1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/18 16:23:52 $
7 Version: $Revision: 1.197 $
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 "gdcmException.h"
28 #include "gdcmDictSet.h"
29 #include "gdcmRLEFramesInfo.h"
30 #include "gdcmJPEGFragmentsInfo.h"
31 #include "gdcmDocEntrySet.h"
32 #include "gdcmSQItem.h"
38 #if defined(_MSC_VER) || defined(__BORLANDC__) || defined(__MINGW32__)
42 #ifdef CMAKE_HAVE_NETINET_IN_H
43 #include <netinet/in.h>
49 //-----------------------------------------------------------------------------
50 // Refer to Document::CheckSwap()
51 //const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
53 // Refer to Document::SetMaxSizeLoadEntry()
54 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
55 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
57 //-----------------------------------------------------------------------------
58 // Constructor / Destructor
62 * @param filename file to be opened for parsing
64 Document::Document( std::string const &filename ) : ElementSet(-1)
66 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
76 Group0002Parsed = false;
78 gdcmVerboseMacro( "Starting parsing of file: " << Filename.c_str());
79 // Fp->seekg( 0, std::ios::beg);
81 Fp->seekg(0, std::ios::end);
82 long lgt = Fp->tellg();
84 Fp->seekg( 0, std::ios::beg);
87 long beg = Fp->tellg();
90 ParseDES( this, beg, lgt, false); // Loading is done during parsing
92 Fp->seekg( 0, std::ios::beg);
94 // Load 'non string' values
96 std::string PhotometricInterpretation = GetEntry(0x0028,0x0004);
97 if( PhotometricInterpretation == "PALETTE COLOR " )
99 LoadEntryBinArea(0x0028,0x1200); // gray LUT
100 /// FIXME FIXME FIXME
101 /// The tags refered by the three following lines used to be CORRECTLY
102 /// defined as having an US Value Representation in the public
103 /// dictionary. BUT the semantics implied by the three following
104 /// lines state that the corresponding tag contents are in fact
105 /// the ones of a BinEntry.
106 /// In order to fix things "Quick and Dirty" the dictionary was
107 /// altered on PURPOSE but now contains a WRONG value.
108 /// In order to fix things and restore the dictionary to its
109 /// correct value, one needs to decided of the semantics by deciding
110 /// whether the following tags are either:
111 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
112 /// also used as BinEntry, which requires the proper conversion,
113 /// - OW, and hence loaded as BinEntry, but afterwards also used
114 /// as ValEntry, which requires the proper conversion.
115 LoadEntryBinArea(0x0028,0x1201); // R LUT
116 LoadEntryBinArea(0x0028,0x1202); // G LUT
117 LoadEntryBinArea(0x0028,0x1203); // B LUT
119 // Segmented Red Palette Color LUT Data
120 LoadEntryBinArea(0x0028,0x1221);
121 // Segmented Green Palette Color LUT Data
122 LoadEntryBinArea(0x0028,0x1222);
123 // Segmented Blue Palette Color LUT Data
124 LoadEntryBinArea(0x0028,0x1223);
126 //FIXME later : how to use it?
127 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
131 // --------------------------------------------------------------
132 // Specific code to allow gdcm to read ACR-LibIDO formated images
133 // Note: ACR-LibIDO is an extension of the ACR standard that was
134 // used at CREATIS. For the time being (say a couple years)
135 // we keep this kludge to allow a smooth move to gdcm for
136 // CREATIS developpers (sorry folks).
138 // if recognition code tells us we deal with a LibIDO image
139 // we switch lineNumber and columnNumber
142 RecCode = GetEntry(0x0008, 0x0010); // recognition code (RET)
143 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
144 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
145 // with "little-endian strings"
147 Filetype = ACR_LIBIDO;
148 std::string rows = GetEntry(0x0028, 0x0010);
149 std::string columns = GetEntry(0x0028, 0x0011);
150 SetEntry(columns, 0x0028, 0x0010);
151 SetEntry(rows , 0x0028, 0x0011);
153 // ----------------- End of ACR-LibIDO kludge ------------------
157 * \brief This default constructor doesn't parse the file. You should
158 * then invoke \ref Document::SetFileName and then the parsing.
160 Document::Document() : ElementSet(-1)
164 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
167 Filetype = ExplicitVR;
168 Group0002Parsed = false;
172 * \brief Canonical destructor.
174 Document::~Document ()
183 //-----------------------------------------------------------------------------
187 * \brief Prints The Dict Entries of THE public Dicom Dictionary
190 void Document::PrintPubDict(std::ostream &os)
192 RefPubDict->SetPrintLevel(PrintLevel);
193 RefPubDict->Print(os);
197 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
200 void Document::PrintShaDict(std::ostream &os)
202 RefShaDict->SetPrintLevel(PrintLevel);
203 RefShaDict->Print(os);
206 //-----------------------------------------------------------------------------
209 * \brief Get the public dictionary used
211 Dict *Document::GetPubDict()
217 * \brief Get the shadow dictionary used
219 Dict *Document::GetShaDict()
225 * \brief Set the shadow dictionary used
226 * @param dict dictionary to use in shadow
228 bool Document::SetShaDict(Dict *dict)
235 * \brief Set the shadow dictionary used
236 * @param dictName name of the dictionary to use in shadow
238 bool Document::SetShaDict(DictKey const &dictName)
240 RefShaDict = Global::GetDicts()->GetDict(dictName);
245 * \brief This predicate, based on hopefully reasonable heuristics,
246 * decides whether or not the current Document was properly parsed
247 * and contains the mandatory information for being considered as
248 * a well formed and usable Dicom/Acr File.
249 * @return true when Document is the one of a reasonable Dicom/Acr file,
252 bool Document::IsReadable()
254 if( Filetype == Unknown)
256 gdcmVerboseMacro( "Wrong filetype");
262 gdcmVerboseMacro( "No tag in internal hash table.");
270 * \brief Accessor to the Transfer Syntax (when present) of the
271 * current document (it internally handles reading the
272 * value from disk when only parsing occured).
273 * @return The encountered Transfer Syntax of the current document.
275 std::string Document::GetTransferSyntax()
277 DocEntry *entry = GetDocEntry(0x0002, 0x0010);
283 // The entry might be present but not loaded (parsing and loading
284 // happen at different stages): try loading and proceed with check...
285 LoadDocEntrySafe(entry);
286 if (ValEntry *valEntry = dynamic_cast< ValEntry* >(entry) )
288 std::string transfer = valEntry->GetValue();
289 // The actual transfer (as read from disk) might be padded. We
290 // first need to remove the potential padding. We can make the
291 // weak assumption that padding was not executed with digits...
292 if ( transfer.length() == 0 )
294 // for brain damaged headers
297 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
299 transfer.erase(transfer.length()-1, 1);
307 * \brief Predicate for dicom version 3 file.
308 * @return True when the file is a dicom version 3.
310 bool Document::IsDicomV3()
312 // Checking if Transfer Syntax exists is enough
313 // Anyway, it's to late check if the 'Preamble' was found ...
314 // And ... would it be a rich idea to check ?
315 // (some 'no Preamble' DICOM images exist !)
316 return GetDocEntry(0x0002, 0x0010) != NULL;
320 * \brief returns the File Type
321 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
322 * @return the FileType code
324 FileType Document::GetFileType()
330 * \brief Tries to open the file \ref Document::Filename and
331 * checks the preamble when existing.
332 * @return The FILE pointer on success.
334 std::ifstream *Document::OpenFile()
337 HasDCMPreamble = false;
338 if (Filename.length() == 0)
345 gdcmVerboseMacro( "File already open: " << Filename.c_str());
348 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
351 gdcmDebugMacro( "Cannot open file: " << Filename.c_str());
358 Fp->read((char*)&zero, (size_t)2);
365 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
367 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
368 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
369 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
370 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
373 = Util::Format("ACR/DICOM with no preamble: (%04x)\n", zero);
374 gdcmVerboseMacro( msg.c_str() );
379 Fp->seekg(126L, std::ios::cur);
381 Fp->read(dicm, (size_t)4);
387 if( memcmp(dicm, "DICM", 4) == 0 )
389 HasDCMPreamble = true;
394 gdcmVerboseMacro( "Not DICOM/ACR (missing preamble)" << Filename.c_str());
400 * \brief closes the file
401 * @return TRUE if the close was successfull
403 bool Document::CloseFile()
412 return true; //FIXME how do we detect a non-close ifstream ?
416 * \brief Writes in a file all the Header Entries (Dicom Elements)
417 * @param fp file pointer on an already open file (actually: Output File Stream)
418 * @param filetype Type of the File to be written
419 * (ACR-NEMA, ExplicitVR, ImplicitVR)
420 * \return Always true.
422 void Document::WriteContent(std::ofstream *fp, FileType filetype)
424 /// \todo move the following lines (and a lot of others, to be written)
425 /// to a future function CheckAndCorrectHeader
426 /// (necessary if user wants to write a DICOM V3 file
427 /// starting from an ACR-NEMA (V2) Header
429 if ( filetype == ImplicitVR || filetype == ExplicitVR )
431 // writing Dicom File Preamble
432 char filePreamble[128];
433 memset(filePreamble, 0, 128);
434 fp->write(filePreamble, 128);
435 fp->write("DICM", 4);
439 * \todo rewrite later, if really usefull
440 * - 'Group Length' element is optional in DICOM
441 * - but un-updated odd groups lengthes can causes pb
444 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
445 * UpdateGroupLength(false,filetype);
446 * if ( filetype == ACR)
447 * UpdateGroupLength(true,ACR);
450 ElementSet::WriteContent(fp, filetype); // This one is recursive
454 * \brief Modifies the value of a given Doc Entry (Dicom Element)
455 * when it exists. Create it with the given value when unexistant.
456 * @param value (string) Value to be set
457 * @param group Group number of the Entry
458 * @param elem Element number of the Entry
459 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
460 * \return pointer to the modified/created Header Entry (NULL when creation
463 ValEntry *Document::ReplaceOrCreate(std::string const &value,
468 ValEntry *valEntry = 0;
469 DocEntry *currentEntry = GetDocEntry( group, elem);
473 valEntry = dynamic_cast< ValEntry* >(currentEntry);
477 if( valEntry->GetVR()!=vr )
480 // if currentEntry doesn't correspond to the requested valEntry
483 if (!RemoveEntry(currentEntry))
485 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
492 // Create a new valEntry if necessary
495 valEntry = NewValEntry(group, elem, vr);
497 if ( !AddEntry(valEntry))
499 gdcmVerboseMacro("AddEntry failed although this is a creation.");
506 // Set the binEntry value
507 SetEntry(value, valEntry);
512 * \brief Modifies the value of a given Header Entry (Dicom Element)
513 * when it exists. Create it with the given value when unexistant.
514 * A copy of the binArea is made to be kept in the Document.
515 * @param binArea (binary) value to be set
516 * @param group Group number of the Entry
517 * @param elem Element number of the Entry
518 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
519 * \return pointer to the modified/created Header Entry (NULL when creation
522 BinEntry *Document::ReplaceOrCreate(uint8_t *binArea,
528 BinEntry *binEntry = 0;
529 DocEntry *currentEntry = GetDocEntry( group, elem);
531 // Verify the currentEntry
534 binEntry = dynamic_cast< BinEntry* >(currentEntry);
538 if( binEntry->GetVR()!=vr )
541 // if currentEntry doesn't correspond to the requested valEntry
544 if (!RemoveEntry(currentEntry))
546 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
553 // Create a new binEntry if necessary
556 binEntry = NewBinEntry(group, elem, vr);
558 if ( !AddEntry(binEntry))
560 gdcmVerboseMacro( "AddEntry failed allthough this is a creation.");
567 // Set the binEntry value
569 if (lgth>0 && binArea)
571 tmpArea = new uint8_t[lgth];
572 memcpy(tmpArea,binArea,lgth);
578 if (!SetEntry(tmpArea,lgth,binEntry))
590 * \brief Modifies the value of a given Header Entry (Dicom Element)
591 * when it exists. Create it when unexistant.
592 * @param group Group number of the Entry
593 * @param elem Element number of the Entry
594 * \return pointer to the modified/created SeqEntry (NULL when creation
597 SeqEntry *Document::ReplaceOrCreate( uint16_t group, uint16_t elem)
599 SeqEntry *seqEntry = 0;
600 DocEntry *currentEntry = GetDocEntry( group, elem);
602 // Verify the currentEntry
605 seqEntry = dynamic_cast< SeqEntry* >(currentEntry);
609 if( seqEntry->GetVR()!="SQ" )
612 // if currentEntry doesn't correspond to the requested valEntry
615 if (!RemoveEntry(currentEntry))
617 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
624 // Create a new seqEntry if necessary
627 seqEntry = NewSeqEntry(group, elem);
629 if ( !AddEntry(seqEntry))
631 gdcmVerboseMacro( "AddEntry failed allthough this is a creation.");
642 * \brief Set a new value if the invoked element exists
643 * Seems to be useless !!!
644 * @param value new element value
645 * @param group group number of the Entry
646 * @param elem element number of the Entry
649 bool Document::ReplaceIfExist(std::string const &value,
650 uint16_t group, uint16_t elem )
652 SetEntry(value, group, elem);
657 //-----------------------------------------------------------------------------
661 * \brief Checks if a given Dicom Element exists within the H table
662 * @param group Group number of the searched Dicom Element
663 * @param elem Element number of the searched Dicom Element
664 * @return true is found
666 bool Document::CheckIfEntryExist(uint16_t group, uint16_t elem )
668 const std::string &key = DictEntry::TranslateToKey(group, elem );
669 return TagHT.count(key) != 0;
674 * \brief Searches within Header Entries (Dicom Elements) parsed with
675 * the public and private dictionaries
676 * for the element value representation of a given tag.
677 * @param group Group number of the searched tag.
678 * @param elem Element number of the searched tag.
679 * @return Corresponding element value representation when it exists,
680 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
682 std::string Document::GetEntry(uint16_t group, uint16_t elem)
684 TagKey key = DictEntry::TranslateToKey(group, elem);
685 if ( !TagHT.count(key))
690 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
694 * \brief Searches within Header Entries (Dicom Elements) parsed with
695 * the public and private dictionaries
696 * for the element value representation of a given tag..
697 * Obtaining the VR (Value Representation) might be needed by caller
698 * to convert the string typed content to caller's native type
699 * (think of C++ vs Python). The VR is actually of a higher level
700 * of semantics than just the native C++ type.
701 * @param group Group number of the searched tag.
702 * @param elem Element number of the searched tag.
703 * @return Corresponding element value representation when it exists,
704 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
706 std::string Document::GetEntryVR(uint16_t group, uint16_t elem)
708 DocEntry *element = GetDocEntry(group, elem);
713 return element->GetVR();
717 * \brief Searches within Header Entries (Dicom Elements) parsed with
718 * the public and private dictionaries
719 * for the value length of a given tag..
720 * @param group Group number of the searched tag.
721 * @param elem Element number of the searched tag.
722 * @return Corresponding element length; -2 if not found
724 int Document::GetEntryLength(uint16_t group, uint16_t elem)
726 DocEntry *element = GetDocEntry(group, elem);
729 return -2; //magic number
731 return element->GetLength();
735 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
736 * through it's (group, element) and modifies it's content with
738 * @param content new value (string) to substitute with
739 * @param group group number of the Dicom Element to modify
740 * @param elem element number of the Dicom Element to modify
742 bool Document::SetEntry(std::string const& content,
743 uint16_t group, uint16_t elem)
745 ValEntry *entry = GetValEntry(group, elem);
748 gdcmVerboseMacro( "No corresponding ValEntry (try promotion first).");
751 return SetEntry(content,entry);
755 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
756 * through it's (group, element) and modifies it's content with
758 * @param content new value (void* -> uint8_t*) to substitute with
759 * @param lgth new value length
760 * @param group group number of the Dicom Element to modify
761 * @param elem element number of the Dicom Element to modify
763 bool Document::SetEntry(uint8_t*content, int lgth,
764 uint16_t group, uint16_t elem)
766 BinEntry *entry = GetBinEntry(group, elem);
769 gdcmVerboseMacro( "No corresponding ValEntry (try promotion first).");
773 return SetEntry(content,lgth,entry);
777 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
778 * and modifies it's content with the given value.
779 * @param content new value (string) to substitute with
780 * @param entry Entry to be modified
782 bool Document::SetEntry(std::string const &content, ValEntry *entry)
786 entry->SetValue(content);
793 * \brief Accesses an existing BinEntry (i.e. a Dicom Element)
794 * and modifies it's content with the given value.
795 * @param content new value (void* -> uint8_t*) to substitute with
796 * @param entry Entry to be modified
797 * @param lgth new value length
799 bool Document::SetEntry(uint8_t *content, int lgth, BinEntry *entry)
803 // Hope Binary field length is *never* wrong
804 /*if(lgth%2) // Non even length are padded with a space (020H).
807 //content = content + '\0'; // fing a trick to enlarge a binary field?
810 entry->SetBinArea(content);
811 entry->SetLength(lgth);
812 entry->SetValue(GDCM_BINLOADED);
819 * \brief Gets (from Header) a 'non string' element value
820 * (LoadElementValues has already be executed)
821 * @param group group number of the Entry
822 * @param elem element number of the Entry
823 * @return Pointer to the 'non string' area
825 void *Document::GetEntryBinArea(uint16_t group, uint16_t elem)
827 DocEntry *entry = GetDocEntry(group, elem);
830 gdcmVerboseMacro( "No entry");
833 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(entry) )
835 return binEntry->GetBinArea();
842 * \brief Loads (from disk) the element content
843 * when a string is not suitable
844 * @param group group number of the Entry
845 * @param elem element number of the Entry
847 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
849 // Search the corresponding DocEntry
850 DocEntry *docElement = GetDocEntry(group, elem);
854 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
858 LoadEntryBinArea(binElement);
862 * \brief Loads (from disk) the element content
863 * when a string is not suitable
864 * @param elem Entry whose binArea is going to be loaded
866 void Document::LoadEntryBinArea(BinEntry *elem)
868 if(elem->GetBinArea())
875 size_t o =(size_t)elem->GetOffset();
876 Fp->seekg(o, std::ios::beg);
878 size_t l = elem->GetLength();
879 uint8_t *a = new uint8_t[l];
882 gdcmVerboseMacro( "Cannot allocate BinEntry content");
886 /// \todo check the result
887 Fp->read((char*)a, l);
888 if( Fp->fail() || Fp->eof())
901 * \brief Sets a 'non string' value to a given Dicom Element
902 * @param area area containing the 'non string' value
903 * @param group Group number of the searched Dicom Element
904 * @param elem Element number of the searched Dicom Element
907 /*bool Document::SetEntryBinArea(uint8_t *area,
908 uint16_t group, uint16_t elem)
910 DocEntry *currentEntry = GetDocEntry(group, elem);
916 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(currentEntry) )
918 binEntry->SetBinArea( area );
926 * \brief retrieves a Dicom Element (the first one) using (group, element)
927 * \warning (group, element) IS NOT an identifier inside the Dicom Header
928 * if you think it's NOT UNIQUE, check the count number
929 * and use iterators to retrieve ALL the Dicoms Elements within
930 * a given couple (group, element)
931 * @param group Group number of the searched Dicom Element
932 * @param elem Element number of the searched Dicom Element
935 DocEntry *Document::GetDocEntry(uint16_t group, uint16_t elem)
937 TagKey key = DictEntry::TranslateToKey(group, elem);
938 if ( !TagHT.count(key))
942 return TagHT.find(key)->second;
946 * \brief Same as \ref Document::GetDocEntry except it only
947 * returns a result when the corresponding entry is of type
949 * @param group Group number of the searched Dicom Element
950 * @param elem Element number of the searched Dicom Element
951 * @return When present, the corresponding ValEntry.
953 ValEntry *Document::GetValEntry(uint16_t group, uint16_t elem)
955 DocEntry *currentEntry = GetDocEntry(group, elem);
960 if ( ValEntry *entry = dynamic_cast<ValEntry*>(currentEntry) )
964 gdcmVerboseMacro( "Unfound ValEntry.");
970 * \brief Same as \ref Document::GetDocEntry except it only
971 * returns a result when the corresponding entry is of type
973 * @param group Group number of the searched Dicom Element
974 * @param elem Element number of the searched Dicom Element
975 * @return When present, the corresponding BinEntry.
977 BinEntry *Document::GetBinEntry(uint16_t group, uint16_t elem)
979 DocEntry *currentEntry = GetDocEntry(group, elem);
984 if ( BinEntry *entry = dynamic_cast<BinEntry*>(currentEntry) )
988 gdcmVerboseMacro( "Unfound BinEntry.");
994 * \brief Loads the element while preserving the current
995 * underlying file position indicator as opposed to
996 * to LoadDocEntry that modifies it.
997 * @param entry Header Entry whose value will be loaded.
1000 void Document::LoadDocEntrySafe(DocEntry *entry)
1004 long PositionOnEntry = Fp->tellg();
1005 LoadDocEntry(entry);
1006 Fp->seekg(PositionOnEntry, std::ios::beg);
1011 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1013 * @return The properly swaped 32 bits integer.
1015 uint32_t Document::SwapLong(uint32_t a)
1022 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1023 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1026 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1029 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1032 gdcmErrorMacro( "Unset swap code:" << SwapCode );
1039 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1041 * @return The properly unswaped 32 bits integer.
1043 uint32_t Document::UnswapLong(uint32_t a)
1049 * \brief Swaps the bytes so they agree with the processor order
1050 * @return The properly swaped 16 bits integer.
1052 uint16_t Document::SwapShort(uint16_t a)
1054 if ( SwapCode == 4321 || SwapCode == 2143 )
1056 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1062 * \brief Unswaps the bytes so they agree with the processor order
1063 * @return The properly unswaped 16 bits integer.
1065 uint16_t Document::UnswapShort(uint16_t a)
1067 return SwapShort(a);
1070 //-----------------------------------------------------------------------------
1074 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1075 * @return length of the parsed set.
1077 void Document::ParseDES(DocEntrySet *set, long offset,
1078 long l_max, bool delim_mode)
1080 DocEntry *newDocEntry = 0;
1081 ValEntry *newValEntry;
1082 BinEntry *newBinEntry;
1083 SeqEntry *newSeqEntry;
1089 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1095 newDocEntry = ReadNextDocEntry( );
1102 vr = newDocEntry->GetVR();
1103 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
1104 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
1105 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1107 if ( newValEntry || newBinEntry )
1111 if ( Filetype == ExplicitVR && ! Global::GetVR()->IsVROfBinaryRepresentable(vr) )
1113 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1114 gdcmVerboseMacro( std::hex << newDocEntry->GetGroup()
1115 << "|" << newDocEntry->GetElement()
1116 << " : Neither Valentry, nor BinEntry."
1117 "Probably unknown VR.");
1120 //////////////////// BinEntry or UNKOWN VR:
1121 // When "this" is a Document the Key is simply of the
1122 // form ( group, elem )...
1123 if ( dynamic_cast< Document* > ( set ) )
1125 newBinEntry->SetKey( newBinEntry->GetKey() );
1127 // but when "this" is a SQItem, we are inserting this new
1128 // valEntry in a sequence item, and the key has the
1129 // generalized form (refer to \ref BaseTagKey):
1130 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1132 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1133 + newBinEntry->GetKey() );
1136 LoadDocEntry( newBinEntry );
1137 if( !set->AddEntry( newBinEntry ) )
1139 //Expect big troubles if here
1140 //delete newBinEntry;
1146 /////////////////////// ValEntry
1147 // When "set" is a Document, then we are at the top of the
1148 // hierarchy and the Key is simply of the form ( group, elem )...
1149 if ( dynamic_cast< Document* > ( set ) )
1151 newValEntry->SetKey( newValEntry->GetKey() );
1153 // ...but when "set" is a SQItem, we are inserting this new
1154 // valEntry in a sequence item. Hence the key has the
1155 // generalized form (refer to \ref BaseTagKey):
1156 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1158 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1159 + newValEntry->GetKey() );
1162 LoadDocEntry( newValEntry );
1163 bool delimitor=newValEntry->IsItemDelimitor();
1164 if( !set->AddEntry( newValEntry ) )
1166 // If here expect big troubles
1167 //delete newValEntry; //otherwise mem leak
1177 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1185 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1186 && ( newDocEntry->GetElement() == 0x0010 ) )
1188 std::string ts = GetTransferSyntax();
1189 if ( Global::GetTS()->IsRLELossless(ts) )
1191 long positionOnEntry = Fp->tellg();
1192 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1194 Fp->seekg( positionOnEntry, std::ios::beg );
1196 else if ( Global::GetTS()->IsJPEG(ts) )
1198 long positionOnEntry = Fp->tellg();
1199 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1200 ComputeJPEGFragmentInfo();
1201 Fp->seekg( positionOnEntry, std::ios::beg );
1205 // Just to make sure we are at the beginning of next entry.
1206 SkipToNextDocEntry(newDocEntry);
1211 unsigned long l = newDocEntry->GetReadLength();
1212 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1214 if ( l == 0xffffffff )
1223 // no other way to create it ...
1224 newSeqEntry->SetDelimitorMode( delim_mode );
1226 // At the top of the hierarchy, stands a Document. When "set"
1227 // is a Document, then we are building the first depth level.
1228 // Hence the SeqEntry we are building simply has a depth
1230 if (/*Document *dummy =*/ dynamic_cast< Document* > ( set ) )
1233 newSeqEntry->SetDepthLevel( 1 );
1234 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1236 // But when "set" is already a SQItem, we are building a nested
1237 // sequence, and hence the depth level of the new SeqEntry
1238 // we are building, is one level deeper:
1239 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1241 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1242 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1243 + newSeqEntry->GetKey() );
1247 { // Don't try to parse zero-length sequences
1248 ParseSQ( newSeqEntry,
1249 newDocEntry->GetOffset(),
1252 set->AddEntry( newSeqEntry );
1253 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1265 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1266 * @return parsed length for this level
1268 void Document::ParseSQ( SeqEntry *seqEntry,
1269 long offset, long l_max, bool delim_mode)
1271 int SQItemNumber = 0;
1273 long offsetStartCurrentSQItem = offset;
1277 // the first time, we read the fff0,e000 of the first SQItem
1278 DocEntry *newDocEntry = ReadNextDocEntry();
1282 // FIXME Should warn user
1287 if ( newDocEntry->IsSequenceDelimitor() )
1289 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1293 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1298 // create the current SQItem
1299 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1300 std::ostringstream newBase;
1301 newBase << seqEntry->GetKey()
1305 itemSQ->SetBaseTagKey( newBase.str() );
1306 unsigned int l = newDocEntry->GetReadLength();
1308 if ( l == 0xffffffff )
1317 // when we're here, element fffe,e000 is already passed.
1318 // it's lost for the SQItem we're going to process !!
1320 //ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1321 //delete newDocEntry; // FIXME well ... it's too late to use it !
1323 // Let's try :------------
1324 // remove fff0,e000, created out of the SQItem
1326 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg);
1327 // fill up the current SQItem, starting at the beginning of fff0,e000
1328 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
1329 offsetStartCurrentSQItem = Fp->tellg();
1330 // end try -----------------
1332 seqEntry->AddEntry( itemSQ, SQItemNumber );
1334 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1342 * \brief Loads the element content if its length doesn't exceed
1343 * the value specified with Document::SetMaxSizeLoadEntry()
1344 * @param entry Header Entry (Dicom Element) to be dealt with
1346 void Document::LoadDocEntry(DocEntry *entry)
1348 uint16_t group = entry->GetGroup();
1349 std::string vr = entry->GetVR();
1350 uint32_t length = entry->GetLength();
1352 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1354 // A SeQuence "contains" a set of Elements.
1355 // (fffe e000) tells us an Element is beginning
1356 // (fffe e00d) tells us an Element just ended
1357 // (fffe e0dd) tells us the current SeQuence just ended
1358 if( group == 0xfffe )
1360 // NO more value field for SQ !
1364 // When the length is zero things are easy:
1367 ((ValEntry *)entry)->SetValue("");
1371 // The elements whose length is bigger than the specified upper bound
1372 // are not loaded. Instead we leave a short notice of the offset of
1373 // the element content and it's length.
1375 std::ostringstream s;
1376 if (length > MaxSizeLoadEntry)
1378 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1380 //s << "gdcm::NotLoaded (BinEntry)";
1381 s << GDCM_NOTLOADED;
1382 s << " Address:" << (long)entry->GetOffset();
1383 s << " Length:" << entry->GetLength();
1384 s << " x(" << std::hex << entry->GetLength() << ")";
1385 binEntryPtr->SetValue(s.str());
1387 // Be carefull : a BinEntry IS_A ValEntry ...
1388 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1390 // s << "gdcm::NotLoaded. (ValEntry)";
1391 s << GDCM_NOTLOADED;
1392 s << " Address:" << (long)entry->GetOffset();
1393 s << " Length:" << entry->GetLength();
1394 s << " x(" << std::hex << entry->GetLength() << ")";
1395 valEntryPtr->SetValue(s.str());
1400 gdcmErrorMacro( "MaxSizeLoadEntry exceeded, neither a BinEntry "
1401 << "nor a ValEntry ?! Should never print that !" );
1404 // to be sure we are at the end of the value ...
1405 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1410 // When we find a BinEntry not very much can be done :
1411 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1413 s << GDCM_BINLOADED;
1414 binEntryPtr->SetValue(s.str());
1415 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1419 /// \todo Any compacter code suggested (?)
1420 if ( IsDocEntryAnInteger(entry) )
1424 // When short integer(s) are expected, read and convert the following
1425 // n *two characters properly i.e. consider them as short integers as
1426 // opposed to strings.
1427 // Elements with Value Multiplicity > 1
1428 // contain a set of integers (not a single one)
1429 if (vr == "US" || vr == "SS")
1432 NewInt = ReadInt16();
1436 for (int i=1; i < nbInt; i++)
1439 NewInt = ReadInt16();
1444 // See above comment on multiple integers (mutatis mutandis).
1445 else if (vr == "UL" || vr == "SL")
1448 NewInt = ReadInt32();
1452 for (int i=1; i < nbInt; i++)
1455 NewInt = ReadInt32();
1460 #ifdef GDCM_NO_ANSI_STRING_STREAM
1461 s << std::ends; // to avoid oddities on Solaris
1462 #endif //GDCM_NO_ANSI_STRING_STREAM
1464 ((ValEntry *)entry)->SetValue(s.str());
1468 // FIXME: We need an additional byte for storing \0 that is not on disk
1469 char *str = new char[length+1];
1470 Fp->read(str, (size_t)length);
1471 str[length] = '\0'; //this is only useful when length is odd
1472 // Special DicomString call to properly handle \0 and even length
1473 std::string newValue;
1476 newValue = Util::DicomString(str, length+1);
1477 gdcmVerboseMacro("Warning: bad length: " << length <<
1478 ",For string :" << newValue.c_str());
1479 // Since we change the length of string update it length
1480 //entry->SetReadLength(length+1);
1484 newValue = Util::DicomString(str, length);
1488 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1490 if ( Fp->fail() || Fp->eof())
1492 gdcmVerboseMacro("Unread element value");
1493 valEntry->SetValue(GDCM_UNREAD);
1499 // Because of correspondance with the VR dic
1500 valEntry->SetValue(newValue);
1504 valEntry->SetValue(newValue);
1509 gdcmErrorMacro( "Should have a ValEntry, here !");
1515 * \brief Find the value Length of the passed Header Entry
1516 * @param entry Header Entry whose length of the value shall be loaded.
1518 void Document::FindDocEntryLength( DocEntry *entry )
1519 throw ( FormatError )
1521 std::string vr = entry->GetVR();
1524 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1526 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1528 // The following reserved two bytes (see PS 3.5-2003, section
1529 // "7.1.2 Data element structure with explicit vr", p 27) must be
1530 // skipped before proceeding on reading the length on 4 bytes.
1531 Fp->seekg( 2L, std::ios::cur);
1532 uint32_t length32 = ReadInt32();
1534 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1539 lengthOB = FindDocEntryLengthOBOrOW();
1541 catch ( FormatUnexpected )
1543 // Computing the length failed (this happens with broken
1544 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1545 // chance to get the pixels by deciding the element goes
1546 // until the end of the file. Hence we artificially fix the
1547 // the length and proceed.
1548 long currentPosition = Fp->tellg();
1549 Fp->seekg(0L,std::ios::end);
1551 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1552 Fp->seekg(currentPosition, std::ios::beg);
1554 entry->SetReadLength(lengthUntilEOF);
1555 entry->SetLength(lengthUntilEOF);
1558 entry->SetReadLength(lengthOB);
1559 entry->SetLength(lengthOB);
1562 FixDocEntryFoundLength(entry, length32);
1566 // Length is encoded on 2 bytes.
1567 length16 = ReadInt16();
1569 // FIXME : This heuristic supposes that the first group following
1570 // group 0002 *has* and element 0000.
1571 // BUT ... Element 0000 is optionnal :-(
1574 // Fixed using : HandleOutOfGroup0002()
1575 // (first hereafter strategy ...)
1577 // We can tell the current file is encoded in big endian (like
1578 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1579 // and it's value is the one of the encoding of a big endian file.
1580 // In order to deal with such big endian encoded files, we have
1581 // (at least) two strategies:
1582 // * when we load the "Transfer Syntax" tag with value of big endian
1583 // encoding, we raise the proper flags. Then we wait for the end
1584 // of the META group (0x0002) among which is "Transfer Syntax",
1585 // before switching the swap code to big endian. We have to postpone
1586 // the switching of the swap code since the META group is fully encoded
1587 // in little endian, and big endian coding only starts at the next
1588 // group. The corresponding code can be hard to analyse and adds
1589 // many additional unnecessary tests for regular tags.
1590 // * the second strategy consists in waiting for trouble, that shall
1591 // appear when we find the first group with big endian encoding. This
1592 // is easy to detect since the length of a "Group Length" tag (the
1593 // ones with zero as element number) has to be of 4 (0x0004). When we
1594 // encounter 1024 (0x0400) chances are the encoding changed and we
1595 // found a group with big endian encoding.
1596 //---> Unfortunately, element 0000 is optional.
1597 //---> This will not work when missing!
1598 // We shall use this second strategy. In order to make sure that we
1599 // can interpret the presence of an apparently big endian encoded
1600 // length of a "Group Length" without committing a big mistake, we
1601 // add an additional check: we look in the already parsed elements
1602 // for the presence of a "Transfer Syntax" whose value has to be "big
1603 // endian encoding". When this is the case, chances are we have got our
1604 // hands on a big endian encoded file: we switch the swap code to
1605 // big endian and proceed...
1608 // if ( element == 0x0000 && length16 == 0x0400 )
1610 // std::string ts = GetTransferSyntax();
1611 // if ( Global::GetTS()->GetSpecialTransferSyntax(ts)
1612 // != TS::ExplicitVRBigEndian )
1614 // throw FormatError( "Document::FindDocEntryLength()",
1615 // " not explicit VR." );
1619 // SwitchByteSwapCode();
1621 // Restore the unproperly loaded values i.e. the group, the element
1622 // and the dictionary entry depending on them.
1623 // uint16_t correctGroup = SwapShort( entry->GetGroup() );
1624 // uint16_t correctElem = SwapShort( entry->GetElement() );
1625 // DictEntry *newTag = GetDictEntry( correctGroup, correctElem ); if ( !newTag )
1627 // This correct tag is not in the dictionary. Create a new one.
1628 // newTag = NewVirtualDictEntry(correctGroup, correctElem);
1630 // FIXME this can create a memory leaks on the old entry that be
1631 // left unreferenced.
1632 // entry->SetDictEntry( newTag );
1636 // 0xffff means that we deal with 'No Length' Sequence
1637 // or 'No Length' SQItem
1638 if ( length16 == 0xffff)
1642 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1647 // Either implicit VR or a non DICOM conformal (see note below) explicit
1648 // VR that ommited the VR of (at least) this element. Farts happen.
1649 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1650 // on Data elements "Implicit and Explicit VR Data Elements shall
1651 // not coexist in a Data Set and Data Sets nested within it".]
1652 // Length is on 4 bytes.
1654 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1655 // even if Transfer Syntax is 'Implicit VR ...'
1657 FixDocEntryFoundLength( entry, ReadInt32() );
1663 * \brief Find the Value Representation of the current Dicom Element.
1664 * @return Value Representation of the current Entry
1666 std::string Document::FindDocEntryVR()
1668 if ( Filetype != ExplicitVR )
1669 return GDCM_UNKNOWN;
1671 long positionOnEntry = Fp->tellg();
1672 // Warning: we believe this is explicit VR (Value Representation) because
1673 // we used a heuristic that found "UL" in the first tag. Alas this
1674 // doesn't guarantee that all the tags will be in explicit VR. In some
1675 // cases (see e-film filtered files) one finds implicit VR tags mixed
1676 // within an explicit VR file. Hence we make sure the present tag
1677 // is in explicit VR and try to fix things if it happens not to be
1681 Fp->read (vr, (size_t)2);
1684 if( !CheckDocEntryVR(vr) )
1686 Fp->seekg(positionOnEntry, std::ios::beg);
1687 return GDCM_UNKNOWN;
1693 * \brief Check the correspondance between the VR of the header entry
1694 * and the taken VR. If they are different, the header entry is
1695 * updated with the new VR.
1696 * @param vr Dicom Value Representation
1697 * @return false if the VR is incorrect of if the VR isn't referenced
1698 * otherwise, it returns true
1700 bool Document::CheckDocEntryVR(VRKey vr)
1702 // CLEANME searching the dicom_vr at each occurence is expensive.
1703 // PostPone this test in an optional integrity check at the end
1704 // of parsing or only in debug mode.
1705 if ( !Global::GetVR()->IsValidVR(vr) )
1712 * \brief Get the transformed value of the header entry. The VR value
1713 * is used to define the transformation to operate on the value
1714 * \warning NOT end user intended method !
1715 * @param entry entry to tranform
1716 * @return Transformed entry value
1718 std::string Document::GetDocEntryValue(DocEntry *entry)
1720 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1722 std::string val = ((ValEntry *)entry)->GetValue();
1723 std::string vr = entry->GetVR();
1724 uint32_t length = entry->GetLength();
1725 std::ostringstream s;
1728 // When short integer(s) are expected, read and convert the following
1729 // n * 2 bytes properly i.e. as a multivaluated strings
1730 // (each single value is separated fromthe next one by '\'
1731 // as usual for standard multivaluated filels
1732 // Elements with Value Multiplicity > 1
1733 // contain a set of short integers (not a single one)
1735 if( vr == "US" || vr == "SS" )
1740 for (int i=0; i < nbInt; i++)
1746 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1747 newInt16 = SwapShort( newInt16 );
1752 // When integer(s) are expected, read and convert the following
1753 // n * 4 bytes properly i.e. as a multivaluated strings
1754 // (each single value is separated fromthe next one by '\'
1755 // as usual for standard multivaluated filels
1756 // Elements with Value Multiplicity > 1
1757 // contain a set of integers (not a single one)
1758 else if( vr == "UL" || vr == "SL" )
1763 for (int i=0; i < nbInt; i++)
1769 newInt32 = ( val[4*i+0] & 0xFF )
1770 + (( val[4*i+1] & 0xFF ) << 8 )
1771 + (( val[4*i+2] & 0xFF ) << 16 )
1772 + (( val[4*i+3] & 0xFF ) << 24 );
1773 newInt32 = SwapLong( newInt32 );
1777 #ifdef GDCM_NO_ANSI_STRING_STREAM
1778 s << std::ends; // to avoid oddities on Solaris
1779 #endif //GDCM_NO_ANSI_STRING_STREAM
1783 return ((ValEntry *)entry)->GetValue();
1787 * \brief Get the reverse transformed value of the header entry. The VR
1788 * value is used to define the reverse transformation to operate on
1790 * \warning NOT end user intended method !
1791 * @param entry Entry to reverse transform
1792 * @return Reverse transformed entry value
1794 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1796 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1798 std::string vr = entry->GetVR();
1799 std::vector<std::string> tokens;
1800 std::ostringstream s;
1802 if ( vr == "US" || vr == "SS" )
1806 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1807 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1808 for (unsigned int i=0; i<tokens.size(); i++)
1810 newInt16 = atoi(tokens[i].c_str());
1811 s << ( newInt16 & 0xFF )
1812 << (( newInt16 >> 8 ) & 0xFF );
1816 if ( vr == "UL" || vr == "SL")
1820 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1821 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1822 for (unsigned int i=0; i<tokens.size();i++)
1824 newInt32 = atoi(tokens[i].c_str());
1825 s << (char)( newInt32 & 0xFF )
1826 << (char)(( newInt32 >> 8 ) & 0xFF )
1827 << (char)(( newInt32 >> 16 ) & 0xFF )
1828 << (char)(( newInt32 >> 24 ) & 0xFF );
1833 #ifdef GDCM_NO_ANSI_STRING_STREAM
1834 s << std::ends; // to avoid oddities on Solaris
1835 #endif //GDCM_NO_ANSI_STRING_STREAM
1839 return ((ValEntry *)entry)->GetValue();
1843 * \brief Skip a given Header Entry
1844 * \warning NOT end user intended method !
1845 * @param entry entry to skip
1847 void Document::SkipDocEntry(DocEntry *entry)
1849 SkipBytes(entry->GetLength());
1853 * \brief Skips to the begining of the next Header Entry
1854 * \warning NOT end user intended method !
1855 * @param currentDocEntry entry to skip
1857 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1859 Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg);
1860 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1861 Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur);
1865 * \brief When the length of an element value is obviously wrong (because
1866 * the parser went Jabberwocky) one can hope improving things by
1867 * applying some heuristics.
1868 * @param entry entry to check
1869 * @param foundLength first assumption about length
1871 void Document::FixDocEntryFoundLength(DocEntry *entry,
1872 uint32_t foundLength)
1874 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1875 if ( foundLength == 0xffffffff)
1880 uint16_t gr = entry->GetGroup();
1881 uint16_t elem = entry->GetElement();
1883 if ( foundLength % 2)
1885 gdcmVerboseMacro( "Warning : Tag with uneven length " << foundLength
1886 << " in x(" << std::hex << gr << "," << elem <<")");
1889 //////// Fix for some naughty General Electric images.
1890 // Allthough not recent many such GE corrupted images are still present
1891 // on Creatis hard disks. Hence this fix shall remain when such images
1892 // are no longer in use (we are talking a few years, here)...
1893 // Note: XMedCom probably uses such a trick since it is able to read
1894 // those pesky GE images ...
1895 if ( foundLength == 13)
1897 // Only happens for this length !
1898 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1901 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
1905 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1906 // Occurence of such images is quite low (unless one leaves close to a
1907 // 'Leonardo' source. Hence, one might consider commenting out the
1908 // following fix on efficiency reasons.
1909 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1912 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
1915 else if ( entry->GetVR() == "SQ" )
1917 foundLength = 0; // ReadLength is unchanged
1920 //////// We encountered a 'delimiter' element i.e. a tag of the form
1921 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1922 // taken into account.
1923 else if( gr == 0xfffe )
1925 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1926 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1927 // causes extra troubles...
1928 if( entry->GetElement() != 0x0000 )
1934 entry->SetLength(foundLength);
1938 * \brief Apply some heuristics to predict whether the considered
1939 * element value contains/represents an integer or not.
1940 * @param entry The element value on which to apply the predicate.
1941 * @return The result of the heuristical predicate.
1943 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1945 uint16_t elem = entry->GetElement();
1946 uint16_t group = entry->GetGroup();
1947 const std::string &vr = entry->GetVR();
1948 uint32_t length = entry->GetLength();
1950 // When we have some semantics on the element we just read, and if we
1951 // a priori know we are dealing with an integer, then we shall be
1952 // able to swap it's element value properly.
1953 if ( elem == 0 ) // This is the group length of the group
1961 // Allthough this should never happen, still some images have a
1962 // corrupted group length [e.g. have a glance at offset x(8336) of
1963 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
1964 // Since for dicom compliant and well behaved headers, the present
1965 // test is useless (and might even look a bit paranoid), when we
1966 // encounter such an ill-formed image, we simply display a warning
1967 // message and proceed on parsing (while crossing fingers).
1968 long filePosition = Fp->tellg();
1969 gdcmVerboseMacro( "Erroneous Group Length element length on : ("
1970 << std::hex << group << " , " << elem
1971 << ") -before- position x(" << filePosition << ")"
1972 << "lgt : " << length );
1976 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1984 * \brief Find the Length till the next sequence delimiter
1985 * \warning NOT end user intended method !
1989 uint32_t Document::FindDocEntryLengthOBOrOW()
1990 throw( FormatUnexpected )
1992 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1993 long positionOnEntry = Fp->tellg();
1994 bool foundSequenceDelimiter = false;
1995 uint32_t totalLength = 0;
1997 while ( !foundSequenceDelimiter )
2003 group = ReadInt16();
2006 catch ( FormatError )
2008 throw FormatError("Unexpected end of file encountered during ",
2009 "Document::FindDocEntryLengthOBOrOW()");
2012 // We have to decount the group and element we just read
2015 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2017 long filePosition = Fp->tellg();
2018 gdcmVerboseMacro( "Neither an Item tag nor a Sequence delimiter tag on :"
2019 << std::hex << group << " , " << elem
2020 << ") -before- position x(" << filePosition << ")" );
2022 Fp->seekg(positionOnEntry, std::ios::beg);
2023 throw FormatUnexpected( "Neither an Item tag nor a Sequence delimiter tag.");
2026 if ( elem == 0xe0dd )
2028 foundSequenceDelimiter = true;
2031 uint32_t itemLength = ReadInt32();
2032 // We add 4 bytes since we just read the ItemLength with ReadInt32
2033 totalLength += itemLength + 4;
2034 SkipBytes(itemLength);
2036 if ( foundSequenceDelimiter )
2041 Fp->seekg( positionOnEntry, std::ios::beg);
2046 * \brief Reads a supposed to be 16 Bits integer
2047 * (swaps it depending on processor endianity)
2048 * @return read value
2050 uint16_t Document::ReadInt16()
2051 throw( FormatError )
2054 Fp->read ((char*)&g, (size_t)2);
2057 throw FormatError( "Document::ReadInt16()", " file error." );
2061 throw FormatError( "Document::ReadInt16()", "EOF." );
2068 * \brief Reads a supposed to be 32 Bits integer
2069 * (swaps it depending on processor endianity)
2070 * @return read value
2072 uint32_t Document::ReadInt32()
2073 throw( FormatError )
2076 Fp->read ((char*)&g, (size_t)4);
2079 throw FormatError( "Document::ReadInt32()", " file error." );
2083 throw FormatError( "Document::ReadInt32()", "EOF." );
2090 * \brief skips bytes inside the source file
2091 * \warning NOT end user intended method !
2094 void Document::SkipBytes(uint32_t nBytes)
2096 //FIXME don't dump the returned value
2097 Fp->seekg((long)nBytes, std::ios::cur);
2101 * \brief Loads all the needed Dictionaries
2102 * \warning NOT end user intended method !
2104 void Document::Initialise()
2106 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2108 RLEInfo = new RLEFramesInfo;
2109 JPEGInfo = new JPEGFragmentsInfo;
2114 * \brief Discover what the swap code is (among little endian, big endian,
2115 * bad little endian, bad big endian).
2117 * @return false when we are absolutely sure
2118 * it's neither ACR-NEMA nor DICOM
2119 * true when we hope ours assuptions are OK
2121 bool Document::CheckSwap()
2123 // The only guaranted way of finding the swap code is to find a
2124 // group tag since we know it's length has to be of four bytes i.e.
2125 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2126 // occurs when we can't find such group...
2128 uint32_t x = 4; // x : for ntohs
2129 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2135 // First, compare HostByteOrder and NetworkByteOrder in order to
2136 // determine if we shall need to swap bytes (i.e. the Endian type).
2137 if ( x == ntohs(x) )
2146 // The easiest case is the one of a 'true' DICOM header, we just have
2147 // to look for the string "DICM" inside the file preamble.
2150 char *entCur = deb + 128;
2151 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2153 gdcmVerboseMacro( "Looks like DICOM Version3 (preamble + DCM)" );
2155 // Group 0002 should always be VR, and the first element 0000
2156 // Let's be carefull (so many wrong headers ...)
2157 // and determine the value representation (VR) :
2158 // Let's skip to the first element (0002,0000) and check there if we find
2159 // "UL" - or "OB" if the 1st one is (0002,0001) -,
2160 // in which case we (almost) know it is explicit VR.
2161 // WARNING: if it happens to be implicit VR then what we will read
2162 // is the length of the group. If this ascii representation of this
2163 // length happens to be "UL" then we shall believe it is explicit VR.
2164 // We need to skip :
2165 // * the 128 bytes of File Preamble (often padded with zeroes),
2166 // * the 4 bytes of "DICM" string,
2167 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2168 // i.e. a total of 136 bytes.
2171 // group 0x0002 *is always* Explicit VR Sometimes ,
2172 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
2173 // *Implicit* VR (see former 'gdcmData/icone.dcm')
2175 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2176 memcmp(entCur, "OB", (size_t)2) == 0 ||
2177 memcmp(entCur, "UI", (size_t)2) == 0 ||
2178 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2179 // when Write DCM *adds*
2181 // Use Document::dicom_vr to test all the possibilities
2182 // instead of just checking for UL, OB and UI !? group 0000
2184 Filetype = ExplicitVR;
2185 gdcmVerboseMacro( "Group 0002 : Explicit Value Representation");
2189 Filetype = ImplicitVR;
2190 gdcmVerboseMacro( "Group 0002 :Not an explicit Value Representation;"
2191 << "Looks like a bugged Header!");
2197 gdcmVerboseMacro( "HostByteOrder != NetworkByteOrder");
2202 gdcmVerboseMacro( "HostByteOrder = NetworkByteOrder");
2205 // Position the file position indicator at first tag
2206 // (i.e. after the file preamble and the "DICM" string).
2207 Fp->seekg(0, std::ios::beg);
2208 Fp->seekg ( 132L, std::ios::beg);
2212 // Alas, this is not a DicomV3 file and whatever happens there is no file
2213 // preamble. We can reset the file position indicator to where the data
2214 // is (i.e. the beginning of the file).
2215 gdcmVerboseMacro( "Not a DICOM Version3 file");
2216 Fp->seekg(0, std::ios::beg);
2218 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2219 // By clean we mean that the length of the first tag is written down.
2220 // If this is the case and since the length of the first group HAS to be
2221 // four (bytes), then determining the proper swap code is straightforward.
2224 // We assume the array of char we are considering contains the binary
2225 // representation of a 32 bits integer. Hence the following dirty
2227 s32 = *((uint32_t *)(entCur));
2248 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2249 // It is time for despaired wild guesses.
2250 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2251 // i.e. the 'group length' element is not present :
2253 // check the supposed-to-be 'group number'
2254 // in ( 0x0001 .. 0x0008 )
2255 // to determine ' SwapCode' value .
2256 // Only 0 or 4321 will be possible
2257 // (no oportunity to check for the formerly well known
2258 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2259 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2260 // the file IS NOT ACR-NEMA nor DICOM V3
2261 // Find a trick to tell it the caller...
2263 s16 = *((uint16_t *)(deb));
2290 gdcmVerboseMacro( "ACR/NEMA unfound swap info (Really hopeless !)");
2298 * \brief Change the Byte Swap code.
2300 void Document::SwitchByteSwapCode()
2302 gdcmVerboseMacro( "Switching Byte Swap code from "<< SwapCode);
2303 if ( SwapCode == 1234 )
2307 else if ( SwapCode == 4321 )
2311 else if ( SwapCode == 3412 )
2315 else if ( SwapCode == 2143 )
2322 * \brief during parsing, Header Elements too long are not loaded in memory
2325 void Document::SetMaxSizeLoadEntry(long newSize)
2331 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2333 MaxSizeLoadEntry = 0xffffffff;
2336 MaxSizeLoadEntry = newSize;
2341 * \brief Header Elements too long will not be printed
2342 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2345 void Document::SetMaxSizePrintEntry(long newSize)
2351 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2353 MaxSizePrintEntry = 0xffffffff;
2356 MaxSizePrintEntry = newSize;
2362 * \brief Handle broken private tag from Philips NTSCAN
2363 * where the endianess is being switch to BigEndian for no
2367 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
2369 // Endian reversion. Some files contain groups of tags with reversed endianess.
2370 static int reversedEndian = 0;
2371 // try to fix endian switching in the middle of headers
2372 if ((group == 0xfeff) && (elem == 0x00e0))
2374 // start endian swap mark for group found
2376 SwitchByteSwapCode();
2381 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2383 // end of reversed endian group
2385 SwitchByteSwapCode();
2390 * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS
2391 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
2393 std::string Document::GetTransferSyntaxName()
2395 // use the TS (TS : Transfer Syntax)
2396 std::string transferSyntax = GetEntry(0x0002,0x0010);
2398 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
2400 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
2401 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
2402 return "Uncompressed ACR-NEMA";
2404 if ( transferSyntax == GDCM_UNFOUND )
2406 gdcmVerboseMacro( "Unfound Transfer Syntax (0002,0010)");
2407 return "Uncompressed ACR-NEMA";
2410 // we do it only when we need it
2411 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
2413 // Global::GetTS() is a global static you shall never try to delete it!
2418 * \brief Group 0002 is always coded Little Endian
2419 * whatever Transfer Syntax is
2422 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2424 // Endian reversion. Some files contain groups of tags with reversed endianess.
2425 if ( !Group0002Parsed && group != 0x0002)
2427 Group0002Parsed = true;
2428 // we just came out of group 0002
2429 // if Transfer syntax is Big Endian we have to change CheckSwap
2431 std::string ts = GetTransferSyntax();
2432 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2434 gdcmVerboseMacro("True DICOM File, with NO Tansfer Syntax: " << ts );
2438 // Group 0002 is always 'Explicit ...' enven when Transfer Syntax says 'Implicit ..."
2440 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ImplicitVRLittleEndian )
2442 Filetype = ImplicitVR;
2445 // FIXME Strangely, this works with
2446 //'Implicit VR Transfer Syntax (GE Private)
2447 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ExplicitVRBigEndian )
2449 gdcmVerboseMacro("Transfer Syntax Name = ["
2450 << GetTransferSyntaxName() << "]" );
2451 SwitchByteSwapCode();
2452 group = SwapShort(group);
2453 elem = SwapShort(elem);
2459 * \brief Read the next tag but WITHOUT loading it's value
2460 * (read the 'Group Number', the 'Element Number',
2461 * gets the Dict Entry
2462 * gets the VR, gets the length, gets the offset value)
2463 * @return On succes the newly created DocEntry, NULL on failure.
2465 DocEntry *Document::ReadNextDocEntry()
2472 group = ReadInt16();
2475 catch ( FormatError e )
2477 // We reached the EOF (or an error occured) therefore
2478 // header parsing has to be considered as finished.
2483 // Sometimes file contains groups of tags with reversed endianess.
2484 HandleBrokenEndian(group, elem);
2486 // In 'true DICOM' files Group 0002 is always little endian
2487 if ( HasDCMPreamble )
2488 HandleOutOfGroup0002(group, elem);
2490 std::string vr = FindDocEntryVR();
2491 std::string realVR = vr;
2493 if( vr == GDCM_UNKNOWN)
2495 DictEntry *dictEntry = GetDictEntry(group,elem);
2497 realVR = dictEntry->GetVR();
2501 if( Global::GetVR()->IsVROfSequence(realVR) )
2502 newEntry = NewSeqEntry(group, elem);
2503 else if( Global::GetVR()->IsVROfStringRepresentable(realVR) )
2504 newEntry = NewValEntry(group, elem,vr);
2506 newEntry = NewBinEntry(group, elem,vr);
2508 if( vr == GDCM_UNKNOWN )
2510 if( Filetype == ExplicitVR )
2512 // We thought this was explicit VR, but we end up with an
2513 // implicit VR tag. Let's backtrack.
2514 if ( newEntry->GetGroup() != 0xfffe )
2517 msg = Util::Format("Entry (%04x,%04x) should be Explicit VR\n",
2518 newEntry->GetGroup(), newEntry->GetElement());
2519 gdcmVerboseMacro( msg.c_str() );
2522 newEntry->SetImplicitVR();
2527 FindDocEntryLength(newEntry);
2529 catch ( FormatError e )
2537 newEntry->SetOffset(Fp->tellg());
2544 * \brief Generate a free TagKey i.e. a TagKey that is not present
2545 * in the TagHt dictionary.
2546 * @param group The generated tag must belong to this group.
2547 * @return The element of tag with given group which is fee.
2549 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2551 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2553 TagKey key = DictEntry::TranslateToKey(group, elem);
2554 if (TagHT.count(key) == 0)
2563 * \brief Assuming the internal file pointer \ref Document::Fp
2564 * is placed at the beginning of a tag check whether this
2565 * tag is (TestGroup, TestElement).
2566 * \warning On success the internal file pointer \ref Document::Fp
2567 * is modified to point after the tag.
2568 * On failure (i.e. when the tag wasn't the expected tag
2569 * (TestGroup, TestElement) the internal file pointer
2570 * \ref Document::Fp is restored to it's original position.
2571 * @param testGroup The expected group of the tag.
2572 * @param testElement The expected Element of the tag.
2573 * @return True on success, false otherwise.
2575 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2577 long positionOnEntry = Fp->tellg();
2578 long currentPosition = Fp->tellg(); // On debugging purposes
2580 //// Read the Item Tag group and element, and make
2581 // sure they are what we expected:
2582 uint16_t itemTagGroup;
2583 uint16_t itemTagElement;
2586 itemTagGroup = ReadInt16();
2587 itemTagElement = ReadInt16();
2589 catch ( FormatError e )
2591 //std::cerr << e << std::endl;
2594 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2596 gdcmVerboseMacro( "Wrong Item Tag found:"
2597 << " We should have found tag ("
2598 << std::hex << testGroup << "," << testElement << ")" << std::endl
2599 << " but instead we encountered tag ("
2600 << std::hex << itemTagGroup << "," << itemTagElement << ")"
2601 << " at address: " << " 0x(" << (unsigned int)currentPosition << ")"
2603 Fp->seekg(positionOnEntry, std::ios::beg);
2611 * \brief Assuming the internal file pointer \ref Document::Fp
2612 * is placed at the beginning of a tag (TestGroup, TestElement),
2613 * read the length associated to the Tag.
2614 * \warning On success the internal file pointer \ref Document::Fp
2615 * is modified to point after the tag and it's length.
2616 * On failure (i.e. when the tag wasn't the expected tag
2617 * (TestGroup, TestElement) the internal file pointer
2618 * \ref Document::Fp is restored to it's original position.
2619 * @param testGroup The expected group of the tag.
2620 * @param testElement The expected Element of the tag.
2621 * @return On success returns the length associated to the tag. On failure
2624 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2627 if ( !ReadTag(testGroup, testElement) )
2632 //// Then read the associated Item Length
2633 long currentPosition = Fp->tellg();
2634 uint32_t itemLength = ReadInt32();
2636 gdcmVerboseMacro( "Basic Item Length is: "
2637 << itemLength << std::endl
2638 << " at address: " << std::hex << (unsigned int)currentPosition);
2644 * \brief When parsing the Pixel Data of an encapsulated file, read
2645 * the basic offset table (when present, and BTW dump it).
2647 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2649 //// Read the Basic Offset Table Item Tag length...
2650 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2652 // When present, read the basic offset table itself.
2653 // Notes: - since the presence of this basic offset table is optional
2654 // we can't rely on it for the implementation, and we will simply
2655 // trash it's content (when present).
2656 // - still, when present, we could add some further checks on the
2657 // lengths, but we won't bother with such fuses for the time being.
2658 if ( itemLength != 0 )
2660 char *basicOffsetTableItemValue = new char[itemLength + 1];
2661 Fp->read(basicOffsetTableItemValue, itemLength);
2664 for (unsigned int i=0; i < itemLength; i += 4 )
2666 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2668 gdcmVerboseMacro( "Read one length: " <<
2669 std::hex << individualLength );
2673 delete[] basicOffsetTableItemValue;
2678 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2679 * Compute the RLE extra information and store it in \ref RLEInfo
2680 * for later pixel retrieval usage.
2682 void Document::ComputeRLEInfo()
2684 std::string ts = GetTransferSyntax();
2685 if ( !Global::GetTS()->IsRLELossless(ts) )
2690 // Encoded pixel data: for the time being we are only concerned with
2691 // Jpeg or RLE Pixel data encodings.
2692 // As stated in PS 3.5-2003, section 8.2 p44:
2693 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2694 // value representation OB is used".
2695 // Hence we expect an OB value representation. Concerning OB VR,
2696 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2697 // "For the Value Representations OB and OW, the encoding shall meet the
2698 // following specifications depending on the Data element tag:"
2700 // - the first item in the sequence of items before the encoded pixel
2701 // data stream shall be basic offset table item. The basic offset table
2702 // item value, however, is not required to be present"
2704 ReadAndSkipEncapsulatedBasicOffsetTable();
2706 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2707 // Loop on the individual frame[s] and store the information
2708 // on the RLE fragments in a RLEFramesInfo.
2709 // Note: - when only a single frame is present, this is a
2711 // - when more than one frame are present, then we are in
2712 // the case of a multi-frame image.
2714 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2716 // Parse the RLE Header and store the corresponding RLE Segment
2717 // Offset Table information on fragments of this current Frame.
2718 // Note that the fragment pixels themselves are not loaded
2719 // (but just skipped).
2720 long frameOffset = Fp->tellg();
2722 uint32_t nbRleSegments = ReadInt32();
2723 if ( nbRleSegments > 16 )
2725 // There should be at most 15 segments (refer to RLEFrame class)
2726 gdcmVerboseMacro( "Too many segments.");
2729 uint32_t rleSegmentOffsetTable[16];
2730 for( int k = 1; k <= 15; k++ )
2732 rleSegmentOffsetTable[k] = ReadInt32();
2735 // Deduce from both the RLE Header and the frameLength the
2736 // fragment length, and again store this info in a
2738 long rleSegmentLength[15];
2739 // skipping (not reading) RLE Segments
2740 if ( nbRleSegments > 1)
2742 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2744 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2745 - rleSegmentOffsetTable[k];
2746 SkipBytes(rleSegmentLength[k]);
2750 rleSegmentLength[nbRleSegments] = frameLength
2751 - rleSegmentOffsetTable[nbRleSegments];
2752 SkipBytes(rleSegmentLength[nbRleSegments]);
2754 // Store the collected info
2755 RLEFrame *newFrameInfo = new RLEFrame;
2756 newFrameInfo->NumberFragments = nbRleSegments;
2757 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2759 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2760 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2762 RLEInfo->Frames.push_back( newFrameInfo );
2765 // Make sure that at the end of the item we encounter a 'Sequence
2767 if ( !ReadTag(0xfffe, 0xe0dd) )
2769 gdcmVerboseMacro( "No sequence delimiter item at end of RLE item sequence");
2774 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2775 * Compute the jpeg extra information (fragment[s] offset[s] and
2776 * length) and store it[them] in \ref JPEGInfo for later pixel
2779 void Document::ComputeJPEGFragmentInfo()
2781 // If you need to, look for comments of ComputeRLEInfo().
2782 std::string ts = GetTransferSyntax();
2783 if ( ! Global::GetTS()->IsJPEG(ts) )
2788 ReadAndSkipEncapsulatedBasicOffsetTable();
2790 // Loop on the fragments[s] and store the parsed information in a
2792 long fragmentLength;
2793 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2795 long fragmentOffset = Fp->tellg();
2797 // Store the collected info
2798 JPEGFragment *newFragment = new JPEGFragment;
2799 newFragment->Offset = fragmentOffset;
2800 newFragment->Length = fragmentLength;
2801 JPEGInfo->Fragments.push_back( newFragment );
2803 SkipBytes( fragmentLength );
2806 // Make sure that at the end of the item we encounter a 'Sequence
2808 if ( !ReadTag(0xfffe, 0xe0dd) )
2810 gdcmVerboseMacro( "No sequence delimiter item at end of JPEG item sequence");
2815 * \brief Walk recursively the given \ref DocEntrySet, and feed
2816 * the given hash table (\ref TagDocEntryHT) with all the
2817 * \ref DocEntry (Dicom entries) encountered.
2818 * This method does the job for \ref BuildFlatHashTable.
2819 * @param builtHT Where to collect all the \ref DocEntry encountered
2820 * when recursively walking the given set.
2821 * @param set The structure to be traversed (recursively).
2823 /*void Document::BuildFlatHashTableRecurse( TagDocEntryHT &builtHT,
2826 if (ElementSet *elementSet = dynamic_cast< ElementSet* > ( set ) )
2828 TagDocEntryHT const ¤tHT = elementSet->GetTagHT();
2829 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2830 i != currentHT.end();
2833 DocEntry *entry = i->second;
2834 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2836 const ListSQItem& items = seqEntry->GetSQItems();
2837 for( ListSQItem::const_iterator item = items.begin();
2838 item != items.end();
2841 BuildFlatHashTableRecurse( builtHT, *item );
2845 builtHT[entry->GetKey()] = entry;
2850 if (SQItem *SQItemSet = dynamic_cast< SQItem* > ( set ) )
2852 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
2853 for (ListDocEntry::const_iterator i = currentList.begin();
2854 i != currentList.end();
2857 DocEntry *entry = *i;
2858 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2860 const ListSQItem& items = seqEntry->GetSQItems();
2861 for( ListSQItem::const_iterator item = items.begin();
2862 item != items.end();
2865 BuildFlatHashTableRecurse( builtHT, *item );
2869 builtHT[entry->GetKey()] = entry;
2876 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2879 * The structure used by a Document (through \ref ElementSet),
2880 * in order to hold the parsed entries of a Dicom header, is a recursive
2881 * one. This is due to the fact that the sequences (when present)
2882 * can be nested. Additionaly, the sequence items (represented in
2883 * gdcm as \ref SQItem) add an extra complexity to the data
2884 * structure. Hence, a gdcm user whishing to visit all the entries of
2885 * a Dicom header will need to dig in the gdcm internals (which
2886 * implies exposing all the internal data structures to the API).
2887 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2888 * recursively builds a temporary hash table, which holds all the
2889 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
2891 * \warning Of course there is NO integrity constrain between the
2892 * returned \ref TagDocEntryHT and the \ref ElementSet used
2893 * to build it. Hence if the underlying \ref ElementSet is
2894 * altered, then it is the caller responsability to invoke
2895 * \ref BuildFlatHashTable again...
2896 * @return The flat std::map<> we juste build.
2898 /*TagDocEntryHT *Document::BuildFlatHashTable()
2900 TagDocEntryHT *FlatHT = new TagDocEntryHT;
2901 BuildFlatHashTableRecurse( *FlatHT, this );
2908 * \brief Compares two documents, according to \ref DicomDir rules
2909 * \warning Does NOT work with ACR-NEMA files
2910 * \todo Find a trick to solve the pb (use RET fields ?)
2912 * @return true if 'smaller'
2914 bool Document::operator<(Document &document)
2917 std::string s1 = GetEntry(0x0010,0x0010);
2918 std::string s2 = document.GetEntry(0x0010,0x0010);
2930 s1 = GetEntry(0x0010,0x0020);
2931 s2 = document.GetEntry(0x0010,0x0020);
2942 // Study Instance UID
2943 s1 = GetEntry(0x0020,0x000d);
2944 s2 = document.GetEntry(0x0020,0x000d);
2955 // Serie Instance UID
2956 s1 = GetEntry(0x0020,0x000e);
2957 s2 = document.GetEntry(0x0020,0x000e);
2974 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
2975 * @param filetype Type of the File to be written
2977 int Document::ComputeGroup0002Length( FileType filetype )
2982 int groupLength = 0;
2983 bool found0002 = false;
2985 // for each zero-level Tag in the DCM Header
2986 DocEntry *entry = GetFirstEntry();
2989 gr = entry->GetGroup();
2995 el = entry->GetElement();
2996 vr = entry->GetVR();
2998 if (filetype == ExplicitVR)
3000 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
3002 groupLength += 4; // explicit VR AND OB, OW, SQ : 4 more bytes
3005 groupLength += 2 + 2 + 4 + entry->GetLength();
3007 else if (found0002 )
3010 entry = GetNextEntry();
3015 } // end namespace gdcm
3017 //-----------------------------------------------------------------------------