1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/02/11 19:19:05 $
7 Version: $Revision: 1.228 $
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 "gdcmValEntry.h"
21 #include "gdcmBinEntry.h"
22 #include "gdcmSeqEntry.h"
23 #include "gdcmGlobal.h"
25 #include "gdcmDebug.h"
27 #include "gdcmDictSet.h"
28 #include "gdcmDocEntrySet.h"
29 #include "gdcmSQItem.h"
36 #if defined(_MSC_VER) || defined(__BORLANDC__) || defined(__MINGW32__)
40 #ifdef CMAKE_HAVE_NETINET_IN_H
41 #include <netinet/in.h>
46 //-----------------------------------------------------------------------------
47 // Refer to Document::CheckSwap()
48 //const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
50 // Refer to Document::SetMaxSizeLoadEntry()
51 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
52 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
54 //-----------------------------------------------------------------------------
55 // Constructor / Destructor
56 // Constructors and destructors are protected to avoid user to invoke directly
59 * @param filename 'Document' (File or DicomDir) to be opened for parsing
61 Document::Document( std::string const &filename )
64 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
74 Group0002Parsed = false;
76 gdcmWarningMacro( "Starting parsing of file: " << Filename.c_str());
78 Fp->seekg(0, std::ios::end);
79 long lgt = Fp->tellg();
81 Fp->seekg(0, std::ios::beg);
84 long beg = Fp->tellg();
87 ParseDES( this, beg, lgt, false); // Loading is done during parsing
89 Fp->seekg( 0, std::ios::beg);
91 // Load 'non string' values
93 std::string PhotometricInterpretation = GetEntryValue(0x0028,0x0004);
94 if( PhotometricInterpretation == "PALETTE COLOR " )
96 LoadEntryBinArea(0x0028,0x1200); // gray LUT
98 /// The tags refered by the three following lines used to be CORRECTLY
99 /// defined as having an US Value Representation in the public
100 /// dictionary. BUT the semantics implied by the three following
101 /// lines state that the corresponding tag contents are in fact
102 /// the ones of a BinEntry.
103 /// In order to fix things "Quick and Dirty" the dictionary was
104 /// altered on PURPOSE but now contains a WRONG value.
105 /// In order to fix things and restore the dictionary to its
106 /// correct value, one needs to decided of the semantics by deciding
107 /// whether the following tags are either:
108 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
109 /// also used as BinEntry, which requires the proper conversion,
110 /// - OW, and hence loaded as BinEntry, but afterwards also used
111 /// as ValEntry, which requires the proper conversion.
112 LoadEntryBinArea(0x0028,0x1201); // R LUT
113 LoadEntryBinArea(0x0028,0x1202); // G LUT
114 LoadEntryBinArea(0x0028,0x1203); // B LUT
116 // Segmented Red Palette Color LUT Data
117 LoadEntryBinArea(0x0028,0x1221);
118 // Segmented Green Palette Color LUT Data
119 LoadEntryBinArea(0x0028,0x1222);
120 // Segmented Blue Palette Color LUT Data
121 LoadEntryBinArea(0x0028,0x1223);
123 //FIXME later : how to use it?
124 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
128 // ----------------------------
129 // Specific code to allow gdcm to read ACR-LibIDO formated images
130 // Note: ACR-LibIDO is an extension of the ACR standard that was
131 // used at CREATIS. For the time being (say a couple years)
132 // we keep this kludge to allow a smooth move to gdcm for
133 // CREATIS developpers (sorry folks).
135 // if recognition code tells us we deal with a LibIDO image
136 // we switch lineNumber and columnNumber
139 RecCode = GetEntryValue(0x0008, 0x0010); // recognition code (RET)
140 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
141 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
142 // with "little-endian strings"
144 Filetype = ACR_LIBIDO;
145 std::string rows = GetEntryValue(0x0028, 0x0010);
146 std::string columns = GetEntryValue(0x0028, 0x0011);
147 SetValEntry(columns, 0x0028, 0x0010);
148 SetValEntry(rows , 0x0028, 0x0011);
150 // --- End of ACR-LibIDO kludge ---
154 * \brief This default constructor doesn't parse the file. You should
155 * then invoke \ref Document::SetFileName and then the parsing.
162 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
165 Filetype = ExplicitVR;
166 Group0002Parsed = false;
170 * \brief Canonical destructor.
172 Document::~Document ()
178 //-----------------------------------------------------------------------------
181 * \brief Get the public dictionary used
183 Dict *Document::GetPubDict()
189 * \brief Get the shadow dictionary used
191 Dict *Document::GetShaDict()
197 * \brief Set the shadow dictionary used
198 * @param dict dictionary to use in shadow
200 bool Document::SetShaDict(Dict *dict)
207 * \brief Set the shadow dictionary used
208 * @param dictName name of the dictionary to use in shadow
210 bool Document::SetShaDict(DictKey const &dictName)
212 RefShaDict = Global::GetDicts()->GetDict(dictName);
217 * \brief This predicate, based on hopefully reasonable heuristics,
218 * decides whether or not the current Document was properly parsed
219 * and contains the mandatory information for being considered as
220 * a well formed and usable Dicom/Acr File.
221 * @return true when Document is the one of a reasonable Dicom/Acr file,
224 bool Document::IsReadable()
226 if( Filetype == Unknown)
228 gdcmWarningMacro( "Wrong filetype");
234 gdcmWarningMacro( "No tag in internal hash table.");
242 * \brief Predicate for dicom version 3 file.
243 * @return True when the file is a dicom version 3.
245 bool Document::IsDicomV3()
247 // Checking if Transfer Syntax exists is enough
248 // Anyway, it's to late check if the 'Preamble' was found ...
249 // And ... would it be a rich idea to check ?
250 // (some 'no Preamble' DICOM images exist !)
251 return GetDocEntry(0x0002, 0x0010) != NULL;
255 * \brief Predicate for Papyrus file
256 * Dedicated to whomsoever it may concern
257 * @return True when the file is a Papyrus file.
259 bool Document::IsPapyrus()
261 // check for Papyrus private Sequence
262 DocEntry *e = GetDocEntry(0x0041, 0x1050);
265 // check if it's actually a Sequence
266 if ( !dynamic_cast<SeqEntry*>(e) )
272 * \brief returns the File Type
273 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
274 * @return the FileType code
276 FileType Document::GetFileType()
282 * \brief Accessor to the Transfer Syntax (when present) of the
283 * current document (it internally handles reading the
284 * value from disk when only parsing occured).
285 * @return The encountered Transfer Syntax of the current document.
287 std::string Document::GetTransferSyntax()
289 DocEntry *entry = GetDocEntry(0x0002, 0x0010);
295 // The entry might be present but not loaded (parsing and loading
296 // happen at different stages): try loading and proceed with check...
297 LoadDocEntrySafe(entry);
298 if (ValEntry *valEntry = dynamic_cast< ValEntry* >(entry) )
300 std::string transfer = valEntry->GetValue();
301 // The actual transfer (as read from disk) might be padded. We
302 // first need to remove the potential padding. We can make the
303 // weak assumption that padding was not executed with digits...
304 if ( transfer.length() == 0 )
306 // for brain damaged headers
309 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
311 transfer.erase(transfer.length()-1, 1);
319 * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS
320 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
322 std::string Document::GetTransferSyntaxName()
324 // use the TS (TS : Transfer Syntax)
325 std::string transferSyntax = GetEntryValue(0x0002,0x0010);
327 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
329 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
330 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
331 return "Uncompressed ACR-NEMA";
333 if ( transferSyntax == GDCM_UNFOUND )
335 gdcmWarningMacro( "Unfound Transfer Syntax (0002,0010)");
336 return "Uncompressed ACR-NEMA";
339 // we do it only when we need it
340 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
342 // Global::GetTS() is a global static you shall never try to delete it!
346 // --------------- Swap Code ------------------
348 * \brief Swaps the bytes so they agree with the processor order
349 * @return The properly swaped 16 bits integer.
351 uint16_t Document::SwapShort(uint16_t a)
353 if ( SwapCode == 4321 || SwapCode == 2143 )
355 a = ((( a << 8 ) & 0xff00 ) | (( a >> 8 ) & 0x00ff ) );
361 * \brief Swaps back the bytes of 4-byte long integer accordingly to
363 * @return The properly swaped 32 bits integer.
365 uint32_t Document::SwapLong(uint32_t a)
372 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
373 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
376 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
379 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
382 gdcmErrorMacro( "Unset swap code:" << SwapCode );
389 // -----------------File I/O ---------------
391 * \brief Tries to open the file \ref Document::Filename and
392 * checks the preamble when existing.
393 * @return The FILE pointer on success.
395 std::ifstream *Document::OpenFile()
397 HasDCMPreamble = false;
398 if (Filename.length() == 0)
405 gdcmWarningMacro( "File already open: " << Filename.c_str());
409 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
412 gdcmDebugMacro( "Cannot open file: " << Filename.c_str());
419 Fp->read((char*)&zero, (size_t)2);
426 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
428 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
429 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
430 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
431 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
434 = Util::Format("ACR/DICOM with no preamble: (%04x)\n", zero);
435 gdcmWarningMacro( msg.c_str() );
440 Fp->seekg(126L, std::ios::cur);
441 char dicm[4] = {' ',' ',' ',' '};
442 Fp->read(dicm, (size_t)4);
448 if( memcmp(dicm, "DICM", 4) == 0 )
450 HasDCMPreamble = true;
455 gdcmWarningMacro( "Not DICOM/ACR (missing preamble)" << Filename.c_str());
461 * \brief closes the file
462 * @return TRUE if the close was successfull
464 bool Document::CloseFile()
476 * \brief Writes in a file all the Header Entries (Dicom Elements)
477 * @param fp file pointer on an already open file (actually: Output File Stream)
478 * @param filetype Type of the File to be written
479 * (ACR-NEMA, ExplicitVR, ImplicitVR)
480 * @return Always true.
482 void Document::WriteContent(std::ofstream *fp, FileType filetype)
484 // \TODO move the following lines (and a lot of others, to be written)
485 // to a future function CheckAndCorrectHeader
487 // (necessary if user wants to write a DICOM V3 file
488 // starting from an ACR-NEMA (V2) Header
490 if ( filetype == ImplicitVR || filetype == ExplicitVR )
492 // writing Dicom File Preamble
493 char filePreamble[128];
494 memset(filePreamble, 0, 128);
495 fp->write(filePreamble, 128);
496 fp->write("DICM", 4);
500 * \todo rewrite later, if really usefull
501 * - 'Group Length' element is optional in DICOM
502 * - but un-updated odd groups lengthes can causes pb
505 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
506 * UpdateGroupLength(false,filetype);
507 * if ( filetype == ACR)
508 * UpdateGroupLength(true,ACR);
511 ElementSet::WriteContent(fp, filetype); // This one is recursive
514 // -----------------------------------------
517 * \brief Loads (from disk) the element content
518 * when a string is not suitable
519 * @param group group number of the Entry
520 * @param elem element number of the Entry
522 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
524 // Search the corresponding DocEntry
525 DocEntry *docElement = GetDocEntry(group, elem);
529 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
533 LoadEntryBinArea(binElement);
537 * \brief Loads (from disk) the element content
538 * when a string is not suitable
539 * @param elem Entry whose binArea is going to be loaded
541 void Document::LoadEntryBinArea(BinEntry *elem)
543 if(elem->GetBinArea())
550 size_t o =(size_t)elem->GetOffset();
551 Fp->seekg(o, std::ios::beg);
553 size_t l = elem->GetLength();
554 uint8_t *a = new uint8_t[l];
557 gdcmWarningMacro( "Cannot allocate BinEntry content");
561 Fp->read((char*)a, l);
562 if( Fp->fail() || Fp->eof())
575 * \brief Loads the element while preserving the current
576 * underlying file position indicator as opposed to
577 * LoadDocEntry that modifies it.
578 * @param entry DocEntry whose value will be loaded.
580 void Document::LoadDocEntrySafe(DocEntry *entry)
584 long PositionOnEntry = Fp->tellg();
586 Fp->seekg(PositionOnEntry, std::ios::beg);
591 * \brief Compares two documents, according to \ref DicomDir rules
592 * \warning Does NOT work with ACR-NEMA files
593 * \todo Find a trick to solve the pb (use RET fields ?)
594 * @param document to compare with current one
595 * @return true if 'smaller'
597 bool Document::operator<(Document &document)
600 std::string s1 = GetEntryValue(0x0010,0x0010);
601 std::string s2 = document.GetEntryValue(0x0010,0x0010);
613 s1 = GetEntryValue(0x0010,0x0020);
614 s2 = document.GetEntryValue(0x0010,0x0020);
625 // Study Instance UID
626 s1 = GetEntryValue(0x0020,0x000d);
627 s2 = document.GetEntryValue(0x0020,0x000d);
638 // Serie Instance UID
639 s1 = GetEntryValue(0x0020,0x000e);
640 s2 = document.GetEntryValue(0x0020,0x000e);
655 //-----------------------------------------------------------------------------
658 * \brief Reads a supposed to be 16 Bits integer
659 * (swaps it depending on processor endianness)
662 uint16_t Document::ReadInt16()
666 Fp->read ((char*)&g, (size_t)2);
669 throw FormatError( "Document::ReadInt16()", " file error." );
673 throw FormatError( "Document::ReadInt16()", "EOF." );
680 * \brief Reads a supposed to be 32 Bits integer
681 * (swaps it depending on processor endianness)
684 uint32_t Document::ReadInt32()
688 Fp->read ((char*)&g, (size_t)4);
691 throw FormatError( "Document::ReadInt32()", " file error." );
695 throw FormatError( "Document::ReadInt32()", "EOF." );
702 * \brief skips bytes inside the source file
703 * \warning NOT end user intended method !
706 void Document::SkipBytes(uint32_t nBytes)
708 //FIXME don't dump the returned value
709 Fp->seekg((long)nBytes, std::ios::cur);
713 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
714 * @param filetype Type of the File to be written
716 int Document::ComputeGroup0002Length( FileType filetype )
722 bool found0002 = false;
724 // for each zero-level Tag in the DCM Header
725 DocEntry *entry = GetFirstEntry();
728 gr = entry->GetGroup();
734 if( entry->GetElement() != 0x0000 )
738 if( filetype == ExplicitVR )
740 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
742 // explicit VR AND OB, OW, SQ : 4 more bytes
746 groupLength += 2 + 2 + 4 + entry->GetLength();
752 entry = GetNextEntry();
757 //-----------------------------------------------------------------------------
760 * \brief Loads all the needed Dictionaries
761 * \warning NOT end user intended method !
763 void Document::Initialize()
765 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
771 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
772 * @return length of the parsed set.
774 void Document::ParseDES(DocEntrySet *set, long offset,
775 long l_max, bool delim_mode)
777 DocEntry *newDocEntry = 0;
778 ValEntry *newValEntry;
779 BinEntry *newBinEntry;
780 SeqEntry *newSeqEntry;
786 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
792 newDocEntry = ReadNextDocEntry( );
799 vr = newDocEntry->GetVR();
800 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
801 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
802 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
804 if ( newValEntry || newBinEntry )
808 if ( Filetype == ExplicitVR &&
809 !Global::GetVR()->IsVROfBinaryRepresentable(vr) )
811 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
812 gdcmWarningMacro( std::hex << newDocEntry->GetGroup()
813 << "|" << newDocEntry->GetElement()
814 << " : Neither Valentry, nor BinEntry."
815 "Probably unknown VR.");
818 //////////////////// BinEntry or UNKOWN VR:
819 // When "this" is a Document the Key is simply of the
820 // form ( group, elem )...
821 if ( dynamic_cast< Document* > ( set ) )
823 newBinEntry->SetKey( newBinEntry->GetKey() );
825 // but when "this" is a SQItem, we are inserting this new
826 // valEntry in a sequence item, and the key has the
827 // generalized form (refer to \ref BaseTagKey):
828 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
830 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
831 + newBinEntry->GetKey() );
834 LoadDocEntry( newBinEntry );
835 if( !set->AddEntry( newBinEntry ) )
837 //Expect big troubles if here
838 //delete newBinEntry;
844 /////////////////////// ValEntry
845 // When "set" is a Document, then we are at the top of the
846 // hierarchy and the Key is simply of the form ( group, elem )...
847 if ( dynamic_cast< Document* > ( set ) )
849 newValEntry->SetKey( newValEntry->GetKey() );
851 // ...but when "set" is a SQItem, we are inserting this new
852 // valEntry in a sequence item. Hence the key has the
853 // generalized form (refer to \ref BaseTagKey):
854 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
856 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
857 + newValEntry->GetKey() );
860 LoadDocEntry( newValEntry );
861 bool delimitor=newValEntry->IsItemDelimitor();
862 if( !set->AddEntry( newValEntry ) )
864 // If here expect big troubles
865 //delete newValEntry; //otherwise mem leak
875 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
883 // Just to make sure we are at the beginning of next entry.
884 SkipToNextDocEntry(newDocEntry);
889 unsigned long l = newDocEntry->GetReadLength();
890 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
892 if ( l == 0xffffffff )
901 // no other way to create it ...
902 newSeqEntry->SetDelimitorMode( delim_mode );
904 // At the top of the hierarchy, stands a Document. When "set"
905 // is a Document, then we are building the first depth level.
906 // Hence the SeqEntry we are building simply has a depth
908 if (/*Document *dummy =*/ dynamic_cast< Document* > ( set ) )
911 newSeqEntry->SetDepthLevel( 1 );
912 newSeqEntry->SetKey( newSeqEntry->GetKey() );
914 // But when "set" is already a SQItem, we are building a nested
915 // sequence, and hence the depth level of the new SeqEntry
916 // we are building, is one level deeper:
917 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
919 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
920 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
921 + newSeqEntry->GetKey() );
925 { // Don't try to parse zero-length sequences
926 ParseSQ( newSeqEntry,
927 newDocEntry->GetOffset(),
930 if( !set->AddEntry( newSeqEntry ) )
934 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
948 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
949 * @return parsed length for this level
951 void Document::ParseSQ( SeqEntry *seqEntry,
952 long offset, long l_max, bool delim_mode)
954 int SQItemNumber = 0;
956 long offsetStartCurrentSQItem = offset;
960 // the first time, we read the fff0,e000 of the first SQItem
961 DocEntry *newDocEntry = ReadNextDocEntry();
965 // FIXME Should warn user
970 if ( newDocEntry->IsSequenceDelimitor() )
972 seqEntry->SetDelimitationItem( newDocEntry );
976 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
981 // create the current SQItem
982 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
983 std::ostringstream newBase;
984 newBase << seqEntry->GetKey()
988 itemSQ->SetBaseTagKey( newBase.str() );
989 unsigned int l = newDocEntry->GetReadLength();
991 if ( l == 0xffffffff )
1000 // Let's try :------------
1001 // remove fff0,e000, created out of the SQItem
1003 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg);
1004 // fill up the current SQItem, starting at the beginning of fff0,e000
1005 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
1006 offsetStartCurrentSQItem = Fp->tellg();
1007 // end try -----------------
1009 seqEntry->AddSQItem( itemSQ, SQItemNumber );
1011 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1019 * \brief Loads the element content if its length doesn't exceed
1020 * the value specified with Document::SetMaxSizeLoadEntry()
1021 * @param entry Header Entry (Dicom Element) to be dealt with
1023 void Document::LoadDocEntry(DocEntry *entry)
1025 uint16_t group = entry->GetGroup();
1026 std::string vr = entry->GetVR();
1027 uint32_t length = entry->GetLength();
1029 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1031 // A SeQuence "contains" a set of Elements.
1032 // (fffe e000) tells us an Element is beginning
1033 // (fffe e00d) tells us an Element just ended
1034 // (fffe e0dd) tells us the current SeQuence just ended
1035 if( group == 0xfffe )
1037 // NO more value field for SQ !
1041 // When the length is zero things are easy:
1044 ((ValEntry *)entry)->SetValue("");
1048 // The elements whose length is bigger than the specified upper bound
1049 // are not loaded. Instead we leave a short notice of the offset of
1050 // the element content and it's length.
1052 std::ostringstream s;
1053 if (length > MaxSizeLoadEntry)
1055 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1057 //s << "gdcm::NotLoaded (BinEntry)";
1058 s << GDCM_NOTLOADED;
1059 s << " Address:" << (long)entry->GetOffset();
1060 s << " Length:" << entry->GetLength();
1061 s << " x(" << std::hex << entry->GetLength() << ")";
1062 binEntryPtr->SetValue(s.str());
1064 // Be carefull : a BinEntry IS_A ValEntry ...
1065 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1067 // s << "gdcm::NotLoaded. (ValEntry)";
1068 s << GDCM_NOTLOADED;
1069 s << " Address:" << (long)entry->GetOffset();
1070 s << " Length:" << entry->GetLength();
1071 s << " x(" << std::hex << entry->GetLength() << ")";
1072 valEntryPtr->SetValue(s.str());
1077 gdcmErrorMacro( "MaxSizeLoadEntry exceeded, neither a BinEntry "
1078 << "nor a ValEntry ?! Should never print that !" );
1081 // to be sure we are at the end of the value ...
1082 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1087 // When we find a BinEntry not very much can be done :
1088 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1090 s << GDCM_BINLOADED;
1091 binEntryPtr->SetValue(s.str());
1092 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1096 if ( IsDocEntryAnInteger(entry) )
1100 // When short integer(s) are expected, read and convert the following
1101 // n *two characters properly i.e. consider them as short integers as
1102 // opposed to strings.
1103 // Elements with Value Multiplicity > 1
1104 // contain a set of integers (not a single one)
1105 if (vr == "US" || vr == "SS")
1108 NewInt = ReadInt16();
1112 for (int i=1; i < nbInt; i++)
1115 NewInt = ReadInt16();
1120 // See above comment on multiple integers (mutatis mutandis).
1121 else if (vr == "UL" || vr == "SL")
1124 NewInt = ReadInt32();
1128 for (int i=1; i < nbInt; i++)
1131 NewInt = ReadInt32();
1136 #ifdef GDCM_NO_ANSI_STRING_STREAM
1137 s << std::ends; // to avoid oddities on Solaris
1138 #endif //GDCM_NO_ANSI_STRING_STREAM
1140 ((ValEntry *)entry)->SetValue(s.str());
1144 // FIXME: We need an additional byte for storing \0 that is not on disk
1145 char *str = new char[length+1];
1146 Fp->read(str, (size_t)length);
1147 str[length] = '\0'; //this is only useful when length is odd
1148 // Special DicomString call to properly handle \0 and even length
1149 std::string newValue;
1152 newValue = Util::DicomString(str, length+1);
1153 gdcmWarningMacro("Warning: bad length: " << length <<
1154 ",For string :" << newValue.c_str());
1155 // Since we change the length of string update it length
1156 //entry->SetReadLength(length+1);
1160 newValue = Util::DicomString(str, length);
1164 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1166 if ( Fp->fail() || Fp->eof())
1168 gdcmWarningMacro("Unread element value");
1169 valEntry->SetValue(GDCM_UNREAD);
1175 // Because of correspondance with the VR dic
1176 valEntry->SetValue(newValue);
1180 valEntry->SetValue(newValue);
1185 gdcmErrorMacro( "Should have a ValEntry, here !");
1190 * \brief Find the value Length of the passed Header Entry
1191 * @param entry Header Entry whose length of the value shall be loaded.
1193 void Document::FindDocEntryLength( DocEntry *entry )
1194 throw ( FormatError )
1196 std::string vr = entry->GetVR();
1199 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1201 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1203 // The following reserved two bytes (see PS 3.5-2003, section
1204 // "7.1.2 Data element structure with explicit vr", p 27) must be
1205 // skipped before proceeding on reading the length on 4 bytes.
1206 Fp->seekg( 2L, std::ios::cur);
1207 uint32_t length32 = ReadInt32();
1209 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1214 lengthOB = FindDocEntryLengthOBOrOW();
1216 catch ( FormatUnexpected )
1218 // Computing the length failed (this happens with broken
1219 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1220 // chance to get the pixels by deciding the element goes
1221 // until the end of the file. Hence we artificially fix the
1222 // the length and proceed.
1223 long currentPosition = Fp->tellg();
1224 Fp->seekg(0L,std::ios::end);
1226 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1227 Fp->seekg(currentPosition, std::ios::beg);
1229 entry->SetReadLength(lengthUntilEOF);
1230 entry->SetLength(lengthUntilEOF);
1233 entry->SetReadLength(lengthOB);
1234 entry->SetLength(lengthOB);
1237 FixDocEntryFoundLength(entry, length32);
1241 // Length is encoded on 2 bytes.
1242 length16 = ReadInt16();
1244 // FIXME : This heuristic supposes that the first group following
1245 // group 0002 *has* and element 0000.
1246 // BUT ... Element 0000 is optionnal :-(
1249 // Fixed using : HandleOutOfGroup0002()
1250 // (first hereafter strategy ...)
1252 // We can tell the current file is encoded in big endian (like
1253 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1254 // and it's value is the one of the encoding of a big endian file.
1255 // In order to deal with such big endian encoded files, we have
1256 // (at least) two strategies:
1257 // * when we load the "Transfer Syntax" tag with value of big endian
1258 // encoding, we raise the proper flags. Then we wait for the end
1259 // of the META group (0x0002) among which is "Transfer Syntax",
1260 // before switching the swap code to big endian. We have to postpone
1261 // the switching of the swap code since the META group is fully encoded
1262 // in little endian, and big endian coding only starts at the next
1263 // group. The corresponding code can be hard to analyse and adds
1264 // many additional unnecessary tests for regular tags.
1265 // * the second strategy consists in waiting for trouble, that shall
1266 // appear when we find the first group with big endian encoding. This
1267 // is easy to detect since the length of a "Group Length" tag (the
1268 // ones with zero as element number) has to be of 4 (0x0004). When we
1269 // encounter 1024 (0x0400) chances are the encoding changed and we
1270 // found a group with big endian encoding.
1271 //---> Unfortunately, element 0000 is optional.
1272 //---> This will not work when missing!
1273 // We shall use this second strategy. In order to make sure that we
1274 // can interpret the presence of an apparently big endian encoded
1275 // length of a "Group Length" without committing a big mistake, we
1276 // add an additional check: we look in the already parsed elements
1277 // for the presence of a "Transfer Syntax" whose value has to be "big
1278 // endian encoding". When this is the case, chances are we have got our
1279 // hands on a big endian encoded file: we switch the swap code to
1280 // big endian and proceed...
1282 // if ( element == 0x0000 && length16 == 0x0400 )
1284 // std::string ts = GetTransferSyntax();
1285 // if ( Global::GetTS()->GetSpecialTransferSyntax(ts)
1286 // != TS::ExplicitVRBigEndian )
1288 // throw FormatError( "Document::FindDocEntryLength()",
1289 // " not explicit VR." );
1293 // SwitchByteSwapCode();
1295 // // Restore the unproperly loaded values i.e. the group, the element
1296 // // and the dictionary entry depending on them.
1297 // uint16_t correctGroup = SwapShort( entry->GetGroup() );
1298 // uint16_t correctElem = SwapShort( entry->GetElement() );
1299 // DictEntry *newTag = GetDictEntry( correctGroup, correctElem );
1302 // // This correct tag is not in the dictionary. Create a new one.
1303 // newTag = NewVirtualDictEntry(correctGroup, correctElem);
1305 // // FIXME this can create a memory leaks on the old entry that be
1306 // // left unreferenced.
1307 // entry->SetDictEntry( newTag );
1310 // 0xffff means that we deal with 'No Length' Sequence
1311 // or 'No Length' SQItem
1312 if ( length16 == 0xffff)
1316 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1321 // Either implicit VR or a non DICOM conformal (see note below) explicit
1322 // VR that ommited the VR of (at least) this element. Farts happen.
1323 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1324 // on Data elements "Implicit and Explicit VR Data Elements shall
1325 // not coexist in a Data Set and Data Sets nested within it".]
1326 // Length is on 4 bytes.
1328 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1329 // even if Transfer Syntax is 'Implicit VR ...'
1331 FixDocEntryFoundLength( entry, ReadInt32() );
1337 * \brief Find the Length till the next sequence delimiter
1338 * \warning NOT end user intended method !
1341 uint32_t Document::FindDocEntryLengthOBOrOW()
1342 throw( FormatUnexpected )
1344 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1345 long positionOnEntry = Fp->tellg();
1346 bool foundSequenceDelimiter = false;
1347 uint32_t totalLength = 0;
1349 while ( !foundSequenceDelimiter )
1355 group = ReadInt16();
1358 catch ( FormatError )
1360 throw FormatError("Unexpected end of file encountered during ",
1361 "Document::FindDocEntryLengthOBOrOW()");
1363 // We have to decount the group and element we just read
1365 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
1367 long filePosition = Fp->tellg();
1368 gdcmWarningMacro( "Neither an Item tag nor a Sequence delimiter tag on :"
1369 << std::hex << group << " , " << elem
1370 << ") -before- position x(" << filePosition << ")" );
1372 Fp->seekg(positionOnEntry, std::ios::beg);
1373 throw FormatUnexpected( "Neither an Item tag nor a Sequence delimiter tag.");
1375 if ( elem == 0xe0dd )
1377 foundSequenceDelimiter = true;
1379 uint32_t itemLength = ReadInt32();
1380 // We add 4 bytes since we just read the ItemLength with ReadInt32
1381 totalLength += itemLength + 4;
1382 SkipBytes(itemLength);
1384 if ( foundSequenceDelimiter )
1389 Fp->seekg( positionOnEntry, std::ios::beg);
1394 * \brief Find the Value Representation of the current Dicom Element.
1395 * @return Value Representation of the current Entry
1397 std::string Document::FindDocEntryVR()
1399 if ( Filetype != ExplicitVR )
1400 return GDCM_UNKNOWN;
1402 long positionOnEntry = Fp->tellg();
1403 // Warning: we believe this is explicit VR (Value Representation) because
1404 // we used a heuristic that found "UL" in the first tag. Alas this
1405 // doesn't guarantee that all the tags will be in explicit VR. In some
1406 // cases (see e-film filtered files) one finds implicit VR tags mixed
1407 // within an explicit VR file. Hence we make sure the present tag
1408 // is in explicit VR and try to fix things if it happens not to be
1412 Fp->read (vr, (size_t)2);
1415 if( !CheckDocEntryVR(vr) )
1417 Fp->seekg(positionOnEntry, std::ios::beg);
1418 return GDCM_UNKNOWN;
1424 * \brief Check the correspondance between the VR of the header entry
1425 * and the taken VR. If they are different, the header entry is
1426 * updated with the new VR.
1427 * @param vr Dicom Value Representation
1428 * @return false if the VR is incorrect of if the VR isn't referenced
1429 * otherwise, it returns true
1431 bool Document::CheckDocEntryVR(VRKey vr)
1433 if ( !Global::GetVR()->IsValidVR(vr) )
1440 * \brief Get the transformed value of the header entry. The VR value
1441 * is used to define the transformation to operate on the value
1442 * \warning NOT end user intended method !
1443 * @param entry entry to tranform
1444 * @return Transformed entry value
1446 std::string Document::GetDocEntryValue(DocEntry *entry)
1448 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1450 std::string val = ((ValEntry *)entry)->GetValue();
1451 std::string vr = entry->GetVR();
1452 uint32_t length = entry->GetLength();
1453 std::ostringstream s;
1456 // When short integer(s) are expected, read and convert the following
1457 // n * 2 bytes properly i.e. as a multivaluated strings
1458 // (each single value is separated fromthe next one by '\'
1459 // as usual for standard multivaluated filels
1460 // Elements with Value Multiplicity > 1
1461 // contain a set of short integers (not a single one)
1463 if( vr == "US" || vr == "SS" )
1468 for (int i=0; i < nbInt; i++)
1474 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1475 newInt16 = SwapShort( newInt16 );
1480 // When integer(s) are expected, read and convert the following
1481 // n * 4 bytes properly i.e. as a multivaluated strings
1482 // (each single value is separated fromthe next one by '\'
1483 // as usual for standard multivaluated filels
1484 // Elements with Value Multiplicity > 1
1485 // contain a set of integers (not a single one)
1486 else if( vr == "UL" || vr == "SL" )
1491 for (int i=0; i < nbInt; i++)
1497 newInt32 = ( val[4*i+0] & 0xFF )
1498 + (( val[4*i+1] & 0xFF ) << 8 )
1499 + (( val[4*i+2] & 0xFF ) << 16 )
1500 + (( val[4*i+3] & 0xFF ) << 24 );
1501 newInt32 = SwapLong( newInt32 );
1505 #ifdef GDCM_NO_ANSI_STRING_STREAM
1506 s << std::ends; // to avoid oddities on Solaris
1507 #endif //GDCM_NO_ANSI_STRING_STREAM
1510 return ((ValEntry *)entry)->GetValue();
1514 * \brief Get the reverse transformed value of the header entry. The VR
1515 * value is used to define the reverse transformation to operate on
1517 * \warning NOT end user intended method !
1518 * @param entry Entry to reverse transform
1519 * @return Reverse transformed entry value
1521 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1523 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1525 std::string vr = entry->GetVR();
1526 std::vector<std::string> tokens;
1527 std::ostringstream s;
1529 if ( vr == "US" || vr == "SS" )
1533 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1534 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1535 for (unsigned int i=0; i<tokens.size(); i++)
1537 newInt16 = atoi(tokens[i].c_str());
1538 s << ( newInt16 & 0xFF )
1539 << (( newInt16 >> 8 ) & 0xFF );
1543 if ( vr == "UL" || vr == "SL")
1547 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1548 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1549 for (unsigned int i=0; i<tokens.size();i++)
1551 newInt32 = atoi(tokens[i].c_str());
1552 s << (char)( newInt32 & 0xFF )
1553 << (char)(( newInt32 >> 8 ) & 0xFF )
1554 << (char)(( newInt32 >> 16 ) & 0xFF )
1555 << (char)(( newInt32 >> 24 ) & 0xFF );
1560 #ifdef GDCM_NO_ANSI_STRING_STREAM
1561 s << std::ends; // to avoid oddities on Solaris
1562 #endif //GDCM_NO_ANSI_STRING_STREAM
1566 return ((ValEntry *)entry)->GetValue();
1570 * \brief Skip a given Header Entry
1571 * \warning NOT end user intended method !
1572 * @param entry entry to skip
1574 void Document::SkipDocEntry(DocEntry *entry)
1576 SkipBytes(entry->GetLength());
1580 * \brief Skips to the beginning of the next Header Entry
1581 * \warning NOT end user intended method !
1582 * @param currentDocEntry entry to skip
1584 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1586 Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg);
1587 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1588 Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur);
1592 * \brief When the length of an element value is obviously wrong (because
1593 * the parser went Jabberwocky) one can hope improving things by
1594 * applying some heuristics.
1595 * @param entry entry to check
1596 * @param foundLength first assumption about length
1598 void Document::FixDocEntryFoundLength(DocEntry *entry,
1599 uint32_t foundLength)
1601 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1602 if ( foundLength == 0xffffffff)
1607 uint16_t gr = entry->GetGroup();
1608 uint16_t elem = entry->GetElement();
1610 if ( foundLength % 2)
1612 gdcmWarningMacro( "Warning : Tag with uneven length " << foundLength
1613 << " in x(" << std::hex << gr << "," << elem <<")");
1616 //////// Fix for some naughty General Electric images.
1617 // Allthough not recent many such GE corrupted images are still present
1618 // on Creatis hard disks. Hence this fix shall remain when such images
1619 // are no longer in use (we are talking a few years, here)...
1620 // Note: XMedCom probably uses such a trick since it is able to read
1621 // those pesky GE images ...
1622 if ( foundLength == 13)
1624 // Only happens for this length !
1625 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1628 entry->SetReadLength(10); // a bug is to be fixed !?
1632 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1633 // Occurence of such images is quite low (unless one leaves close to a
1634 // 'Leonardo' source. Hence, one might consider commenting out the
1635 // following fix on efficiency reasons.
1636 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1639 entry->SetReadLength(4); // a bug is to be fixed !?
1642 else if ( entry->GetVR() == "SQ" )
1644 foundLength = 0; // ReadLength is unchanged
1647 //////// We encountered a 'delimiter' element i.e. a tag of the form
1648 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1649 // taken into account.
1650 else if( gr == 0xfffe )
1652 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1653 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1654 // causes extra troubles...
1655 if( entry->GetElement() != 0x0000 )
1660 entry->SetLength(foundLength);
1664 * \brief Apply some heuristics to predict whether the considered
1665 * element value contains/represents an integer or not.
1666 * @param entry The element value on which to apply the predicate.
1667 * @return The result of the heuristical predicate.
1669 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1671 uint16_t elem = entry->GetElement();
1672 uint16_t group = entry->GetGroup();
1673 const std::string &vr = entry->GetVR();
1674 uint32_t length = entry->GetLength();
1676 // When we have some semantics on the element we just read, and if we
1677 // a priori know we are dealing with an integer, then we shall be
1678 // able to swap it's element value properly.
1679 if ( elem == 0 ) // This is the group length of the group
1687 // Allthough this should never happen, still some images have a
1688 // corrupted group length [e.g. have a glance at offset x(8336) of
1689 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
1690 // Since for dicom compliant and well behaved headers, the present
1691 // test is useless (and might even look a bit paranoid), when we
1692 // encounter such an ill-formed image, we simply display a warning
1693 // message and proceed on parsing (while crossing fingers).
1694 long filePosition = Fp->tellg();
1695 gdcmWarningMacro( "Erroneous Group Length element length on : ("
1696 << std::hex << group << " , " << elem
1697 << ") -before- position x(" << filePosition << ")"
1698 << "lgt : " << length );
1702 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1710 * \brief Discover what the swap code is (among little endian, big endian,
1711 * bad little endian, bad big endian).
1713 * @return false when we are absolutely sure
1714 * it's neither ACR-NEMA nor DICOM
1715 * true when we hope ours assuptions are OK
1717 bool Document::CheckSwap()
1719 // The only guaranted way of finding the swap code is to find a
1720 // group tag since we know it's length has to be of four bytes i.e.
1721 // 0x00000004. Finding the swap code in then straigthforward. Trouble
1722 // occurs when we can't find such group...
1729 // First, compare HostByteOrder and NetworkByteOrder in order to
1730 // determine if we shall need to swap bytes (i.e. the Endian type).
1732 uint32_t x = 4; // x : for ntohs
1733 bool net2host = (x == ntohs(x));// true when HostByteOrder is the same as NetworkByteOrder
1735 // The easiest case is the one of a 'true' DICOM header, we just have
1736 // to look for the string "DICM" inside the file preamble.
1739 char *entCur = deb + 128;
1740 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
1742 gdcmWarningMacro( "Looks like DICOM Version3 (preamble + DCM)" );
1744 // Group 0002 should always be VR, and the first element 0000
1745 // Let's be carefull (so many wrong headers ...)
1746 // and determine the value representation (VR) :
1747 // Let's skip to the first element (0002,0000) and check there if we find
1748 // "UL" - or "OB" if the 1st one is (0002,0001) -,
1749 // in which case we (almost) know it is explicit VR.
1750 // WARNING: if it happens to be implicit VR then what we will read
1751 // is the length of the group. If this ascii representation of this
1752 // length happens to be "UL" then we shall believe it is explicit VR.
1753 // We need to skip :
1754 // * the 128 bytes of File Preamble (often padded with zeroes),
1755 // * the 4 bytes of "DICM" string,
1756 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
1757 // i.e. a total of 136 bytes.
1760 // group 0x0002 *is always* Explicit VR Sometimes ,
1761 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
1762 // *Implicit* VR (see former 'gdcmData/icone.dcm')
1764 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
1765 memcmp(entCur, "OB", (size_t)2) == 0 ||
1766 memcmp(entCur, "UI", (size_t)2) == 0 ||
1767 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
1768 // when Write DCM *adds*
1770 // Use Document::dicom_vr to test all the possibilities
1771 // instead of just checking for UL, OB and UI !? group 0000
1773 Filetype = ExplicitVR;
1774 gdcmWarningMacro( "Group 0002 : Explicit Value Representation");
1778 Filetype = ImplicitVR;
1779 gdcmWarningMacro( "Group 0002 :Not an explicit Value Representation;"
1780 << "Looks like a bugged Header!");
1786 gdcmWarningMacro( "HostByteOrder != NetworkByteOrder");
1791 gdcmWarningMacro( "HostByteOrder = NetworkByteOrder");
1794 // Position the file position indicator at first tag
1795 // (i.e. after the file preamble and the "DICM" string).
1796 Fp->seekg(0, std::ios::beg);
1797 Fp->seekg ( 132L, std::ios::beg);
1801 // Alas, this is not a DicomV3 file and whatever happens there is no file
1802 // preamble. We can reset the file position indicator to where the data
1803 // is (i.e. the beginning of the file).
1804 gdcmWarningMacro( "Not a DICOM Version3 file");
1805 Fp->seekg(0, std::ios::beg);
1807 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
1808 // By clean we mean that the length of the first tag is written down.
1809 // If this is the case and since the length of the first group HAS to be
1810 // four (bytes), then determining the proper swap code is straightforward.
1813 // We assume the array of char we are considering contains the binary
1814 // representation of a 32 bits integer. Hence the following dirty
1816 s32 = *((uint32_t *)(entCur));
1837 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
1838 // It is time for despaired wild guesses.
1839 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
1840 // i.e. the 'group length' element is not present :
1842 // check the supposed-to-be 'group number'
1843 // in ( 0x0001 .. 0x0008 )
1844 // to determine ' SwapCode' value .
1845 // Only 0 or 4321 will be possible
1846 // (no oportunity to check for the formerly well known
1847 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
1848 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
1849 // the file IS NOT ACR-NEMA nor DICOM V3
1850 // Find a trick to tell it the caller...
1852 s16 = *((uint16_t *)(deb));
1879 gdcmWarningMacro( "ACR/NEMA unfound swap info (Really hopeless !)");
1887 * \brief Change the Byte Swap code.
1889 void Document::SwitchByteSwapCode()
1891 gdcmWarningMacro( "Switching Byte Swap code from "<< SwapCode);
1892 if ( SwapCode == 1234 )
1896 else if ( SwapCode == 4321 )
1900 else if ( SwapCode == 3412 )
1904 else if ( SwapCode == 2143 )
1911 * \brief during parsing, Header Elements too long are not loaded in memory
1912 * @param newSize new size
1914 void Document::SetMaxSizeLoadEntry(long newSize)
1920 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
1922 MaxSizeLoadEntry = 0xffffffff;
1925 MaxSizeLoadEntry = newSize;
1929 * \brief Header Elements too long will not be printed
1930 * See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
1931 * @param newSize new size
1933 void Document::SetMaxSizePrintEntry(long newSize)
1939 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
1941 MaxSizePrintEntry = 0xffffffff;
1944 MaxSizePrintEntry = newSize;
1949 * \brief Read the next tag but WITHOUT loading it's value
1950 * (read the 'Group Number', the 'Element Number',
1951 * gets the Dict Entry
1952 * gets the VR, gets the length, gets the offset value)
1953 * @return On succes the newly created DocEntry, NULL on failure.
1955 DocEntry *Document::ReadNextDocEntry()
1962 group = ReadInt16();
1965 catch ( FormatError e )
1967 // We reached the EOF (or an error occured) therefore
1968 // header parsing has to be considered as finished.
1972 // Sometimes file contains groups of tags with reversed endianess.
1973 HandleBrokenEndian(group, elem);
1975 // In 'true DICOM' files Group 0002 is always little endian
1976 if ( HasDCMPreamble )
1977 HandleOutOfGroup0002(group, elem);
1979 std::string vr = FindDocEntryVR();
1980 std::string realVR = vr;
1982 if( vr == GDCM_UNKNOWN)
1984 DictEntry *dictEntry = GetDictEntry(group,elem);
1986 realVR = dictEntry->GetVR();
1990 if( Global::GetVR()->IsVROfSequence(realVR) )
1991 newEntry = NewSeqEntry(group, elem);
1992 else if( Global::GetVR()->IsVROfStringRepresentable(realVR) )
1993 newEntry = NewValEntry(group, elem,vr);
1995 newEntry = NewBinEntry(group, elem,vr);
1997 if( vr == GDCM_UNKNOWN )
1999 if( Filetype == ExplicitVR )
2001 // We thought this was explicit VR, but we end up with an
2002 // implicit VR tag. Let's backtrack.
2003 if ( newEntry->GetGroup() != 0xfffe )
2006 msg = Util::Format("Entry (%04x,%04x) should be Explicit VR\n",
2007 newEntry->GetGroup(), newEntry->GetElement());
2008 gdcmWarningMacro( msg.c_str() );
2011 newEntry->SetImplicitVR();
2016 FindDocEntryLength(newEntry);
2018 catch ( FormatError e )
2026 newEntry->SetOffset(Fp->tellg());
2032 * \brief Handle broken private tag from Philips NTSCAN
2033 * where the endianess is being switch to BigEndian for no
2037 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
2039 // Endian reversion. Some files contain groups of tags with reversed endianess.
2040 static int reversedEndian = 0;
2041 // try to fix endian switching in the middle of headers
2042 if ((group == 0xfeff) && (elem == 0x00e0))
2044 // start endian swap mark for group found
2046 SwitchByteSwapCode();
2051 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2053 // end of reversed endian group
2055 SwitchByteSwapCode();
2060 * \brief Group 0002 is always coded Little Endian
2061 * whatever Transfer Syntax is
2064 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2066 // Endian reversion. Some files contain groups of tags with reversed endianess.
2067 if ( !Group0002Parsed && group != 0x0002)
2069 Group0002Parsed = true;
2070 // we just came out of group 0002
2071 // if Transfer syntax is Big Endian we have to change CheckSwap
2073 std::string ts = GetTransferSyntax();
2074 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2076 gdcmWarningMacro("True DICOM File, with NO Tansfer Syntax: " << ts );
2080 // Group 0002 is always 'Explicit ...' enven when Transfer Syntax says 'Implicit ..."
2082 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ImplicitVRLittleEndian )
2084 Filetype = ImplicitVR;
2087 // FIXME Strangely, this works with
2088 //'Implicit VR Transfer Syntax (GE Private)
2089 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ExplicitVRBigEndian )
2091 gdcmWarningMacro("Transfer Syntax Name = ["
2092 << GetTransferSyntaxName() << "]" );
2093 SwitchByteSwapCode();
2094 group = SwapShort(group);
2095 elem = SwapShort(elem);
2100 //-----------------------------------------------------------------------------
2103 //-----------------------------------------------------------------------------
2104 } // end namespace gdcm