1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/18 18:03:16 $
7 Version: $Revision: 1.198 $
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 using (group, element)
927 * @param group Group number of the searched Dicom Element
928 * @param elem Element number of the searched Dicom Element
931 DocEntry *Document::GetDocEntry(uint16_t group, uint16_t elem)
933 TagKey key = DictEntry::TranslateToKey(group, elem);
934 if ( !TagHT.count(key))
938 return TagHT.find(key)->second;
942 * \brief Same as \ref Document::GetDocEntry except it only
943 * returns a result when the corresponding entry is of type
945 * @param group Group number of the searched Dicom Element
946 * @param elem Element number of the searched Dicom Element
947 * @return When present, the corresponding ValEntry.
949 ValEntry *Document::GetValEntry(uint16_t group, uint16_t elem)
951 DocEntry *currentEntry = GetDocEntry(group, elem);
956 if ( ValEntry *entry = dynamic_cast<ValEntry*>(currentEntry) )
960 gdcmVerboseMacro( "Unfound ValEntry.");
966 * \brief Same as \ref Document::GetDocEntry except it only
967 * returns a result when the corresponding entry is of type
969 * @param group Group number of the searched Dicom Element
970 * @param elem Element number of the searched Dicom Element
971 * @return When present, the corresponding BinEntry.
973 BinEntry *Document::GetBinEntry(uint16_t group, uint16_t elem)
975 DocEntry *currentEntry = GetDocEntry(group, elem);
980 if ( BinEntry *entry = dynamic_cast<BinEntry*>(currentEntry) )
984 gdcmVerboseMacro( "Unfound BinEntry.");
990 * \brief Same as \ref Document::GetDocEntry except it only
991 * returns a result when the corresponding entry is of type
993 * @param group Group number of the searched Dicom Element
994 * @param elem Element number of the searched Dicom Element
995 * @return When present, the corresponding SeqEntry.
997 SeqEntry *Document::GetSeqEntry(uint16_t group, uint16_t elem)
999 DocEntry *currentEntry = GetDocEntry(group, elem);
1000 if ( !currentEntry )
1004 if ( SeqEntry *entry = dynamic_cast<SeqEntry*>(currentEntry) )
1008 gdcmVerboseMacro( "Unfound SeqEntry.");
1015 * \brief Loads the element while preserving the current
1016 * underlying file position indicator as opposed to
1017 * LoadDocEntry that modifies it.
1018 * @param entry Header Entry whose value will be loaded.
1021 void Document::LoadDocEntrySafe(DocEntry *entry)
1025 long PositionOnEntry = Fp->tellg();
1026 LoadDocEntry(entry);
1027 Fp->seekg(PositionOnEntry, std::ios::beg);
1032 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1034 * @return The properly swaped 32 bits integer.
1036 uint32_t Document::SwapLong(uint32_t a)
1043 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1044 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1047 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1050 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1053 gdcmErrorMacro( "Unset swap code:" << SwapCode );
1060 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1062 * @return The properly unswaped 32 bits integer.
1064 uint32_t Document::UnswapLong(uint32_t a)
1070 * \brief Swaps the bytes so they agree with the processor order
1071 * @return The properly swaped 16 bits integer.
1073 uint16_t Document::SwapShort(uint16_t a)
1075 if ( SwapCode == 4321 || SwapCode == 2143 )
1077 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1083 * \brief Unswaps the bytes so they agree with the processor order
1084 * @return The properly unswaped 16 bits integer.
1086 uint16_t Document::UnswapShort(uint16_t a)
1088 return SwapShort(a);
1091 //-----------------------------------------------------------------------------
1095 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1096 * @return length of the parsed set.
1098 void Document::ParseDES(DocEntrySet *set, long offset,
1099 long l_max, bool delim_mode)
1101 DocEntry *newDocEntry = 0;
1102 ValEntry *newValEntry;
1103 BinEntry *newBinEntry;
1104 SeqEntry *newSeqEntry;
1110 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1116 newDocEntry = ReadNextDocEntry( );
1123 vr = newDocEntry->GetVR();
1124 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
1125 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
1126 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1128 if ( newValEntry || newBinEntry )
1132 if ( Filetype == ExplicitVR && ! Global::GetVR()->IsVROfBinaryRepresentable(vr) )
1134 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1135 gdcmVerboseMacro( std::hex << newDocEntry->GetGroup()
1136 << "|" << newDocEntry->GetElement()
1137 << " : Neither Valentry, nor BinEntry."
1138 "Probably unknown VR.");
1141 //////////////////// BinEntry or UNKOWN VR:
1142 // When "this" is a Document the Key is simply of the
1143 // form ( group, elem )...
1144 if ( dynamic_cast< Document* > ( set ) )
1146 newBinEntry->SetKey( newBinEntry->GetKey() );
1148 // but when "this" is a SQItem, we are inserting this new
1149 // valEntry in a sequence item, and the key has the
1150 // generalized form (refer to \ref BaseTagKey):
1151 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1153 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1154 + newBinEntry->GetKey() );
1157 LoadDocEntry( newBinEntry );
1158 if( !set->AddEntry( newBinEntry ) )
1160 //Expect big troubles if here
1161 //delete newBinEntry;
1167 /////////////////////// ValEntry
1168 // When "set" is a Document, then we are at the top of the
1169 // hierarchy and the Key is simply of the form ( group, elem )...
1170 if ( dynamic_cast< Document* > ( set ) )
1172 newValEntry->SetKey( newValEntry->GetKey() );
1174 // ...but when "set" is a SQItem, we are inserting this new
1175 // valEntry in a sequence item. Hence the key has the
1176 // generalized form (refer to \ref BaseTagKey):
1177 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1179 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1180 + newValEntry->GetKey() );
1183 LoadDocEntry( newValEntry );
1184 bool delimitor=newValEntry->IsItemDelimitor();
1185 if( !set->AddEntry( newValEntry ) )
1187 // If here expect big troubles
1188 //delete newValEntry; //otherwise mem leak
1198 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1206 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1207 && ( newDocEntry->GetElement() == 0x0010 ) )
1209 std::string ts = GetTransferSyntax();
1210 if ( Global::GetTS()->IsRLELossless(ts) )
1212 long positionOnEntry = Fp->tellg();
1213 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1215 Fp->seekg( positionOnEntry, std::ios::beg );
1217 else if ( Global::GetTS()->IsJPEG(ts) )
1219 long positionOnEntry = Fp->tellg();
1220 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1221 ComputeJPEGFragmentInfo();
1222 Fp->seekg( positionOnEntry, std::ios::beg );
1226 // Just to make sure we are at the beginning of next entry.
1227 SkipToNextDocEntry(newDocEntry);
1232 unsigned long l = newDocEntry->GetReadLength();
1233 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1235 if ( l == 0xffffffff )
1244 // no other way to create it ...
1245 newSeqEntry->SetDelimitorMode( delim_mode );
1247 // At the top of the hierarchy, stands a Document. When "set"
1248 // is a Document, then we are building the first depth level.
1249 // Hence the SeqEntry we are building simply has a depth
1251 if (/*Document *dummy =*/ dynamic_cast< Document* > ( set ) )
1254 newSeqEntry->SetDepthLevel( 1 );
1255 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1257 // But when "set" is already a SQItem, we are building a nested
1258 // sequence, and hence the depth level of the new SeqEntry
1259 // we are building, is one level deeper:
1260 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1262 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1263 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1264 + newSeqEntry->GetKey() );
1268 { // Don't try to parse zero-length sequences
1269 ParseSQ( newSeqEntry,
1270 newDocEntry->GetOffset(),
1273 set->AddEntry( newSeqEntry );
1274 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1286 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1287 * @return parsed length for this level
1289 void Document::ParseSQ( SeqEntry *seqEntry,
1290 long offset, long l_max, bool delim_mode)
1292 int SQItemNumber = 0;
1294 long offsetStartCurrentSQItem = offset;
1298 // the first time, we read the fff0,e000 of the first SQItem
1299 DocEntry *newDocEntry = ReadNextDocEntry();
1303 // FIXME Should warn user
1308 if ( newDocEntry->IsSequenceDelimitor() )
1310 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1314 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1319 // create the current SQItem
1320 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1321 std::ostringstream newBase;
1322 newBase << seqEntry->GetKey()
1326 itemSQ->SetBaseTagKey( newBase.str() );
1327 unsigned int l = newDocEntry->GetReadLength();
1329 if ( l == 0xffffffff )
1338 // when we're here, element fffe,e000 is already passed.
1339 // it's lost for the SQItem we're going to process !!
1341 //ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1342 //delete newDocEntry; // FIXME well ... it's too late to use it !
1344 // Let's try :------------
1345 // remove fff0,e000, created out of the SQItem
1347 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg);
1348 // fill up the current SQItem, starting at the beginning of fff0,e000
1349 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
1350 offsetStartCurrentSQItem = Fp->tellg();
1351 // end try -----------------
1353 seqEntry->AddEntry( itemSQ, SQItemNumber );
1355 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1363 * \brief Loads the element content if its length doesn't exceed
1364 * the value specified with Document::SetMaxSizeLoadEntry()
1365 * @param entry Header Entry (Dicom Element) to be dealt with
1367 void Document::LoadDocEntry(DocEntry *entry)
1369 uint16_t group = entry->GetGroup();
1370 std::string vr = entry->GetVR();
1371 uint32_t length = entry->GetLength();
1373 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1375 // A SeQuence "contains" a set of Elements.
1376 // (fffe e000) tells us an Element is beginning
1377 // (fffe e00d) tells us an Element just ended
1378 // (fffe e0dd) tells us the current SeQuence just ended
1379 if( group == 0xfffe )
1381 // NO more value field for SQ !
1385 // When the length is zero things are easy:
1388 ((ValEntry *)entry)->SetValue("");
1392 // The elements whose length is bigger than the specified upper bound
1393 // are not loaded. Instead we leave a short notice of the offset of
1394 // the element content and it's length.
1396 std::ostringstream s;
1397 if (length > MaxSizeLoadEntry)
1399 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1401 //s << "gdcm::NotLoaded (BinEntry)";
1402 s << GDCM_NOTLOADED;
1403 s << " Address:" << (long)entry->GetOffset();
1404 s << " Length:" << entry->GetLength();
1405 s << " x(" << std::hex << entry->GetLength() << ")";
1406 binEntryPtr->SetValue(s.str());
1408 // Be carefull : a BinEntry IS_A ValEntry ...
1409 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1411 // s << "gdcm::NotLoaded. (ValEntry)";
1412 s << GDCM_NOTLOADED;
1413 s << " Address:" << (long)entry->GetOffset();
1414 s << " Length:" << entry->GetLength();
1415 s << " x(" << std::hex << entry->GetLength() << ")";
1416 valEntryPtr->SetValue(s.str());
1421 gdcmErrorMacro( "MaxSizeLoadEntry exceeded, neither a BinEntry "
1422 << "nor a ValEntry ?! Should never print that !" );
1425 // to be sure we are at the end of the value ...
1426 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1431 // When we find a BinEntry not very much can be done :
1432 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1434 s << GDCM_BINLOADED;
1435 binEntryPtr->SetValue(s.str());
1436 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1440 /// \todo Any compacter code suggested (?)
1441 if ( IsDocEntryAnInteger(entry) )
1445 // When short integer(s) are expected, read and convert the following
1446 // n *two characters properly i.e. consider them as short integers as
1447 // opposed to strings.
1448 // Elements with Value Multiplicity > 1
1449 // contain a set of integers (not a single one)
1450 if (vr == "US" || vr == "SS")
1453 NewInt = ReadInt16();
1457 for (int i=1; i < nbInt; i++)
1460 NewInt = ReadInt16();
1465 // See above comment on multiple integers (mutatis mutandis).
1466 else if (vr == "UL" || vr == "SL")
1469 NewInt = ReadInt32();
1473 for (int i=1; i < nbInt; i++)
1476 NewInt = ReadInt32();
1481 #ifdef GDCM_NO_ANSI_STRING_STREAM
1482 s << std::ends; // to avoid oddities on Solaris
1483 #endif //GDCM_NO_ANSI_STRING_STREAM
1485 ((ValEntry *)entry)->SetValue(s.str());
1489 // FIXME: We need an additional byte for storing \0 that is not on disk
1490 char *str = new char[length+1];
1491 Fp->read(str, (size_t)length);
1492 str[length] = '\0'; //this is only useful when length is odd
1493 // Special DicomString call to properly handle \0 and even length
1494 std::string newValue;
1497 newValue = Util::DicomString(str, length+1);
1498 gdcmVerboseMacro("Warning: bad length: " << length <<
1499 ",For string :" << newValue.c_str());
1500 // Since we change the length of string update it length
1501 //entry->SetReadLength(length+1);
1505 newValue = Util::DicomString(str, length);
1509 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1511 if ( Fp->fail() || Fp->eof())
1513 gdcmVerboseMacro("Unread element value");
1514 valEntry->SetValue(GDCM_UNREAD);
1520 // Because of correspondance with the VR dic
1521 valEntry->SetValue(newValue);
1525 valEntry->SetValue(newValue);
1530 gdcmErrorMacro( "Should have a ValEntry, here !");
1536 * \brief Find the value Length of the passed Header Entry
1537 * @param entry Header Entry whose length of the value shall be loaded.
1539 void Document::FindDocEntryLength( DocEntry *entry )
1540 throw ( FormatError )
1542 std::string vr = entry->GetVR();
1545 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1547 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1549 // The following reserved two bytes (see PS 3.5-2003, section
1550 // "7.1.2 Data element structure with explicit vr", p 27) must be
1551 // skipped before proceeding on reading the length on 4 bytes.
1552 Fp->seekg( 2L, std::ios::cur);
1553 uint32_t length32 = ReadInt32();
1555 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1560 lengthOB = FindDocEntryLengthOBOrOW();
1562 catch ( FormatUnexpected )
1564 // Computing the length failed (this happens with broken
1565 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1566 // chance to get the pixels by deciding the element goes
1567 // until the end of the file. Hence we artificially fix the
1568 // the length and proceed.
1569 long currentPosition = Fp->tellg();
1570 Fp->seekg(0L,std::ios::end);
1572 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1573 Fp->seekg(currentPosition, std::ios::beg);
1575 entry->SetReadLength(lengthUntilEOF);
1576 entry->SetLength(lengthUntilEOF);
1579 entry->SetReadLength(lengthOB);
1580 entry->SetLength(lengthOB);
1583 FixDocEntryFoundLength(entry, length32);
1587 // Length is encoded on 2 bytes.
1588 length16 = ReadInt16();
1590 // FIXME : This heuristic supposes that the first group following
1591 // group 0002 *has* and element 0000.
1592 // BUT ... Element 0000 is optionnal :-(
1595 // Fixed using : HandleOutOfGroup0002()
1596 // (first hereafter strategy ...)
1598 // We can tell the current file is encoded in big endian (like
1599 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1600 // and it's value is the one of the encoding of a big endian file.
1601 // In order to deal with such big endian encoded files, we have
1602 // (at least) two strategies:
1603 // * when we load the "Transfer Syntax" tag with value of big endian
1604 // encoding, we raise the proper flags. Then we wait for the end
1605 // of the META group (0x0002) among which is "Transfer Syntax",
1606 // before switching the swap code to big endian. We have to postpone
1607 // the switching of the swap code since the META group is fully encoded
1608 // in little endian, and big endian coding only starts at the next
1609 // group. The corresponding code can be hard to analyse and adds
1610 // many additional unnecessary tests for regular tags.
1611 // * the second strategy consists in waiting for trouble, that shall
1612 // appear when we find the first group with big endian encoding. This
1613 // is easy to detect since the length of a "Group Length" tag (the
1614 // ones with zero as element number) has to be of 4 (0x0004). When we
1615 // encounter 1024 (0x0400) chances are the encoding changed and we
1616 // found a group with big endian encoding.
1617 //---> Unfortunately, element 0000 is optional.
1618 //---> This will not work when missing!
1619 // We shall use this second strategy. In order to make sure that we
1620 // can interpret the presence of an apparently big endian encoded
1621 // length of a "Group Length" without committing a big mistake, we
1622 // add an additional check: we look in the already parsed elements
1623 // for the presence of a "Transfer Syntax" whose value has to be "big
1624 // endian encoding". When this is the case, chances are we have got our
1625 // hands on a big endian encoded file: we switch the swap code to
1626 // big endian and proceed...
1629 // if ( element == 0x0000 && length16 == 0x0400 )
1631 // std::string ts = GetTransferSyntax();
1632 // if ( Global::GetTS()->GetSpecialTransferSyntax(ts)
1633 // != TS::ExplicitVRBigEndian )
1635 // throw FormatError( "Document::FindDocEntryLength()",
1636 // " not explicit VR." );
1640 // SwitchByteSwapCode();
1642 // Restore the unproperly loaded values i.e. the group, the element
1643 // and the dictionary entry depending on them.
1644 // uint16_t correctGroup = SwapShort( entry->GetGroup() );
1645 // uint16_t correctElem = SwapShort( entry->GetElement() );
1646 // DictEntry *newTag = GetDictEntry( correctGroup, correctElem ); if ( !newTag )
1648 // This correct tag is not in the dictionary. Create a new one.
1649 // newTag = NewVirtualDictEntry(correctGroup, correctElem);
1651 // FIXME this can create a memory leaks on the old entry that be
1652 // left unreferenced.
1653 // entry->SetDictEntry( newTag );
1657 // 0xffff means that we deal with 'No Length' Sequence
1658 // or 'No Length' SQItem
1659 if ( length16 == 0xffff)
1663 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1668 // Either implicit VR or a non DICOM conformal (see note below) explicit
1669 // VR that ommited the VR of (at least) this element. Farts happen.
1670 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1671 // on Data elements "Implicit and Explicit VR Data Elements shall
1672 // not coexist in a Data Set and Data Sets nested within it".]
1673 // Length is on 4 bytes.
1675 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1676 // even if Transfer Syntax is 'Implicit VR ...'
1678 FixDocEntryFoundLength( entry, ReadInt32() );
1684 * \brief Find the Value Representation of the current Dicom Element.
1685 * @return Value Representation of the current Entry
1687 std::string Document::FindDocEntryVR()
1689 if ( Filetype != ExplicitVR )
1690 return GDCM_UNKNOWN;
1692 long positionOnEntry = Fp->tellg();
1693 // Warning: we believe this is explicit VR (Value Representation) because
1694 // we used a heuristic that found "UL" in the first tag. Alas this
1695 // doesn't guarantee that all the tags will be in explicit VR. In some
1696 // cases (see e-film filtered files) one finds implicit VR tags mixed
1697 // within an explicit VR file. Hence we make sure the present tag
1698 // is in explicit VR and try to fix things if it happens not to be
1702 Fp->read (vr, (size_t)2);
1705 if( !CheckDocEntryVR(vr) )
1707 Fp->seekg(positionOnEntry, std::ios::beg);
1708 return GDCM_UNKNOWN;
1714 * \brief Check the correspondance between the VR of the header entry
1715 * and the taken VR. If they are different, the header entry is
1716 * updated with the new VR.
1717 * @param vr Dicom Value Representation
1718 * @return false if the VR is incorrect of if the VR isn't referenced
1719 * otherwise, it returns true
1721 bool Document::CheckDocEntryVR(VRKey vr)
1723 // CLEANME searching the dicom_vr at each occurence is expensive.
1724 // PostPone this test in an optional integrity check at the end
1725 // of parsing or only in debug mode.
1726 if ( !Global::GetVR()->IsValidVR(vr) )
1733 * \brief Get the transformed value of the header entry. The VR value
1734 * is used to define the transformation to operate on the value
1735 * \warning NOT end user intended method !
1736 * @param entry entry to tranform
1737 * @return Transformed entry value
1739 std::string Document::GetDocEntryValue(DocEntry *entry)
1741 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1743 std::string val = ((ValEntry *)entry)->GetValue();
1744 std::string vr = entry->GetVR();
1745 uint32_t length = entry->GetLength();
1746 std::ostringstream s;
1749 // When short integer(s) are expected, read and convert the following
1750 // n * 2 bytes properly i.e. as a multivaluated strings
1751 // (each single value is separated fromthe next one by '\'
1752 // as usual for standard multivaluated filels
1753 // Elements with Value Multiplicity > 1
1754 // contain a set of short integers (not a single one)
1756 if( vr == "US" || vr == "SS" )
1761 for (int i=0; i < nbInt; i++)
1767 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1768 newInt16 = SwapShort( newInt16 );
1773 // When integer(s) are expected, read and convert the following
1774 // n * 4 bytes properly i.e. as a multivaluated strings
1775 // (each single value is separated fromthe next one by '\'
1776 // as usual for standard multivaluated filels
1777 // Elements with Value Multiplicity > 1
1778 // contain a set of integers (not a single one)
1779 else if( vr == "UL" || vr == "SL" )
1784 for (int i=0; i < nbInt; i++)
1790 newInt32 = ( val[4*i+0] & 0xFF )
1791 + (( val[4*i+1] & 0xFF ) << 8 )
1792 + (( val[4*i+2] & 0xFF ) << 16 )
1793 + (( val[4*i+3] & 0xFF ) << 24 );
1794 newInt32 = SwapLong( newInt32 );
1798 #ifdef GDCM_NO_ANSI_STRING_STREAM
1799 s << std::ends; // to avoid oddities on Solaris
1800 #endif //GDCM_NO_ANSI_STRING_STREAM
1804 return ((ValEntry *)entry)->GetValue();
1808 * \brief Get the reverse transformed value of the header entry. The VR
1809 * value is used to define the reverse transformation to operate on
1811 * \warning NOT end user intended method !
1812 * @param entry Entry to reverse transform
1813 * @return Reverse transformed entry value
1815 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1817 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1819 std::string vr = entry->GetVR();
1820 std::vector<std::string> tokens;
1821 std::ostringstream s;
1823 if ( vr == "US" || vr == "SS" )
1827 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1828 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1829 for (unsigned int i=0; i<tokens.size(); i++)
1831 newInt16 = atoi(tokens[i].c_str());
1832 s << ( newInt16 & 0xFF )
1833 << (( newInt16 >> 8 ) & 0xFF );
1837 if ( vr == "UL" || vr == "SL")
1841 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1842 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1843 for (unsigned int i=0; i<tokens.size();i++)
1845 newInt32 = atoi(tokens[i].c_str());
1846 s << (char)( newInt32 & 0xFF )
1847 << (char)(( newInt32 >> 8 ) & 0xFF )
1848 << (char)(( newInt32 >> 16 ) & 0xFF )
1849 << (char)(( newInt32 >> 24 ) & 0xFF );
1854 #ifdef GDCM_NO_ANSI_STRING_STREAM
1855 s << std::ends; // to avoid oddities on Solaris
1856 #endif //GDCM_NO_ANSI_STRING_STREAM
1860 return ((ValEntry *)entry)->GetValue();
1864 * \brief Skip a given Header Entry
1865 * \warning NOT end user intended method !
1866 * @param entry entry to skip
1868 void Document::SkipDocEntry(DocEntry *entry)
1870 SkipBytes(entry->GetLength());
1874 * \brief Skips to the begining of the next Header Entry
1875 * \warning NOT end user intended method !
1876 * @param currentDocEntry entry to skip
1878 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1880 Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg);
1881 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1882 Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur);
1886 * \brief When the length of an element value is obviously wrong (because
1887 * the parser went Jabberwocky) one can hope improving things by
1888 * applying some heuristics.
1889 * @param entry entry to check
1890 * @param foundLength first assumption about length
1892 void Document::FixDocEntryFoundLength(DocEntry *entry,
1893 uint32_t foundLength)
1895 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1896 if ( foundLength == 0xffffffff)
1901 uint16_t gr = entry->GetGroup();
1902 uint16_t elem = entry->GetElement();
1904 if ( foundLength % 2)
1906 gdcmVerboseMacro( "Warning : Tag with uneven length " << foundLength
1907 << " in x(" << std::hex << gr << "," << elem <<")");
1910 //////// Fix for some naughty General Electric images.
1911 // Allthough not recent many such GE corrupted images are still present
1912 // on Creatis hard disks. Hence this fix shall remain when such images
1913 // are no longer in use (we are talking a few years, here)...
1914 // Note: XMedCom probably uses such a trick since it is able to read
1915 // those pesky GE images ...
1916 if ( foundLength == 13)
1918 // Only happens for this length !
1919 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1922 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
1926 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1927 // Occurence of such images is quite low (unless one leaves close to a
1928 // 'Leonardo' source. Hence, one might consider commenting out the
1929 // following fix on efficiency reasons.
1930 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1933 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
1936 else if ( entry->GetVR() == "SQ" )
1938 foundLength = 0; // ReadLength is unchanged
1941 //////// We encountered a 'delimiter' element i.e. a tag of the form
1942 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1943 // taken into account.
1944 else if( gr == 0xfffe )
1946 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1947 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1948 // causes extra troubles...
1949 if( entry->GetElement() != 0x0000 )
1955 entry->SetLength(foundLength);
1959 * \brief Apply some heuristics to predict whether the considered
1960 * element value contains/represents an integer or not.
1961 * @param entry The element value on which to apply the predicate.
1962 * @return The result of the heuristical predicate.
1964 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1966 uint16_t elem = entry->GetElement();
1967 uint16_t group = entry->GetGroup();
1968 const std::string &vr = entry->GetVR();
1969 uint32_t length = entry->GetLength();
1971 // When we have some semantics on the element we just read, and if we
1972 // a priori know we are dealing with an integer, then we shall be
1973 // able to swap it's element value properly.
1974 if ( elem == 0 ) // This is the group length of the group
1982 // Allthough this should never happen, still some images have a
1983 // corrupted group length [e.g. have a glance at offset x(8336) of
1984 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
1985 // Since for dicom compliant and well behaved headers, the present
1986 // test is useless (and might even look a bit paranoid), when we
1987 // encounter such an ill-formed image, we simply display a warning
1988 // message and proceed on parsing (while crossing fingers).
1989 long filePosition = Fp->tellg();
1990 gdcmVerboseMacro( "Erroneous Group Length element length on : ("
1991 << std::hex << group << " , " << elem
1992 << ") -before- position x(" << filePosition << ")"
1993 << "lgt : " << length );
1997 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2005 * \brief Find the Length till the next sequence delimiter
2006 * \warning NOT end user intended method !
2010 uint32_t Document::FindDocEntryLengthOBOrOW()
2011 throw( FormatUnexpected )
2013 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2014 long positionOnEntry = Fp->tellg();
2015 bool foundSequenceDelimiter = false;
2016 uint32_t totalLength = 0;
2018 while ( !foundSequenceDelimiter )
2024 group = ReadInt16();
2027 catch ( FormatError )
2029 throw FormatError("Unexpected end of file encountered during ",
2030 "Document::FindDocEntryLengthOBOrOW()");
2033 // We have to decount the group and element we just read
2036 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2038 long filePosition = Fp->tellg();
2039 gdcmVerboseMacro( "Neither an Item tag nor a Sequence delimiter tag on :"
2040 << std::hex << group << " , " << elem
2041 << ") -before- position x(" << filePosition << ")" );
2043 Fp->seekg(positionOnEntry, std::ios::beg);
2044 throw FormatUnexpected( "Neither an Item tag nor a Sequence delimiter tag.");
2047 if ( elem == 0xe0dd )
2049 foundSequenceDelimiter = true;
2052 uint32_t itemLength = ReadInt32();
2053 // We add 4 bytes since we just read the ItemLength with ReadInt32
2054 totalLength += itemLength + 4;
2055 SkipBytes(itemLength);
2057 if ( foundSequenceDelimiter )
2062 Fp->seekg( positionOnEntry, std::ios::beg);
2067 * \brief Reads a supposed to be 16 Bits integer
2068 * (swaps it depending on processor endianity)
2069 * @return read value
2071 uint16_t Document::ReadInt16()
2072 throw( FormatError )
2075 Fp->read ((char*)&g, (size_t)2);
2078 throw FormatError( "Document::ReadInt16()", " file error." );
2082 throw FormatError( "Document::ReadInt16()", "EOF." );
2089 * \brief Reads a supposed to be 32 Bits integer
2090 * (swaps it depending on processor endianity)
2091 * @return read value
2093 uint32_t Document::ReadInt32()
2094 throw( FormatError )
2097 Fp->read ((char*)&g, (size_t)4);
2100 throw FormatError( "Document::ReadInt32()", " file error." );
2104 throw FormatError( "Document::ReadInt32()", "EOF." );
2111 * \brief skips bytes inside the source file
2112 * \warning NOT end user intended method !
2115 void Document::SkipBytes(uint32_t nBytes)
2117 //FIXME don't dump the returned value
2118 Fp->seekg((long)nBytes, std::ios::cur);
2122 * \brief Loads all the needed Dictionaries
2123 * \warning NOT end user intended method !
2125 void Document::Initialise()
2127 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2129 RLEInfo = new RLEFramesInfo;
2130 JPEGInfo = new JPEGFragmentsInfo;
2135 * \brief Discover what the swap code is (among little endian, big endian,
2136 * bad little endian, bad big endian).
2138 * @return false when we are absolutely sure
2139 * it's neither ACR-NEMA nor DICOM
2140 * true when we hope ours assuptions are OK
2142 bool Document::CheckSwap()
2144 // The only guaranted way of finding the swap code is to find a
2145 // group tag since we know it's length has to be of four bytes i.e.
2146 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2147 // occurs when we can't find such group...
2149 uint32_t x = 4; // x : for ntohs
2150 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2156 // First, compare HostByteOrder and NetworkByteOrder in order to
2157 // determine if we shall need to swap bytes (i.e. the Endian type).
2158 if ( x == ntohs(x) )
2167 // The easiest case is the one of a 'true' DICOM header, we just have
2168 // to look for the string "DICM" inside the file preamble.
2171 char *entCur = deb + 128;
2172 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2174 gdcmVerboseMacro( "Looks like DICOM Version3 (preamble + DCM)" );
2176 // Group 0002 should always be VR, and the first element 0000
2177 // Let's be carefull (so many wrong headers ...)
2178 // and determine the value representation (VR) :
2179 // Let's skip to the first element (0002,0000) and check there if we find
2180 // "UL" - or "OB" if the 1st one is (0002,0001) -,
2181 // in which case we (almost) know it is explicit VR.
2182 // WARNING: if it happens to be implicit VR then what we will read
2183 // is the length of the group. If this ascii representation of this
2184 // length happens to be "UL" then we shall believe it is explicit VR.
2185 // We need to skip :
2186 // * the 128 bytes of File Preamble (often padded with zeroes),
2187 // * the 4 bytes of "DICM" string,
2188 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2189 // i.e. a total of 136 bytes.
2192 // group 0x0002 *is always* Explicit VR Sometimes ,
2193 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
2194 // *Implicit* VR (see former 'gdcmData/icone.dcm')
2196 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2197 memcmp(entCur, "OB", (size_t)2) == 0 ||
2198 memcmp(entCur, "UI", (size_t)2) == 0 ||
2199 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2200 // when Write DCM *adds*
2202 // Use Document::dicom_vr to test all the possibilities
2203 // instead of just checking for UL, OB and UI !? group 0000
2205 Filetype = ExplicitVR;
2206 gdcmVerboseMacro( "Group 0002 : Explicit Value Representation");
2210 Filetype = ImplicitVR;
2211 gdcmVerboseMacro( "Group 0002 :Not an explicit Value Representation;"
2212 << "Looks like a bugged Header!");
2218 gdcmVerboseMacro( "HostByteOrder != NetworkByteOrder");
2223 gdcmVerboseMacro( "HostByteOrder = NetworkByteOrder");
2226 // Position the file position indicator at first tag
2227 // (i.e. after the file preamble and the "DICM" string).
2228 Fp->seekg(0, std::ios::beg);
2229 Fp->seekg ( 132L, std::ios::beg);
2233 // Alas, this is not a DicomV3 file and whatever happens there is no file
2234 // preamble. We can reset the file position indicator to where the data
2235 // is (i.e. the beginning of the file).
2236 gdcmVerboseMacro( "Not a DICOM Version3 file");
2237 Fp->seekg(0, std::ios::beg);
2239 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2240 // By clean we mean that the length of the first tag is written down.
2241 // If this is the case and since the length of the first group HAS to be
2242 // four (bytes), then determining the proper swap code is straightforward.
2245 // We assume the array of char we are considering contains the binary
2246 // representation of a 32 bits integer. Hence the following dirty
2248 s32 = *((uint32_t *)(entCur));
2269 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2270 // It is time for despaired wild guesses.
2271 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2272 // i.e. the 'group length' element is not present :
2274 // check the supposed-to-be 'group number'
2275 // in ( 0x0001 .. 0x0008 )
2276 // to determine ' SwapCode' value .
2277 // Only 0 or 4321 will be possible
2278 // (no oportunity to check for the formerly well known
2279 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2280 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2281 // the file IS NOT ACR-NEMA nor DICOM V3
2282 // Find a trick to tell it the caller...
2284 s16 = *((uint16_t *)(deb));
2311 gdcmVerboseMacro( "ACR/NEMA unfound swap info (Really hopeless !)");
2319 * \brief Change the Byte Swap code.
2321 void Document::SwitchByteSwapCode()
2323 gdcmVerboseMacro( "Switching Byte Swap code from "<< SwapCode);
2324 if ( SwapCode == 1234 )
2328 else if ( SwapCode == 4321 )
2332 else if ( SwapCode == 3412 )
2336 else if ( SwapCode == 2143 )
2343 * \brief during parsing, Header Elements too long are not loaded in memory
2346 void Document::SetMaxSizeLoadEntry(long newSize)
2352 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2354 MaxSizeLoadEntry = 0xffffffff;
2357 MaxSizeLoadEntry = newSize;
2362 * \brief Header Elements too long will not be printed
2363 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2366 void Document::SetMaxSizePrintEntry(long newSize)
2372 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2374 MaxSizePrintEntry = 0xffffffff;
2377 MaxSizePrintEntry = newSize;
2383 * \brief Handle broken private tag from Philips NTSCAN
2384 * where the endianess is being switch to BigEndian for no
2388 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
2390 // Endian reversion. Some files contain groups of tags with reversed endianess.
2391 static int reversedEndian = 0;
2392 // try to fix endian switching in the middle of headers
2393 if ((group == 0xfeff) && (elem == 0x00e0))
2395 // start endian swap mark for group found
2397 SwitchByteSwapCode();
2402 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2404 // end of reversed endian group
2406 SwitchByteSwapCode();
2411 * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS
2412 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
2414 std::string Document::GetTransferSyntaxName()
2416 // use the TS (TS : Transfer Syntax)
2417 std::string transferSyntax = GetEntry(0x0002,0x0010);
2419 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
2421 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
2422 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
2423 return "Uncompressed ACR-NEMA";
2425 if ( transferSyntax == GDCM_UNFOUND )
2427 gdcmVerboseMacro( "Unfound Transfer Syntax (0002,0010)");
2428 return "Uncompressed ACR-NEMA";
2431 // we do it only when we need it
2432 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
2434 // Global::GetTS() is a global static you shall never try to delete it!
2439 * \brief Group 0002 is always coded Little Endian
2440 * whatever Transfer Syntax is
2443 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2445 // Endian reversion. Some files contain groups of tags with reversed endianess.
2446 if ( !Group0002Parsed && group != 0x0002)
2448 Group0002Parsed = true;
2449 // we just came out of group 0002
2450 // if Transfer syntax is Big Endian we have to change CheckSwap
2452 std::string ts = GetTransferSyntax();
2453 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2455 gdcmVerboseMacro("True DICOM File, with NO Tansfer Syntax: " << ts );
2459 // Group 0002 is always 'Explicit ...' enven when Transfer Syntax says 'Implicit ..."
2461 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ImplicitVRLittleEndian )
2463 Filetype = ImplicitVR;
2466 // FIXME Strangely, this works with
2467 //'Implicit VR Transfer Syntax (GE Private)
2468 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ExplicitVRBigEndian )
2470 gdcmVerboseMacro("Transfer Syntax Name = ["
2471 << GetTransferSyntaxName() << "]" );
2472 SwitchByteSwapCode();
2473 group = SwapShort(group);
2474 elem = SwapShort(elem);
2480 * \brief Read the next tag but WITHOUT loading it's value
2481 * (read the 'Group Number', the 'Element Number',
2482 * gets the Dict Entry
2483 * gets the VR, gets the length, gets the offset value)
2484 * @return On succes the newly created DocEntry, NULL on failure.
2486 DocEntry *Document::ReadNextDocEntry()
2493 group = ReadInt16();
2496 catch ( FormatError e )
2498 // We reached the EOF (or an error occured) therefore
2499 // header parsing has to be considered as finished.
2504 // Sometimes file contains groups of tags with reversed endianess.
2505 HandleBrokenEndian(group, elem);
2507 // In 'true DICOM' files Group 0002 is always little endian
2508 if ( HasDCMPreamble )
2509 HandleOutOfGroup0002(group, elem);
2511 std::string vr = FindDocEntryVR();
2512 std::string realVR = vr;
2514 if( vr == GDCM_UNKNOWN)
2516 DictEntry *dictEntry = GetDictEntry(group,elem);
2518 realVR = dictEntry->GetVR();
2522 if( Global::GetVR()->IsVROfSequence(realVR) )
2523 newEntry = NewSeqEntry(group, elem);
2524 else if( Global::GetVR()->IsVROfStringRepresentable(realVR) )
2525 newEntry = NewValEntry(group, elem,vr);
2527 newEntry = NewBinEntry(group, elem,vr);
2529 if( vr == GDCM_UNKNOWN )
2531 if( Filetype == ExplicitVR )
2533 // We thought this was explicit VR, but we end up with an
2534 // implicit VR tag. Let's backtrack.
2535 if ( newEntry->GetGroup() != 0xfffe )
2538 msg = Util::Format("Entry (%04x,%04x) should be Explicit VR\n",
2539 newEntry->GetGroup(), newEntry->GetElement());
2540 gdcmVerboseMacro( msg.c_str() );
2543 newEntry->SetImplicitVR();
2548 FindDocEntryLength(newEntry);
2550 catch ( FormatError e )
2558 newEntry->SetOffset(Fp->tellg());
2565 * \brief Generate a free TagKey i.e. a TagKey that is not present
2566 * in the TagHt dictionary.
2567 * @param group The generated tag must belong to this group.
2568 * @return The element of tag with given group which is fee.
2570 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2572 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2574 TagKey key = DictEntry::TranslateToKey(group, elem);
2575 if (TagHT.count(key) == 0)
2584 * \brief Assuming the internal file pointer \ref Document::Fp
2585 * is placed at the beginning of a tag check whether this
2586 * tag is (TestGroup, TestElement).
2587 * \warning On success the internal file pointer \ref Document::Fp
2588 * is modified to point after the tag.
2589 * On failure (i.e. when the tag wasn't the expected tag
2590 * (TestGroup, TestElement) the internal file pointer
2591 * \ref Document::Fp is restored to it's original position.
2592 * @param testGroup The expected group of the tag.
2593 * @param testElement The expected Element of the tag.
2594 * @return True on success, false otherwise.
2596 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2598 long positionOnEntry = Fp->tellg();
2599 long currentPosition = Fp->tellg(); // On debugging purposes
2601 //// Read the Item Tag group and element, and make
2602 // sure they are what we expected:
2603 uint16_t itemTagGroup;
2604 uint16_t itemTagElement;
2607 itemTagGroup = ReadInt16();
2608 itemTagElement = ReadInt16();
2610 catch ( FormatError e )
2612 //std::cerr << e << std::endl;
2615 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2617 gdcmVerboseMacro( "Wrong Item Tag found:"
2618 << " We should have found tag ("
2619 << std::hex << testGroup << "," << testElement << ")" << std::endl
2620 << " but instead we encountered tag ("
2621 << std::hex << itemTagGroup << "," << itemTagElement << ")"
2622 << " at address: " << " 0x(" << (unsigned int)currentPosition << ")"
2624 Fp->seekg(positionOnEntry, std::ios::beg);
2632 * \brief Assuming the internal file pointer \ref Document::Fp
2633 * is placed at the beginning of a tag (TestGroup, TestElement),
2634 * read the length associated to the Tag.
2635 * \warning On success the internal file pointer \ref Document::Fp
2636 * is modified to point after the tag and it's length.
2637 * On failure (i.e. when the tag wasn't the expected tag
2638 * (TestGroup, TestElement) the internal file pointer
2639 * \ref Document::Fp is restored to it's original position.
2640 * @param testGroup The expected group of the tag.
2641 * @param testElement The expected Element of the tag.
2642 * @return On success returns the length associated to the tag. On failure
2645 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2648 if ( !ReadTag(testGroup, testElement) )
2653 //// Then read the associated Item Length
2654 long currentPosition = Fp->tellg();
2655 uint32_t itemLength = ReadInt32();
2657 gdcmVerboseMacro( "Basic Item Length is: "
2658 << itemLength << std::endl
2659 << " at address: " << std::hex << (unsigned int)currentPosition);
2665 * \brief When parsing the Pixel Data of an encapsulated file, read
2666 * the basic offset table (when present, and BTW dump it).
2668 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2670 //// Read the Basic Offset Table Item Tag length...
2671 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2673 // When present, read the basic offset table itself.
2674 // Notes: - since the presence of this basic offset table is optional
2675 // we can't rely on it for the implementation, and we will simply
2676 // trash it's content (when present).
2677 // - still, when present, we could add some further checks on the
2678 // lengths, but we won't bother with such fuses for the time being.
2679 if ( itemLength != 0 )
2681 char *basicOffsetTableItemValue = new char[itemLength + 1];
2682 Fp->read(basicOffsetTableItemValue, itemLength);
2685 for (unsigned int i=0; i < itemLength; i += 4 )
2687 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2689 gdcmVerboseMacro( "Read one length: " <<
2690 std::hex << individualLength );
2694 delete[] basicOffsetTableItemValue;
2699 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2700 * Compute the RLE extra information and store it in \ref RLEInfo
2701 * for later pixel retrieval usage.
2703 void Document::ComputeRLEInfo()
2705 std::string ts = GetTransferSyntax();
2706 if ( !Global::GetTS()->IsRLELossless(ts) )
2711 // Encoded pixel data: for the time being we are only concerned with
2712 // Jpeg or RLE Pixel data encodings.
2713 // As stated in PS 3.5-2003, section 8.2 p44:
2714 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2715 // value representation OB is used".
2716 // Hence we expect an OB value representation. Concerning OB VR,
2717 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2718 // "For the Value Representations OB and OW, the encoding shall meet the
2719 // following specifications depending on the Data element tag:"
2721 // - the first item in the sequence of items before the encoded pixel
2722 // data stream shall be basic offset table item. The basic offset table
2723 // item value, however, is not required to be present"
2725 ReadAndSkipEncapsulatedBasicOffsetTable();
2727 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2728 // Loop on the individual frame[s] and store the information
2729 // on the RLE fragments in a RLEFramesInfo.
2730 // Note: - when only a single frame is present, this is a
2732 // - when more than one frame are present, then we are in
2733 // the case of a multi-frame image.
2735 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2737 // Parse the RLE Header and store the corresponding RLE Segment
2738 // Offset Table information on fragments of this current Frame.
2739 // Note that the fragment pixels themselves are not loaded
2740 // (but just skipped).
2741 long frameOffset = Fp->tellg();
2743 uint32_t nbRleSegments = ReadInt32();
2744 if ( nbRleSegments > 16 )
2746 // There should be at most 15 segments (refer to RLEFrame class)
2747 gdcmVerboseMacro( "Too many segments.");
2750 uint32_t rleSegmentOffsetTable[16];
2751 for( int k = 1; k <= 15; k++ )
2753 rleSegmentOffsetTable[k] = ReadInt32();
2756 // Deduce from both the RLE Header and the frameLength the
2757 // fragment length, and again store this info in a
2759 long rleSegmentLength[15];
2760 // skipping (not reading) RLE Segments
2761 if ( nbRleSegments > 1)
2763 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2765 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2766 - rleSegmentOffsetTable[k];
2767 SkipBytes(rleSegmentLength[k]);
2771 rleSegmentLength[nbRleSegments] = frameLength
2772 - rleSegmentOffsetTable[nbRleSegments];
2773 SkipBytes(rleSegmentLength[nbRleSegments]);
2775 // Store the collected info
2776 RLEFrame *newFrameInfo = new RLEFrame;
2777 newFrameInfo->NumberFragments = nbRleSegments;
2778 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2780 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2781 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2783 RLEInfo->Frames.push_back( newFrameInfo );
2786 // Make sure that at the end of the item we encounter a 'Sequence
2788 if ( !ReadTag(0xfffe, 0xe0dd) )
2790 gdcmVerboseMacro( "No sequence delimiter item at end of RLE item sequence");
2795 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2796 * Compute the jpeg extra information (fragment[s] offset[s] and
2797 * length) and store it[them] in \ref JPEGInfo for later pixel
2800 void Document::ComputeJPEGFragmentInfo()
2802 // If you need to, look for comments of ComputeRLEInfo().
2803 std::string ts = GetTransferSyntax();
2804 if ( ! Global::GetTS()->IsJPEG(ts) )
2809 ReadAndSkipEncapsulatedBasicOffsetTable();
2811 // Loop on the fragments[s] and store the parsed information in a
2813 long fragmentLength;
2814 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2816 long fragmentOffset = Fp->tellg();
2818 // Store the collected info
2819 JPEGFragment *newFragment = new JPEGFragment;
2820 newFragment->Offset = fragmentOffset;
2821 newFragment->Length = fragmentLength;
2822 JPEGInfo->Fragments.push_back( newFragment );
2824 SkipBytes( fragmentLength );
2827 // Make sure that at the end of the item we encounter a 'Sequence
2829 if ( !ReadTag(0xfffe, 0xe0dd) )
2831 gdcmVerboseMacro( "No sequence delimiter item at end of JPEG item sequence");
2836 * \brief Walk recursively the given \ref DocEntrySet, and feed
2837 * the given hash table (\ref TagDocEntryHT) with all the
2838 * \ref DocEntry (Dicom entries) encountered.
2839 * This method does the job for \ref BuildFlatHashTable.
2840 * @param builtHT Where to collect all the \ref DocEntry encountered
2841 * when recursively walking the given set.
2842 * @param set The structure to be traversed (recursively).
2844 /*void Document::BuildFlatHashTableRecurse( TagDocEntryHT &builtHT,
2847 if (ElementSet *elementSet = dynamic_cast< ElementSet* > ( set ) )
2849 TagDocEntryHT const ¤tHT = elementSet->GetTagHT();
2850 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2851 i != currentHT.end();
2854 DocEntry *entry = i->second;
2855 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2857 const ListSQItem& items = seqEntry->GetSQItems();
2858 for( ListSQItem::const_iterator item = items.begin();
2859 item != items.end();
2862 BuildFlatHashTableRecurse( builtHT, *item );
2866 builtHT[entry->GetKey()] = entry;
2871 if (SQItem *SQItemSet = dynamic_cast< SQItem* > ( set ) )
2873 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
2874 for (ListDocEntry::const_iterator i = currentList.begin();
2875 i != currentList.end();
2878 DocEntry *entry = *i;
2879 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2881 const ListSQItem& items = seqEntry->GetSQItems();
2882 for( ListSQItem::const_iterator item = items.begin();
2883 item != items.end();
2886 BuildFlatHashTableRecurse( builtHT, *item );
2890 builtHT[entry->GetKey()] = entry;
2897 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2900 * The structure used by a Document (through \ref ElementSet),
2901 * in order to hold the parsed entries of a Dicom header, is a recursive
2902 * one. This is due to the fact that the sequences (when present)
2903 * can be nested. Additionaly, the sequence items (represented in
2904 * gdcm as \ref SQItem) add an extra complexity to the data
2905 * structure. Hence, a gdcm user whishing to visit all the entries of
2906 * a Dicom header will need to dig in the gdcm internals (which
2907 * implies exposing all the internal data structures to the API).
2908 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2909 * recursively builds a temporary hash table, which holds all the
2910 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
2912 * \warning Of course there is NO integrity constrain between the
2913 * returned \ref TagDocEntryHT and the \ref ElementSet used
2914 * to build it. Hence if the underlying \ref ElementSet is
2915 * altered, then it is the caller responsability to invoke
2916 * \ref BuildFlatHashTable again...
2917 * @return The flat std::map<> we juste build.
2919 /*TagDocEntryHT *Document::BuildFlatHashTable()
2921 TagDocEntryHT *FlatHT = new TagDocEntryHT;
2922 BuildFlatHashTableRecurse( *FlatHT, this );
2929 * \brief Compares two documents, according to \ref DicomDir rules
2930 * \warning Does NOT work with ACR-NEMA files
2931 * \todo Find a trick to solve the pb (use RET fields ?)
2933 * @return true if 'smaller'
2935 bool Document::operator<(Document &document)
2938 std::string s1 = GetEntry(0x0010,0x0010);
2939 std::string s2 = document.GetEntry(0x0010,0x0010);
2951 s1 = GetEntry(0x0010,0x0020);
2952 s2 = document.GetEntry(0x0010,0x0020);
2963 // Study Instance UID
2964 s1 = GetEntry(0x0020,0x000d);
2965 s2 = document.GetEntry(0x0020,0x000d);
2976 // Serie Instance UID
2977 s1 = GetEntry(0x0020,0x000e);
2978 s2 = document.GetEntry(0x0020,0x000e);
2995 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
2996 * @param filetype Type of the File to be written
2998 int Document::ComputeGroup0002Length( FileType filetype )
3003 int groupLength = 0;
3004 bool found0002 = false;
3006 // for each zero-level Tag in the DCM Header
3007 DocEntry *entry = GetFirstEntry();
3010 gr = entry->GetGroup();
3016 el = entry->GetElement();
3017 vr = entry->GetVR();
3019 if (filetype == ExplicitVR)
3021 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
3023 groupLength += 4; // explicit VR AND OB, OW, SQ : 4 more bytes
3026 groupLength += 2 + 2 + 4 + entry->GetLength();
3028 else if (found0002 )
3031 entry = GetNextEntry();
3036 } // end namespace gdcm
3038 //-----------------------------------------------------------------------------