; } +/** + * \brief Swaps back the bytes of 8-byte long 'double' accordingly to + * processor order. + * @return The properly swaped 64 bits double. + */ +double Document::SwapDouble(double a) +{ + switch (SwapCode) + { + // There were no 'double' at ACR-NEMA time. + // We just have to deal with 'straight Little Endian' and + // 'straight Big Endian' + case 1234 : + break; + case 4321 : + { + char *beg = (char *)&a; + char *end = beg + 7; + char t; + for (unsigned int i = 0; i<7; i++) + { + t = *beg; + *beg = *end; + *end = t; + beg++, + end--; + } + } + break; + default : + gdcmErrorMacro( "Unexpected swap code:" << SwapCode ); + a = 0.; + } + return 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() { @@ -570,7 +602,7 @@ std::ifstream *Document::OpenFile() if ( Fp ) { - gdcmWarningMacro( "File already open: " << Filename.c_str()); + gdcmDebugMacro( "File already open: " << Filename.c_str()); CloseFile(); } @@ -581,6 +613,10 @@ std::ifstream *Document::OpenFile() // a spurious message will appear when you use, for instance // gdcm::FileHelper *fh = new gdcm::FileHelper( outputFileName ); // to create outputFileName. + + // FIXME : if the upper comment is still usefull + // --> the constructor is not so good ... + gdcmWarningMacro( "Cannot open file: " << Filename.c_str()); delete Fp; Fp = 0; @@ -597,21 +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 at the beginning of the file:(%04x)\n", zero); - 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() ) @@ -619,15 +642,37 @@ 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(); - gdcmWarningMacro( "Neither ACR/No Preamble Dicom nor DICOMV3 file: " + // 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()); return 0; } @@ -653,11 +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 ) + if ( filetype == ImplicitVR || filetype == ExplicitVR || + filetype == JPEG || filetype == JPEG2000 ) { // writing Dicom File Preamble char filePreamble[128]; @@ -665,7 +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 @@ -676,9 +721,15 @@ void Document::WriteContent(std::ofstream *fp, FileType filetype) * UpdateGroupLength(false,filetype); * if ( filetype == ACR) * UpdateGroupLength(true,ACR); + * + * --> Computing group length for groups with embeded Sequences + * --> 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); } // ----------------------------------------- @@ -692,81 +743,122 @@ void Document::WriteContent(std::ofstream *fp, FileType filetype) void Document::LoadEntryBinArea(uint16_t group, uint16_t elem) { // Search the corresponding DocEntry - DocEntry *docElement = GetDocEntry(group, elem); - if ( !docElement ) + DocEntry *docEntry = GetDocEntry(group, elem); + if ( !docEntry ) { - gdcmWarningMacro(std::hex << group << "|" << elem - << "doesn't exist" ); + gdcmDebugMacro(std::hex << group << "|" << elem + << " doesn't exist" ); return; } - BinEntry *binElement = dynamic_cast(docElement); - if ( !binElement ) + DataEntry *dataEntry = dynamic_cast(docEntry); + if ( !dataEntry ) { gdcmWarningMacro(std::hex << group << "|" << elem - << "is NOT a BinEntry"); + << " is NOT a DataEntry"); return; } - LoadEntryBinArea(binElement); + LoadEntryBinArea(dataEntry); } /** * \brief Loads (from disk) the element content * when a string is not suitable - * @param elem Entry whose binArea is going to be loaded + * @param entry Entry whose binArea is going to be loaded */ -void Document::LoadEntryBinArea(BinEntry *elem) -{ - if (elem->GetBinArea() ) +void Document::LoadEntryBinArea(DataEntry *entry) +{ + if( entry->GetBinArea() ) return; bool openFile = !Fp; if ( openFile ) OpenFile(); - size_t o =(size_t)elem->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 = elem->GetLength(); - uint8_t *a = new uint8_t[l]; - if ( !a ) + size_t l = entry->GetLength(); + uint8_t *data = new uint8_t[l]; + if ( !data ) { - gdcmWarningMacro( "Cannot allocate BinEntry content for : " - << std::hex << elem->GetGroup() - << "|" << elem->GetElement() ); + gdcmWarningMacro( "Cannot allocate DataEntry content for : " + << std::hex << entry->GetGroup() + << "|" << entry->GetElement() ); return; } - Fp->read((char*)a, l); + // Read the data + Fp->read((char*)data, l); if ( Fp->fail() || Fp->eof() ) { - delete[] a; + delete[] data; + entry->SetState(DataEntry::STATE_UNREAD); return; } - elem->SetBinArea(a); + // Swap the data content if necessary + uint32_t i; + unsigned short vrLgth = + Global::GetVR()->GetAtomicElementLength(entry->GetVR()); - if ( openFile ) +// FIXME : trouble expected if we read an ... OW Entry (LUT, etc ..) +// if( entry->GetVR() == "OW" ) +// vrLgth = 1; + + switch(vrLgth) + { + case 1: + { + break; + } + case 2: + { + uint16_t *data16 = (uint16_t *)data; + for(i=0;iSetBinArea(data); + + 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 @@ -775,8 +867,8 @@ void Document::LoadDocEntrySafe(DocEntry *entry) bool Document::operator<(Document &document) { // Patient Name - std::string s1 = GetEntryValue(0x0010,0x0010); - std::string s2 = document.GetEntryValue(0x0010,0x0010); + std::string s1 = GetEntryString(0x0010,0x0010); + std::string s2 = document.GetEntryString(0x0010,0x0010); if (s1 < s2) { return true; @@ -788,8 +880,8 @@ bool Document::operator<(Document &document) else { // Patient ID - s1 = GetEntryValue(0x0010,0x0020); - s2 = document.GetEntryValue(0x0010,0x0020); + s1 = GetEntryString(0x0010,0x0020); + s2 = document.GetEntryString(0x0010,0x0020); if ( s1 < s2 ) { return true; @@ -801,8 +893,8 @@ bool Document::operator<(Document &document) else { // Study Instance UID - s1 = GetEntryValue(0x0020,0x000d); - s2 = document.GetEntryValue(0x0020,0x000d); + s1 = GetEntryString(0x0020,0x000d); + s2 = document.GetEntryString(0x0020,0x000d); if ( s1 < s2 ) { return true; @@ -814,8 +906,8 @@ bool Document::operator<(Document &document) else { // Serie Instance UID - s1 = GetEntryValue(0x0020,0x000e); - s2 = document.GetEntryValue(0x0020,0x000e); + s1 = GetEntryString(0x0020,0x000e); + s2 = document.GetEntryString(0x0020,0x000e); if ( s1 < s2 ) { return true; @@ -832,6 +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) @@ -854,6 +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) @@ -877,23 +1002,25 @@ 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( /*FileType filetype*/ ) +int Document::ComputeGroup0002Length( ) { uint16_t gr; - std::string vr; + VRKey vr; int groupLength = 0; bool found0002 = false; @@ -912,18 +1039,13 @@ int Document::ComputeGroup0002Length( /*FileType filetype*/ ) { vr = entry->GetVR(); - // FIXME : group 0x0002 is *always* Explicit VR! - - //if ( filetype == ExplicitVR ) - //{ - // if ( (vr == "OB") || (vr == "OW") || (vr == "UT") || (vr == "SQ") ) - // (no SQ, OW, UT in group 0x0002;) + //if ( (vr == "OB")||(vr == "OW")||(vr == "UT")||(vr == "SQ")) + // (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(); } } @@ -935,11 +1057,38 @@ int Document::ComputeGroup0002Length( /*FileType filetype*/ ) 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() { @@ -959,52 +1108,55 @@ void Document::ParseDES(DocEntrySet *set, long offset, long l_max, bool delim_mode) { DocEntry *newDocEntry; - ValEntry *newValEntry; - BinEntry *newBinEntry; + 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; - gdcmWarningMacro( "Enter in ParseDES, delim-mode " << delim_mode - << " at offset " << std::hex << offset ); + gdcmDebugMacro( "Enter in ParseDES, delim-mode " << delim_mode + << " at offset " << std::hex << "0x(" << offset << ")" ); while (true) { - if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) - { - break; - } - - newDocEntry = ReadNextDocEntry( ); - - // FIXME : - // Private tag, in IMplicit VR are defaulted as a BinEntry, - // Very often they are only composed of Printable characters, - // and could be defaulted as a ValEntry. - // It's too late to do the Job - // (we should check the value, but we know it after LoadDocEntry ...) - // --> in next gdcm major release let's unify ValEntry and BinEntry ! + + ///\todo FIXME : On 64 bits processors, tellg gives unexpected results after a while ? + /// Probabely a bug in gdcm code somewhere (some memory erased ?) - // Uncoment this printf line to be able to 'follow' the DocEntries - // when something *very* strange happens +// 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; + } - //printf( "%04x|%04x %s\n",newDocEntry->GetGroup(), - // newDocEntry->GetElement(), - // newDocEntry->GetVR().c_str() ); + newDocEntry = ReadNextDocEntry( ); if ( !newDocEntry ) { break; } + + // Uncoment this cerr line to be able to 'follow' the DocEntries + // when something *very* strange happens + if( Debug::GetDebugFlag() ) + std::cerr<GetKey()<<" "<GetVR()<IsItemStarter() ) { - newDocEntry = Backtrack(newDocEntry); + // Debug message within the method ! + newDocEntry = Backtrack(newDocEntry, set); } else { @@ -1012,134 +1164,73 @@ void Document::ParseDES(DocEntrySet *set, long offset, } used = true; - newValEntry = dynamic_cast(newDocEntry); - newBinEntry = dynamic_cast(newDocEntry); + newDataEntry = dynamic_cast(newDocEntry); - if ( newValEntry || newBinEntry ) + if ( newDataEntry ) { - //////////////////////////// ContentEntry - if ( newBinEntry ) - { - vr = newDocEntry->GetVR(); - if ( Filetype == ExplicitVR && - !Global::GetVR()->IsVROfBinaryRepresentable(vr) ) - { - ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR - gdcmWarningMacro( std::hex << newDocEntry->GetGroup() - << "|" << newDocEntry->GetElement() - << " : Neither Valentry, nor BinEntry." - "Probably unknown VR."); - } + //////////////////////////// DataEntry + + //vr = newDocEntry->GetVR(); // useless ? - //////////////////// BinEntry or UNKOWN VR: - - // When "this" is a Document the Key is simply of the - // form ( group, elem )... - //if ( set == this ) // ( dynamic_cast< Document* > ( set ) ) - //{ - // newBinEntry->SetKey( newBinEntry->GetKey() ); - //} - // but when "this" is a SQItem, we are inserting this new - // valEntry in a sequence item, and the key has the - // generalized form (refer to \ref BaseTagKey): - - // time waste hunting - //if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) ) - //{ - // newBinEntry->SetKey( parentSQItem->GetBaseTagKey() - // + newBinEntry->GetKey() ); - //} - - if ( !set->AddEntry( newBinEntry ) ) - { - gdcmWarningMacro( "in ParseDES : cannot add a BinEntry " - << newBinEntry->GetKey() - << " (at offset : " - << newBinEntry->GetOffset() << " )" ); - used=false; - } - else - { - // Load only if we can add (not a duplicate key) - LoadDocEntry( newBinEntry ); - } - } // end BinEntry + if ( !set->AddEntry( newDataEntry ) ) + { + gdcmDebugMacro( "in ParseDES : cannot add a DataEntry " + << newDataEntry->GetKey() + << " (at offset : 0x(" + << newDataEntry->GetOffset() << ") )" ); + used=false; + } else { - /////////////////////// ValEntry - - // When "set" is a Document, then we are at the top of the - // hierarchy and the Key is simply of the form ( group, elem )... - //if ( set == this ) // ( dynamic_cast< Document* > ( set ) ) - //{ - // newValEntry->SetKey( newValEntry->GetKey() ); - //} - // ...but when "set" is a SQItem, we are inserting this new - // valEntry in a sequence item. Hence the key has the - // generalized form (refer to \ref BaseTagKey): - - // time waste hunting - //if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) ) - //{ - // newValEntry->SetKey( parentSQItem->GetBaseTagKey() - // + newValEntry->GetKey() ); - //} - - if ( !set->AddEntry( newValEntry ) ) - { - gdcmWarningMacro( "in ParseDES : cannot add a ValEntry " - << newValEntry->GetKey() - << " (at offset : " - << newValEntry->GetOffset() << " )" ); - used=false; - } - else - { - // Load only if we can add (not a duplicate key) - LoadDocEntry( newValEntry ); - } - - if ( newValEntry->GetElement() == 0x0000 ) // if on group length + newDataEntry->Delete(); + // 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 { - if ( newValEntry->GetGroup()%2 != 0 ) // if Shadow Group + if ( LoadMode & LD_NOSHADOW ) // if user asked to skip shad.gr { - if ( LoadMode & LD_NOSHADOW ) // if user asked to skip shad.gr + std::string strLgrGroup = newDataEntry->GetString(); + + int lgrGroup; + //if ( newDataEntry->IsUnfound() ) /?!? JPR { - std::string strLgrGroup = newValEntry->GetValue(); - int lgrGroup; - if ( strLgrGroup != GDCM_UNFOUND) - { - lgrGroup = atoi(strLgrGroup.c_str()); - Fp->seekg(lgrGroup, std::ios::cur); - //used = false; // never used - RemoveEntry( newDocEntry ); // Remove and delete - // bcc 5.5 is right "assigned a value that's never used" - // newDocEntry = 0; - continue; - } + lgrGroup = atoi(strLgrGroup.c_str()); + Fp->seekg(lgrGroup, std::ios::cur); // Once per Shadow group, when NOSHADOW + RemoveEntry( newDocEntry ); // Remove and delete + continue; } } } + } - bool delimitor = newValEntry->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) ) - { - if ( !used ) - delete newDocEntry; - break; - } + // '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(); + break; } // 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 ) @@ -1157,18 +1248,16 @@ 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); - RemoveEntry( newDocEntry ); // Remove and delete - //used = false; // never used + Fp->seekg( l, std::ios::cur); // once per SQITEM, when NOSHADOWSEQ + newDocEntry->Delete(); // Delete, not in the set continue; } } if ( (LoadMode & LD_NOSEQ) && ! delim_mode_intern ) { // User asked to skip *any* SeQuence - Fp->seekg( l, std::ios::cur); - //used = false; // never used - RemoveEntry( newDocEntry ); // Remove and delete + Fp->seekg( l, std::ios::cur); // Once per SQ, when NOSEQ + newDocEntry->Delete(); // Delete, not in the set continue; } // delay the dynamic cast as late as possible @@ -1181,11 +1270,9 @@ void Document::ParseDES(DocEntrySet *set, long offset, // is a Document, then we are building the first depth level. // Hence the SeqEntry we are building simply has a depth // level of one: -// SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ); if ( set == this ) // ( dynamic_cast< Document* > ( set ) ) { newSeqEntry->SetDepthLevel( 1 ); - // newSeqEntry->SetKey( newSeqEntry->GetKey() ); } // But when "set" is already a SQItem, we are building a nested // sequence, and hence the depth level of the new SeqEntry @@ -1195,56 +1282,65 @@ void Document::ParseDES(DocEntrySet *set, long offset, else if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) ) { newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 ); - - // newSeqEntry->SetKey( parentSQItem->GetBaseTagKey() - // + newSeqEntry->GetKey() ); } if ( l != 0 ) { // Don't try to parse zero-length sequences - gdcmWarningMacro( "Entry in ParseSQ, delim " << delim_mode_intern - << " at offset " << std::hex - << newDocEntry->GetOffset() ); + gdcmDebugMacro( "Entry in ParseSQ, delim " << delim_mode_intern + << " 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); - gdcmWarningMacro( "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 ) - delete newDocEntry; - break; + newDocEntry->Delete(); + break; } } // end SeqEntry : VR = "SQ" if ( !used ) { - delete newDocEntry; + newDocEntry->Delete(); } first = false; + + if (UnexpectedEOF) // some terminator was missing + { + break; + } } // end While - gdcmWarningMacro( "Exit from ParseDES, delim-mode " << delim_mode ); + 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; @@ -1257,34 +1353,31 @@ void Document::ParseSQ( SeqEntry *seqEntry, DocEntry *newDocEntry = ReadNextDocEntry(); if ( !newDocEntry ) - { - // FIXME Should warn user + { + // 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 ) { if ( newDocEntry->IsSequenceDelimitor() ) { seqEntry->SetDelimitationItem( newDocEntry ); + newDocEntry->Delete(); break; } } - if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) + else // ! delim_mode { - delete newDocEntry; - break; + if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per SQItem when no delim mode + { + newDocEntry->Delete(); + break; + } } // create the current SQItem - SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() ); -/* - std::ostringstream newBase; - newBase << seqEntry->GetKey() - << "/" - << SQItemNumber - << "#"; - itemSQ->SetBaseTagKey( newBase.str() ); -*/ + SQItem *itemSQ = SQItem::New( seqEntry->GetDepthLevel() ); unsigned int l = newDocEntry->GetReadLength(); if ( l == 0xffffffff ) @@ -1295,38 +1388,52 @@ 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; + } - // Let's try :------------ - // remove fff0,e000, created out of the SQItem - delete newDocEntry; - 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(); - // end try ----------------- + 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 - delete docEntry; + docEntry->Delete(); // Get all info we can from PreviousDocEntry uint16_t group = PreviousDocEntry->GetGroup(); @@ -1334,10 +1441,11 @@ DocEntry *Document::Backtrack(DocEntry *docEntry) uint32_t lgt = PreviousDocEntry->GetLength(); long offset = PreviousDocEntry->GetOffset(); - gdcmWarningMacro( "Backtrack :" << std::hex << group - << "|" << elem - << " at offset " << offset ); - RemoveEntry( PreviousDocEntry ); + gdcmDebugMacro( "Backtrack :" << std::hex << group + << "|" << elem + << " at offset 0x(" <RemoveEntry( PreviousDocEntry ); // forge the Seq Entry DocEntry *newEntry = NewSeqEntry(group, elem); @@ -1345,10 +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! } /** @@ -1359,182 +1466,64 @@ return newEntry; */ void Document::LoadDocEntry(DocEntry *entry, bool forceLoad) { - uint16_t group = entry->GetGroup(); - std::string vr = entry->GetVR(); - uint32_t length = entry->GetLength(); + uint16_t group = entry->GetGroup(); + uint16_t elem = entry->GetElement(); + const VRKey &vr = entry->GetVR(); + 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 - if ( group == 0xfffe ) + // (fffe 0000) is an 'impossible' tag value, + // found in MR-PHILIPS-16-Multi-Seq.dcm + + if ( (group == 0xfffe && elem != 0x0000 ) || vr == "SQ" ) { // NO more value field for SQ ! return; } + DataEntry *dataEntryPtr = dynamic_cast< DataEntry* >(entry); + if( !dataEntryPtr ) + { + return; + } + // When the length is zero things are easy: if ( length == 0 ) { - ((ValEntry *)entry)->SetValue(""); + dataEntryPtr->SetBinArea(NULL,true); return; } // The elements whose length is bigger than the specified upper bound - // are not loaded. Instead we leave a short notice on the offset of - // the element content and it's length. - - std::ostringstream s; + // are not loaded. if (!forceLoad) { if (length > MaxSizeLoadEntry) { - if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) ) - { - s << GDCM_NOTLOADED; - s << " Ad.:" << (long)entry->GetOffset(); - s << " x(" << std::hex << entry->GetOffset() << ")"; - s << std::dec; - s << " Lgt:" << entry->GetLength(); - s << " x(" << std::hex << entry->GetLength() << ")"; - binEntryPtr->SetValue(s.str()); - } - else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) ) - { - s << GDCM_NOTLOADED; - s << " Address:" << (long)entry->GetOffset(); - s << " Length:" << entry->GetLength(); - s << " x(" << std::hex << entry->GetLength() << ")"; - valEntryPtr->SetValue(s.str()); - } - else - { - // fusible - gdcmErrorMacro( "MaxSizeLoadEntry exceeded, neither a BinEntry " - << "nor a ValEntry ?! Should never print that !" ); - } + 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; } } - - // When we find a BinEntry not very much can be done : - if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) ) - { - s << GDCM_BINLOADED; - binEntryPtr->SetValue(s.str()); - LoadEntryBinArea(binEntryPtr); // last one, not to erase length ! - return; - } - - if ( IsDocEntryAnInteger(entry) ) - { - uint32_t NewInt; - int nbInt; - // When short integer(s) are expected, read and convert the following - // (n * 2) characters properly i.e. consider them as short integers as - // opposed to strings. - // Elements with Value Multiplicity > 1 - // contain a set of integers (not a single one) - if (vr == "US" || vr == "SS") - { - nbInt = length / 2; - NewInt = ReadInt16(); - s << NewInt; - if (nbInt > 1) - { - for (int i=1; i < nbInt; i++) - { - s << '\\'; - NewInt = ReadInt16(); - s << NewInt; - } - } - } - // See above comment on multiple integers (mutatis mutandis). - else if (vr == "UL" || vr == "SL") - { - nbInt = length / 4; - NewInt = ReadInt32(); - s << NewInt; - if (nbInt > 1) - { - for (int i=1; i < nbInt; i++) - { - s << '\\'; - NewInt = ReadInt32(); - s << NewInt; - } - } - } -#ifdef GDCM_NO_ANSI_STRING_STREAM - s << std::ends; // to avoid oddities on Solaris -#endif //GDCM_NO_ANSI_STRING_STREAM - - ((ValEntry *)entry)->SetValue(s.str()); - return; - } - // FIXME: We need an additional byte for storing \0 that is not on disk - char *str = new char[length+1]; - Fp->read(str, (size_t)length); - str[length] = '\0'; //this is only useful when length is odd - // Special DicomString call to properly handle \0 and even length - std::string newValue; - if ( length % 2 ) - { - newValue = Util::DicomString(str, length+1); - gdcmWarningMacro("Warning: bad length: " << length << - " For string :" << newValue.c_str()); - // Since we change the length of string update it length - //entry->SetReadLength(length+1); - } - else - { - newValue = Util::DicomString(str, length); - } - delete[] str; - - if ( ValEntry *valEntry = dynamic_cast(entry) ) - { - if ( Fp->fail() || Fp->eof()) - { - if ( Fp->fail() ) - gdcmWarningMacro("--> fail"); - - gdcmWarningMacro("Unread element value " << valEntry->GetKey() - << " lgt : " << valEntry->GetReadLength() - << " at " << std::hex << valEntry->GetOffset()); - valEntry->SetValue(GDCM_UNREAD); - return; - } - -// if ( vr == "UI" ) -// { -// // Because of correspondance with the VR dic -// valEntry->SetValue(newValue); -// } -// else -// { -// valEntry->SetValue(newValue); -// } - -// Anybody remembers the genesis of strange previous (commented out) code? - valEntry->SetValue(newValue); - - } - else - { - gdcmWarningMacro("Should have a ValEntry, here ! " << valEntry->GetKey() - << " lgt : " << valEntry->GetReadLength() - << " at " << std::hex << valEntry->GetOffset()); - } + /// \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 ! } /** @@ -1544,26 +1533,37 @@ void Document::LoadDocEntry(DocEntry *entry, bool forceLoad) void Document::FindDocEntryLength( DocEntry *entry ) throw ( FormatError ) { - std::string vr = entry->GetVR(); + 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 ) { @@ -1575,11 +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); @@ -1592,10 +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) @@ -1615,9 +1614,11 @@ void Document::FindDocEntryLength( DocEntry *entry ) // Length is on 4 bytes. // Well ... group 0002 is always coded in 'Explicit VR Litle Endian' - // even if Transfer Syntax is 'Implicit VR ...' - - FixDocEntryFoundLength( entry, ReadInt32() ); + // even if Transfer Syntax is 'Implicit VR ...' + // --> Except for 'Implicit VR Big Endian Transfer Syntax GE Private' + // where Group 0x0002 is *also* encoded in Implicit VR ! + + FixDocEntryFoundLength( entry, GetInt32() /*ReadInt32()*/ ); return; } } @@ -1631,7 +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; @@ -1639,27 +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."); } @@ -1667,7 +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); @@ -1677,7 +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; } @@ -1685,28 +1693,50 @@ uint32_t Document::FindDocEntryLengthOBOrOW() * \brief Find the Value Representation of the current Dicom Element. * @return Value Representation of the current Entry */ -std::string Document::FindDocEntryVR() +VRKey Document::FindDocEntryVR() { if ( Filetype != ExplicitVR ) - return GDCM_UNKNOWN; + { + 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. Alas this - // doesn't guarantee that all the tags will be in explicit VR. In some - // cases (see e-film filtered files) one finds implicit VR tags mixed - // within an explicit VR file. Hence we make sure the present tag - // is in explicit VR and try to fix things if it happens not to be - // the case. + // 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 + // 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. - char vr[3]; - Fp->read (vr, (size_t)2); - vr[2] = 0; + VRKey vr; + //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) ) - { - Fp->seekg(positionOnEntry, std::ios::beg); - return GDCM_UNKNOWN; + 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; } @@ -1719,147 +1749,13 @@ std::string Document::FindDocEntryVR() * @return false if the VR is incorrect or if the VR isn't referenced * otherwise, it returns true */ -bool Document::CheckDocEntryVR(VRKey vr) +bool Document::CheckDocEntryVR(const VRKey &vr) { - if ( !Global::GetVR()->IsValidVR(vr) ) - return false; - - return true; -} - -/** - * \brief Get the transformed value of the header entry. The VR value - * is used to define the transformation to operate on the value - * \warning NOT end user intended method ! - * @param entry entry to tranform - * @return Transformed entry value - */ -std::string Document::GetDocEntryValue(DocEntry *entry) -{ - if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() ) - { - std::string val = ((ValEntry *)entry)->GetValue(); - std::string vr = entry->GetVR(); - uint32_t length = entry->GetLength(); - std::ostringstream s; - int nbInt; - - // When short integer(s) are expected, read and convert the following - // n * 2 bytes properly i.e. as a multivaluated strings - // (each single value is separated fromthe next one by '\' - // as usual for standard multivaluated filels - // Elements with Value Multiplicity > 1 - // contain a set of short integers (not a single one) - - if ( vr == "US" || vr == "SS" ) - { - uint16_t newInt16; - - nbInt = length / 2; - for (int i=0; i < nbInt; i++) - { - if ( i != 0 ) - { - s << '\\'; - } - newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8); - newInt16 = SwapShort( newInt16 ); - s << newInt16; - } - } - - // When integer(s) are expected, read and convert the following - // n * 4 bytes properly i.e. as a multivaluated strings - // (each single value is separated fromthe next one by '\' - // as usual for standard multivaluated filels - // Elements with Value Multiplicity > 1 - // contain a set of integers (not a single one) - else if ( vr == "UL" || vr == "SL" ) - { - uint32_t newInt32; - - nbInt = length / 4; - for (int i=0; i < nbInt; i++) - { - if ( i != 0) - { - s << '\\'; - } - newInt32 = ( val[4*i+0] & 0xFF ) - + (( val[4*i+1] & 0xFF ) << 8 ) - + (( val[4*i+2] & 0xFF ) << 16 ) - + (( val[4*i+3] & 0xFF ) << 24 ); - newInt32 = SwapLong( newInt32 ); - s << newInt32; - } - } -#ifdef GDCM_NO_ANSI_STRING_STREAM - s << std::ends; // to avoid oddities on Solaris -#endif //GDCM_NO_ANSI_STRING_STREAM - return s.str(); - } - return ((ValEntry *)entry)->GetValue(); -} - -/** - * \brief Get the reverse transformed value of the header entry. The VR - * value is used to define the reverse transformation to operate on - * the value - * \warning NOT end user intended method ! - * @param entry Entry to reverse transform - * @return Reverse transformed entry value - */ -std::string Document::GetDocEntryUnvalue(DocEntry *entry) -{ - if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() ) - { - std::string vr = entry->GetVR(); - std::vector tokens; - std::ostringstream s; - - if ( vr == "US" || vr == "SS" ) - { - uint16_t newInt16; - - tokens.erase( tokens.begin(), tokens.end()); // clean any previous value - Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\"); - for (unsigned int i=0; i> 8 ) & 0xFF ); - } - tokens.clear(); - } - if ( vr == "UL" || vr == "SL") - { - uint32_t newInt32; - - tokens.erase(tokens.begin(),tokens.end()); // clean any previous value - Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\"); - for (unsigned int i=0; i> 8 ) & 0xFF ) - << (char)(( newInt32 >> 16 ) & 0xFF ) - << (char)(( newInt32 >> 24 ) & 0xFF ); - } - tokens.clear(); - } - -#ifdef GDCM_NO_ANSI_STRING_STREAM - s << std::ends; // to avoid oddities on Solaris -#endif //GDCM_NO_ANSI_STRING_STREAM - return s.str(); - } - - return ((ValEntry *)entry)->GetValue(); + return Global::GetVR()->IsValidVR(vr); } /** * \brief Skip a given Header Entry - * \warning NOT end user intended method ! * @param entry entry to skip */ void Document::SkipDocEntry(DocEntry *entry) @@ -1869,19 +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 } } @@ -1890,26 +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 + 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 @@ -1930,30 +1831,45 @@ 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; } - } + else + { + foundLength=12; // to skip the mess that follows this bugged Tag ! + } + } entry->SetLength(foundLength); } @@ -1967,12 +1883,12 @@ bool Document::IsDocEntryAnInteger(DocEntry *entry) { uint16_t elem = entry->GetElement(); uint16_t group = entry->GetGroup(); - const std::string &vr = entry->GetVR(); + const VRKey &vr = entry->GetVR(); uint32_t length = entry->GetLength(); // 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 ) @@ -1981,14 +1897,15 @@ bool Document::IsDocEntryAnInteger(DocEntry *entry) } else { - // Allthough this should never happen, still some images have a + // Although this should never happen, still some images have a // corrupted group length [e.g. have a glance at offset x(8336) of - // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm]. + // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm. // Since for dicom compliant and well behaved headers, the present // 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 << ")" @@ -2029,7 +1946,7 @@ bool Document::CheckSwap() char *entCur = deb + 128; if ( memcmp(entCur, "DICM", (size_t)4) == 0 ) { - gdcmWarningMacro( "Looks like DICOM Version3 (preamble + DCM)" ); + gdcmDebugMacro( "Looks like DICOM Version3 (preamble + DCM)" ); // Group 0002 should always be VR, and the first element 0000 // Let's be carefull (so many wrong headers ...) @@ -2047,7 +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') @@ -2061,7 +1978,7 @@ bool Document::CheckSwap() // instead of just checking for UL, OB and UI !? group 0000 { Filetype = ExplicitVR; - gdcmWarningMacro( "Group 0002 : Explicit Value Representation"); + gdcmDebugMacro( "Group 0002 : Explicit Value Representation"); } else { @@ -2070,23 +1987,24 @@ 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; - gdcmWarningMacro( "HostByteOrder != NetworkByteOrder"); + gdcmDebugMacro( "HostByteOrder != NetworkByteOrder, SwapCode = 4321"); } else { SwapCode = 1234; - gdcmWarningMacro( "HostByteOrder = NetworkByteOrder"); + gdcmDebugMacro( "HostByteOrder = NetworkByteOrder, SwapCode = 1234"); } // 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 ---------------- @@ -2096,8 +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 @@ -2124,8 +2042,9 @@ bool Document::CheckSwap() memcmp(entCur, "AE", (size_t)2) == 0 || memcmp(entCur, "OB", (size_t)2) == 0 ) { - Filetype = ExplicitVR; - gdcmWarningMacro( "Group 0002 : Explicit Value Representation"); + Filetype = ExplicitVR; // FIXME : not enough to say it's Explicit + // Wait untill reading Transfer Syntax + gdcmDebugMacro( "Group 0002 : Explicit Value Representation"); return true; } } @@ -2171,12 +2090,14 @@ bool Document::CheckSwap() // Only 0 or 4321 will be possible // (no oportunity to check for the formerly well known // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian' - // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-) + // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc-3, 4, ..., 8-) // the file IS NOT ACR-NEMA nor DICOM V3 // 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 : @@ -2187,6 +2108,7 @@ bool Document::CheckSwap() case 0x0006 : case 0x0007 : case 0x0008 : + case 0x0028 : SwapCode = 1234; Filetype = ACR; return true; @@ -2198,13 +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 (Really 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() ); + gdcmDebugMacro( "Switching Byte Swap code from "<< SwapCode + << " at: 0x" << std::hex << Fp->tellg() ); // Only when DEBUG if ( SwapCode == 1234 ) { SwapCode = 4321; @@ -2232,13 +2194,14 @@ void Document::SwitchByteSwapCode() { SwapCode = 3412; } + gdcmDebugMacro( " Into: "<< SwapCode ); } /** * \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 ) { @@ -2257,78 +2220,143 @@ 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); - - std::string vr = FindDocEntryVR(); - std::string realVR = vr; + { + 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_UNKNOWN ) + 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(); // GetDictEntry registered it } } } + // 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()->IsVROfStringRepresentable(realVR) ) - newEntry = NewValEntry(group, elem, realVR); + //if ( Global::GetVR()->IsVROfSequence(realVR) ) + if (realVR == "SQ") + { + newEntry = NewSeqEntry(CurrentGroup, CurrentElem); + } else - newEntry = NewBinEntry(group, elem, realVR); + { + newEntry = NewDataEntry(CurrentGroup, CurrentElem, realVR); + static_cast(newEntry)->SetState(DataEntry::STATE_NOTLOADED); + } - if ( vr == GDCM_UNKNOWN ) + if ( vr == GDCM_VRUNKNOWN ) { if ( Filetype == ExplicitVR ) { // 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 0x(%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(); @@ -2341,12 +2369,11 @@ DocEntry *Document::ReadNextDocEntry() catch ( FormatError ) { // Call it quits - delete newEntry; + newEntry->Delete(); return 0; } - newEntry->SetOffset(Fp->tellg()); - + newEntry->SetOffset(Fp->tellg()); // for each DocEntry return newEntry; } @@ -2358,12 +2385,31 @@ DocEntry *Document::ReadNextDocEntry() */ void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem) { - // Endian reversion. Some files contain groups of tags with reversed endianess. + // 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; // try to fix endian switching in the middle of headers if ((group == 0xfeff) && (elem == 0x00e0)) { // start endian swap mark for group found + gdcmDebugMacro( "Start endian swap mark found." ); reversedEndian++; SwitchByteSwapCode(); // fix the tag @@ -2373,6 +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(); } @@ -2383,14 +2430,14 @@ void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem) // Do what you want, it breaks ! //reversedEndian--; //SwitchByteSwapCode(); - gdcmWarningMacro( "Should never get here! reversed Sequence Terminator!" ); + gdcmWarningMacro( "Should never get here! reversed Sequence Terminator!" ); // fix the tag group = 0xfffe; elem = 0xe0dd; } else if (group == 0xfffe && elem == 0xe0dd) { - gdcmWarningMacro( "Straight Sequence Terminator." ); + gdcmDebugMacro( "Straight Sequence Terminator." ); } } @@ -2401,38 +2448,86 @@ void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem) */ void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem) { - // Endian reversion. Some files contain groups of tags with reversed endianess. - if ( !Group0002Parsed && group != 0x0002) - { + // Endian reversion. + // Some files contain groups of tags with reversed endianess. + Group0002Parsed = true; // we just came out of group 0002 - // if Transfer syntax is Big Endian we have to change CheckSwap + // if Transfer Syntax is Big Endian we have to change CheckSwap std::string ts = GetTransferSyntax(); - if ( !Global::GetTS()->IsTransferSyntax(ts) ) - { - gdcmWarningMacro("True DICOM File, with NO Tansfer Syntax: " << ts ); - return; - } + TS::SpecialType s = Global::GetTS()->GetSpecialTransferSyntax(ts); - // Group 0002 is always 'Explicit ...' enven when Transfer Syntax says 'Implicit ..." + // Group 0002 is always 'Explicit ...' + // even when Transfer Syntax says 'Implicit ..." - if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ImplicitVRLittleEndian ) - { - Filetype = ImplicitVR; - } + if ( s == TS::ImplicitVRLittleEndian + || + s == TS::ImplicitVRBigEndianPrivateGE + ) + { + Filetype = ImplicitVR; + } // FIXME Strangely, this works with - //'Implicit VR Transfer Syntax (GE Private) - if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ExplicitVRBigEndian ) + //'Implicit VR BigEndian Transfer Syntax' (GE Private) + // + // --> Probabely normal, since we considered we never have + // to trust manufacturers. + // (we often find 'Implicit VR' tag, + // even when Transfer Syntax tells us it's Explicit ... + + // NEVER trust the meta elements! + // (see what ezDICOM does ...) + + /* + if ( s == TS::ExplicitVRBigEndian ) { - gdcmWarningMacro("Transfer Syntax Name = [" + gdcmDebugMacro("Transfer Syntax Name = [" << GetTransferSyntaxName() << "]" ); SwitchByteSwapCode(); 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; + } } //-----------------------------------------------------------------------------