X-Git-Url: https://git.creatis.insa-lyon.fr/pubgit/?a=blobdiff_plain;f=src%2FgdcmDocument.cxx;h=c60284b6d51e09afadbd75d4239f6e89a27723e5;hb=6278320cc85da00d2d56ffbf07806e84966892c3;hp=550aa731d16a7d903272ca104f8ae3d962a7e37c;hpb=2f27193f6cbd42d1143738e40563b4a4bd4a0c8a;p=gdcm.git diff --git a/src/gdcmDocument.cxx b/src/gdcmDocument.cxx index 550aa731..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: 2004/10/25 04:47:43 $ - Version: $Revision: 1.114 $ + 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 @@ -17,121 +17,203 @@ =========================================================================*/ #include "gdcmDocument.h" -#include "gdcmValEntry.h" -#include "gdcmBinEntry.h" #include "gdcmSeqEntry.h" #include "gdcmGlobal.h" #include "gdcmUtil.h" #include "gdcmDebug.h" +#include "gdcmTS.h" +#include "gdcmDictSet.h" +#include "gdcmDocEntrySet.h" +#include "gdcmSQItem.h" +#include "gdcmDataEntry.h" #include #include +#include +#include // for isdigit +#include // for atoi -// For nthos: -#ifdef _MSC_VER - #include -#else - #include -#endif +#if defined(__BORLANDC__) + #include // for memset +#endif -namespace gdcm +namespace GDCM_NAME_SPACE { -static const char *TransferSyntaxStrings[] = { - // Implicit VR Little Endian - "1.2.840.10008.1.2", - // Explicit VR Little Endian - "1.2.840.10008.1.2.1", - // Deflated Explicit VR Little Endian - "1.2.840.10008.1.2.1.99", - // Explicit VR Big Endian - "1.2.840.10008.1.2.2", - // JPEG Baseline (Process 1) - "1.2.840.10008.1.2.4.50", - // JPEG Extended (Process 2 & 4) - "1.2.840.10008.1.2.4.51", - // JPEG Extended (Process 3 & 5) - "1.2.840.10008.1.2.4.52", - // JPEG Spectral Selection, Non-Hierarchical (Process 6 & 8) - "1.2.840.10008.1.2.4.53", - // JPEG Full Progression, Non-Hierarchical (Process 10 & 12) - "1.2.840.10008.1.2.4.55", - // JPEG Lossless, Non-Hierarchical (Process 14) - "1.2.840.10008.1.2.4.57", - // JPEG Lossless, Hierarchical, First-Order Prediction (Process 14, [Selection Value 1]) - "1.2.840.10008.1.2.4.70", - // JPEG 2000 Lossless - "1.2.840.10008.1.2.4.90", - // JPEG 2000 - "1.2.840.10008.1.2.4.91", - // RLE Lossless - "1.2.840.10008.1.2.5", - // Unknown - "Unknown Transfer Syntax" -}; - //----------------------------------------------------------------------------- -// Refer to Document::CheckSwap() -const unsigned int Document::HEADER_LENGTH_TO_READ = 256; // Refer to Document::SetMaxSizeLoadEntry() const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096 -const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff; //----------------------------------------------------------------------------- // Constructor / Destructor +// Constructors and destructors are protected to avoid user to invoke directly + +/** + * \brief This default constructor neither loads nor parses the file. + * You should then invoke Document::Load. + * + */ +Document::Document() + :ElementSet() +{ + Fp = 0; + + SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE); + 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 Canonical destructor. + */ +Document::~Document () +{ + CloseFile(); +} + +//----------------------------------------------------------------------------- +// Public +/** + * \brief Loader. use SetLoadMode(), SetFileName() before ! + * @return false if file cannot be open or no swap info was found, + * or no tag was found. + */ +bool Document::Load( ) +{ + if ( GetFileName() == "" ) + { + gdcmWarningMacro( "Use SetFileName, before !" ); + 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 constructor - * @param filename file to be opened for parsing + * \brief Performs the Loading Job (internal use only) + * @return false if file cannot be open or no swap info was found, + * or no tag was found. */ -Document::Document( std::string const & filename ) : ElementSet(-1) +bool Document::DoTheLoadingDocumentJob( ) { - SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE); - Filename = filename; - Initialise(); + if ( ! IsDocumentModified ) // Nothing to do ! + return true; + ClearEntry(); + + Fp = 0; if ( !OpenFile() ) { - return; + // warning already performed in OpenFile() + Filetype = Unknown; + return false; } - dbg.Verbose(0, "Document::Document: starting parsing of file: ", - Filename.c_str()); - Fp->seekg( 0, std::ios_base::beg); - - Fp->seekg(0, std::ios_base::end); - long lgt = Fp->tellg(); - - Fp->seekg( 0, std::ios_base::beg); - CheckSwap(); - long beg = Fp->tellg(); - lgt -= beg; - - ParseDES( this, beg, lgt, false); // le Load sera fait a la volee + Group0002Parsed = false; + + gdcmDebugMacro( "Starting parsing of file: " << Filename.c_str()); + + // 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) + if (! CheckSwap() ) + { + gdcmWarningMacro( "Neither a DICOM V3 nor an ACR-NEMA file: " + << Filename.c_str()); + CloseFile(); + return false; + } + + long beg = Fp->tellg(); // just after DICOM preamble (if any) + + lgt -= beg; // remaining length to parse + + // Recursive call. + // Loading is done during parsing + OffsetOfPreviousParseDES = beg; + ParseDES( this, beg, lgt, false); // delim_mode is first defaulted to false + + if ( IsEmpty() ) + { + gdcmErrorMacro( "No tag in internal hash table for: " + << Filename.c_str()); + CloseFile(); + return false; + } + IsDocumentAlreadyLoaded = true; - Fp->seekg( 0, std::ios_base::beg); + //Fp->seekg(0, std::ios::beg); // Once per Document! // Load 'non string' values - std::string PhotometricInterpretation = GetEntryByNumber(0x0028,0x0004); - if( PhotometricInterpretation == "PALETTE COLOR " ) + std::string PhotometricInterpretation = GetEntryString(0x0028,0x0004); + if ( PhotometricInterpretation == "PALETTE COLOR " ) { - LoadEntryBinArea(0x0028,0x1200); // gray LUT - /// FIXME FIXME FIXME + // FIXME + // Probabely this line should be outside the 'if' + // Try to find an image sample holding a 'gray LUT' + 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 - /// dictionnary. BUT the semantics implied by the three following + /// dictionary. BUT the semantics implied by the three following /// lines state that the corresponding tag contents are in fact /// the ones of a BinEntry. - /// In order to fix things "Quick and Dirty" the dictionnary was - /// altered on PURPOUS but now contains a WRONG value. + /// 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 - /// wether the following tags are either: + /// 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 @@ -142,94 +224,99 @@ Document::Document( std::string const & filename ) : ElementSet(-1) LoadEntryBinArea(0x0028,0x1222); // Segmented Blue Palette Color LUT Data LoadEntryBinArea(0x0028,0x1223); - } + } + //FIXME later : how to use it? - LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent) + SeqEntry *modLutSeq = GetSeqEntry(0x0028,0x3000); // Modality LUT Sequence + if ( modLutSeq !=0 ) + { + SQItem *sqi= modLutSeq->GetFirstSQItem(); + if ( sqi != 0 ) + { + DataEntry *dataEntry = sqi->GetDataEntry(0x0028,0x3006); // LUT Data + if ( dataEntry != 0 ) + { + if ( dataEntry->GetLength() != 0 ) + { + // FIXME : CTX dependent means : contexted dependant. + // see upper comment. + LoadEntryBinArea(dataEntry); //LUT Data (CTX dependent) + } + } + } + } + + // Force Loading some more elements if user asked to. + + GDCM_NAME_SPACE::DocEntry *d; + for (ListElements::iterator it = UserForceLoadList.begin(); + it != UserForceLoadList.end(); + ++it) + { + gdcmDebugMacro( "Force Load " << std::hex + << (*it).Group << "|" <<(*it).Elem ); + + d = GetDocEntry( (*it).Group, (*it).Elem); + + if ( d == NULL) + { + gdcmWarningMacro( "You asked to ForceLoad " << std::hex + << (*it).Group <<"|"<< (*it).Elem + << " that doesn't exist" ); + continue; + } + + LoadDocEntry(d, true); + } - CloseFile(); + CloseFile(); - // -------------------------------------------------------------- + // ---------------------------- // Specific code to allow gdcm to read ACR-LibIDO formated images // Note: ACR-LibIDO is an extension of the ACR standard that was - // used at CREATIS. For the time being (say a couple years) - // we keep this kludge to allow a smooth move to gdcm for - // CREATIS developpers (sorry folks). + // used at CREATIS. For the time being (say a couple of years) + // we keep this kludge to allow CREATIS users + // reading their old images. // // if recognition code tells us we deal with a LibIDO image // we switch lineNumber and columnNumber // std::string RecCode; - RecCode = GetEntryByNumber(0x0008, 0x0010); // recognition code - 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 = GetEntryByNumber(0x0028, 0x0010); - std::string columns = GetEntryByNumber(0x0028, 0x0011); - SetEntryByNumber(columns, 0x0028, 0x0010); - SetEntryByNumber(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 ------------------ - - PrintLevel = 1; // 'Medium' print level by default + // --- End of ACR-LibIDO kludge --- + return true; } -/** - * \brief This default constructor doesn't parse the file. You should - * then invoke \ref Document::SetFileName and then the parsing. - */ -Document::Document() : ElementSet(-1) -{ - SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE); - Initialise(); - PrintLevel = 1; // 'Medium' print level by default -} /** - * \brief Canonical destructor. + * \brief Adds a new element we want to load anyway + * @param group Group number of the target tag. + * @param elem Element number of the target tag. */ -Document::~Document () -{ - RefPubDict = NULL; - RefShaDict = NULL; - - // Recursive clean up of sequences - for (TagDocEntryHT::const_iterator it = TagHT.begin(); - it != TagHT.end(); ++it ) - { - //delete it->second; //temp remove - } - TagHT.clear(); -} - -//----------------------------------------------------------------------------- -// Print - -/** - * \brief Prints The Dict Entries of THE public Dicom Dictionary - * @return - */ -void Document::PrintPubDict(std::ostream & os) -{ - RefPubDict->Print(os); -} - -/** - * \brief Prints The Dict Entries of THE shadow Dicom Dictionary - * @return - */ -void Document::PrintShaDict(std::ostream & os) -{ - RefShaDict->Print(os); +void Document::AddForceLoadElement (uint16_t group, uint16_t elem) +{ + DicomElement el; + el.Group = group; + el.Elem = elem; + UserForceLoadList.push_back(el); } - -//----------------------------------------------------------------------------- -// Public /** * \brief Get the public dictionary used */ -Dict* Document::GetPubDict() +Dict *Document::GetPubDict() { return RefPubDict; } @@ -237,14 +324,14 @@ Dict* Document::GetPubDict() /** * \brief Get the shadow dictionary used */ -Dict* Document::GetShaDict() +Dict *Document::GetShaDict() { return RefShaDict; } /** * \brief Set the shadow dictionary used - * \param dict dictionary to use in shadow + * @param dict dictionary to use in shadow */ bool Document::SetShaDict(Dict *dict) { @@ -254,196 +341,339 @@ bool Document::SetShaDict(Dict *dict) /** * \brief Set the shadow dictionary used - * \param dictName name of the dictionary to use in shadow + * @param dictName name of the dictionary to use in shadow */ -bool Document::SetShaDict(DictKey const & dictName) +bool Document::SetShaDict(DictKey const &dictName) { RefShaDict = Global::GetDicts()->GetDict(dictName); return !RefShaDict; } /** - * \brief This predicate, based on hopefully reasonable heuristics, - * decides whether or not the current Document was properly parsed - * and contains the mandatory information for being considered as - * a well formed and usable Dicom/Acr File. - * @return true when Document is the one of a reasonable Dicom/Acr file, - * false otherwise. + * \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() +bool Document::IsParsable() { - if( Filetype == Unknown) + if ( Filetype == Unknown ) { - dbg.Verbose(0, "Document::IsReadable: wrong filetype"); + gdcmWarningMacro( "Wrong filetype for " << GetFileName()); return false; } - if( TagHT.empty() ) - { - dbg.Verbose(0, "Document::IsReadable: no tags in internal" - " hash table."); + if ( IsEmpty() ) + { + 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. + * @return True when the file is a dicom version 3. + */ +bool Document::IsDicomV3() +{ + // Checking if Transfer Syntax exists is enough + // Anyway, it's too late check if the 'Preamble' was found ... + // And ... would it be a rich idea to check ? + // (some 'no Preamble' DICOM images exist !) + return GetDocEntry(0x0002, 0x0010) != NULL; +} + +/** + * \brief Predicate for Papyrus file + * Dedicated to whomsoever it may concern + * @return True when the file is a Papyrus file. + */ +bool Document::IsPapyrus() +{ + // check for Papyrus private Sequence + DocEntry *e = GetDocEntry(0x0041, 0x1050); + if ( !e ) + return false; + // check if it's actually a Sequence + if ( !dynamic_cast(e) ) + return false; + return true; +} +/** + * \brief returns the File Type + * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown) + * @return the FileType code + */ +FileType Document::GetFileType() +{ + return Filetype; +} /** - * \brief Internal function that checks whether the Transfer Syntax given - * as argument is the one present in the current document. - * @param syntaxToCheck The transfert syntax we need to check against. - * @return True when SyntaxToCheck corresponds to the Transfer Syntax of - * the current document. False either when the document contains - * no Transfer Syntax, or when the Tranfer Syntaxes doesn't match. + * \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, if DICOM. + * GDCM_UNKNOWN for ACR-NEMA files (or broken headers ...) */ -TransferSyntaxType Document::GetTransferSyntax() +std::string Document::GetTransferSyntax() { - DocEntry *entry = GetDocEntryByNumber(0x0002, 0x0010); + DocEntry *entry = GetDocEntry(0x0002, 0x0010); if ( !entry ) { - return UnknownTS; + return GDCM_UNKNOWN; } // 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 - return UnknownTS; + gdcmWarningMacro( "Transfer Syntax has length = 0."); + return GDCM_UNKNOWN; } - while ( !isdigit(transfer[transfer.length()-1]) ) + while ( !isdigit((unsigned char)transfer[transfer.length()-1]) ) { transfer.erase(transfer.length()-1, 1); - } - for (int i = 0; TransferSyntaxStrings[i] != NULL; i++) - { - if ( TransferSyntaxStrings[i] == transfer ) + if ( transfer.length() == 0 ) { - return TransferSyntaxType(i); + // for brain damaged headers + gdcmWarningMacro( "Transfer Syntax contains no valid character."); + return GDCM_UNKNOWN; } } + return transfer; } - return UnknownTS; + return GDCM_UNKNOWN; } -bool Document::IsJPEGLossless() -{ - TransferSyntaxType r = GetTransferSyntax(); - return r == JPEGFullProgressionProcess10_12 - || r == JPEGLosslessProcess14 - || r == JPEGLosslessProcess14_1; -} - /** - * \brief Determines if the Transfer Syntax was already encountered - * and if it corresponds to a JPEG2000 one - * @return True when JPEG2000 (Lossly or LossLess) found. False in all - * other cases. + * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS + * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID) */ -bool Document::IsJPEG2000() +std::string Document::GetTransferSyntaxName() { - TransferSyntaxType r = GetTransferSyntax(); - return r == JPEG2000Lossless || r == JPEG2000; -} + // use the TS (TS : Transfer Syntax) + std::string transferSyntax = GetEntryString(0x0002,0x0010); -/** - * \brief Determines if the Transfer Syntax corresponds to any form - * of Jpeg encoded Pixel data. - * @return True when any form of JPEG found. False otherwise. - */ -bool Document::IsJPEG() -{ - TransferSyntaxType r = GetTransferSyntax(); - return r == JPEGBaselineProcess1 - || r == JPEGExtendedProcess2_4 - || r == JPEGExtendedProcess3_5 - || r == JPEGSpectralSelectionProcess6_8 - || IsJPEGLossless() - || IsJPEG2000(); -} + if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) ) + { + gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl + << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" ); + return "Uncompressed ACR-NEMA"; + } + if ( transferSyntax == GDCM_UNFOUND ) + { + gdcmDebugMacro( "Unfound Transfer Syntax (0002,0010)"); + return "Uncompressed ACR-NEMA"; + } + + // we do it only when we need it + const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax ); + // Global::GetTS() is a global static you shall never try to delete it! + return tsName; +} +// +// --------------- Swap Code ------------------ /** - * \brief Determines if the Transfer Syntax corresponds to encapsulated - * of encoded Pixel Data (as opposed to native). - * @return True when encapsulated. False when native. + * \brief Swaps the bytes so they agree with the processor order + * @return The properly swaped 16 bits integer. */ -bool Document::IsEncapsulate() +uint16_t Document::SwapShort(uint16_t a) { - TransferSyntaxType r = GetTransferSyntax(); - return IsJPEG() || r == RLELossless; + if ( SwapCode == 4321 || SwapCode == 2143 ) + { + //a = ((( a << 8 ) & 0xff00 ) | (( a >> 8 ) & 0x00ff ) ); + // Save CPU time + a = ( a << 8 ) | ( a >> 8 ); + } + return a; } /** - * \brief Predicate for dicom version 3 file. - * @return True when the file is a dicom version 3. + * \brief Swaps back the bytes of 4-byte long integer accordingly to + * processor order. + * @return The properly swaped 32 bits integer. */ -bool Document::IsDicomV3() +uint32_t Document::SwapLong(uint32_t a) { - // Checking if Transfert Syntax exists is enough - // Anyway, it's to late check if the 'Preamble' was found ... - // And ... would it be a rich idea to check ? - // (some 'no Preamble' DICOM images exist !) - return GetDocEntryByNumber(0x0002, 0x0010) != NULL; -} + switch (SwapCode) + { + case 1234 : + break; + case 4321 : +// a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) | +// ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) ); +// save CPU time + a=( ( a<<24) | ((a<<8) & 0x00ff0000) | + ((a>>8) & 0x0000ff00) | (a>>24) ); + break; + case 3412 : +// a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) ); + a=( (a<<16) | (a>>16) ); + break; + case 2143 : + a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) ); + break; + default : + gdcmErrorMacro( "Unexpected swap code:" << SwapCode ); + a = 0; + } + return a; +} /** - * \brief returns the File Type - * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown) - * @return the FileType code + * \brief Swaps back the bytes of 8-byte long 'double' accordingly to + * processor order. + * @return The properly swaped 64 bits double. */ -FileType Document::GetFileType() +double Document::SwapDouble(double a) { - return Filetype; -} + 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() +std::ifstream *Document::OpenFile() { - Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary); + HasDCMPreamble = false; + if (Filename.length() == 0) + { + return 0; + } + + if ( Fp ) + { + gdcmDebugMacro( "File already open: " << Filename.c_str()); + CloseFile(); + } - if(!Fp) + Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary); + if ( ! *Fp ) { - dbg.Verbose( 0, - "Document::OpenFile cannot open file: ", - Filename.c_str()); + // Don't user gdcmErrorMacro : + // 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; return 0; + //exit(1); // No function is allowed to leave the application instead + // of warning the caller } - uint16_t zero; - Fp->read((char*)&zero, (size_t)2 ); - - //ACR -- or DICOM with no Preamble -- - if( zero == 0x0008 || zero == 0x0800 || zero == 0x0002 || zero == 0x0200 ) + uint16_t zero = 0; + Fp->read((char*)&zero, (size_t)2); + if ( Fp->eof() ) { - return Fp; + CloseFile(); + return 0; } - //DICOM - Fp->seekg(126L, std::ios_base::cur); - char dicm[4]; + //-- DICOM -- + Fp->seekg(126L, std::ios::cur); // Once per Document + char dicm[4]; // = {' ',' ',' ',' '}; Fp->read(dicm, (size_t)4); - if( memcmp(dicm, "DICM", 4) == 0 ) + if ( Fp->eof() ) { + CloseFile(); + return 0; + } + + if ( memcmp(dicm, "DICM", 4) == 0 ) + { + HasDCMPreamble = true; return Fp; } - - Fp->close(); - dbg.Verbose( 0, - "Document::OpenFile not DICOM/ACR (missing preamble)", - Filename.c_str()); - + + //-- 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 pollute the output + // while directory recursive parsing ... + gdcmDebugMacro( "Neither ACR/No Preamble Dicom nor DICOMV3 file: " + << Filename.c_str()); return 0; } @@ -453,1053 +683,701 @@ std::ifstream* Document::OpenFile() */ bool Document::CloseFile() { - Fp->close(); - delete Fp; - Fp = 0; - - return true; //FIXME how do we detect a non-close ifstream ? + if ( Fp ) + { + Fp->close(); + delete Fp; + Fp = 0; + } + return true; } /** - * \brief Writes in a file all the Header Entries (Dicom Elements) - * @param fp file pointer on an already open file + * \brief Writes in a file all the Entries (Dicom Elements) + * @param fp file pointer on an already open file (actually: Output File Stream) * @param filetype Type of the File to be written * (ACR-NEMA, ExplicitVR, ImplicitVR) - * \return Always true. */ -void Document::Write(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 == JPEG2000 ) + { + // writing Dicom File Preamble + char filePreamble[128]; + memset(filePreamble, 0, 128); + fp->write(filePreamble, 128); + fp->write("DICM", 4); + } + /* + * \todo rewrite later, if really usefull + * - 'Group Length' element is optional in DICOM + * - but un-updated odd groups lengthes can causes pb + * (xmedcon breaker) + * + * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) ) + * 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) + */ + // This one is recursive + // false : outside MetaElements + // false : outside Sequence + ElementSet::WriteContent(fp, filetype, false, false); +} + +// ----------------------------------------- +// Content entries +/** + * \brief Loads (from disk) the element content + * when a string is not suitable + * @param group group number of the Entry + * @param elem element number of the Entry + */ +void Document::LoadEntryBinArea(uint16_t group, uint16_t elem) { - /// \todo move the following lines (and a lot of others, to be written) - /// to a future function CheckAndCorrectHeader - /// (necessary if user wants to write a DICOM V3 file - /// starting from an ACR-NEMA (V2) Header - - if (filetype == ImplicitVR) + // Search the corresponding DocEntry + DocEntry *docEntry = GetDocEntry(group, elem); + if ( !docEntry ) { - std::string ts = TransferSyntaxStrings[ImplicitVRLittleEndian]; - ReplaceOrCreateByNumber(ts, 0x0002, 0x0010); - - /// \todo Refer to standards on page 21, chapter 6.2 - /// "Value representation": values with a VR of UI shall be - /// padded with a single trailing null - /// in the following case we have to padd manually with a 0 - - SetEntryLengthByNumber(18, 0x0002, 0x0010); - } - - if (filetype == ExplicitVR) + gdcmDebugMacro(std::hex << group << "|" << elem + << " doesn't exist" ); + return; + } + DataEntry *dataEntry = dynamic_cast(docEntry); + if ( !dataEntry ) { - std::string ts = TransferSyntaxStrings[ExplicitVRLittleEndian]; - ReplaceOrCreateByNumber(ts, 0x0002, 0x0010); - - /// \todo Refer to standards on page 21, chapter 6.2 - /// "Value representation": values with a VR of UI shall be - /// padded with a single trailing null - /// Dans le cas suivant on doit pader manuellement avec un 0 - - SetEntryLengthByNumber(20, 0x0002, 0x0010); + gdcmWarningMacro(std::hex << group << "|" << elem + << " is NOT a DataEntry"); + return; } - + LoadEntryBinArea(dataEntry); +} + /** - * \todo rewrite later, if really usefull - * - 'Group Length' element is optional in DICOM - * - but un-updated odd groups lengthes can causes pb - * (xmedcon breaker) - * - * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) ) - * UpdateGroupLength(false,filetype); - * if ( filetype == ACR) - * UpdateGroupLength(true,ACR); + * \brief Loads (from disk) the element content + * when a string is not suitable + * @param entry Entry whose binArea is going to be loaded */ - - ElementSet::Write(fp, filetype); // This one is recursive +void Document::LoadEntryBinArea(DataEntry *entry) +{ + if( entry->GetBinArea() ) + return; -} + bool openFile = !Fp; + if ( openFile ) + OpenFile(); -/** - * \brief Modifies the value of a given Header Entry (Dicom Element) - * when it exists. Create it with the given value when unexistant. - * @param value (string) Value to be set - * @param group Group number of the Entry - * @param elem Element number of the Entry - * @param VR V(alue) R(epresentation) of the Entry -if private Entry- - * \return pointer to the modified/created Header Entry (NULL when creation - * failed). - */ -ValEntry* Document::ReplaceOrCreateByNumber( - std::string const & value, - uint16_t group, - uint16_t elem, - TagName const & vr ) -{ - ValEntry* valEntry = 0; - DocEntry* currentEntry = GetDocEntryByNumber( group, elem); - - if (!currentEntry) - { - // check if (group,element) DictEntry exists - // if it doesn't, create an entry in DictSet::VirtualEntry - // and use it + //size_t o =(size_t)entry->GetOffset(); + Fp->seekg((size_t)entry->GetOffset(), std::ios::beg); // FIXME : for each DataEntry ! - // Find out if the tag we received is in the dictionaries: - Dict *pubDict = Global::GetDicts()->GetDefaultPubDict(); - DictEntry* dictEntry = pubDict->GetDictEntryByNumber(group, elem); - if (!dictEntry) - { - currentEntry = NewDocEntryByNumber(group, elem, vr); - } - else - { - currentEntry = NewDocEntryByNumber(group, elem); - } + size_t l = entry->GetLength(); + uint8_t *data = new uint8_t[l]; + if ( !data ) + { + gdcmWarningMacro( "Cannot allocate DataEntry content for : " + << std::hex << entry->GetGroup() + << "|" << entry->GetElement() ); + return; + } + + // Read the data + Fp->read((char*)data, l); + if ( Fp->fail() || Fp->eof() ) + { + delete[] data; + entry->SetState(DataEntry::STATE_UNREAD); + return; + } + + // Swap the data content if necessary + uint32_t i; + unsigned short vrLgth = + Global::GetVR()->GetAtomicElementLength(entry->GetVR()); - if (!currentEntry) +// FIXME : trouble expected if we read an ... OW Entry (LUT, etc ..) +// if( entry->GetVR() == "OW" ) +// vrLgth = 1; + + switch(vrLgth) + { + case 1: + { + break; + } + case 2: { - dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to" - " NewDocEntryByNumber failed."); - return NULL; + uint16_t *data16 = (uint16_t *)data; + for(i=0;i(currentEntry); - if ( !valEntry ) // Euuuuh? It wasn't a ValEntry - // then we change it to a ValEntry ? - // Shouldn't it be considered as an error ? + case 8: { - // We need to promote the DocEntry to a ValEntry: - valEntry = new ValEntry(currentEntry); - if (!RemoveEntry(currentEntry)) - { - dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal" - " of previous DocEntry failed."); - return NULL; - } - if ( !AddEntry(valEntry)) - { - dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding" - " promoted ValEntry failed."); - return NULL; - } + double *data64 = (double *)data; + for(i=0;iSetBinArea(data); - SetEntryByNumber(value, group, elem); + if ( openFile ) // The file is left in the state (open/close) it was at entrance + CloseFile(); +} - return valEntry; -} +/** + * \brief Loads the element while preserving the current + * underlying file position indicator as opposed to + * 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(); // LoadDocEntrySafe is not used +// LoadDocEntry(entry); +// Fp->seekg(PositionOnEntry, std::ios::beg); // LoadDocEntrySafe is not used +// } +//} -/* - * \brief Modifies the value of a given Header Entry (Dicom Element) - * when it exists. Create it with the given value when unexistant. - * @param binArea (binary) value to be set - * @param Group Group number of the Entry - * @param Elem Element number of the Entry - * \return pointer to the modified/created Header Entry (NULL when creation - * failed). +/** + * \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 + * @return true if 'smaller' */ -BinEntry* Document::ReplaceOrCreateByNumber( - uint8_t* binArea, - int lgth, - uint16_t group, - uint16_t elem, - TagName const & vr ) +bool Document::operator<(Document &document) { - BinEntry* binEntry = 0; - DocEntry* currentEntry = GetDocEntryByNumber( group, elem); - if (!currentEntry) + // Patient Name + std::string s1 = GetEntryString(0x0010,0x0010); + std::string s2 = document.GetEntryString(0x0010,0x0010); + if (s1 < s2) { - - // check if (group,element) DictEntry exists - // if it doesn't, create an entry in DictSet::VirtualEntry - // and use it - - // Find out if the tag we received is in the dictionaries: - Dict *pubDict = Global::GetDicts()->GetDefaultPubDict(); - DictEntry *dictEntry = pubDict->GetDictEntryByNumber(group, elem); - - if (!dictEntry) - { - currentEntry = NewDocEntryByNumber(group, elem, vr); - } - else - { - currentEntry = NewDocEntryByNumber(group, elem); - } - if (!currentEntry) + return true; + } + else if ( s1 > s2 ) + { + return false; + } + else + { + // Patient ID + s1 = GetEntryString(0x0010,0x0020); + s2 = document.GetEntryString(0x0010,0x0020); + if ( s1 < s2 ) { - dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to" - " NewDocEntryByNumber failed."); - return NULL; + return true; } - binEntry = new BinEntry(currentEntry); - if ( !AddEntry(binEntry)) + else if ( s1 > s2 ) { - dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry" - " failed allthough this is a creation."); + return false; } - } - else - { - binEntry = dynamic_cast< BinEntry* >(currentEntry); - if ( !binEntry ) // Euuuuh? It wasn't a BinEntry - // then we change it to a BinEntry ? - // Shouldn't it be considered as an error ? + else { - // We need to promote the DocEntry to a BinEntry: - binEntry = new BinEntry(currentEntry); - if (!RemoveEntry(currentEntry)) + // Study Instance UID + s1 = GetEntryString(0x0020,0x000d); + s2 = document.GetEntryString(0x0020,0x000d); + if ( s1 < s2 ) + { + return true; + } + else if ( s1 > s2 ) { - dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal" - " of previous DocEntry failed."); - return NULL; + return false; } - if ( !AddEntry(binEntry)) + else { - dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding" - " promoted BinEntry failed."); - return NULL; + // Serie Instance UID + s1 = GetEntryString(0x0020,0x000e); + s2 = document.GetEntryString(0x0020,0x000e); + if ( s1 < s2 ) + { + return true; + } + else if ( s1 > s2 ) + { + return false; + } } } } - - SetEntryByNumber(binArea, lgth, group, elem); - - return binEntry; -} - - -/* - * \brief Modifies the value of a given Header Entry (Dicom Element) - * when it exists. Create it when unexistant. - * @param Group Group number of the Entry - * @param Elem Element number of the Entry - * \return pointer to the modified/created SeqEntry (NULL when creation - * failed). - */ -SeqEntry* Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem) -{ - SeqEntry* b = 0; - DocEntry* a = GetDocEntryByNumber( group, elem); - if (!a) - { - a = NewSeqEntryByNumber(group, elem); - if (!a) - { - return 0; - } - - b = new SeqEntry(a, 1); // FIXME : 1 (Depth) - AddEntry(b); - } - return b; -} - -/** - * \brief Set a new value if the invoked element exists - * Seems to be useless !!! - * @param value new element value - * @param group group number of the Entry - * @param elem element number of the Entry - * \return boolean - */ -bool Document::ReplaceIfExistByNumber(std::string const & value, - uint16_t group, uint16_t elem ) -{ - SetEntryByNumber(value, group, elem); - - return true; -} + return false; +} //----------------------------------------------------------------------------- // Protected /** - * \brief Checks if a given Dicom Element exists within the H table - * @param group Group number of the searched Dicom Element - * @param element Element number of the searched Dicom Element - * @return true is found + * \brief Reads a given length of bytes + * (in order to avoid to many CPU time-consuming fread-s) + * @param l length to read */ -bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element ) -{ - const std::string &key = DictEntry::TranslateToKey(group, element ); - return TagHT.count(key); -} - -/** - * \brief Searches within Header Entries (Dicom Elements) parsed with - * the public and private dictionaries - * for the element value of a given tag. - * \warning Don't use any longer : use GetPubEntryByName - * @param tagName name of the searched element. - * @return Corresponding element value when it exists, - * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise. - */ -std::string Document::GetEntryByName(TagName const & tagName) -{ - DictEntry* dictEntry = RefPubDict->GetDictEntryByName(tagName); - if( !dictEntry ) +void Document::ReadBegBuffer(size_t l) + throw( FormatError ) +{ + Fp->read (BegBuffer, (size_t)l); + if ( Fp->fail() ) { - return GDCM_UNFOUND; + throw FormatError( "Document::ReadBegBuffer()", " file error." ); } - - return GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement()); -} - -/** - * \brief Searches within Header Entries (Dicom Elements) parsed with - * the public and private dictionaries - * for the element value representation of a given tag. - * - * Obtaining the VR (Value Representation) might be needed by caller - * to convert the string typed content to caller's native type - * (think of C++ vs Python). The VR is actually of a higher level - * of semantics than just the native C++ type. - * @param tagName name of the searched element. - * @return Corresponding element value representation when it exists, - * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise. - */ -std::string Document::GetEntryVRByName(TagName const& tagName) -{ - DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName); - if( dictEntry == NULL) + if ( Fp->eof() ) { - return GDCM_UNFOUND; + throw FormatError( "Document::ReadBegBuffer()", "EOF." ); } - - DocEntry* elem = GetDocEntryByNumber(dictEntry->GetGroup(), - dictEntry->GetElement()); - return elem->GetVR(); + PtrBegBuffer = BegBuffer; + CurrentOffsetPosition+=l; } - /** - * \brief Searches within Header Entries (Dicom Elements) parsed with - * the public and private dictionaries - * for the element value representation of a given tag. - * @param group Group number of the searched tag. - * @param element Element number of the searched tag. - * @return Corresponding element value representation when it exists, - * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise. + * \brief Reads a supposed to be 16 Bits integer + * (swaps it depending on processor endianness) + * @return read value */ -std::string Document::GetEntryByNumber(uint16_t group, uint16_t element) +uint16_t Document::ReadInt16() + throw( FormatError ) { - TagKey key = DictEntry::TranslateToKey(group, element); - /// \todo use map methods, instead of multimap JPR - if ( !TagHT.count(key)) + uint16_t g; + Fp->read ((char*)&g, (size_t)2); + if ( Fp->fail() ) { - return GDCM_UNFOUND; + throw FormatError( "Document::ReadInt16()", " file error." ); } - - return ((ValEntry *)TagHT.find(key)->second)->GetValue(); -} - -/** - * \brief Searches within Header Entries (Dicom Elements) parsed with - * the public and private dictionaries - * for the element value representation of a given tag.. - * - * Obtaining the VR (Value Representation) might be needed by caller - * to convert the string typed content to caller's native type - * (think of C++ vs Python). The VR is actually of a higher level - * of semantics than just the native C++ type. - * @param group Group number of the searched tag. - * @param element Element number of the searched tag. - * @return Corresponding element value representation when it exists, - * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise. - */ -std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element) -{ - DocEntry* elem = GetDocEntryByNumber(group, element); - if ( !elem ) + if ( Fp->eof() ) { - return GDCM_UNFOUND; + throw FormatError( "Document::ReadInt16()", "EOF." ); } - return elem->GetVR(); + g = SwapShort(g); + return g; } /** - * \brief Searches within Header Entries (Dicom Elements) parsed with - * the public and private dictionaries - * for the value length of a given tag.. - * @param group Group number of the searched tag. - * @param element Element number of the searched tag. - * @return Corresponding element length; -2 if not found + * \brief Gets from BegBuffer a supposed to be 16 Bits integer + * (swaps it depending on processor endianness) + * @return read value */ -int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element) +uint16_t Document::GetInt16() { - DocEntry* elem = GetDocEntryByNumber(group, element); - if ( !elem ) - { - return -2; //magic number - } - return elem->GetLength(); + uint16_t g = *((uint16_t*)PtrBegBuffer); + g = SwapShort(g); + PtrBegBuffer+=2; + return g; } /** - * \brief Sets the value (string) of the Header Entry (Dicom Element) - * @param content string value of the Dicom Element - * @param tagName name of the searched Dicom Element. - * @return true when found + * \brief Reads a supposed to be 32 Bits integer + * (swaps it depending on processor endianness) + * @return read value */ -bool Document::SetEntryByName(std::string const & content, - TagName const & tagName) +uint32_t Document::ReadInt32() + throw( FormatError ) { - DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName); - if( !dictEntry ) + uint32_t g; + Fp->read ((char*)&g, (size_t)4); + if ( Fp->fail() ) { - return false; + throw FormatError( "Document::ReadInt32()", " file error." ); } - - return SetEntryByNumber(content,dictEntry->GetGroup(), - dictEntry->GetElement()); + if ( Fp->eof() ) + { + throw FormatError( "Document::ReadInt32()", "EOF." ); + } + g = SwapLong(g); + return g; } /** - * \brief Accesses an existing DocEntry (i.e. a Dicom Element) - * through it's (group, element) and modifies it's content with - * the given value. - * @param content new value (string) to substitute with - * @param group group number of the Dicom Element to modify - * @param element element number of the Dicom Element to modify + * \brief Gets from BegBuffer a supposed to be 32 Bits integer + * (swaps it depending on processor endianness) + * @return read value */ -bool Document::SetEntryByNumber(std::string const& content, - uint16_t group, uint16_t element) +uint32_t Document::GetInt32() { - int c; - int l; - - ValEntry* valEntry = GetValEntryByNumber(group, element); - if (!valEntry ) - { - dbg.Verbose(0, "Document::SetEntryByNumber: no corresponding", - " ValEntry (try promotion first)."); - return false; - } - // Non even content must be padded with a space (020H)... - std::string finalContent = content; - if( finalContent.length() % 2 ) - { - finalContent += '\0'; // ... therefore we padd with (000H) .!?! - } - valEntry->SetValue(finalContent); - - // Integers have a special treatement for their length: - - l = finalContent.length(); - if ( l != 0) // To avoid to be cheated by 'zero length' integers - { - VRKey vr = valEntry->GetVR(); - if( vr == "US" || vr == "SS" ) - { - // for multivaluated items - c = Util::CountSubstring(content, "\\") + 1; - l = c*2; - } - else if( vr == "UL" || vr == "SL" ) - { - // for multivaluated items - c = Util::CountSubstring(content, "\\") + 1; - l = c*4;; - } - } - valEntry->SetLength(l); - return true; -} + uint32_t g = *((uint32_t*)PtrBegBuffer); + g = SwapLong(g); + PtrBegBuffer+=4; + return g; +} /** - * \brief Accesses an existing DocEntry (i.e. a Dicom Element) - * through it's (group, element) and modifies it's content with - * the given value. - * @param content new value (void* -> uint8_t*) to substitute with - * @param lgth new value length - * @param group group number of the Dicom Element to modify - * @param element element number of the Dicom Element to modify + * \brief Re-computes the length of the Dicom group 0002. */ -bool Document::SetEntryByNumber(uint8_t*content, int lgth, - uint16_t group, uint16_t element) +int Document::ComputeGroup0002Length( ) { - (void)lgth; //not used - TagKey key = DictEntry::TranslateToKey(group, element); - if ( !TagHT.count(key)) + uint16_t gr; + VRKey vr; + + int groupLength = 0; + bool found0002 = false; + + // for each zero-level Tag in the DCM Header + DocEntry *entry = GetFirstEntry(); + while( entry ) { - return false; - } + gr = entry->GetGroup(); -/* Hope Binary field length is *never* wrong - if(lgth%2) // Non even length are padded with a space (020H). - { - lgth++; - //content = content + '\0'; // fing a trick to enlarge a binary field? - } -*/ - BinEntry* a = (BinEntry *)TagHT[key]; - a->SetBinArea(content); - a->SetLength(lgth); - a->SetValue(GDCM_BINLOADED); + if ( gr == 0x0002 ) + { + found0002 = true; - return true; -} + if ( entry->GetElement() != 0x0000 ) + { + vr = entry->GetVR(); + + //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, OL, SQ, UT, UN) : 4 more bytes + groupLength += 4; + } + groupLength += 2 + 2 + 4 + entry->GetLength(); + } + } + else if (found0002 ) + break; -/** - * \brief Accesses an existing DocEntry (i.e. a Dicom Element) - * in the PubDocEntrySet of this instance - * through it's (group, element) and modifies it's length with - * the given value. - * \warning Use with extreme caution. - * @param l new length to substitute with - * @param group group number of the Entry to modify - * @param element element number of the Entry to modify - * @return true on success, false otherwise. - */ -bool Document::SetEntryLengthByNumber(uint32_t l, - uint16_t group, uint16_t element) -{ - /// \todo use map methods, instead of multimap JPR - TagKey key = DictEntry::TranslateToKey(group, element); - if ( !TagHT.count(key) ) - { - return false; - } - if ( l % 2 ) - { - l++; // length must be even + entry = GetNextEntry(); } - ( ((TagHT.equal_range(key)).first)->second )->SetLength(l); - - return true ; + return groupLength; } /** - * \brief Gets (from Header) the offset of a 'non string' element value - * (LoadElementValues has already be executed) - * @param group group number of the Entry - * @param elem element number of the Entry - * @return File Offset of the Element Value + * \brief CallStartMethod */ -size_t Document::GetEntryOffsetByNumber(uint16_t group, uint16_t elem) +void Document::CallStartMethod() { - DocEntry* entry = GetDocEntryByNumber(group, elem); - if (!entry) - { - dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry present."); - return 0; - } - return entry->GetOffset(); + Progress = 0.0f; + Abort = false; + CommandManager::ExecuteCommand(this,CMD_STARTPROGRESS); } /** - * \brief Gets (from Header) a 'non string' element value - * (LoadElementValues has already be executed) - * @param group group number of the Entry - * @param elem element number of the Entry - * @return Pointer to the 'non string' area + * \brief CallProgressMethod */ -void* Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem) +void Document::CallProgressMethod() { - DocEntry* entry = GetDocEntryByNumber(group, elem); - if (!entry) - { - dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry"); - return 0; - } - if ( BinEntry* binEntry = dynamic_cast(entry) ) - { - return binEntry->GetBinArea(); - } - - return 0; + CommandManager::ExecuteCommand(this,CMD_PROGRESS); } /** - * \brief Loads (from disk) the element content - * when a string is not suitable - * @param group group number of the Entry - * @param elem element number of the Entry - */ -void* Document::LoadEntryBinArea(uint16_t group, uint16_t elem) -{ - DocEntry *docElement = GetDocEntryByNumber(group, elem); - if ( !docElement ) - { - return NULL; - } - size_t o =(size_t)docElement->GetOffset(); - Fp->seekg( o, std::ios_base::beg); - size_t l = docElement->GetLength(); - uint8_t* a = new uint8_t[l]; - if(!a) - { - dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a"); - return NULL; - } - Fp->read((char*)a, l); - if( Fp->fail() || Fp->eof() )//Fp->gcount() == 1 - { - delete[] a; - return NULL; - } - /// \todo Drop any already existing void area! JPR - if( !SetEntryBinAreaByNumber( a, group, elem ) ) - { - dbg.Verbose(0, "Document::LoadEntryBinArea setting failed."); - } - return a; -} -/** - * \brief Loads (from disk) the element content - * when a string is not suitable - * @param element Entry whose binArea is going to be loaded + * \brief CallEndMethod */ -void* Document::LoadEntryBinArea(BinEntry* element) +void Document::CallEndMethod() { - size_t o =(size_t)element->GetOffset(); - Fp->seekg(o, std::ios_base::beg); - size_t l = element->GetLength(); - uint8_t* a = new uint8_t[l]; - if( !a ) - { - dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a"); - return NULL; - } - element->SetBinArea((uint8_t*)a); - /// \todo check the result - Fp->read((char*)a, l); - if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1 - { - delete[] a; - return NULL; - } - - return a; + Progress = 1.0f; + CommandManager::ExecuteCommand(this,CMD_ENDPROGRESS); } +//----------------------------------------------------------------------------- +// Private /** - * \brief Sets a 'non string' value to a given Dicom Element - * @param area area containing the 'non string' value - * @param group Group number of the searched Dicom Element - * @param element Element number of the searched Dicom Element - * @return + * \brief Loads all the needed Dictionaries + * \warning NOT end user intended method ! */ -bool Document::SetEntryBinAreaByNumber(uint8_t* area, - uint16_t group, uint16_t element) +void Document::Initialize() { - DocEntry* currentEntry = GetDocEntryByNumber(group, element); - if ( !currentEntry ) - { - return false; - } - if ( BinEntry* binEntry = dynamic_cast(currentEntry) ) - { - binEntry->SetBinArea( area ); - return true; - } - return true; + RefPubDict = Global::GetDicts()->GetDefaultPubDict(); + RefShaDict = NULL; + Filetype = Unknown; } /** - * \brief Update the entries with the shadow dictionary. - * Only non even entries are analyzed - */ -void Document::UpdateShaEntries() + * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries) + * @param set DocEntrySet we are going to parse ('zero level' or a SQItem) + * @param offset start of parsing + * @param l_max length to parse (meaningless when we are in 'delimitor mode') + * @param delim_mode : whether we are in 'delimitor mode' (l=0xffffff) or not + */ +void Document::ParseDES(DocEntrySet *set, long offset, + long l_max, bool delim_mode) { - //DictEntry *entry; - std::string vr; + DocEntry *newDocEntry; + DataEntry *newDataEntry; + SeqEntry *newSeqEntry; + //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 << "0x(" << offset << ")" ); + while (true) + { - /// \todo TODO : still any use to explore recursively the whole structure? -/* - for(ListTag::iterator it=listEntries.begin(); - it!=listEntries.end(); - ++it) - { - // Odd group => from public dictionary - if((*it)->GetGroup()%2==0) - continue; - - // Peer group => search the corresponding dict entry - if(RefShaDict) - entry=RefShaDict->GetDictEntryByNumber((*it)->GetGroup(),(*it)->GetElement()); - else - entry=NULL; + ///\todo FIXME : On 64 bits processors, tellg gives unexpected results after a while ? + /// Probabely a bug in gdcm code somewhere (some memory erased ?) - if((*it)->IsImplicitVR()) - vr="Implicit"; - else - vr=(*it)->GetVR(); +// 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; + } - (*it)->SetValue(GetDocEntryUnvalue(*it)); // to go on compiling - if(entry){ - // Set the new entry and the new value - (*it)->SetDictEntry(entry); - CheckDocEntryVR(*it,vr); + newDocEntry = ReadNextDocEntry( ); - (*it)->SetValue(GetDocEntryValue(*it)); // to go on compiling - - } - else + if ( !newDocEntry ) { - // Remove precedent value transformation - (*it)->SetDictEntry(NewVirtualDictEntry((*it)->GetGroup(),(*it)->GetElement(),vr)); + break; } - } -*/ -} - -/** - * \brief Searches within the Header Entries for a Dicom Element of - * a given tag. - * @param tagName name of the searched Dicom Element. - * @return Corresponding Dicom Element when it exists, and NULL - * otherwise. - */ -DocEntry* Document::GetDocEntryByName(TagName const & tagName) -{ - DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName); - if( !dictEntry ) - { - return NULL; - } - - return GetDocEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement()); -} - -/** - * \brief retrieves a Dicom Element (the first one) using (group, element) - * \warning (group, element) IS NOT an identifier inside the Dicom Header - * if you think it's NOT UNIQUE, check the count number - * and use iterators to retrieve ALL the Dicoms Elements within - * a given couple (group, element) - * @param group Group number of the searched Dicom Element - * @param element Element number of the searched Dicom Element - * @return - */ -DocEntry* Document::GetDocEntryByNumber(uint16_t group, uint16_t element) -{ - TagKey key = DictEntry::TranslateToKey(group, element); - if ( !TagHT.count(key)) - { - return NULL; - } - return TagHT.find(key)->second; -} - -/** - * \brief Same as \ref Document::GetDocEntryByNumber except it only - * returns a result when the corresponding entry is of type - * ValEntry. - * @return When present, the corresponding ValEntry. - */ -ValEntry* Document::GetValEntryByNumber(uint16_t group, uint16_t element) -{ - DocEntry* currentEntry = GetDocEntryByNumber(group, element); - if ( !currentEntry ) - { - return 0; - } - if ( ValEntry* valEntry = dynamic_cast(currentEntry) ) - { - return valEntry; - } - dbg.Verbose(0, "Document::GetValEntryByNumber: unfound ValEntry."); - - return 0; -} - -/** - * \brief Loads the element while preserving the current - * underlying file position indicator as opposed to - * to LoadDocEntry that modifies it. - * @param entry Header Entry whose value shall be loaded. - * @return - */ -void Document::LoadDocEntrySafe(DocEntry * entry) -{ - long PositionOnEntry = Fp->tellg(); - LoadDocEntry(entry); - Fp->seekg(PositionOnEntry, std::ios_base::beg); -} - -/** - * \brief Swaps back the bytes of 4-byte long integer accordingly to - * processor order. - * @return The properly swaped 32 bits integer. - */ -uint32_t Document::SwapLong(uint32_t a) -{ - switch (SwapCode) - { - case 0 : - break; - case 4321 : - a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) | - ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) ); - break; - - case 3412 : - a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) ); - break; - - case 2143 : - a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) ); - break; - default : - //std::cout << "swapCode= " << SwapCode << std::endl; - dbg.Error(" Document::SwapLong : unset swap code"); - a = 0; - } - return a; -} - -/** - * \brief Unswaps back the bytes of 4-byte long integer accordingly to - * processor order. - * @return The properly unswaped 32 bits integer. - */ -uint32_t Document::UnswapLong(uint32_t a) -{ - return SwapLong(a); -} - -/** - * \brief Swaps the bytes so they agree with the processor order - * @return The properly swaped 16 bits integer. - */ -uint16_t Document::SwapShort(uint16_t a) -{ - if ( SwapCode == 4321 || SwapCode == 2143 ) - { - a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) ); - } - return a; -} - -/** - * \brief Unswaps the bytes so they agree with the processor order - * @return The properly unswaped 16 bits integer. - */ -uint16_t Document::UnswapShort(uint16_t a) -{ - return SwapShort(a); -} - -//----------------------------------------------------------------------------- -// Private + + // Uncoment this cerr line to be able to 'follow' the DocEntries + // when something *very* strange happens + if( Debug::GetDebugFlag() ) + std::cerr<GetKey()<<" "<GetVR()<tellg()-offset) >= l_max) + // an Item Starter found elsewhere but in the first position + // of a SeqEntry means previous entry was a Sequence + // but we didn't get it (private Sequence + Implicit VR) + // we have to backtrack. + if ( !first && newDocEntry->IsItemStarter() ) { - break; + // Debug message within the method ! + newDocEntry = Backtrack(newDocEntry, set); } - newDocEntry = ReadNextDocEntry( ); - if ( !newDocEntry ) - { - break; + else + { + PreviousDocEntry = newDocEntry; } + + used = true; + newDataEntry = dynamic_cast(newDocEntry); - VRKey vr = newDocEntry->GetVR(); - if ( vr != "SQ" ) + if ( newDataEntry ) { - - if ( Global::GetVR()->IsVROfGdcmStringRepresentable(vr) ) + //////////////////////////// DataEntry + + //vr = newDocEntry->GetVR(); // useless ? + + if ( !set->AddEntry( newDataEntry ) ) { - /////////////////////// ValEntry - ValEntry* newValEntry = - new ValEntry( newDocEntry->GetDictEntry() ); - newValEntry->Copy( newDocEntry ); - - // 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 (Document* dummy = dynamic_cast< Document* > ( set ) ) - { - (void)dummy; - 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): - if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) ) - { - newValEntry->SetKey( parentSQItem->GetBaseTagKey() - + newValEntry->GetKey() ); - } - - set->AddEntry( newValEntry ); - LoadDocEntry( newValEntry ); - if (newValEntry->IsItemDelimitor()) - { - break; - } - if ( !delim_mode && (Fp->tellg()-offset) >= l_max) - { - break; - } + gdcmDebugMacro( "in ParseDES : cannot add a DataEntry " + << newDataEntry->GetKey() + << " (at offset : 0x(" + << newDataEntry->GetOffset() << ") )" ); + used=false; } else { - if ( ! Global::GetVR()->IsVROfGdcmBinaryRepresentable(vr) ) - { - ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR - dbg.Verbose(0, "Document::ParseDES: neither Valentry, " - "nor BinEntry. Probably unknown VR."); - } - - //////////////////// BinEntry or UNKOWN VR: - BinEntry* newBinEntry = - new BinEntry( newDocEntry->GetDictEntry() ); - newBinEntry->Copy( newDocEntry ); - - // When "this" is a Document the Key is simply of the - // form ( group, elem )... - if (Document* dummy = dynamic_cast< Document* > ( set ) ) - { - (void)dummy; - newBinEntry->SetKey( newBinEntry->GetKey() ); - } - // but when "this" is a SQItem, we are inserting this new - // valEntry in a sequence item, and the kay has the - // generalized form (refer to \ref BaseTagKey): - if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) ) + 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 { - newBinEntry->SetKey( parentSQItem->GetBaseTagKey() - + newBinEntry->GetKey() ); + if ( LoadMode & LD_NOSHADOW ) // if user asked to skip shad.gr + { + std::string strLgrGroup = newDataEntry->GetString(); + + int lgrGroup; + //if ( newDataEntry->IsUnfound() ) /?!? JPR + { + lgrGroup = atoi(strLgrGroup.c_str()); + Fp->seekg(lgrGroup, std::ios::cur); // Once per Shadow group, when NOSHADOW + RemoveEntry( newDocEntry ); // Remove and delete + continue; + } + } } - - set->AddEntry( newBinEntry ); - LoadDocEntry( newBinEntry ); } - if ( ( newDocEntry->GetGroup() == 0x7fe0 ) - && ( newDocEntry->GetElement() == 0x0010 ) ) + 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; + + // '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 ) { - TransferSyntaxType ts = GetTransferSyntax(); - if ( ts == RLELossless ) - { - long PositionOnEntry = Fp->tellg(); - Fp->seekg( newDocEntry->GetOffset(), std::ios_base::beg ); - ComputeRLEInfo(); - Fp->seekg( PositionOnEntry, std::ios_base::beg ); - } - else if ( IsJPEG() ) - { - long PositionOnEntry = Fp->tellg(); - Fp->seekg( newDocEntry->GetOffset(), std::ios_base::beg ); - ComputeJPEGFragmentInfo(); - Fp->seekg( PositionOnEntry, std::ios_base::beg ); - } + 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 { - // VR = "SQ" - unsigned long l = newDocEntry->GetReadLength(); - if ( l != 0 ) // don't mess the delim_mode for zero-length sequence + /////////////////////// SeqEntry : VR = "SQ" + + unsigned long l = newDocEntry->GetReadLength(); + if ( l != 0 ) // don't mess the delim_mode for 'zero-length sequence' { if ( l == 0xffffffff ) { - delim_mode = true; + delim_mode_intern = true; } else { - delim_mode = false; + delim_mode_intern = false; } } - // no other way to create it ... - SeqEntry* newSeqEntry = - new SeqEntry( newDocEntry->GetDictEntry() ); - newSeqEntry->Copy( newDocEntry ); - newSeqEntry->SetDelimitorMode( delim_mode ); + + if ( (LoadMode & LD_NOSHADOWSEQ) && ! delim_mode_intern ) + { + // User asked to skip SeQuences *only* if they belong to Shadow Group + if ( newDocEntry->GetGroup()%2 != 0 ) + { + 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); // Once per SQ, when NOSEQ + newDocEntry->Delete(); // Delete, not in the set + continue; + } + // delay the dynamic cast as late as possible + newSeqEntry = dynamic_cast(newDocEntry); + + // no other way to create the Delimitor ... + newSeqEntry->SetDelimitorMode( delim_mode_intern ); // At the top of the hierarchy, stands a Document. When "set" // is a Document, then we are building the first depth level. // Hence the SeqEntry we are building simply has a depth // level of one: - if (Document* dummy = dynamic_cast< Document* > ( set ) ) + if ( set == this ) // ( dynamic_cast< Document* > ( set ) ) { - (void)dummy; newSeqEntry->SetDepthLevel( 1 ); - newSeqEntry->SetKey( newSeqEntry->GetKey() ); } - // But when "set" is allready a SQItem, we are building a nested + // But when "set" is already a SQItem, we are building a nested // sequence, and hence the depth level of the new SeqEntry // we are building, is one level deeper: - if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) ) + + // time waste hunting + 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 - ParseSQ( newSeqEntry, - newDocEntry->GetOffset(), - l, delim_mode); + + gdcmDebugMacro( "Entry in ParseSQ, delim " << delim_mode_intern + << " at offset 0x(" << std::hex + << newDocEntry->GetOffset() << ")"); + + bool res = ParseSQ( newSeqEntry, + newDocEntry->GetOffset(), + l, 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 : 0x(" + << newSeqEntry->GetOffset() << ") )" ); + used = false; } - set->AddEntry( newSeqEntry ); - if ( !delim_mode && (Fp->tellg()-offset) >= l_max) + else + { + newDocEntry->Delete(); + } + + // if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) // Once per SeqEntry + + 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(); break; } + } // end SeqEntry : VR = "SQ" + + if ( !used ) + { + newDocEntry->Delete(); } - delete newDocEntry; - } + 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; bool dlm_mod; + long offsetStartCurrentSQItem = offset; while (true) { - DocEntry* newDocEntry = ReadNextDocEntry(); + // the first time, we read the fff0,e000 of the first SQItem + DocEntry *newDocEntry = ReadNextDocEntry(); + if ( !newDocEntry ) - { - // FIXME Should warn user - break; + { + // The most frequent is when a SQ terminator is missing (?!?) + gdcmWarningMacro("in ParseSQ : should never get here!"); + UnexpectedEOF = true; + return false; } - if( delim_mode ) + if ( delim_mode ) { if ( newDocEntry->IsSequenceDelimitor() ) { - seqEntry->SetSequenceDelimitationItem( newDocEntry ); + seqEntry->SetDelimitationItem( newDocEntry ); + newDocEntry->Delete(); break; } } - if ( !delim_mode && (Fp->tellg()-offset) >= l_max) + else // ! delim_mode { - break; + if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per SQItem when no delim mode + { + newDocEntry->Delete(); + break; + } } - - SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() ); - std::ostringstream newBase; - newBase << seqEntry->GetKey() - << "/" - << SQItemNumber - << "#"; - itemSQ->SetBaseTagKey( newBase.str() ); + // create the current SQItem + SQItem *itemSQ = SQItem::New( seqEntry->GetDepthLevel() ); unsigned int l = newDocEntry->GetReadLength(); if ( l == 0xffffffff ) @@ -1510,208 +1388,182 @@ void Document::ParseSQ( SeqEntry* seqEntry, { dlm_mod = false; } - - ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod); - - seqEntry->AddEntry( itemSQ, SQItemNumber ); + + // avoid infinite loop when Bad assumption was made on illegal 'unknown length' UN //JPRx + + if (offsetStartCurrentSQItem <= OffsetOfPreviousParseDES) + { + gdcmWarningMacro("Bad assumption was made on illegal 'unknown length' UN!" << std::endl << + "OffsetOfPreviousParseDES " << std::hex << OffsetOfPreviousParseDES + << " offsetStartCurrentSQItem " << offsetStartCurrentSQItem); + /// \todo when "Bad assumption (SQ) on illegal 'unknown length' UN", Backtrack again + try OB + return false; + } + else + { + OffsetOfPreviousParseDES = offsetStartCurrentSQItem; + } + + // fill up the current SQItem, starting at the beginning of fff0,e000 + Fp->seekg(offsetStartCurrentSQItem, std::ios::beg); // Once per SQItem + ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod); + offsetStartCurrentSQItem = Fp->tellg(); // Once per SQItem + + seqEntry->AddSQItem( itemSQ, SQItemNumber ); + itemSQ->Delete(); + newDocEntry->Delete(); SQItemNumber++; - if ( !delim_mode && ( 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 set DocEntrySet (ElementSet/SQItem) the DocEntry will belong + */ +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 + docEntry->Delete(); + + // Get all info we can from PreviousDocEntry + uint16_t group = PreviousDocEntry->GetGroup(); + uint16_t elem = PreviousDocEntry->GetElement(); + uint32_t lgt = PreviousDocEntry->GetLength(); + long offset = PreviousDocEntry->GetOffset(); + + gdcmDebugMacro( "Backtrack :" << std::hex << group + << "|" << elem + << " at offset 0x(" <RemoveEntry( PreviousDocEntry ); + + // forge the Seq Entry + DocEntry *newEntry = NewSeqEntry(group, elem); + newEntry->SetLength(lgt); + newEntry->SetOffset(offset); + + // Move back to the beginning of the Sequence + + Fp->seekg(offset, std::ios::beg); // Only for Shadow Implicit VR SQ + return newEntry; // It will added where it has to be! } /** - * \brief Loads the element content if its length doesn't exceed - * the value specified with Document::SetMaxSizeLoadEntry() - * @param entry Header Entry (Dicom Element) to be dealt with + * \brief Loads (or not) the element content depending if its length exceeds + * or not the value specified with Document::SetMaxSizeLoadEntry() + * @param entry Header Entry (Dicom Element) to be dealt with + * @param forceLoad whether you want to force loading of 'long' elements */ -void Document::LoadDocEntry(DocEntry* entry) +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_base::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 of the offset of - // the element content and it's length. - - std::ostringstream s; - if (length > MaxSizeLoadEntry) - { - if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) ) - { - //s << "gdcm::NotLoaded (BinEntry)"; - s << GDCM_NOTLOADED; - s << " Address:" << (long)entry->GetOffset(); - s << " Length:" << entry->GetLength(); - s << " x(" << std::hex << entry->GetLength() << ")"; - binEntryPtr->SetValue(s.str()); - } - // Be carefull : a BinEntry IS_A ValEntry ... - else if (ValEntry* valEntryPtr = dynamic_cast< ValEntry* >(entry) ) - { - // s << "gdcm::NotLoaded. (ValEntry)"; - s << GDCM_NOTLOADED; - s << " Address:" << (long)entry->GetOffset(); - s << " Length:" << entry->GetLength(); - s << " x(" << std::hex << entry->GetLength() << ")"; - valEntryPtr->SetValue(s.str()); - } - else - { - // fusible - std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry " - << "nor a ValEntry ?! Should never print that !" << std::endl; - } + // are not loaded. - // to be sure we are at the end of the value ... - Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),std::ios_base::beg); - return; - } - - // When we find a BinEntry not very much can be done : - if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) ) + if (!forceLoad) { - s << GDCM_BINLOADED; - binEntryPtr->SetValue(s.str()); - LoadEntryBinArea(binEntryPtr); // last one, not to erase length ! - return; - } - - /// \todo Any compacter code suggested (?) - if ( IsDocEntryAnInteger(entry) ) - { - uint32_t NewInt; - int nbInt; - // When short integer(s) are expected, read and convert the following - // n *two 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") + if (length > MaxSizeLoadEntry) { - 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 + dataEntryPtr->SetBinArea(NULL,true); + dataEntryPtr->SetState(DataEntry::STATE_NOTLOADED); - ((ValEntry *)entry)->SetValue(s.str()); - return; - } - - // 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'; - std::string newValue = str; - delete[] str; - - if ( ValEntry* valEntry = dynamic_cast(entry) ) - { - if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1 - { - dbg.Verbose(1, "Document::LoadDocEntry", - "unread element value"); - valEntry->SetValue(GDCM_UNREAD); + // to be sure we are at the end of the value ... + // Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(), + // std::ios::beg); //JPRx return; } - - if( vr == "UI" ) - { - // Because of correspondance with the VR dic - valEntry->SetValue(newValue); - } - else - { - valEntry->SetValue(newValue); - } - } - else - { - dbg.Error(true, "Document::LoadDocEntry" - "Should have a ValEntry, here !"); } + + /// \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 ! } - /** - * \brief Find the value Length of the passed Header Entry + * \brief Find the value Length of the passed Doc Entry * @param entry Header Entry whose length of the value shall be loaded. */ void Document::FindDocEntryLength( DocEntry *entry ) throw ( FormatError ) { - uint16_t element = entry->GetElement(); - 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 == "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_base::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 { - /// \todo rename that to FindDocEntryLengthOBOrOW since - /// the above test is on both OB and OW... - lengthOB = FindDocEntryLengthOB(); + lengthOB = FindDocEntryLengthOBOrOW();// for encapsulation of encoded pixel } catch ( FormatUnexpected ) { @@ -1720,81 +1572,33 @@ void Document::FindDocEntryLength( DocEntry *entry ) // chance to get the pixels by deciding the element goes // until the end of the file. Hence we artificially fix the // the length and proceed. - long currentPosition = Fp->tellg(); - Fp->seekg(0L,std::ios_base::end); - long lengthUntilEOF = Fp->tellg() - currentPosition; - Fp->seekg(currentPosition, std::ios_base::beg); + gdcmWarningMacro( " Computing the length failed for " << + entry->GetKey() <<" in " <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; // 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); return; } + entry->SetReadLength(lengthOB); entry->SetLength(lengthOB); return; } FixDocEntryFoundLength(entry, length32); return; } - // Length is encoded on 2 bytes. - length16 = ReadInt16(); - - // We can tell the current file is encoded in big endian (like - // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag - // and it's value is the one of the encoding of a big endian file. - // In order to deal with such big endian encoded files, we have - // (at least) two strategies: - // * when we load the "Transfer Syntax" tag with value of big endian - // encoding, we raise the proper flags. Then we wait for the end - // of the META group (0x0002) among which is "Transfer Syntax", - // before switching the swap code to big endian. We have to postpone - // the switching of the swap code since the META group is fully encoded - // in little endian, and big endian coding only starts at the next - // group. The corresponding code can be hard to analyse and adds - // many additional unnecessary tests for regular tags. - // * the second strategy consists in waiting for trouble, that shall - // appear when we find the first group with big endian encoding. This - // is easy to detect since the length of a "Group Length" tag (the - // ones with zero as element number) has to be of 4 (0x0004). When we - // encounter 1024 (0x0400) chances are the encoding changed and we - // found a group with big endian encoding. - // We shall use this second strategy. In order to make sure that we - // can interpret the presence of an apparently big endian encoded - // length of a "Group Length" without committing a big mistake, we - // add an additional check: we look in the already parsed elements - // for the presence of a "Transfer Syntax" whose value has to be "big - // endian encoding". When this is the case, chances are we have got our - // hands on a big endian encoded file: we switch the swap code to - // big endian and proceed... - if ( element == 0x0000 && length16 == 0x0400 ) - { - TransferSyntaxType ts = GetTransferSyntax(); - if ( ts != ExplicitVRBigEndian ) - { - throw FormatError( "Document::FindDocEntryLength()", - " not explicit VR." ); - return; - } - length16 = 4; - SwitchSwapToBigEndian(); - // Restore the unproperly loaded values i.e. the group, the element - // and the dictionary entry depending on them. - uint16_t correctGroup = SwapShort( entry->GetGroup() ); - uint16_t correctElem = SwapShort( entry->GetElement() ); - DictEntry* newTag = GetDictEntryByNumber( correctGroup, - correctElem ); - if ( !newTag ) - { - // This correct tag is not in the dictionary. Create a new one. - newTag = NewVirtualDictEntry(correctGroup, correctElem); - } - // FIXME this can create a memory leaks on the old entry that be - // left unreferenced. - entry->SetDictEntry( newTag ); - } - - // Heuristic: well, some files are really ill-formed. + //length16 = ReadInt16(); + length16 = GetInt16(); + // 0xffff means that we deal with 'No Length' Sequence + // or 'No Length' SQItem if ( length16 == 0xffff) - { - // 0xffff means that we deal with 'Unknown Length' Sequence + { length16 = 0; } FixDocEntryFoundLength( entry, (uint32_t)length16 ); @@ -1808,270 +1612,150 @@ void Document::FindDocEntryLength( DocEntry *entry ) // on Data elements "Implicit and Explicit VR Data Elements shall // not coexist in a Data Set and Data Sets nested within it".] // Length is on 4 bytes. - - FixDocEntryFoundLength( entry, ReadInt32() ); + + // 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' + // where Group 0x0002 is *also* encoded in Implicit VR ! + + FixDocEntryFoundLength( entry, GetInt32() /*ReadInt32()*/ ); return; } } /** - * \brief Find the Value Representation of the current Dicom Element. - * @param entry + * \brief Find the Length till the next sequence delimiter + * \warning NOT end user intended method ! + * @return */ -void Document::FindDocEntryVR( DocEntry *entry ) +uint32_t Document::FindDocEntryLengthOBOrOW() + throw( FormatUnexpected ) { - if ( Filetype != ExplicitVR ) - { - return; - } - - char vr[3]; - - long positionOnEntry = Fp->tellg(); - // 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. + // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data. - Fp->read (vr, (size_t)2); - vr[2] = 0; - - if( !CheckDocEntryVR(entry, vr) ) - { - Fp->seekg(positionOnEntry, std::ios_base::beg); - // When this element is known in the dictionary we shall use, e.g. for - // the semantics (see the usage of IsAnInteger), the VR proposed by the - // dictionary entry. Still we have to flag the element as implicit since - // we know now our assumption on expliciteness is not furfilled. - // avoid . - if ( entry->IsVRUnknown() ) - { - entry->SetVR("Implicit"); - } - entry->SetImplicitVR(); - } -} + long positionOnEntry = Fp->tellg(); // Only for OB,OW DataElements -/** - * \brief Check the correspondance between the VR of the header entry - * and the taken VR. If they are different, the header entry is - * updated with the new VR. - * @param entry Header Entry to check - * @param vr Dicom Value Representation - * @return false if the VR is incorrect of if the VR isn't referenced - * otherwise, it returns true -*/ -bool Document::CheckDocEntryVR(DocEntry *entry, VRKey vr) -{ - std::string msg; - bool realExplicit = true; - - // Assume we are reading a falsely explicit VR file i.e. we reached - // a tag where we expect reading a VR but are in fact we read the - // first to bytes of the length. Then we will interogate (through find) - // the dicom_vr dictionary with oddities like "\004\0" which crashes - // both GCC and VC++ implementations of the STL map. Hence when the - // expected VR read happens to be non-ascii characters we consider - // we hit falsely explicit VR tag. - - if ( !isalpha(vr[0]) && !isalpha(vr[1]) ) - { - realExplicit = false; - } - - // CLEANME searching the dicom_vr at each occurence is expensive. - // PostPone this test in an optional integrity check at the end - // of parsing or only in debug mode. - if ( realExplicit && !Global::GetVR()->Count(vr) ) - { - realExplicit = false; - } + bool foundSequenceDelimiter = false; + uint32_t totalLength = 0; - if ( !realExplicit ) + while ( !foundSequenceDelimiter ) { - // We thought this was explicit VR, but we end up with an - // implicit VR tag. Let's backtrack. - msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n", - entry->GetGroup(), entry->GetElement()); - dbg.Verbose(1, "Document::FindVR: ", msg.c_str()); + uint16_t group; + uint16_t elem; - if( entry->GetGroup() % 2 && entry->GetElement() == 0x0000) - { - // Group length is UL ! - DictEntry* newEntry = NewVirtualDictEntry( - entry->GetGroup(), entry->GetElement(), - "UL", "FIXME", "Group Length"); - entry->SetDictEntry( newEntry ); + try + { ///\todo make sure there is never OL encoded pixel data! + + //group = ReadInt16(); // Once per fragment (if any) of OB,OW DataElements + //elem = ReadInt16(); // Once per fragment (if any) of OB,OW DataElements + ReadBegBuffer(4); // Once per fragment (if any) of OB,OW DataElements } - return false; - } - - if ( entry->IsVRUnknown() ) - { - // When not a dictionary entry, we can safely overwrite the VR. - if( entry->GetElement() == 0x0000 ) + catch ( FormatError ) { - // Group length is UL ! - entry->SetVR("UL"); + throw FormatError("Unexpected end of file encountered during ", + "Document::FindDocEntryLengthOBOrOW()"); } - else + group = GetInt16(); + elem = GetInt16(); + + // We have to decount the group and element we just read + totalLength += 4; + if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) ) { - entry->SetVR(vr); + gdcmWarningMacro( + "Neither an Item tag nor a Sequence delimiter tag on :" + << std::hex << group << "|" << elem << ") Pos. on entry was 0x(" <seekg(positionOnEntry, std::ios::beg); // Once per fragment (if any) of OB,OW DataElements + throw FormatUnexpected( + "Neither an Item tag nor a Sequence delimiter tag."); } - } - else if ( entry->GetVR() != vr ) - { - // The VR present in the file and the dictionary disagree. We assume - // the file writer knew best and use the VR of the file. Since it would - // be unwise to overwrite the VR of a dictionary (since it would - // compromise it's next user), we need to clone the actual DictEntry - // and change the VR for the read one. - DictEntry* newEntry = NewVirtualDictEntry( - entry->GetGroup(), entry->GetElement(), - vr, "FIXME", entry->GetName()); - entry->SetDictEntry(newEntry); - } - - 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" ) + if ( elem == 0xe0dd ) { - 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; - } + foundSequenceDelimiter = true; } - - // 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 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); + + if ( foundSequenceDelimiter ) { - 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; - } + break; } -#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(); + Fp->seekg( positionOnEntry, std::ios::beg); // Only once for OB,OW DataElements + return totalLength; } /** - * \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 + * \brief Find the Value Representation of the current Dicom Element. + * @return Value Representation of the current Entry */ -std::string Document::GetDocEntryUnvalue(DocEntry* entry) +VRKey Document::FindDocEntryVR() { - if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() ) + if ( Filetype != ExplicitVR ) { - std::string vr = entry->GetVR(); - std::vector tokens; - std::ostringstream s; - - if ( vr == "US" || vr == "SS" ) - { - uint16_t newInt16; + return GDCM_VRUNKNOWN; + } - 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; + // 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 + // 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. - 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(); - } + 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) ) + { -#ifdef GDCM_NO_ANSI_STRING_STREAM - s << std::ends; // to avoid oddities on Solaris -#endif //GDCM_NO_ANSI_STRING_STREAM - return s.str(); + 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; +} - return ((ValEntry *)entry)->GetValue(); +/** + * \brief Check the correspondance between the VR of the header entry + * and the taken VR. If they are different, the header entry is + * updated with the new VR. + * @param vr Dicom Value Representation + * @return false if the VR is incorrect or if the VR isn't referenced + * otherwise, it returns true +*/ +bool Document::CheckDocEntryVR(const VRKey &vr) +{ + 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) @@ -2080,14 +1764,20 @@ void Document::SkipDocEntry(DocEntry *entry) } /** - * \brief Skips to the begining of the next Header Entry - * \warning NOT end user intended method ! - * @param entry entry to skip + * \brief Skips to the beginning of the next Header Entry + * @param currentDocEntry entry to skip */ -void Document::SkipToNextDocEntry(DocEntry *entry) +void Document::SkipToNextDocEntry(DocEntry *currentDocEntry) { - Fp->seekg((long)(entry->GetOffset()), std::ios_base::beg); - Fp->seekg( (long)(entry->GetReadLength()), std::ios_base::cur); + long l = currentDocEntry->GetReadLength(); + if ( (uint32_t) l == (uint32_t)-1 ) // length = 0xffff shouldn't appear here ... + // ... but PMS imagers happen ! + return; + Fp->seekg((size_t)(currentDocEntry->GetOffset()), std::ios::beg); //FIXME :each DocEntry + if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb + { + Fp->seekg( l,std::ios::cur); //FIXME :each DocEntry + } } /** @@ -2095,44 +1785,45 @@ void Document::SkipToNextDocEntry(DocEntry *entry) * the parser went Jabberwocky) one can hope improving things by * applying some heuristics. * @param entry entry to check - * @param foundLength fist 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 el = entry->GetElement(); + + uint16_t gr = entry->GetGroup(); + uint16_t elem = entry->GetElement(); if ( foundLength % 2) { - std::ostringstream s; - s << "Warning : Tag with uneven length " - << foundLength - << " in x(" << std::hex << gr << "," << el <<")" << std::dec; - dbg.Verbose(0, s.str().c_str()); + 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 - // are no longer in user (we are talking a few years, here)... - // Note: XMedCom probably uses such a trick since it is able to read + // are no longer in use (we are talking a few years, here)... + // Note: XMedCon probably uses such a trick since it is able to read // those pesky GE images ... if ( foundLength == 13) { // Only happens for this length ! - if ( entry->GetGroup() != 0x0008 - || ( entry->GetElement() != 0x0070 - && entry->GetElement() != 0x0080 ) ) + if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) ) { foundLength = 10; - entry->SetReadLength(10); /// \todo a bug is to be fixed !? + entry->SetReadLength(10); // a bug is to be fixed !? } } @@ -2140,211 +1831,93 @@ 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 ( entry->GetGroup() == 0x0009 - && ( entry->GetElement() == 0x1113 - || entry->GetElement() == 0x1114 ) ) - { - foundLength = 4; - entry->SetReadLength(4); /// \todo 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( entry->GetGroup() == 0xfffe ) - { + 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; } - } - - entry->SetUsableLength(foundLength); -} - -/** - * \brief Apply some heuristics to predict whether the considered - * element value contains/represents an integer or not. - * @param entry The element value on which to apply the predicate. - * @return The result of the heuristical predicate. - */ -bool Document::IsDocEntryAnInteger(DocEntry *entry) -{ - uint16_t element = entry->GetElement(); - uint16_t group = entry->GetGroup(); - std::string 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. - if ( element == 0 ) // This is the group length of the group - { - if ( length == 4 ) - { - return true; - } - else - { - // Allthough 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]. - // 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). - std::ostringstream s; - long filePosition = Fp->tellg(); - s << "Erroneous Group Length element length on : (" \ - << std::hex << group << " , " << element - << ") -before- position x(" << filePosition << ")" - << "lgt : " << length; - dbg.Verbose(0, "Document::IsDocEntryAnInteger", s.str().c_str() ); - } - } - - if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" ) - { - return true; - } - - return false; -} - -/** - * \brief Find the Length till the next sequence delimiter - * \warning NOT end user intended method ! - * @return - */ - -uint32_t Document::FindDocEntryLengthOB() - throw( FormatUnexpected ) -{ - // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data. - long positionOnEntry = Fp->tellg(); - bool foundSequenceDelimiter = false; - uint32_t totalLength = 0; - - while ( !foundSequenceDelimiter ) - { - uint16_t group; - uint16_t elem; - try - { - group = ReadInt16(); - elem = ReadInt16(); - } - catch ( FormatError ) - { - throw FormatError("Document::FindDocEntryLengthOB()", - " group or element not present."); - } - - // We have to decount the group and element we just read - totalLength += 4; - - if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) ) - { - dbg.Verbose(1, "Document::FindDocEntryLengthOB: neither an Item " - "tag nor a Sequence delimiter tag."); - Fp->seekg(positionOnEntry, std::ios_base::beg); - throw FormatUnexpected("Document::FindDocEntryLengthOB()", - "Neither an Item tag nor a Sequence " - "delimiter tag."); - } - - if ( elem == 0xe0dd ) - { - foundSequenceDelimiter = true; - } - - uint32_t itemLength = ReadInt32(); - // We add 4 bytes since we just read the ItemLength with ReadInt32 - totalLength += itemLength + 4; - SkipBytes(itemLength); - - if ( foundSequenceDelimiter ) - { - break; - } - } - Fp->seekg( positionOnEntry, std::ios_base::beg); - return totalLength; -} - -/** - * \brief Reads a supposed to be 16 Bits integer - * (swaps it depending on processor endianity) - * @return read value - */ -uint16_t Document::ReadInt16() - throw( FormatError ) -{ - uint16_t g; - Fp->read ((char*)&g, (size_t)2); - if ( Fp->fail() ) - { - throw FormatError( "Document::ReadInt16()", " file error." ); - } - if( Fp->eof() ) - { - throw FormatError( "Document::ReadInt16()", "EOF." ); - } - g = SwapShort(g); - return g; -} - -/** - * \brief Reads a supposed to be 32 Bits integer - * (swaps it depending on processor endianity) - * @return read value - */ -uint32_t Document::ReadInt32() - throw( FormatError ) -{ - uint32_t g; - Fp->read ((char*)&g, (size_t)4); - if ( Fp->fail() ) - { - throw FormatError( "Document::ReadInt32()", " file error." ); - } - if( Fp->eof() ) - { - throw FormatError( "Document::ReadInt32()", "EOF." ); + else + { + foundLength=12; // to skip the mess that follows this bugged Tag ! + } } - g = SwapLong(g); - return g; -} - -/** - * \brief skips bytes inside the source file - * \warning NOT end user intended method ! - * @return - */ -void Document::SkipBytes(uint32_t nBytes) -{ - //FIXME don't dump the returned value - Fp->seekg((long)nBytes, std::ios_base::cur); + entry->SetLength(foundLength); } /** - * \brief Loads all the needed Dictionaries - * \warning NOT end user intended method ! + * \brief Apply some heuristics to predict whether the considered + * element value contains/represents an integer or not. + * @param entry The element value on which to apply the predicate. + * @return The result of the heuristical predicate. */ -void Document::Initialise() +bool Document::IsDocEntryAnInteger(DocEntry *entry) { - RefPubDict = Global::GetDicts()->GetDefaultPubDict(); - RefShaDict = NULL; - RLEInfo = new RLEFramesInfo; - JPEGInfo = new JPEGFragmentsInfo; + uint16_t elem = entry->GetElement(); + uint16_t group = entry->GetGroup(); + 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 its element value properly. + if ( elem == 0 ) // This is the group length of the group + { + if ( length == 4 ) + { + return true; + } + else + { + // 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. + // 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(); // 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 << ")" + << "lgt : " << length ); + } + } + + if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" ) + { + return true; + } + return false; } /** @@ -2356,49 +1929,34 @@ void Document::Initialise() * true when we hope ours assuptions are OK */ bool Document::CheckSwap() -{ - // The only guaranted way of finding the swap code is to find a - // group tag since we know it's length has to be of four bytes i.e. - // 0x00000004. Finding the swap code in then straigthforward. Trouble - // occurs when we can't find such group... - - uint32_t x = 4; // x : for ntohs - bool net2host; // true when HostByteOrder is the same as NetworkByteOrder +{ uint32_t s32; uint16_t s16; - char deb[HEADER_LENGTH_TO_READ]; + char deb[256]; // First, compare HostByteOrder and NetworkByteOrder in order to // determine if we shall need to swap bytes (i.e. the Endian type). - if ( x == ntohs(x) ) - { - net2host = true; - } - else - { - net2host = false; - } + bool net2host = Util::IsCurrentProcessorBigEndian(); - // The easiest case is the one of a DICOM header, since it possesses a - // file preamble where it suffice to look for the string "DICM". - Fp->read(deb, HEADER_LENGTH_TO_READ); + // The easiest case is the one of a 'true' DICOM header, we just have + // to look for the string "DICM" inside the file preamble. + Fp->read(deb, 256); char *entCur = deb + 128; - if( memcmp(entCur, "DICM", (size_t)4) == 0 ) + if ( memcmp(entCur, "DICM", (size_t)4) == 0 ) { - dbg.Verbose(1, "Document::CheckSwap:", "looks like DICOM Version3"); + gdcmDebugMacro( "Looks like DICOM Version3 (preamble + DCM)" ); - // Next, determine the value representation (VR). Let's skip to the - // first element (0002, 0000) and check there if we find "UL" - // - or "OB" if the 1st one is (0002,0001) -, + // Group 0002 should always be VR, and the first element 0000 + // Let's be carefull (so many wrong headers ...) + // and determine the value representation (VR) : + // Let's skip to the first element (0002,0000) and check there if we find + // "UL" - or "OB" if the 1st one is (0002,0001) -, // in which case we (almost) know it is explicit VR. // WARNING: if it happens to be implicit VR then what we will read // is the length of the group. If this ascii representation of this // length happens to be "UL" then we shall believe it is explicit VR. - // FIXME: in order to fix the above warning, we could read the next - // element value (or a couple of elements values) in order to make - // sure we are not commiting a big mistake. // We need to skip : // * the 128 bytes of File Preamble (often padded with zeroes), // * the 4 bytes of "DICM" string, @@ -2406,59 +1964,94 @@ bool Document::CheckSwap() // i.e. a total of 136 bytes. entCur = deb + 136; - // FIXME : FIXME: - // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR, - // but elem 0002,0010 (Transfert Syntax) tells us the file is - // *Implicit* VR. -and it is !- + // 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') - if( memcmp(entCur, "UL", (size_t)2) == 0 || - memcmp(entCur, "OB", (size_t)2) == 0 || - memcmp(entCur, "UI", (size_t)2) == 0 || - memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later - // when Write DCM *adds* + if ( memcmp(entCur, "UL", (size_t)2) == 0 || + memcmp(entCur, "OB", (size_t)2) == 0 || + memcmp(entCur, "UI", (size_t)2) == 0 || + memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later + // when Write DCM *adds* // FIXME // Use Document::dicom_vr to test all the possibilities // instead of just checking for UL, OB and UI !? group 0000 { Filetype = ExplicitVR; - dbg.Verbose(1, "Document::CheckSwap:", - "explicit Value Representation"); + gdcmDebugMacro( "Group 0002 : Explicit Value Representation"); } else { Filetype = ImplicitVR; - dbg.Verbose(1, "Document::CheckSwap:", - "not an explicit Value Representation"); + gdcmWarningMacro( "Group 0002 :Not an explicit Value Representation;" + << "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; - dbg.Verbose(1, "Document::CheckSwap:", - "HostByteOrder != NetworkByteOrder"); + gdcmDebugMacro( "HostByteOrder != NetworkByteOrder, SwapCode = 4321"); } else { - SwapCode = 0; - dbg.Verbose(1, "Document::CheckSwap:", - "HostByteOrder = NetworkByteOrder"); + SwapCode = 1234; + 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_base::beg); - Fp->seekg ( 132L, std::ios_base::beg); + // Position the file position indicator at first tag + // (i.e. after the file preamble and the "DICM" string). + + Fp->seekg ( 132L, std::ios::beg); // Once per Document + CurrentOffsetPosition = 132; return true; - } // End of DicomV3 + } // ------------------------------- End of DicomV3 ---------------- // Alas, this is not a DicomV3 file and whatever happens there is no file // preamble. We can reset the file position indicator to where the data // is (i.e. the beginning of the file). - dbg.Verbose(1, "Document::CheckSwap:", "not a DICOM Version3 file"); - Fp->seekg(0, std::ios_base::beg); + + gdcmWarningMacro( "Not a Kosher DICOM Version3 file (no preamble)"); + + 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 + + s16 = *((uint16_t *)(deb)); + SwapCode = 0; + switch ( s16 ) + { + case 0x0002 : + SwapCode = 1234; + entCur = deb + 4; + break; + case 0x0200 : + SwapCode = 4321; + entCur = deb + 6; + } + + if ( SwapCode != 0 ) + { + if ( memcmp(entCur, "UL", (size_t)2) == 0 || + memcmp(entCur, "OB", (size_t)2) == 0 || + memcmp(entCur, "UI", (size_t)2) == 0 || + memcmp(entCur, "SH", (size_t)2) == 0 || + memcmp(entCur, "AE", (size_t)2) == 0 || + memcmp(entCur, "OB", (size_t)2) == 0 ) + { + Filetype = ExplicitVR; // FIXME : not enough to say it's Explicit + // Wait untill reading Transfer Syntax + gdcmDebugMacro( "Group 0002 : Explicit Value Representation"); + return true; + } + } +// ------------------------------- End of 'No Preamble' DicomV3 ------------- // Our next best chance would be to be considering a 'clean' ACR/NEMA file. - // By clean we mean that the length of the first tag is written down. + // By clean we mean that the length of the first group is written down. // If this is the case and since the length of the first group HAS to be // four (bytes), then determining the proper swap code is straightforward. @@ -2467,7 +2060,6 @@ bool Document::CheckSwap() // representation of a 32 bits integer. Hence the following dirty // trick : s32 = *((uint32_t *)(entCur)); - switch( s32 ) { case 0x00040000 : @@ -2483,7 +2075,7 @@ bool Document::CheckSwap() Filetype = ACR; return true; case 0x00000004 : - SwapCode = 0; + SwapCode = 1234; Filetype = ACR; return true; default : @@ -2492,62 +2084,107 @@ bool Document::CheckSwap() // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA, // i.e. the 'group length' element is not present : - // check the supposed to be 'group number' - // 0x0002 or 0x0004 or 0x0008 + // check the supposed-to-be 'group number' + // in ( 0x0001 .. 0x0008 ) // to determine ' SwapCode' value . // 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 -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 : case 0x0002 : + case 0x0003 : case 0x0004 : - case 0x0008 : - SwapCode = 0; + case 0x0005 : + case 0x0006 : + case 0x0007 : + case 0x0008 : + case 0x0028 : + SwapCode = 1234; Filetype = ACR; return true; + case 0x0100 : case 0x0200 : + case 0x0300 : case 0x0400 : - case 0x0800 : + case 0x0500 : + case 0x0600 : + case 0x0700 : + case 0x0800 : + case 0x2800 : SwapCode = 4321; Filetype = ACR; return true; default : - dbg.Verbose(0, "Document::CheckSwap:", - "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() ); // Only when DEBUG + if ( SwapCode == 1234 ) { SwapCode = 4321; } else if ( SwapCode == 4321 ) { - SwapCode = 0; + SwapCode = 1234; } else if ( SwapCode == 3412 ) { @@ -2557,13 +2194,14 @@ void Document::SwitchSwapToBigEndian() { SwapCode = 3412; } + gdcmDebugMacro( " Into: "<< SwapCode ); } /** - * \brief during parsing, Header Elements too long are not loaded in memory - * @param newSize + * \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 ) { @@ -2577,505 +2215,323 @@ void Document::SetMaxSizeLoadEntry(long newSize) MaxSizeLoadEntry = newSize; } - -/** - * \brief Header Elements too long will not be printed - * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE - * @param newSize - */ -void Document::SetMaxSizePrintEntry(long newSize) -{ - //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME - if ( newSize < 0 ) - { - return; - } - if ((uint32_t)newSize >= (uint32_t)0xffffffff ) - { - MaxSizePrintEntry = 0xffffffff; - return; - } - MaxSizePrintEntry = newSize; -} - - - /** - * \brief Read the next tag but WITHOUT loading it's value + * \brief Read the next tag WITHOUT loading it's value * (read the 'Group Number', the 'Element Number', - * gets the Dict Entry + * 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() +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 e ) + 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. - //std::cout << e; return 0; } - - DocEntry *newEntry = NewDocEntryByNumber(group, elem); - FindDocEntryVR(newEntry); - - try - { - FindDocEntryLength(newEntry); - } - catch ( FormatError e ) + + changeFromUN = false; + CurrentGroup = GetInt16(); + CurrentElem = GetInt16(); + + // In 'true DICOM' files Group 0002 is always little endian + if ( HasDCMPreamble ) { - // Call it quits - //std::cout << e; - delete newEntry; - return 0; - } - - newEntry->SetOffset(Fp->tellg()); - - return newEntry; -} + 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; -/** - * \brief Generate a free TagKey i.e. a TagKey that is not present - * in the TagHt dictionary. - * @param group The generated tag must belong to this group. - * @return The element of tag with given group which is fee. - */ -uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group) -{ - for (uint32_t elem = 0; elem < UINT32_MAX; elem++) + if ( vr == GDCM_VRUNKNOWN ) { - TagKey key = DictEntry::TranslateToKey(group, elem); - if (TagHT.count(key) == 0) + if ( CurrentElem == 0x0000 ) // Group Length { - return elem; + realVR = "UL"; // must be UL + } + else if (CurrentGroup == 0xfffe) // Don't get DictEntry for Delimitors + { + realVR = "UL"; } - } - return UINT32_MAX; -} - -/** - * \brief Assuming the internal file pointer \ref Document::Fp - * is placed at the beginning of a tag check whether this - * tag is (TestGroup, TestElement). - * \warning On success the internal file pointer \ref Document::Fp - * is modified to point after the tag. - * On failure (i.e. when the tag wasn't the expected tag - * (TestGroup, TestElement) the internal file pointer - * \ref Document::Fp is restored to it's original position. - * @param testGroup The expected group of the tag. - * @param testElement The expected Element of the tag. - * @return True on success, false otherwise. - */ -bool Document::ReadTag(uint16_t testGroup, uint16_t testElement) -{ - long positionOnEntry = Fp->tellg(); - long currentPosition = Fp->tellg(); // On debugging purposes - - //// Read the Item Tag group and element, and make - // sure they are what we expected: - uint16_t itemTagGroup = ReadInt16(); - uint16_t itemTagElement = ReadInt16(); - if ( itemTagGroup != testGroup || itemTagElement != testElement ) - { - std::ostringstream s; - s << " We should have found tag ("; - s << std::hex << testGroup << "," << testElement << ")" << std::endl; - s << " but instead we encountered tag ("; - s << std::hex << itemTagGroup << "," << itemTagElement << ")" - << std::endl; - s << " at address: " << (unsigned)currentPosition << std::endl; - dbg.Verbose(0, "Document::ReadItemTagLength: wrong Item Tag found:"); - dbg.Verbose(0, s.str().c_str()); - Fp->seekg(positionOnEntry, std::ios_base::beg); - return false; + // 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(CurrentGroup,CurrentElem);//only when ImplicitVR + if ( dictEntry ) + { + realVR = dictEntry->GetVR(); + dictEntry->Unregister(); // GetDictEntry registered it + } + } } - return true; -} - -/** - * \brief Assuming the internal file pointer \ref Document::Fp - * is placed at the beginning of a tag (TestGroup, TestElement), - * read the length associated to the Tag. - * \warning On success the internal file pointer \ref Document::Fp - * is modified to point after the tag and it's length. - * On failure (i.e. when the tag wasn't the expected tag - * (TestGroup, TestElement) the internal file pointer - * \ref Document::Fp is restored to it's original position. - * @param testGroup The expected group of the tag. - * @param testElement The expected Element of the tag. - * @return On success returns the length associated to the tag. On failure - * returns 0. - */ -uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement) -{ - long positionOnEntry = Fp->tellg(); - (void)positionOnEntry; - if ( !ReadTag(testGroup, testElement) ) - { - return 0; - } - - //// Then read the associated Item Length - long currentPosition = Fp->tellg(); - uint32_t itemLength = ReadInt32(); + // if UN found, let's check the dictionary, and trust it! + // (maybe a private dictionary exists?) + else if (vr == "UN") { - std::ostringstream s; - s << "Basic Item Length is: " - << itemLength << std::endl; - s << " at address: " << (unsigned)currentPosition << std::endl; - dbg.Verbose(0, "Document::ReadItemTagLength: ", s.str().c_str()); - } - return itemLength; -} - -/** - * \brief When parsing the Pixel Data of an encapsulated file, read - * the basic offset table (when present, and BTW dump it). - */ -void Document::ReadAndSkipEncapsulatedBasicOffsetTable() -{ - //// Read the Basic Offset Table Item Tag length... - uint32_t itemLength = ReadTagLength(0xfffe, 0xe000); - - // When present, read the basic offset table itself. - // Notes: - since the presence of this basic offset table is optional - // we can't rely on it for the implementation, and we will simply - // trash it's content (when present). - // - still, when present, we could add some further checks on the - // lengths, but we won't bother with such fuses for the time being. - if ( itemLength != 0 ) - { - char* basicOffsetTableItemValue = new char[itemLength + 1]; - Fp->read(basicOffsetTableItemValue, itemLength); - -#ifdef GDCM_DEBUG - for (unsigned int i=0; i < itemLength; i += 4 ) + DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem); + if ( dictEntry ) { - uint32_t individualLength = str2num( &basicOffsetTableItemValue[i], - uint32_t); - std::ostringstream s; - s << " Read one length: "; - s << std::hex << individualLength << std::endl; - dbg.Verbose(0, - "Document::ReadAndSkipEncapsulatedBasicOffsetTable: ", - s.str().c_str()); - } -#endif //GDCM_DEBUG + realVR = dictEntry->GetVR(); + dictEntry->Unregister(); // GetDictEntry registered it - delete[] basicOffsetTableItemValue; + // 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 + + } } -} -/** - * \brief Parse pixel data from disk of [multi-]fragment RLE encoding. - * Compute the RLE extra information and store it in \ref RLEInfo - * for later pixel retrieval usage. - */ -void Document::ComputeRLEInfo() -{ - TransferSyntaxType ts = GetTransferSyntax(); - if ( ts != RLELossless ) + DocEntry *newEntry; + //if ( Global::GetVR()->IsVROfSequence(realVR) ) + if (realVR == "SQ") { - return; + newEntry = NewSeqEntry(CurrentGroup, CurrentElem); + } + else + { + newEntry = NewDataEntry(CurrentGroup, CurrentElem, realVR); + static_cast(newEntry)->SetState(DataEntry::STATE_NOTLOADED); } - // Encoded pixel data: for the time being we are only concerned with - // Jpeg or RLE Pixel data encodings. - // As stated in PS 3.5-2003, section 8.2 p44: - // "If sent in Encapsulated Format (i.e. other than the Native Format) the - // value representation OB is used". - // Hence we expect an OB value representation. Concerning OB VR, - // the section PS 3.5-2003, section A.4.c p 58-59, states: - // "For the Value Representations OB and OW, the encoding shall meet the - // following specifications depending on the Data element tag:" - // [...snip...] - // - the first item in the sequence of items before the encoded pixel - // data stream shall be basic offset table item. The basic offset table - // item value, however, is not required to be present" - - ReadAndSkipEncapsulatedBasicOffsetTable(); - - // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G) - // Loop on the individual frame[s] and store the information - // on the RLE fragments in a RLEFramesInfo. - // Note: - when only a single frame is present, this is a - // classical image. - // - when more than one frame are present, then we are in - // the case of a multi-frame image. - long frameLength; - while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) ) - { - // Parse the RLE Header and store the corresponding RLE Segment - // Offset Table information on fragments of this current Frame. - // Note that the fragment pixels themselves are not loaded - // (but just skipped). - long frameOffset = Fp->tellg(); - - uint32_t nbRleSegments = ReadInt32(); - if ( nbRleSegments > 16 ) + if ( vr == GDCM_VRUNKNOWN ) + { + if ( Filetype == ExplicitVR ) { - // There should be at most 15 segments (refer to RLEFrame class) - dbg.Verbose(0, "Document::ComputeRLEInfo: too many segments."); - } + // We thought this was explicit VR, but we end up with an + // implicit VR tag. Let's backtrack. - uint32_t rleSegmentOffsetTable[15]; - for( int k = 1; k <= 15; k++ ) - { - rleSegmentOffsetTable[k] = ReadInt32(); - } + //if ( newEntry->GetGroup() != 0xfffe ) + if (CurrentGroup != 0xfffe) + { + int offset = Fp->tellg();//Only when heuristic for Explicit/Implicit was wrong - // Deduce from both the RLE Header and the frameLength the - // fragment length, and again store this info in a - // RLEFramesInfo. - long rleSegmentLength[15]; - // skipping (not reading) RLE Segments - if ( nbRleSegments > 1) - { - for(unsigned int k = 1; k <= nbRleSegments-1; k++) - { - rleSegmentLength[k] = rleSegmentOffsetTable[k+1] - - rleSegmentOffsetTable[k]; - SkipBytes(rleSegmentLength[k]); + gdcmWarningMacro("Entry (" << newEntry->GetKey() << ") at x(" + << std::hex << offset << ") should be Explicit VR"); } - } - - rleSegmentLength[nbRleSegments] = frameLength - - rleSegmentOffsetTable[nbRleSegments]; - SkipBytes(rleSegmentLength[nbRleSegments]); - - // Store the collected info - RLEFrame* newFrameInfo = new RLEFrame; - newFrameInfo->NumberFragments = nbRleSegments; - for( unsigned int uk = 1; uk <= nbRleSegments; uk++ ) - { - newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk]; - newFrameInfo->Length[uk] = rleSegmentLength[uk]; - } - RLEInfo->Frames.push_back( newFrameInfo ); + } + newEntry->SetImplicitVR(); } - // Make sure that at the end of the item we encounter a 'Sequence - // Delimiter Item': - if ( !ReadTag(0xfffe, 0xe0dd) ) + try + { + FindDocEntryLength(newEntry); + } + catch ( FormatError ) { - dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter "); - dbg.Verbose(0, " item at end of RLE item sequence"); + // Call it quits + newEntry->Delete(); + return 0; } + + newEntry->SetOffset(Fp->tellg()); // for each DocEntry + return newEntry; } /** - * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding. - * Compute the jpeg extra information (fragment[s] offset[s] and - * length) and store it[them] in \ref JPEGInfo for later pixel - * retrieval usage. + * \brief Handle broken private tag from Philips NTSCAN + * where the endianess is being switched to BigEndian + * for no apparent reason + * @return no return */ -void Document::ComputeJPEGFragmentInfo() +void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem) { - // If you need to, look for comments of ComputeRLEInfo(). - if ( ! IsJPEG() ) + // 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)) { - return; - } - - ReadAndSkipEncapsulatedBasicOffsetTable(); - - // Loop on the fragments[s] and store the parsed information in a - // JPEGInfo. - long fragmentLength; - while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) ) - { - long fragmentOffset = Fp->tellg(); - - // Store the collected info - JPEGFragment* newFragment = new JPEGFragment; - newFragment->Offset = fragmentOffset; - newFragment->Length = fragmentLength; - JPEGInfo->Fragments.push_back( newFragment ); - - SkipBytes( fragmentLength ); + // start endian swap mark for group found + gdcmDebugMacro( " delimiter 0x3f3f found." ); + // fix the tag + group = 0xfffe; + elem = 0xe000; + return; } - - // Make sure that at the end of the item we encounter a 'Sequence - // Delimiter Item': - if ( !ReadTag(0xfffe, 0xe0dd) ) + // --- 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 + group = 0xfffe; + elem = 0xe000; + } + else if (group == 0xfffe && elem == 0xe00d && reversedEndian) { - dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter "); - dbg.Verbose(0, " item at end of JPEG item sequence"); + // end of reversed endian group + gdcmDebugMacro( "End of reversed endian." ); + reversedEndian--; + SwitchByteSwapCode(); } -} - -/** - * \brief Walk recursively the given \ref DocEntrySet, and feed - * the given hash table (\ref TagDocEntryHT) with all the - * \ref DocEntry (Dicom entries) encountered. - * This method does the job for \ref BuildFlatHashTable. - * @param builtHT Where to collect all the \ref DocEntry encountered - * when recursively walking the given set. - * @param set The structure to be traversed (recursively). - */ -void Document::BuildFlatHashTableRecurse( TagDocEntryHT& builtHT, - DocEntrySet* set ) -{ - if (ElementSet* elementSet = dynamic_cast< ElementSet* > ( set ) ) + else if (group == 0xfeff && elem == 0xdde0) { - TagDocEntryHT const & currentHT = elementSet->GetTagHT(); - for( TagDocEntryHT::const_iterator i = currentHT.begin(); - i != currentHT.end(); - ++i) - { - DocEntry* entry = i->second; - if ( SeqEntry* seqEntry = dynamic_cast(entry) ) - { - const ListSQItem& items = seqEntry->GetSQItems(); - for( ListSQItem::const_iterator item = items.begin(); - item != items.end(); - ++item) - { - BuildFlatHashTableRecurse( builtHT, *item ); - } - continue; - } - builtHT[entry->GetKey()] = entry; - } - return; - } - - if (SQItem* SQItemSet = dynamic_cast< SQItem* > ( set ) ) + // reversed Sequence Terminator found + // probabely a bug in the header ! + // Do what you want, it breaks ! + //reversedEndian--; + //SwitchByteSwapCode(); + gdcmWarningMacro( "Should never get here! reversed Sequence Terminator!" ); + // fix the tag + group = 0xfffe; + elem = 0xe0dd; + } + else if (group == 0xfffe && elem == 0xe0dd) { - const ListDocEntry& currentList = SQItemSet->GetDocEntries(); - for (ListDocEntry::const_iterator i = currentList.begin(); - i != currentList.end(); - ++i) - { - DocEntry* entry = *i; - if ( SeqEntry* seqEntry = dynamic_cast(entry) ) - { - const ListSQItem& items = seqEntry->GetSQItems(); - for( ListSQItem::const_iterator item = items.begin(); - item != items.end(); - ++item) - { - BuildFlatHashTableRecurse( builtHT, *item ); - } - continue; - } - builtHT[entry->GetKey()] = entry; - } - + gdcmDebugMacro( "Straight Sequence Terminator." ); } } /** - * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current - * Document. - * - * The structure used by a Document (through \ref ElementSet), - * in order to old the parsed entries of a Dicom header, is a recursive - * one. This is due to the fact that the sequences (when present) - * can be nested. Additionaly, the sequence items (represented in - * gdcm as \ref SQItem) add an extra complexity to the data - * structure. Hence, a gdcm user whishing to visit all the entries of - * a Dicom header will need to dig in the gdcm internals (which - * implies exposing all the internal data structures to the API). - * In order to avoid this burden to the user, \ref BuildFlatHashTable - * recursively builds a temporary hash table, which holds all the - * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a - * std::map<>). - * \warning Of course there is NO integrity constrain between the - * returned \ref TagDocEntryHT and the \ref ElementSet used - * to build it. Hence if the underlying \ref ElementSet is - * altered, then it is the caller responsability to invoke - * \ref BuildFlatHashTable again... - * @return The flat std::map<> we juste build. + * \brief Group 0002 is always coded Little Endian + * whatever Transfer Syntax is + * @return no return */ -TagDocEntryHT* Document::BuildFlatHashTable() +void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem) { - TagDocEntryHT* FlatHT = new TagDocEntryHT; - BuildFlatHashTableRecurse( *FlatHT, this ); - return FlatHT; -} + // 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 + std::string ts = GetTransferSyntax(); + TS::SpecialType s = Global::GetTS()->GetSpecialTransferSyntax(ts); + // Group 0002 is always 'Explicit ...' + // even when Transfer Syntax says 'Implicit ..." -/** - * \brief Compares two documents, according to \ref DicomDir rules - * \warning Does NOT work with ACR-NEMA files - * \todo Find a trick to solve the pb (use RET fields ?) - * @param document - * @return true if 'smaller' - */ -bool Document::operator<(Document &document) -{ - // Patient Name - std::string s1 = GetEntryByNumber(0x0010,0x0010); - std::string s2 = document.GetEntryByNumber(0x0010,0x0010); - if(s1 < s2) - { - return true; - } - else if( s1 > s2 ) - { - return false; - } - else - { - // Patient ID - s1 = GetEntryByNumber(0x0010,0x0020); - s2 = document.GetEntryByNumber(0x0010,0x0020); - if ( s1 < s2 ) + if ( s == TS::ImplicitVRLittleEndian + || + s == TS::ImplicitVRBigEndianPrivateGE + ) { - return true; + Filetype = ImplicitVR; } - else if ( s1 > s2 ) + + // FIXME Strangely, this works with + //'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 ) + { + 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) { - return false; + gdcmWarningMacro("Transfer Syntax [" + << GetTransferSyntaxName() << "] :" + << " not yet dealt with "); + return; } - else + + // The following shouldn't occur very often + // Let's check at the very end. + + if ( ts == GDCM_UNKNOWN ) { - // Study Instance UID - s1 = GetEntryByNumber(0x0020,0x000d); - s2 = document.GetEntryByNumber(0x0020,0x000d); - if ( s1 < s2 ) - { - return true; - } - else if( s1 > s2 ) - { - return false; - } - else - { - // Serie Instance UID - s1 = GetEntryByNumber(0x0020,0x000e); - s2 = document.GetEntryByNumber(0x0020,0x000e); - if ( s1 < s2 ) - { - return true; - } - else if( s1 > s2 ) - { - return false; - } - } + gdcmDebugMacro("True DICOM File, with NO Transfer Syntax (?!) " ); + return; } - } - return false; + + if ( !Global::GetTS()->IsTransferSyntax(ts) ) + { + gdcmWarningMacro("True DICOM File, with illegal Transfer Syntax: [" + << ts << "]"); + return; + } } -} // end namespace gdcm +//----------------------------------------------------------------------------- +// Print //----------------------------------------------------------------------------- +} // end namespace gdcm