1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/11/03 14:21:50 $
7 Version: $Revision: 1.321 $
9 Copyright (c) CREATIS (Centre de Recherche et d'Applications en Traitement de
10 l'Image). All rights reserved. See Doc/License.txt or
11 http://www.creatis.insa-lyon.fr/Public/Gdcm/License.html for details.
13 This software is distributed WITHOUT ANY WARRANTY; without even
14 the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
15 PURPOSE. See the above copyright notices for more information.
17 =========================================================================*/
19 #include "gdcmDocument.h"
20 #include "gdcmSeqEntry.h"
21 #include "gdcmGlobal.h"
23 #include "gdcmDebug.h"
25 #include "gdcmDictSet.h"
26 #include "gdcmDocEntrySet.h"
27 #include "gdcmSQItem.h"
28 #include "gdcmDataEntry.h"
33 #include <ctype.h> // for isdigit
34 #include <stdlib.h> // for atoi
38 //-----------------------------------------------------------------------------
40 // Refer to Document::SetMaxSizeLoadEntry()
41 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
43 //-----------------------------------------------------------------------------
44 // Constructor / Destructor
45 // Constructors and destructors are protected to avoid user to invoke directly
48 * \brief This default constructor neither loads nor parses the file.
49 * You should then invoke \ref Document::Load.
57 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
60 Filetype = ExplicitVR;
61 // Load will set it to true if sucessfull
62 Group0002Parsed = false;
63 IsDocumentAlreadyLoaded = false;
64 IsDocumentModified = true;
65 LoadMode = LD_ALL; // default : load everything, later
70 * \brief Canonical destructor.
72 Document::~Document ()
77 //-----------------------------------------------------------------------------
81 * \brief Loader. use SetLoadMode(), SetFileName() before !
82 * @return false if file cannot be open or no swap info was found,
83 * or no tag was found.
85 bool Document::Load( )
87 if ( GetFileName() == "" )
89 gdcmWarningMacro( "Use SetFileName, before !" );
92 return DoTheLoadingDocumentJob( );
95 * \brief Loader. (DEPRECATED : not to break the API)
96 * @param fileName 'Document' (File or DicomDir) to be open for parsing
97 * @return false if file cannot be open or no swap info was found,
98 * or no tag was found.
100 bool Document::Load( std::string const &fileName )
103 return DoTheLoadingDocumentJob( );
107 * \brief Performs the Loading Job (internal use only)
108 * @return false if file cannot be open or no swap info was found,
109 * or no tag was found.
111 bool Document::DoTheLoadingDocumentJob( )
113 if ( ! IsDocumentModified ) // Nothing to do !
121 // warning already performed in OpenFile()
126 Group0002Parsed = false;
128 gdcmDebugMacro( "Starting parsing of file: " << Filename.c_str());
130 Fp->seekg(0, std::ios::end);
131 long lgt = Fp->tellg(); // total length of the file
133 Fp->seekg(0, std::ios::beg);
135 // CheckSwap returns a boolean
136 // (false if no swap info of any kind was found)
139 gdcmWarningMacro( "Neither a DICOM V3 nor an ACR-NEMA file: "
140 << Filename.c_str());
145 long beg = Fp->tellg(); // just after DICOM preamble (if any)
147 lgt -= beg; // remaining length to parse
150 // Loading is done during parsing
151 ParseDES( this, beg, lgt, false); // delim_mode is first defaulted to false
155 gdcmErrorMacro( "No tag in internal hash table for: "
156 << Filename.c_str());
160 IsDocumentAlreadyLoaded = true;
162 Fp->seekg( 0, std::ios::beg);
164 // Load 'non string' values
166 std::string PhotometricInterpretation = GetEntryString(0x0028,0x0004);
167 if ( PhotometricInterpretation == "PALETTE COLOR " )
170 // Probabely this line should be outside the 'if'
171 // Try to find an image sample holding a 'gray LUT'
172 LoadEntryBinArea(0x0028,0x1200); // gray LUT
175 /// --> FIXME : The difference between BinEntry and DataEntry
176 /// --> no longer exists, but the alteration of Dicom Dictionary remains.
177 /// --> Old comment restored on purpose.
178 /// --> New one (replacing both BinEntry and ValEntry by DataEntry)
179 /// --> had absolutely no meaning.
180 /// --> The whole comment will be removed when the stuff is cleaned !
182 /// The tags refered by the three following lines used to be CORRECTLY
183 /// defined as having an US Value Representation in the public
184 /// dictionary. BUT the semantics implied by the three following
185 /// lines state that the corresponding tag contents are in fact
186 /// the ones of a BinEntry.
187 /// In order to fix things "Quick and Dirty" the dictionary was
188 /// altered on PURPOSE but now contains a WRONG value.
189 /// In order to fix things and restore the dictionary to its
190 /// correct value, one needs to decide of the semantics by deciding
191 /// whether the following tags are either :
192 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
193 /// also used as BinEntry, which requires the proper conversion,
194 /// - OW, and hence loaded as BinEntry, but afterwards also used
195 /// as ValEntry, which requires the proper conversion.
197 // --> OB (byte aray) or OW (short int aray)
198 // The actual VR has to be deduced from other entries.
199 // Our way of loading them may fail in some cases :
200 // We must or not SwapByte depending on other field values.
202 LoadEntryBinArea(0x0028,0x1201); // R LUT
203 LoadEntryBinArea(0x0028,0x1202); // G LUT
204 LoadEntryBinArea(0x0028,0x1203); // B LUT
206 // Segmented Red Palette Color LUT Data
207 LoadEntryBinArea(0x0028,0x1221);
208 // Segmented Green Palette Color LUT Data
209 LoadEntryBinArea(0x0028,0x1222);
210 // Segmented Blue Palette Color LUT Data
211 LoadEntryBinArea(0x0028,0x1223);
214 //FIXME later : how to use it?
215 SeqEntry *modLutSeq = GetSeqEntry(0x0028,0x3000);
218 SQItem *sqi= modLutSeq->GetFirstSQItem();
221 DataEntry *dataEntry = sqi->GetDataEntry(0x0028,0x3006);
222 if ( dataEntry != 0 )
224 if ( dataEntry->GetLength() != 0 )
226 // FIXME : CTX dependent means : contexted dependant.
227 // see upper comment.
228 LoadEntryBinArea(dataEntry); //LUT Data (CTX dependent)
234 // Force Loading some more elements if user asked to.
237 for (ListElements::iterator it = UserForceLoadList.begin();
238 it != UserForceLoadList.end();
241 gdcmDebugMacro( "Force Load " << std::hex
242 << (*it).Group << "|" <<(*it).Elem );
244 d = GetDocEntry( (*it).Group, (*it).Elem);
248 gdcmWarningMacro( "You asked toForce Load " << std::hex
249 << (*it).Group <<"|"<< (*it).Elem
250 << " that doesn't exist" );
254 LoadDocEntry(d, true);
259 // ----------------------------
260 // Specific code to allow gdcm to read ACR-LibIDO formated images
261 // Note: ACR-LibIDO is an extension of the ACR standard that was
262 // used at CREATIS. For the time being (say a couple of years)
263 // we keep this kludge to allow CREATIS users
264 // reading their old images.
266 // if recognition code tells us we deal with a LibIDO image
267 // we switch lineNumber and columnNumber
270 RecCode = GetEntryString(0x0008, 0x0010); // recognition code (RET)
271 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
272 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
273 // with "little-endian strings"
275 Filetype = ACR_LIBIDO;
276 std::string rows = GetEntryString(0x0028, 0x0010);
277 std::string columns = GetEntryString(0x0028, 0x0011);
278 SetEntryString(columns, 0x0028, 0x0010);
279 SetEntryString(rows , 0x0028, 0x0011);
281 // --- End of ACR-LibIDO kludge ---
287 * \brief Adds a new element we want to load anyway
288 * @param group Group number of the target tag.
289 * @param elem Element number of the target tag.
291 void Document::AddForceLoadElement (uint16_t group, uint16_t elem)
296 UserForceLoadList.push_back(el);
299 * \brief Get the public dictionary used
301 Dict *Document::GetPubDict()
307 * \brief Get the shadow dictionary used
309 Dict *Document::GetShaDict()
315 * \brief Set the shadow dictionary used
316 * @param dict dictionary to use in shadow
318 bool Document::SetShaDict(Dict *dict)
325 * \brief Set the shadow dictionary used
326 * @param dictName name of the dictionary to use in shadow
328 bool Document::SetShaDict(DictKey const &dictName)
330 RefShaDict = Global::GetDicts()->GetDict(dictName);
335 * \brief This predicate tells us whether or not the current Document
336 * was properly parsed and contains at least *one* Dicom Element
337 * (and nothing more, sorry).
338 * @return false when we're 150 % sure it's NOT a Dicom/Acr file,
341 bool Document::IsReadable()
343 if ( Filetype == Unknown )
345 gdcmErrorMacro( "Wrong filetype");
351 gdcmErrorMacro( "No tag in internal hash table.");
359 * \brief Predicate for dicom version 3 file.
360 * @return True when the file is a dicom version 3.
362 bool Document::IsDicomV3()
364 // Checking if Transfer Syntax exists is enough
365 // Anyway, it's too late check if the 'Preamble' was found ...
366 // And ... would it be a rich idea to check ?
367 // (some 'no Preamble' DICOM images exist !)
368 return GetDocEntry(0x0002, 0x0010) != NULL;
372 * \brief Predicate for Papyrus file
373 * Dedicated to whomsoever it may concern
374 * @return True when the file is a Papyrus file.
376 bool Document::IsPapyrus()
378 // check for Papyrus private Sequence
379 DocEntry *e = GetDocEntry(0x0041, 0x1050);
382 // check if it's actually a Sequence
383 if ( !dynamic_cast<SeqEntry*>(e) )
389 * \brief returns the File Type
390 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
391 * @return the FileType code
393 FileType Document::GetFileType()
399 * \brief Accessor to the Transfer Syntax (when present) of the
400 * current document (it internally handles reading the
401 * value from disk when only parsing occured).
402 * @return The encountered Transfer Syntax of the current document, if DICOM.
403 * GDCM_UNKNOWN for ACR-NEMA files (or broken headers ...)
405 std::string Document::GetTransferSyntax()
407 DocEntry *entry = GetDocEntry(0x0002, 0x0010);
413 // The entry might be present but not loaded (parsing and loading
414 // happen at different stages): try loading and proceed with check...
415 LoadDocEntrySafe(entry);
416 if (DataEntry *dataEntry = dynamic_cast<DataEntry *>(entry) )
418 std::string transfer = dataEntry->GetString();
419 // The actual transfer (as read from disk) might be padded. We
420 // first need to remove the potential padding. We can make the
421 // weak assumption that padding was not executed with digits...
422 if ( transfer.length() == 0 )
424 // for brain damaged headers
425 gdcmWarningMacro( "Transfer Syntax has length = 0.");
428 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
430 transfer.erase(transfer.length()-1, 1);
431 if ( transfer.length() == 0 )
433 // for brain damaged headers
434 gdcmWarningMacro( "Transfer Syntax contains no valid character.");
444 * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS
445 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
447 std::string Document::GetTransferSyntaxName()
449 // use the TS (TS : Transfer Syntax)
450 std::string transferSyntax = GetEntryString(0x0002,0x0010);
452 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
454 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
455 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
456 return "Uncompressed ACR-NEMA";
458 if ( transferSyntax == GDCM_UNFOUND )
460 gdcmDebugMacro( "Unfound Transfer Syntax (0002,0010)");
461 return "Uncompressed ACR-NEMA";
464 // we do it only when we need it
465 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
467 // Global::GetTS() is a global static you shall never try to delete it!
471 // --------------- Swap Code ------------------
473 * \brief Swaps the bytes so they agree with the processor order
474 * @return The properly swaped 16 bits integer.
476 uint16_t Document::SwapShort(uint16_t a)
478 if ( SwapCode == 4321 || SwapCode == 2143 )
480 //a = ((( a << 8 ) & 0xff00 ) | (( a >> 8 ) & 0x00ff ) );
482 a = ( a << 8 ) | ( a >> 8 );
488 * \brief Swaps back the bytes of 4-byte long integer accordingly to
490 * @return The properly swaped 32 bits integer.
492 uint32_t Document::SwapLong(uint32_t a)
499 // a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
500 // ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
502 a=( ( a<<24) | ((a<<8) & 0x00ff0000) |
503 ((a>>8) & 0x0000ff00) | (a>>24) );
506 // a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
507 a=( (a<<16) | (a>>16) );
510 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
513 gdcmErrorMacro( "Unexpected swap code:" << SwapCode );
520 * \brief Swaps back the bytes of 8-byte long 'double' accordingly to
522 * @return The properly swaped 64 bits double.
524 double Document::SwapDouble(double a)
528 // There were no 'double' at ACR-NEMA time.
529 // We just have to deal with 'straight Little Endian' and
530 // 'straight Big Endian'
535 char *beg = (char *)&a;
538 for (unsigned int i = 0; i<7; i++)
549 gdcmErrorMacro( "Unexpected swap code:" << SwapCode );
556 // -----------------File I/O ---------------
558 * \brief Tries to open the file \ref Document::Filename and
559 * checks the preamble when existing.
560 * @return The FILE pointer on success.
562 std::ifstream *Document::OpenFile()
564 HasDCMPreamble = false;
565 if (Filename.length() == 0)
572 gdcmDebugMacro( "File already open: " << Filename.c_str());
576 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
579 // Don't user gdcmErrorMacro :
580 // a spurious message will appear when you use, for instance
581 // gdcm::FileHelper *fh = new gdcm::FileHelper( outputFileName );
582 // to create outputFileName.
584 // FIXME : if the upper comment is still usefull
585 // --> the constructor is not so good ...
587 gdcmWarningMacro( "Cannot open file: " << Filename.c_str());
591 //exit(1); // No function is allowed to leave the application instead
592 // of warning the caller
596 Fp->read((char*)&zero, (size_t)2);
603 //-- ACR or DICOM with no Preamble; may start with a Shadow Group --
605 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
606 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
607 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
608 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
610 std::string msg = Util::Format(
611 "ACR/DICOM starting by 0x(%04x) at the beginning of the file\n", zero);
612 // FIXME : is it a Warning message, or a Debug message?
613 gdcmWarningMacro( msg.c_str() );
618 Fp->seekg(126L, std::ios::cur);
619 char dicm[4]; // = {' ',' ',' ',' '};
620 Fp->read(dicm, (size_t)4);
626 if ( memcmp(dicm, "DICM", 4) == 0 )
628 HasDCMPreamble = true;
632 // -- Neither ACR/No Preamble Dicom nor DICOMV3 file
634 // Don't user Warning nor Error, not to polute the output
635 // while directory recursive parsing ...
636 gdcmDebugMacro( "Neither ACR/No Preamble Dicom nor DICOMV3 file: "
637 << Filename.c_str());
642 * \brief closes the file
643 * @return TRUE if the close was successfull
645 bool Document::CloseFile()
657 * \brief Writes in a file all the Entries (Dicom Elements)
658 * @param fp file pointer on an already open file (actually: Output File Stream)
659 * @param filetype Type of the File to be written
660 * (ACR-NEMA, ExplicitVR, ImplicitVR)
662 void Document::WriteContent(std::ofstream *fp, FileType filetype)
664 // Skip if user wants to write an ACR-NEMA file
666 if ( filetype == ImplicitVR || filetype == ExplicitVR ||
669 // writing Dicom File Preamble
670 char filePreamble[128];
671 memset(filePreamble, 0, 128);
672 fp->write(filePreamble, 128);
673 fp->write("DICM", 4);
677 * \todo rewrite later, if really usefull
678 * - 'Group Length' element is optional in DICOM
679 * - but un-updated odd groups lengthes can causes pb
682 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
683 * UpdateGroupLength(false,filetype);
684 * if ( filetype == ACR)
685 * UpdateGroupLength(true,ACR);
687 * --> Computing group length for groups with embeded Sequences
688 * --> was too much tricky / we were [in a hurry / too lazy]
689 * --> We don't write the element 0x0000 (group length)
692 ElementSet::WriteContent(fp, filetype); // This one is recursive
695 // -----------------------------------------
698 * \brief Loads (from disk) the element content
699 * when a string is not suitable
700 * @param group group number of the Entry
701 * @param elem element number of the Entry
703 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
705 // Search the corresponding DocEntry
706 DocEntry *docEntry = GetDocEntry(group, elem);
709 gdcmWarningMacro(std::hex << group << "|" << elem
710 << "doesn't exist" );
713 DataEntry *dataEntry = dynamic_cast<DataEntry *>(docEntry);
716 gdcmWarningMacro(std::hex << group << "|" << elem
717 << "is NOT a DataEntry");
720 LoadEntryBinArea(dataEntry);
724 * \brief Loads (from disk) the element content
725 * when a string is not suitable
726 * @param entry Entry whose binArea is going to be loaded
728 void Document::LoadEntryBinArea(DataEntry *entry)
730 if( entry->GetBinArea() )
737 size_t o =(size_t)entry->GetOffset();
738 Fp->seekg(o, std::ios::beg);
740 size_t l = entry->GetLength();
741 uint8_t *data = new uint8_t[l];
744 gdcmWarningMacro( "Cannot allocate DataEntry content for : "
745 << std::hex << entry->GetGroup()
746 << "|" << entry->GetElement() );
751 Fp->read((char*)data, l);
752 if ( Fp->fail() || Fp->eof() )
755 entry->SetState(DataEntry::STATE_UNREAD);
759 // Swap the data content if necessary
761 unsigned short vrLgth =
762 Global::GetVR()->GetAtomicElementLength(entry->GetVR());
764 // FIXME : trouble expected if we read an ... OW Entry (LUT, etc ..)
765 // if( entry->GetVR() == "OW" )
776 uint16_t *data16 = (uint16_t *)data;
777 for(i=0;i<l/vrLgth;i++)
778 data16[i] = SwapShort(data16[i]);
783 uint32_t *data32 = (uint32_t *)data;
784 for(i=0;i<l/vrLgth;i++)
785 data32[i] = SwapLong(data32[i]);
790 double *data64 = (double *)data;
791 for(i=0;i<l/vrLgth;i++)
792 data64[i] = SwapDouble(data64[i]);
797 entry->SetBinArea(data);
804 * \brief Loads the element while preserving the current
805 * underlying file position indicator as opposed to
806 * LoadDocEntry that modifies it.
807 * @param entry DocEntry whose value will be loaded.
809 void Document::LoadDocEntrySafe(DocEntry *entry)
813 long PositionOnEntry = Fp->tellg();
815 Fp->seekg(PositionOnEntry, std::ios::beg);
820 * \brief Compares two documents, according to \ref DicomDir rules
821 * \warning Does NOT work with ACR-NEMA files
822 * \todo Find a trick to solve the pb (use RET fields ?)
823 * @param document to compare with current one
824 * @return true if 'smaller'
826 bool Document::operator<(Document &document)
829 std::string s1 = GetEntryString(0x0010,0x0010);
830 std::string s2 = document.GetEntryString(0x0010,0x0010);
842 s1 = GetEntryString(0x0010,0x0020);
843 s2 = document.GetEntryString(0x0010,0x0020);
854 // Study Instance UID
855 s1 = GetEntryString(0x0020,0x000d);
856 s2 = document.GetEntryString(0x0020,0x000d);
867 // Serie Instance UID
868 s1 = GetEntryString(0x0020,0x000e);
869 s2 = document.GetEntryString(0x0020,0x000e);
884 //-----------------------------------------------------------------------------
887 * \brief Reads a supposed to be 16 Bits integer
888 * (swaps it depending on processor endianness)
891 uint16_t Document::ReadInt16()
895 Fp->read ((char*)&g, (size_t)2);
898 throw FormatError( "Document::ReadInt16()", " file error." );
902 throw FormatError( "Document::ReadInt16()", "EOF." );
909 * \brief Reads a supposed to be 32 Bits integer
910 * (swaps it depending on processor endianness)
913 uint32_t Document::ReadInt32()
917 Fp->read ((char*)&g, (size_t)4);
920 throw FormatError( "Document::ReadInt32()", " file error." );
924 throw FormatError( "Document::ReadInt32()", "EOF." );
931 * \brief skips bytes inside the source file
932 * \warning NOT end user intended method !
935 void Document::SkipBytes(uint32_t nBytes)
937 //FIXME don't dump the returned value
938 Fp->seekg((long)nBytes, std::ios::cur);
942 * \brief Re-computes the length of the Dicom group 0002.
944 int Document::ComputeGroup0002Length( )
950 bool found0002 = false;
952 // for each zero-level Tag in the DCM Header
953 DocEntry *entry = GetFirstEntry();
956 gr = entry->GetGroup();
962 if ( entry->GetElement() != 0x0000 )
966 //if ( (vr == "OB")||(vr == "OW")||(vr == "UT")||(vr == "SQ"))
967 // (no SQ, OW, UT in group 0x0002;)
970 // explicit VR AND (OB, OW, SQ, UT) : 4 more bytes
974 groupLength += 2 + 2 + 4 + entry->GetLength();
980 entry = GetNextEntry();
985 //-----------------------------------------------------------------------------
988 * \brief Loads all the needed Dictionaries
989 * \warning NOT end user intended method !
991 void Document::Initialize()
993 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
999 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1000 * @param set DocEntrySet we are going to parse ('zero level' or a SQItem)
1001 * @param offset start of parsing
1002 * @param l_max length to parse (meaningless when we are in 'delimitor mode')
1003 * @param delim_mode : whether we are in 'delimitor mode' (l=0xffffff) or not
1005 void Document::ParseDES(DocEntrySet *set, long offset,
1006 long l_max, bool delim_mode)
1008 DocEntry *newDocEntry;
1009 DataEntry *newDataEntry;
1010 SeqEntry *newSeqEntry;
1012 bool used; // will be set to false when something wrong happens to an Entry.
1013 // (Entry will then be deleted)
1014 bool delim_mode_intern = delim_mode;
1016 gdcmDebugMacro( "Enter in ParseDES, delim-mode " << delim_mode
1017 << " at offset " << std::hex << offset );
1020 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1025 newDocEntry = ReadNextDocEntry( );
1027 // Uncoment this cerr line to be able to 'follow' the DocEntries
1028 // when something *very* strange happens
1029 if( Debug::GetDebugFlag() )
1030 std::cerr<<newDocEntry->GetKey()<<" "<<newDocEntry->GetVR()<<std::endl;
1037 // an Item Starter found elsewhere but the first position
1038 // of a SeqEntry means previous entry was a Sequence
1039 // but we didn't get it (private Sequence + Implicit VR)
1040 // we have to backtrack.
1041 if ( !first && newDocEntry->IsItemStarter() )
1043 // Debug message within the method !
1044 newDocEntry = Backtrack(newDocEntry);
1048 PreviousDocEntry = newDocEntry;
1052 newDataEntry = dynamic_cast<DataEntry*>(newDocEntry);
1056 //////////////////////////// DataEntry
1058 vr = newDocEntry->GetVR();
1060 if ( !set->AddEntry( newDataEntry ) )
1062 gdcmDebugMacro( "in ParseDES : cannot add a DataEntry "
1063 << newDataEntry->GetKey()
1065 << newDataEntry->GetOffset() << " )" );
1070 newDataEntry->Delete();
1071 // Load only if we can add (not a duplicate key)
1072 LoadDocEntry( newDataEntry );
1075 if ( newDataEntry->GetElement() == 0x0000 ) // if on group length
1077 if ( newDataEntry->GetGroup()%2 != 0 ) // if Shadow Group
1079 if ( LoadMode & LD_NOSHADOW ) // if user asked to skip shad.gr
1081 std::string strLgrGroup = newDataEntry->GetString();
1083 if ( newDataEntry->IsUnfound() )
1085 lgrGroup = atoi(strLgrGroup.c_str());
1086 Fp->seekg(lgrGroup, std::ios::cur);
1087 //used = false; // never used
1088 RemoveEntry( newDocEntry ); // Remove and delete
1089 // bcc 5.5 is right "assigned a value that's never used"
1097 bool delimitor = newDataEntry->IsItemDelimitor();
1100 (!delim_mode && ((long)(Fp->tellg())-offset) >= l_max) )
1103 newDocEntry->Delete();
1107 // Just to make sure we are at the beginning of next entry.
1108 SkipToNextDocEntry(newDocEntry);
1112 /////////////////////// SeqEntry : VR = "SQ"
1114 unsigned long l = newDocEntry->GetReadLength();
1115 if ( l != 0 ) // don't mess the delim_mode for 'zero-length sequence'
1117 if ( l == 0xffffffff )
1119 delim_mode_intern = true;
1123 delim_mode_intern = false;
1127 if ( (LoadMode & LD_NOSHADOWSEQ) && ! delim_mode_intern )
1129 // User asked to skip SeQuences *only* if they belong to Shadow Group
1130 if ( newDocEntry->GetGroup()%2 != 0 )
1132 Fp->seekg( l, std::ios::cur);
1133 newDocEntry->Delete(); // Delete, not in the set
1137 if ( (LoadMode & LD_NOSEQ) && ! delim_mode_intern )
1139 // User asked to skip *any* SeQuence
1140 Fp->seekg( l, std::ios::cur);
1141 newDocEntry->Delete(); // Delete, not in the set
1144 // delay the dynamic cast as late as possible
1145 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1147 // no other way to create the Delimitor ...
1148 newSeqEntry->SetDelimitorMode( delim_mode_intern );
1150 // At the top of the hierarchy, stands a Document. When "set"
1151 // is a Document, then we are building the first depth level.
1152 // Hence the SeqEntry we are building simply has a depth
1154 if ( set == this ) // ( dynamic_cast< Document* > ( set ) )
1156 newSeqEntry->SetDepthLevel( 1 );
1158 // But when "set" is already a SQItem, we are building a nested
1159 // sequence, and hence the depth level of the new SeqEntry
1160 // we are building, is one level deeper:
1162 // time waste hunting
1163 else if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1165 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1169 { // Don't try to parse zero-length sequences
1171 gdcmDebugMacro( "Entry in ParseSQ, delim " << delim_mode_intern
1172 << " at offset " << std::hex
1173 << newDocEntry->GetOffset() );
1175 ParseSQ( newSeqEntry,
1176 newDocEntry->GetOffset(),
1177 l, delim_mode_intern);
1179 gdcmDebugMacro( "Exit from ParseSQ, delim " << delim_mode_intern);
1182 if ( !set->AddEntry( newSeqEntry ) )
1184 gdcmWarningMacro( "in ParseDES : cannot add a SeqEntry "
1185 << newSeqEntry->GetKey()
1187 << newSeqEntry->GetOffset() << " )" );
1192 newDocEntry->Delete();
1195 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1198 newDocEntry->Delete();
1201 } // end SeqEntry : VR = "SQ"
1205 newDocEntry->Delete();
1209 gdcmDebugMacro( "Exit from ParseDES, delim-mode " << delim_mode );
1213 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1214 * @return parsed length for this level
1216 void Document::ParseSQ( SeqEntry *seqEntry,
1217 long offset, long l_max, bool delim_mode)
1219 int SQItemNumber = 0;
1221 long offsetStartCurrentSQItem = offset;
1225 // the first time, we read the fff0,e000 of the first SQItem
1226 DocEntry *newDocEntry = ReadNextDocEntry();
1230 gdcmWarningMacro("in ParseSQ : should never get here!");
1235 if ( newDocEntry->IsSequenceDelimitor() )
1237 seqEntry->SetDelimitationItem( newDocEntry );
1238 newDocEntry->Delete();
1242 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1244 newDocEntry->Delete();
1247 // create the current SQItem
1248 SQItem *itemSQ = SQItem::New( seqEntry->GetDepthLevel() );
1249 unsigned int l = newDocEntry->GetReadLength();
1251 if ( l == 0xffffffff )
1260 // remove fff0,e000, created out of the SQItem
1261 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg);
1262 // fill up the current SQItem, starting at the beginning of fff0,e000
1264 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
1266 offsetStartCurrentSQItem = Fp->tellg();
1268 seqEntry->AddSQItem( itemSQ, SQItemNumber );
1270 newDocEntry->Delete();
1272 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1280 * \brief When a private Sequence + Implicit VR is encountered
1281 * we cannot guess it's a Sequence till we find the first
1282 * Item Starter. We then backtrack to do the job.
1283 * @param docEntry Item Starter that warned us
1285 DocEntry *Document::Backtrack(DocEntry *docEntry)
1287 // delete the Item Starter, built erroneously out of any Sequence
1288 // it's not yet in the HTable/chained list
1291 // Get all info we can from PreviousDocEntry
1292 uint16_t group = PreviousDocEntry->GetGroup();
1293 uint16_t elem = PreviousDocEntry->GetElement();
1294 uint32_t lgt = PreviousDocEntry->GetLength();
1295 long offset = PreviousDocEntry->GetOffset();
1297 gdcmDebugMacro( "Backtrack :" << std::hex << group
1299 << " at offset " << offset );
1300 RemoveEntry( PreviousDocEntry );
1302 // forge the Seq Entry
1303 DocEntry *newEntry = NewSeqEntry(group, elem);
1304 newEntry->SetLength(lgt);
1305 newEntry->SetOffset(offset);
1307 // Move back to the beginning of the Sequence
1308 Fp->seekg( 0, std::ios::beg);
1309 Fp->seekg(offset, std::ios::cur);
1315 * \brief Loads (or not) the element content depending if its length exceeds
1316 * or not the value specified with Document::SetMaxSizeLoadEntry()
1317 * @param entry Header Entry (Dicom Element) to be dealt with
1318 * @param forceLoad whether you want to force loading of 'long' elements
1320 void Document::LoadDocEntry(DocEntry *entry, bool forceLoad)
1322 uint16_t group = entry->GetGroup();
1323 uint16_t elem = entry->GetElement();
1324 const VRKey &vr = entry->GetVR();
1325 uint32_t length = entry->GetLength();
1327 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1329 // A SeQuence "contains" a set of Elements.
1330 // (fffe e000) tells us an Element is beginning
1331 // (fffe e00d) tells us an Element just ended
1332 // (fffe e0dd) tells us the current SeQuence just ended
1334 // (fffe 0000) is an 'impossible' tag value,
1335 // found in MR-PHILIPS-16-Multi-Seq.dcm
1337 if ( (group == 0xfffe && elem != 0x0000 ) || vr == "SQ" )
1339 // NO more value field for SQ !
1343 DataEntry *dataEntryPtr = dynamic_cast< DataEntry* >(entry);
1349 // When the length is zero things are easy:
1352 dataEntryPtr->SetBinArea(NULL,true);
1356 // The elements whose length is bigger than the specified upper bound
1359 std::ostringstream s;
1363 if (length > MaxSizeLoadEntry)
1365 dataEntryPtr->SetBinArea(NULL,true);
1366 dataEntryPtr->SetState(DataEntry::STATE_NOTLOADED);
1368 // to be sure we are at the end of the value ...
1369 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1375 LoadEntryBinArea(dataEntryPtr); // last one, not to erase length !
1379 * \brief Find the value Length of the passed Doc Entry
1380 * @param entry Header Entry whose length of the value shall be loaded.
1382 void Document::FindDocEntryLength( DocEntry *entry )
1383 throw ( FormatError )
1385 const VRKey &vr = entry->GetVR();
1388 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1390 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UT"
1393 // The following reserved two bytes (see PS 3.5-2003, section
1394 // "7.1.2 Data element structure with explicit vr", p 27) must be
1395 // skipped before proceeding on reading the length on 4 bytes.
1396 Fp->seekg( 2L, std::ios::cur);
1397 uint32_t length32 = ReadInt32();
1399 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1404 lengthOB = FindDocEntryLengthOBOrOW();
1406 catch ( FormatUnexpected )
1408 // Computing the length failed (this happens with broken
1409 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1410 // chance to get the pixels by deciding the element goes
1411 // until the end of the file. Hence we artificially fix the
1412 // the length and proceed.
1413 gdcmWarningMacro( " Computing the length failed for " <<
1414 entry->GetKey() <<" in " <<GetFileName());
1416 long currentPosition = Fp->tellg();
1417 Fp->seekg(0L,std::ios::end);
1419 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1420 Fp->seekg(currentPosition, std::ios::beg);
1422 entry->SetReadLength(lengthUntilEOF);
1423 entry->SetLength(lengthUntilEOF);
1426 entry->SetReadLength(lengthOB);
1427 entry->SetLength(lengthOB);
1430 FixDocEntryFoundLength(entry, length32);
1434 // Length is encoded on 2 bytes.
1435 length16 = ReadInt16();
1437 // 0xffff means that we deal with 'No Length' Sequence
1438 // or 'No Length' SQItem
1439 if ( length16 == 0xffff)
1443 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1448 // Either implicit VR or a non DICOM conformal (see note below) explicit
1449 // VR that ommited the VR of (at least) this element. Farts happen.
1450 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1451 // on Data elements "Implicit and Explicit VR Data Elements shall
1452 // not coexist in a Data Set and Data Sets nested within it".]
1453 // Length is on 4 bytes.
1455 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1456 // even if Transfer Syntax is 'Implicit VR ...'
1457 // --> Except for 'Implicit VR Big Endian Transfer Syntax GE Private'
1459 FixDocEntryFoundLength( entry, ReadInt32() );
1465 * \brief Find the Length till the next sequence delimiter
1466 * \warning NOT end user intended method !
1469 uint32_t Document::FindDocEntryLengthOBOrOW()
1470 throw( FormatUnexpected )
1472 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1473 long positionOnEntry = Fp->tellg();
1474 bool foundSequenceDelimiter = false;
1475 uint32_t totalLength = 0;
1477 while ( !foundSequenceDelimiter )
1483 group = ReadInt16();
1486 catch ( FormatError )
1488 throw FormatError("Unexpected end of file encountered during ",
1489 "Document::FindDocEntryLengthOBOrOW()");
1491 // We have to decount the group and element we just read
1493 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
1495 long filePosition = Fp->tellg();
1497 "Neither an Item tag nor a Sequence delimiter tag on :"
1498 << std::hex << group << " , " << elem
1499 << ") -before- position x(" << filePosition << ")" );
1501 Fp->seekg(positionOnEntry, std::ios::beg);
1502 throw FormatUnexpected(
1503 "Neither an Item tag nor a Sequence delimiter tag.");
1505 if ( elem == 0xe0dd )
1507 foundSequenceDelimiter = true;
1509 uint32_t itemLength = ReadInt32();
1510 // We add 4 bytes since we just read the ItemLength with ReadInt32
1511 totalLength += itemLength + 4;
1512 SkipBytes(itemLength);
1514 if ( foundSequenceDelimiter )
1519 Fp->seekg( positionOnEntry, std::ios::beg);
1524 * \brief Find the Value Representation of the current Dicom Element.
1525 * @return Value Representation of the current Entry
1527 VRKey Document::FindDocEntryVR()
1529 if ( Filetype != ExplicitVR )
1530 return GDCM_VRUNKNOWN;
1532 long positionOnEntry = Fp->tellg();
1533 // Warning: we believe this is explicit VR (Value Representation) because
1534 // we used a heuristic that found "UL" in the first tag and/or
1535 // 'Transfer Syntax' told us it is.
1536 // Alas this doesn't guarantee that all the tags will be in explicit VR.
1537 // In some cases one finds implicit VR tags mixed within an explicit VR file.
1538 // Hence we make sure the present tag is in explicit VR and try to fix things
1539 // if it happens not to be the case.
1542 Fp->read(&(vr[0]),(size_t)2);
1544 if ( !CheckDocEntryVR(vr) )
1546 gdcmWarningMacro( "Unknown VR " << std::hex << "0x("
1547 << (unsigned int)vr[0] << "|" << (unsigned int)vr[1]
1548 << ") at offset :" << positionOnEntry );
1549 Fp->seekg(positionOnEntry, std::ios::beg);
1550 return GDCM_VRUNKNOWN;
1556 * \brief Check the correspondance between the VR of the header entry
1557 * and the taken VR. If they are different, the header entry is
1558 * updated with the new VR.
1559 * @param vr Dicom Value Representation
1560 * @return false if the VR is incorrect or if the VR isn't referenced
1561 * otherwise, it returns true
1563 bool Document::CheckDocEntryVR(const VRKey &vr)
1565 return Global::GetVR()->IsValidVR(vr);
1569 * \brief Skip a given Header Entry
1570 * \warning NOT end user intended method !
1571 * @param entry entry to skip
1573 void Document::SkipDocEntry(DocEntry *entry)
1575 SkipBytes(entry->GetLength());
1579 * \brief Skips to the beginning of the next Header Entry
1580 * \warning NOT end user intended method !
1581 * @param currentDocEntry entry to skip
1583 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1585 int l = currentDocEntry->GetReadLength();
1586 if ( l == -1 ) // length = 0xffff shouldn't appear here ...
1587 // ... but PMS imagers happen !
1589 Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg);
1590 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1592 Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur);
1597 * \brief When the length of an element value is obviously wrong (because
1598 * the parser went Jabberwocky) one can hope improving things by
1599 * applying some heuristics.
1600 * @param entry entry to check
1601 * @param foundLength first assumption about length
1603 void Document::FixDocEntryFoundLength(DocEntry *entry,
1604 uint32_t foundLength)
1606 entry->SetReadLength( foundLength );// will be updated only if a bug is found
1607 if ( foundLength == 0xffffffff)
1612 uint16_t gr = entry->GetGroup();
1613 uint16_t elem = entry->GetElement();
1615 if ( foundLength % 2)
1617 gdcmWarningMacro( "Warning : Tag with uneven length " << foundLength
1618 << " in x(" << std::hex << gr << "," << elem <<")");
1621 //////// Fix for some naughty General Electric images.
1622 // Allthough not recent many such GE corrupted images are still present
1623 // on Creatis hard disks. Hence this fix shall remain when such images
1624 // are no longer in use (we are talking a few years, here)...
1625 // Note: XMedCon probably uses such a trick since it is able to read
1626 // those pesky GE images ...
1627 if ( foundLength == 13)
1629 // Only happens for this length !
1630 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1633 entry->SetReadLength(10); // a bug is to be fixed !?
1637 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1638 // Occurence of such images is quite low (unless one leaves close to a
1639 // 'Leonardo' source. Hence, one might consider commenting out the
1640 // following fix on efficiency reasons.
1641 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1644 entry->SetReadLength(4); // a bug is to be fixed !
1647 else if ( entry->GetVR() == "SQ" )
1649 foundLength = 0; // ReadLength is unchanged
1652 //////// We encountered a 'delimiter' element i.e. a tag of the form
1653 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1654 // taken into account.
1655 else if ( gr == 0xfffe )
1657 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1658 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1659 // causes extra troubles...
1660 if ( entry->GetElement() != 0x0000 )
1666 foundLength=12; // to skip the mess that follows this bugged Tag !
1669 entry->SetLength(foundLength);
1673 * \brief Apply some heuristics to predict whether the considered
1674 * element value contains/represents an integer or not.
1675 * @param entry The element value on which to apply the predicate.
1676 * @return The result of the heuristical predicate.
1678 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1680 uint16_t elem = entry->GetElement();
1681 uint16_t group = entry->GetGroup();
1682 const VRKey &vr = entry->GetVR();
1683 uint32_t length = entry->GetLength();
1685 // When we have some semantics on the element we just read, and if we
1686 // a priori know we are dealing with an integer, then we shall be
1687 // able to swap it's element value properly.
1688 if ( elem == 0 ) // This is the group length of the group
1696 // Although this should never happen, still some images have a
1697 // corrupted group length [e.g. have a glance at offset x(8336) of
1698 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm.
1699 // Since for dicom compliant and well behaved headers, the present
1700 // test is useless (and might even look a bit paranoid), when we
1701 // encounter such an ill-formed image, we simply display a warning
1702 // message and proceed on parsing (while crossing fingers).
1703 long filePosition = Fp->tellg();
1704 gdcmWarningMacro( "Erroneous Group Length element length on : ("
1705 << std::hex << group << " , " << elem
1706 << ") -before- position x(" << filePosition << ")"
1707 << "lgt : " << length );
1711 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1719 * \brief Discover what the swap code is (among little endian, big endian,
1720 * bad little endian, bad big endian).
1722 * @return false when we are absolutely sure
1723 * it's neither ACR-NEMA nor DICOM
1724 * true when we hope ours assuptions are OK
1726 bool Document::CheckSwap()
1733 // First, compare HostByteOrder and NetworkByteOrder in order to
1734 // determine if we shall need to swap bytes (i.e. the Endian type).
1735 bool net2host = Util::IsCurrentProcessorBigEndian();
1737 // The easiest case is the one of a 'true' DICOM header, we just have
1738 // to look for the string "DICM" inside the file preamble.
1741 char *entCur = deb + 128;
1742 if ( memcmp(entCur, "DICM", (size_t)4) == 0 )
1744 gdcmDebugMacro( "Looks like DICOM Version3 (preamble + DCM)" );
1746 // Group 0002 should always be VR, and the first element 0000
1747 // Let's be carefull (so many wrong headers ...)
1748 // and determine the value representation (VR) :
1749 // Let's skip to the first element (0002,0000) and check there if we find
1750 // "UL" - or "OB" if the 1st one is (0002,0001) -,
1751 // in which case we (almost) know it is explicit VR.
1752 // WARNING: if it happens to be implicit VR then what we will read
1753 // is the length of the group. If this ascii representation of this
1754 // length happens to be "UL" then we shall believe it is explicit VR.
1755 // We need to skip :
1756 // * the 128 bytes of File Preamble (often padded with zeroes),
1757 // * the 4 bytes of "DICM" string,
1758 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
1759 // i.e. a total of 136 bytes.
1762 // group 0x0002 *is always* Explicit VR Sometimes ,
1763 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
1764 // *Implicit* VR (see former 'gdcmData/icone.dcm')
1766 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
1767 memcmp(entCur, "OB", (size_t)2) == 0 ||
1768 memcmp(entCur, "UI", (size_t)2) == 0 ||
1769 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
1770 // when Write DCM *adds*
1772 // Use Document::dicom_vr to test all the possibilities
1773 // instead of just checking for UL, OB and UI !? group 0000
1775 Filetype = ExplicitVR;
1776 gdcmDebugMacro( "Group 0002 : Explicit Value Representation");
1780 Filetype = ImplicitVR;
1781 gdcmWarningMacro( "Group 0002 :Not an explicit Value Representation;"
1782 << "Looks like a bugged Header!");
1788 gdcmDebugMacro( "HostByteOrder != NetworkByteOrder, SwapCode = 4321");
1793 gdcmDebugMacro( "HostByteOrder = NetworkByteOrder, SwapCode = 1234");
1796 // Position the file position indicator at first tag
1797 // (i.e. after the file preamble and the "DICM" string).
1799 Fp->seekg(0, std::ios::beg); // FIXME : Is it usefull?
1801 Fp->seekg ( 132L, std::ios::beg);
1803 } // ------------------------------- End of DicomV3 ----------------
1805 // Alas, this is not a DicomV3 file and whatever happens there is no file
1806 // preamble. We can reset the file position indicator to where the data
1807 // is (i.e. the beginning of the file).
1809 gdcmWarningMacro( "Not a Kosher DICOM Version3 file (no preamble)");
1811 Fp->seekg(0, std::ios::beg);
1813 // Let's check 'No Preamble Dicom File' :
1814 // Should start with group 0x0002
1815 // and be Explicit Value Representation
1817 s16 = *((uint16_t *)(deb));
1830 if ( SwapCode != 0 )
1832 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
1833 memcmp(entCur, "OB", (size_t)2) == 0 ||
1834 memcmp(entCur, "UI", (size_t)2) == 0 ||
1835 memcmp(entCur, "SH", (size_t)2) == 0 ||
1836 memcmp(entCur, "AE", (size_t)2) == 0 ||
1837 memcmp(entCur, "OB", (size_t)2) == 0 )
1839 Filetype = ExplicitVR;
1840 gdcmDebugMacro( "Group 0002 : Explicit Value Representation");
1844 // ------------------------------- End of 'No Preamble' DicomV3 -------------
1846 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
1847 // By clean we mean that the length of the first group is written down.
1848 // If this is the case and since the length of the first group HAS to be
1849 // four (bytes), then determining the proper swap code is straightforward.
1852 // We assume the array of char we are considering contains the binary
1853 // representation of a 32 bits integer. Hence the following dirty
1855 s32 = *((uint32_t *)(entCur));
1875 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
1876 // It is time for despaired wild guesses.
1877 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
1878 // i.e. the 'group length' element is not present :
1880 // check the supposed-to-be 'group number'
1881 // in ( 0x0001 .. 0x0008 )
1882 // to determine ' SwapCode' value .
1883 // Only 0 or 4321 will be possible
1884 // (no oportunity to check for the formerly well known
1885 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
1886 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc-3, 4, ..., 8-)
1887 // the file IS NOT ACR-NEMA nor DICOM V3
1888 // Find a trick to tell it the caller...
1890 s16 = *((uint16_t *)(deb));
1917 gdcmWarningMacro("ACR/NEMA unfound swap info (Hopeless !)");
1925 * \brief Change the Byte Swap code.
1927 void Document::SwitchByteSwapCode()
1929 gdcmDebugMacro( "Switching Byte Swap code from "<< SwapCode
1930 << " at: 0x" << std::hex << Fp->tellg() );
1931 if ( SwapCode == 1234 )
1935 else if ( SwapCode == 4321 )
1939 else if ( SwapCode == 3412 )
1943 else if ( SwapCode == 2143 )
1947 gdcmDebugMacro( " Into: "<< SwapCode );
1951 * \brief during parsing, Header Elements too long are not loaded in memory
1952 * @param newSize new size
1954 void Document::SetMaxSizeLoadEntry(long newSize)
1960 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
1962 MaxSizeLoadEntry = 0xffffffff;
1965 MaxSizeLoadEntry = newSize;
1969 * \brief Read the next tag WITHOUT loading it's value
1970 * (read the 'Group Number', the 'Element Number',
1971 * gets the Dict Entry
1972 * gets the VR, gets the length, gets the offset value)
1973 * @return On succes : the newly created DocEntry, NULL on failure.
1975 DocEntry *Document::ReadNextDocEntry()
1982 group = ReadInt16();
1985 catch ( FormatError )
1987 // We reached the EOF (or an error occured) therefore
1988 // header parsing has to be considered as finished.
1992 // Sometimes file contains groups of tags with reversed endianess.
1993 HandleBrokenEndian(group, elem);
1995 // In 'true DICOM' files Group 0002 is always little endian
1996 if ( HasDCMPreamble )
1997 HandleOutOfGroup0002(group, elem);
1999 VRKey vr = FindDocEntryVR();
2003 if ( vr == GDCM_VRUNKNOWN )
2005 if ( elem == 0x0000 ) // Group Length
2007 realVR = "UL"; // must be UL
2009 else if (group%2 == 1 && (elem >= 0x0010 && elem <=0x00ff ))
2011 // DICOM PS 3-5 7.8.1 a) states that those
2012 // (gggg-0010->00FF where gggg is odd) attributes have to be LO
2017 DictEntry *dictEntry = GetDictEntry(group,elem);
2020 realVR = dictEntry->GetVR();
2021 dictEntry->Unregister();
2025 // gdcmDebugMacro( "Found VR: " << vr << " / Real VR: " << realVR );
2028 if ( Global::GetVR()->IsVROfSequence(realVR) )
2029 newEntry = NewSeqEntry(group, elem);
2032 newEntry = NewDataEntry(group, elem, realVR);
2033 static_cast<DataEntry *>(newEntry)->SetState(DataEntry::STATE_NOTLOADED);
2036 if ( vr == GDCM_VRUNKNOWN )
2038 if ( Filetype == ExplicitVR )
2040 // We thought this was explicit VR, but we end up with an
2041 // implicit VR tag. Let's backtrack.
2042 if ( newEntry->GetGroup() != 0xfffe )
2045 int offset = Fp->tellg();
2047 "Entry (%04x,%04x) at x(%x) should be Explicit VR\n",
2048 newEntry->GetGroup(), newEntry->GetElement(), offset );
2049 gdcmWarningMacro( msg.c_str() );
2052 newEntry->SetImplicitVR();
2057 FindDocEntryLength(newEntry);
2059 catch ( FormatError )
2066 newEntry->SetOffset(Fp->tellg());
2072 * \brief Handle broken private tag from Philips NTSCAN
2073 * where the endianess is being switched to BigEndian
2074 * for no apparent reason
2077 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
2079 // Endian reversion.
2080 // Some files contain groups of tags with reversed endianess.
2081 static int reversedEndian = 0;
2082 // try to fix endian switching in the middle of headers
2083 if ((group == 0xfeff) && (elem == 0x00e0))
2085 // start endian swap mark for group found
2087 SwitchByteSwapCode();
2092 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2094 // end of reversed endian group
2096 SwitchByteSwapCode();
2098 else if (group == 0xfeff && elem == 0xdde0)
2100 // reversed Sequence Terminator found
2101 // probabely a bug in the header !
2102 // Do what you want, it breaks !
2104 //SwitchByteSwapCode();
2105 gdcmWarningMacro( "Should never get here! reversed Sequence Terminator!" );
2110 else if (group == 0xfffe && elem == 0xe0dd)
2112 gdcmDebugMacro( "Straight Sequence Terminator." );
2117 * \brief Group 0002 is always coded Little Endian
2118 * whatever Transfer Syntax is
2121 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2123 // Endian reversion.
2124 // Some files contain groups of tags with reversed endianess.
2125 if ( !Group0002Parsed && group != 0x0002)
2127 Group0002Parsed = true;
2128 // we just came out of group 0002
2129 // if Transfer Syntax is Big Endian we have to change CheckSwap
2131 std::string ts = GetTransferSyntax();
2132 if ( ts == GDCM_UNKNOWN )
2134 gdcmDebugMacro("True DICOM File, with NO Transfer Syntax (?!) " );
2137 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2139 gdcmWarningMacro("True DICOM File, with illegal Transfer Syntax: ["
2144 // Group 0002 is always 'Explicit ...'
2145 // even when Transfer Syntax says 'Implicit ..."
2147 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) ==
2148 TS::ImplicitVRLittleEndian )
2150 Filetype = ImplicitVR;
2153 // FIXME Strangely, this works with
2154 //'Implicit VR BigEndian Transfer Syntax (GE Private)
2156 // --> Probabely normal, since we considered we never have
2157 // to trust manufacturers.
2158 // (we find very often 'Implicit VR' tag,
2159 // even when Transfer Syntax tells us it's Explicit ...
2160 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) ==
2161 TS::ExplicitVRBigEndian )
2163 gdcmDebugMacro("Transfer Syntax Name = ["
2164 << GetTransferSyntaxName() << "]" );
2165 SwitchByteSwapCode();
2166 group = SwapShort(group);
2167 elem = SwapShort(elem);
2172 //-----------------------------------------------------------------------------
2175 //-----------------------------------------------------------------------------
2176 } // end namespace gdcm