1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/17 17:27:03 $
7 Version: $Revision: 1.193 $
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__)
41 #include <netinet/in.h>
47 //-----------------------------------------------------------------------------
48 // Refer to Document::CheckSwap()
49 //const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
51 // Refer to Document::SetMaxSizeLoadEntry()
52 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
53 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
55 //-----------------------------------------------------------------------------
56 // Constructor / Destructor
60 * @param filename file to be opened for parsing
62 Document::Document( std::string const &filename ) : ElementSet(-1)
64 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
74 Group0002Parsed = false;
76 gdcmVerboseMacro( "Starting parsing of file: " << Filename.c_str());
77 // Fp->seekg( 0, std::ios::beg);
79 Fp->seekg(0, std::ios::end);
80 long lgt = Fp->tellg();
82 Fp->seekg( 0, std::ios::beg);
85 long beg = Fp->tellg();
88 ParseDES( this, beg, lgt, false); // Loading is done during parsing
90 Fp->seekg( 0, std::ios::beg);
92 // Load 'non string' values
94 std::string PhotometricInterpretation = GetEntry(0x0028,0x0004);
95 if( PhotometricInterpretation == "PALETTE COLOR " )
97 LoadEntryBinArea(0x0028,0x1200); // gray LUT
99 /// The tags refered by the three following lines used to be CORRECTLY
100 /// defined as having an US Value Representation in the public
101 /// dictionary. BUT the semantics implied by the three following
102 /// lines state that the corresponding tag contents are in fact
103 /// the ones of a BinEntry.
104 /// In order to fix things "Quick and Dirty" the dictionary was
105 /// altered on PURPOSE but now contains a WRONG value.
106 /// In order to fix things and restore the dictionary to its
107 /// correct value, one needs to decided of the semantics by deciding
108 /// whether the following tags are either:
109 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
110 /// also used as BinEntry, which requires the proper conversion,
111 /// - OW, and hence loaded as BinEntry, but afterwards also used
112 /// as ValEntry, which requires the proper conversion.
113 LoadEntryBinArea(0x0028,0x1201); // R LUT
114 LoadEntryBinArea(0x0028,0x1202); // G LUT
115 LoadEntryBinArea(0x0028,0x1203); // B LUT
117 // Segmented Red Palette Color LUT Data
118 LoadEntryBinArea(0x0028,0x1221);
119 // Segmented Green Palette Color LUT Data
120 LoadEntryBinArea(0x0028,0x1222);
121 // Segmented Blue Palette Color LUT Data
122 LoadEntryBinArea(0x0028,0x1223);
124 //FIXME later : how to use it?
125 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
129 // --------------------------------------------------------------
130 // Specific code to allow gdcm to read ACR-LibIDO formated images
131 // Note: ACR-LibIDO is an extension of the ACR standard that was
132 // used at CREATIS. For the time being (say a couple years)
133 // we keep this kludge to allow a smooth move to gdcm for
134 // CREATIS developpers (sorry folks).
136 // if recognition code tells us we deal with a LibIDO image
137 // we switch lineNumber and columnNumber
140 RecCode = GetEntry(0x0008, 0x0010); // recognition code
141 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
142 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
143 // with "little-endian strings"
145 Filetype = ACR_LIBIDO;
146 std::string rows = GetEntry(0x0028, 0x0010);
147 std::string columns = GetEntry(0x0028, 0x0011);
148 SetEntry(columns, 0x0028, 0x0010);
149 SetEntry(rows , 0x0028, 0x0011);
151 // ----------------- End of ACR-LibIDO kludge ------------------
155 * \brief This default constructor doesn't parse the file. You should
156 * then invoke \ref Document::SetFileName and then the parsing.
158 Document::Document() : ElementSet(-1)
162 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
165 Filetype = ExplicitVR;
166 Group0002Parsed = false;
170 * \brief Canonical destructor.
172 Document::~Document ()
181 //-----------------------------------------------------------------------------
185 * \brief Prints The Dict Entries of THE public Dicom Dictionary
188 void Document::PrintPubDict(std::ostream &os)
190 RefPubDict->SetPrintLevel(PrintLevel);
191 RefPubDict->Print(os);
195 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
198 void Document::PrintShaDict(std::ostream &os)
200 RefShaDict->SetPrintLevel(PrintLevel);
201 RefShaDict->Print(os);
204 //-----------------------------------------------------------------------------
207 * \brief Get the public dictionary used
209 Dict *Document::GetPubDict()
215 * \brief Get the shadow dictionary used
217 Dict *Document::GetShaDict()
223 * \brief Set the shadow dictionary used
224 * @param dict dictionary to use in shadow
226 bool Document::SetShaDict(Dict *dict)
233 * \brief Set the shadow dictionary used
234 * @param dictName name of the dictionary to use in shadow
236 bool Document::SetShaDict(DictKey const &dictName)
238 RefShaDict = Global::GetDicts()->GetDict(dictName);
243 * \brief This predicate, based on hopefully reasonable heuristics,
244 * decides whether or not the current Document was properly parsed
245 * and contains the mandatory information for being considered as
246 * a well formed and usable Dicom/Acr File.
247 * @return true when Document is the one of a reasonable Dicom/Acr file,
250 bool Document::IsReadable()
252 if( Filetype == Unknown)
254 gdcmVerboseMacro( "Wrong filetype");
260 gdcmVerboseMacro( "No tags in internal hash table.");
268 * \brief Accessor to the Transfer Syntax (when present) of the
269 * current document (it internally handles reading the
270 * value from disk when only parsing occured).
271 * @return The encountered Transfer Syntax of the current document.
273 std::string Document::GetTransferSyntax()
275 DocEntry *entry = GetDocEntry(0x0002, 0x0010);
281 // The entry might be present but not loaded (parsing and loading
282 // happen at different stages): try loading and proceed with check...
283 LoadDocEntrySafe(entry);
284 if (ValEntry *valEntry = dynamic_cast< ValEntry* >(entry) )
286 std::string transfer = valEntry->GetValue();
287 // The actual transfer (as read from disk) might be padded. We
288 // first need to remove the potential padding. We can make the
289 // weak assumption that padding was not executed with digits...
290 if ( transfer.length() == 0 )
292 // for brain damaged headers
295 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
297 transfer.erase(transfer.length()-1, 1);
305 * \brief Predicate for dicom version 3 file.
306 * @return True when the file is a dicom version 3.
308 bool Document::IsDicomV3()
310 // Checking if Transfer Syntax exists is enough
311 // Anyway, it's to late check if the 'Preamble' was found ...
312 // And ... would it be a rich idea to check ?
313 // (some 'no Preamble' DICOM images exist !)
314 return GetDocEntry(0x0002, 0x0010) != NULL;
318 * \brief returns the File Type
319 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
320 * @return the FileType code
322 FileType Document::GetFileType()
328 * \brief Tries to open the file \ref Document::Filename and
329 * checks the preamble when existing.
330 * @return The FILE pointer on success.
332 std::ifstream *Document::OpenFile()
335 HasDCMPreamble = false;
336 if (Filename.length() == 0)
343 gdcmVerboseMacro( "File already open: " << Filename.c_str());
346 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
349 gdcmDebugMacro( "Cannot open file: " << Filename.c_str());
356 Fp->read((char*)&zero, (size_t)2);
363 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
365 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
366 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
367 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
368 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
371 = Util::Format("ACR/DICOM with no preamble: (%04x)\n", zero);
372 gdcmVerboseMacro( msg.c_str() );
377 Fp->seekg(126L, std::ios::cur);
379 Fp->read(dicm, (size_t)4);
385 if( memcmp(dicm, "DICM", 4) == 0 )
387 HasDCMPreamble = true;
392 gdcmVerboseMacro( "Not DICOM/ACR (missing preamble)" << Filename.c_str());
398 * \brief closes the file
399 * @return TRUE if the close was successfull
401 bool Document::CloseFile()
410 return true; //FIXME how do we detect a non-close ifstream ?
414 * \brief Writes in a file all the Header Entries (Dicom Elements)
415 * @param fp file pointer on an already open file (actually: Output File Stream)
416 * @param filetype Type of the File to be written
417 * (ACR-NEMA, ExplicitVR, ImplicitVR)
418 * \return Always true.
420 void Document::WriteContent(std::ofstream *fp, FileType filetype)
422 /// \todo move the following lines (and a lot of others, to be written)
423 /// to a future function CheckAndCorrectHeader
424 /// (necessary if user wants to write a DICOM V3 file
425 /// starting from an ACR-NEMA (V2) Header
427 if ( filetype == ImplicitVR || filetype == ExplicitVR )
429 // writing Dicom File Preamble
430 char filePreamble[128];
431 memset(filePreamble, 0, 128);
432 fp->write(filePreamble, 128);
433 fp->write("DICM", 4);
437 * \todo rewrite later, if really usefull
438 * - 'Group Length' element is optional in DICOM
439 * - but un-updated odd groups lengthes can causes pb
442 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
443 * UpdateGroupLength(false,filetype);
444 * if ( filetype == ACR)
445 * UpdateGroupLength(true,ACR);
448 ElementSet::WriteContent(fp, filetype); // This one is recursive
452 * \brief Modifies the value of a given Doc Entry (Dicom Element)
453 * when it exists. Create it with the given value when unexistant.
454 * @param value (string) Value to be set
455 * @param group Group number of the Entry
456 * @param elem Element number of the Entry
457 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
458 * \return pointer to the modified/created Header Entry (NULL when creation
461 ValEntry *Document::ReplaceOrCreate(std::string const &value,
466 ValEntry *valEntry = 0;
467 DocEntry *currentEntry = GetDocEntry( group, elem);
471 valEntry = dynamic_cast< ValEntry* >(currentEntry);
475 if( valEntry->GetVR()!=vr )
478 // if currentEntry doesn't correspond to the requested valEntry
481 if (!RemoveEntry(currentEntry))
483 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
490 // Create a new valEntry if necessary
493 valEntry = NewValEntry(group, elem, vr);
495 if ( !AddEntry(valEntry))
497 gdcmVerboseMacro("AddEntry failed although this is a creation.");
504 // Set the binEntry value
505 SetEntry(value, valEntry);
510 * \brief Modifies the value of a given Header Entry (Dicom Element)
511 * when it exists. Create it with the given value when unexistant.
512 * A copy of the binArea is made to be kept in the Document.
513 * @param binArea (binary) value to be set
514 * @param Group Group number of the Entry
515 * @param Elem Element number of the Entry
516 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
517 * \return pointer to the modified/created Header Entry (NULL when creation
520 BinEntry *Document::ReplaceOrCreate(uint8_t *binArea,
526 BinEntry *binEntry = 0;
527 DocEntry *currentEntry = GetDocEntry( group, elem);
529 // Verify the currentEntry
532 binEntry = dynamic_cast< BinEntry* >(currentEntry);
536 if( binEntry->GetVR()!=vr )
539 // if currentEntry doesn't correspond to the requested valEntry
542 if (!RemoveEntry(currentEntry))
544 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
551 // Create a new binEntry if necessary
554 binEntry = NewBinEntry(group, elem, vr);
556 if ( !AddEntry(binEntry))
558 gdcmVerboseMacro( "AddEntry failed allthough this is a creation.");
565 // Set the binEntry value
567 if (lgth>0 && binArea)
569 tmpArea = new uint8_t[lgth];
570 memcpy(tmpArea,binArea,lgth);
576 if (!SetEntry(tmpArea,lgth,binEntry))
588 * \brief Modifies the value of a given Header Entry (Dicom Element)
589 * when it exists. Create it when unexistant.
590 * @param group Group number of the Entry
591 * @param elem Element number of the Entry
592 * \return pointer to the modified/created SeqEntry (NULL when creation
595 SeqEntry *Document::ReplaceOrCreate( uint16_t group, uint16_t elem)
597 SeqEntry *seqEntry = 0;
598 DocEntry *currentEntry = GetDocEntry( group, elem);
600 // Verify the currentEntry
603 seqEntry = dynamic_cast< SeqEntry* >(currentEntry);
607 if( seqEntry->GetVR()!="SQ" )
610 // if currentEntry doesn't correspond to the requested valEntry
613 if (!RemoveEntry(currentEntry))
615 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
622 // Create a new seqEntry if necessary
625 seqEntry = NewSeqEntry(group, elem);
627 if ( !AddEntry(seqEntry))
629 gdcmVerboseMacro( "AddEntry failed allthough this is a creation.");
640 * \brief Set a new value if the invoked element exists
641 * Seems to be useless !!!
642 * @param value new element value
643 * @param group group number of the Entry
644 * @param elem element number of the Entry
647 bool Document::ReplaceIfExist(std::string const &value,
648 uint16_t group, uint16_t elem )
650 SetEntry(value, group, elem);
655 //-----------------------------------------------------------------------------
659 * \brief Checks if a given Dicom Element exists within the H table
660 * @param group Group number of the searched Dicom Element
661 * @param elem Element number of the searched Dicom Element
662 * @return true is found
664 bool Document::CheckIfEntryExist(uint16_t group, uint16_t elem )
666 const std::string &key = DictEntry::TranslateToKey(group, elem );
667 return TagHT.count(key) != 0;
672 * \brief Searches within Header Entries (Dicom Elements) parsed with
673 * the public and private dictionaries
674 * for the element value representation of a given tag.
675 * @param group Group number of the searched tag.
676 * @param elem Element number of the searched tag.
677 * @return Corresponding element value representation when it exists,
678 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
680 std::string Document::GetEntry(uint16_t group, uint16_t elem)
682 TagKey key = DictEntry::TranslateToKey(group, elem);
683 if ( !TagHT.count(key))
688 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
692 * \brief Searches within Header Entries (Dicom Elements) parsed with
693 * the public and private dictionaries
694 * for the element value representation of a given tag..
695 * Obtaining the VR (Value Representation) might be needed by caller
696 * to convert the string typed content to caller's native type
697 * (think of C++ vs Python). The VR is actually of a higher level
698 * of semantics than just the native C++ type.
699 * @param group Group number of the searched tag.
700 * @param elem Element number of the searched tag.
701 * @return Corresponding element value representation when it exists,
702 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
704 std::string Document::GetEntryVR(uint16_t group, uint16_t elem)
706 DocEntry *elem = GetDocEntry(group, elem);
711 return elem->GetVR();
715 * \brief Searches within Header Entries (Dicom Elements) parsed with
716 * the public and private dictionaries
717 * for the value length of a given tag..
718 * @param group Group number of the searched tag.
719 * @param elem Element number of the searched tag.
720 * @return Corresponding element length; -2 if not found
722 int Document::GetEntryLength(uint16_t group, uint16_t elem)
724 DocEntry *elem = GetDocEntry(group, elem);
727 return -2; //magic number
729 return elem->GetLength();
733 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
734 * through it's (group, element) and modifies it's content with
736 * @param content new value (string) to substitute with
737 * @param group group number of the Dicom Element to modify
738 * @param elem element number of the Dicom Element to modify
740 bool Document::SetEntry(std::string const& content,
741 uint16_t group, uint16_t elem)
743 ValEntry *entry = GetValEntry(group, elem);
746 gdcmVerboseMacro( "No corresponding ValEntry (try promotion first).");
749 return SetEntry(content,entry);
753 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
754 * through it's (group, element) and modifies it's content with
756 * @param content new value (void* -> uint8_t*) to substitute with
757 * @param lgth new value length
758 * @param group group number of the Dicom Element to modify
759 * @param elem element number of the Dicom Element to modify
761 bool Document::SetEntry(uint8_t*content, int lgth,
762 uint16_t group, uint16_t elem)
764 BinEntry *entry = GetBinEntry(group, elem);
767 gdcmVerboseMacro( "No corresponding ValEntry (try promotion first).");
771 return SetEntry(content,lgth,entry);
775 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
776 * and modifies it's content with the given value.
777 * @param content new value (string) to substitute with
778 * @param entry Entry to be modified
780 bool Document::SetEntry(std::string const &content,ValEntry *entry)
784 entry->SetValue(content);
791 * \brief Accesses an existing BinEntry (i.e. a Dicom Element)
792 * and modifies it's content with the given value.
793 * @param content new value (void* -> uint8_t*) to substitute with
794 * @param entry Entry to be modified
795 * @param lgth new value length
797 bool Document::SetEntry(uint8_t *content, int lgth, BinEntry *entry)
801 // Hope Binary field length is *never* wrong
802 /*if(lgth%2) // Non even length are padded with a space (020H).
805 //content = content + '\0'; // fing a trick to enlarge a binary field?
808 entry->SetBinArea(content);
809 entry->SetLength(lgth);
810 entry->SetValue(GDCM_BINLOADED);
817 * \brief Gets (from Header) a 'non string' element value
818 * (LoadElementValues has already be executed)
819 * @param group group number of the Entry
820 * @param elem element number of the Entry
821 * @return Pointer to the 'non string' area
823 void *Document::GetEntryBinArea(uint16_t group, uint16_t elem)
825 DocEntry *entry = GetDocEntry(group, elem);
828 gdcmVerboseMacro( "No entry");
831 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(entry) )
833 return binEntry->GetBinArea();
840 * \brief Loads (from disk) the element content
841 * when a string is not suitable
842 * @param group group number of the Entry
843 * @param elem element number of the Entry
845 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
847 // Search the corresponding DocEntry
848 DocEntry *docElement = GetDocEntry(group, elem);
852 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
856 LoadEntryBinArea(binElement);
860 * \brief Loads (from disk) the element content
861 * when a string is not suitable
862 * @param elem Entry whose binArea is going to be loaded
864 void Document::LoadEntryBinArea(BinEntry *elem)
866 if(elem->GetBinArea())
873 size_t o =(size_t)elem->GetOffset();
874 Fp->seekg(o, std::ios::beg);
876 size_t l = elem->GetLength();
877 uint8_t *a = new uint8_t[l];
880 gdcmVerboseMacro( "Cannot allocate a");
884 /// \todo check the result
885 Fp->read((char*)a, l);
886 if( Fp->fail() || Fp->eof())
899 * \brief Sets a 'non string' value to a given Dicom Element
900 * @param area area containing the 'non string' value
901 * @param group Group number of the searched Dicom Element
902 * @param elem Element number of the searched Dicom Element
905 /*bool Document::SetEntryBinArea(uint8_t *area,
906 uint16_t group, uint16_t elem)
908 DocEntry *currentEntry = GetDocEntry(group, elem);
914 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(currentEntry) )
916 binEntry->SetBinArea( area );
924 * \brief retrieves a Dicom Element (the first one) using (group, element)
925 * \warning (group, element) IS NOT an identifier inside the Dicom Header
926 * if you think it's NOT UNIQUE, check the count number
927 * and use iterators to retrieve ALL the Dicoms Elements within
928 * a given couple (group, element)
929 * @param group Group number of the searched Dicom Element
930 * @param elem Element number of the searched Dicom Element
933 DocEntry *Document::GetDocEntry(uint16_t group, uint16_t elem)
935 TagKey key = DictEntry::TranslateToKey(group, elem);
936 if ( !TagHT.count(key))
940 return TagHT.find(key)->second;
944 * \brief Same as \ref Document::GetDocEntry except it only
945 * returns a result when the corresponding entry is of type
947 * @param group Group number of the searched Dicom Element
948 * @param elem Element number of the searched Dicom Element
949 * @return When present, the corresponding ValEntry.
951 ValEntry *Document::GetValEntry(uint16_t group, uint16_t elem)
953 DocEntry *currentEntry = GetDocEntry(group, elem);
958 if ( ValEntry *entry = dynamic_cast<ValEntry*>(currentEntry) )
962 gdcmVerboseMacro( "Unfound ValEntry.");
968 * \brief Same as \ref Document::GetDocEntry except it only
969 * returns a result when the corresponding entry is of type
971 * @param group Group number of the searched Dicom Element
972 * @param elem Element number of the searched Dicom Element
973 * @return When present, the corresponding BinEntry.
975 BinEntry *Document::GetBinEntry(uint16_t group, uint16_t elem)
977 DocEntry *currentEntry = GetDocEntry(group, elem);
982 if ( BinEntry *entry = dynamic_cast<BinEntry*>(currentEntry) )
986 gdcmVerboseMacro( "Unfound BinEntry.");
992 * \brief Loads the element while preserving the current
993 * underlying file position indicator as opposed to
994 * to LoadDocEntry that modifies it.
995 * @param entry Header Entry whose value will be loaded.
998 void Document::LoadDocEntrySafe(DocEntry *entry)
1002 long PositionOnEntry = Fp->tellg();
1003 LoadDocEntry(entry);
1004 Fp->seekg(PositionOnEntry, std::ios::beg);
1009 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1011 * @return The properly swaped 32 bits integer.
1013 uint32_t Document::SwapLong(uint32_t a)
1020 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1021 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1024 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1027 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1030 gdcmErrorMacro( "Unset swap code:" << SwapCode );
1037 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1039 * @return The properly unswaped 32 bits integer.
1041 uint32_t Document::UnswapLong(uint32_t a)
1047 * \brief Swaps the bytes so they agree with the processor order
1048 * @return The properly swaped 16 bits integer.
1050 uint16_t Document::SwapShort(uint16_t a)
1052 if ( SwapCode == 4321 || SwapCode == 2143 )
1054 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1060 * \brief Unswaps the bytes so they agree with the processor order
1061 * @return The properly unswaped 16 bits integer.
1063 uint16_t Document::UnswapShort(uint16_t a)
1065 return SwapShort(a);
1068 //-----------------------------------------------------------------------------
1072 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1073 * @return length of the parsed set.
1075 void Document::ParseDES(DocEntrySet *set, long offset,
1076 long l_max, bool delim_mode)
1078 DocEntry *newDocEntry = 0;
1079 ValEntry *newValEntry;
1080 BinEntry *newBinEntry;
1081 SeqEntry *newSeqEntry;
1087 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1093 newDocEntry = ReadNextDocEntry( );
1100 vr = newDocEntry->GetVR();
1101 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
1102 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
1103 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1105 if ( newValEntry || newBinEntry )
1109 if ( ! Global::GetVR()->IsVROfBinaryRepresentable(vr) )
1111 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1112 gdcmVerboseMacro( std::hex << newDocEntry->GetGroup()
1113 << "|" << newDocEntry->GetElement()
1114 << " : Neither Valentry, nor BinEntry."
1115 "Probably unknown VR.");
1118 //////////////////// BinEntry or UNKOWN VR:
1119 // When "this" is a Document the Key is simply of the
1120 // form ( group, elem )...
1121 if (/*Document *dummy =*/ dynamic_cast< Document* > ( set ) )
1124 newBinEntry->SetKey( newBinEntry->GetKey() );
1126 // but when "this" is a SQItem, we are inserting this new
1127 // valEntry in a sequence item, and the key has the
1128 // generalized form (refer to \ref BaseTagKey):
1129 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1131 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1132 + newBinEntry->GetKey() );
1135 LoadDocEntry( newBinEntry );
1136 if( !set->AddEntry( newBinEntry ) )
1138 //Expect big troubles if here
1139 //delete newBinEntry;
1145 /////////////////////// ValEntry
1146 // When "set" is a Document, then we are at the top of the
1147 // hierarchy and the Key is simply of the form ( group, elem )...
1148 if (/*Document *dummy =*/ 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 FixDocEntryFoundLength( entry, ReadInt32() );
1660 * \brief Find the Value Representation of the current Dicom Element.
1661 * @return Value Representation of the current Entry
1663 std::string Document::FindDocEntryVR()
1665 if ( Filetype != ExplicitVR )
1666 return GDCM_UNKNOWN;
1668 long positionOnEntry = Fp->tellg();
1669 // Warning: we believe this is explicit VR (Value Representation) because
1670 // we used a heuristic that found "UL" in the first tag. Alas this
1671 // doesn't guarantee that all the tags will be in explicit VR. In some
1672 // cases (see e-film filtered files) one finds implicit VR tags mixed
1673 // within an explicit VR file. Hence we make sure the present tag
1674 // is in explicit VR and try to fix things if it happens not to be
1678 Fp->read (vr, (size_t)2);
1681 if( !CheckDocEntryVR(vr) )
1683 Fp->seekg(positionOnEntry, std::ios::beg);
1684 return GDCM_UNKNOWN;
1690 * \brief Check the correspondance between the VR of the header entry
1691 * and the taken VR. If they are different, the header entry is
1692 * updated with the new VR.
1693 * @param vr Dicom Value Representation
1694 * @return false if the VR is incorrect of if the VR isn't referenced
1695 * otherwise, it returns true
1697 bool Document::CheckDocEntryVR(VRKey vr)
1699 // CLEANME searching the dicom_vr at each occurence is expensive.
1700 // PostPone this test in an optional integrity check at the end
1701 // of parsing or only in debug mode.
1702 if ( !Global::GetVR()->IsValidVR(vr) )
1709 * \brief Get the transformed value of the header entry. The VR value
1710 * is used to define the transformation to operate on the value
1711 * \warning NOT end user intended method !
1712 * @param entry entry to tranform
1713 * @return Transformed entry value
1715 std::string Document::GetDocEntryValue(DocEntry *entry)
1717 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1719 std::string val = ((ValEntry *)entry)->GetValue();
1720 std::string vr = entry->GetVR();
1721 uint32_t length = entry->GetLength();
1722 std::ostringstream s;
1725 // When short integer(s) are expected, read and convert the following
1726 // n * 2 bytes properly i.e. as a multivaluated strings
1727 // (each single value is separated fromthe next one by '\'
1728 // as usual for standard multivaluated filels
1729 // Elements with Value Multiplicity > 1
1730 // contain a set of short integers (not a single one)
1732 if( vr == "US" || vr == "SS" )
1737 for (int i=0; i < nbInt; i++)
1743 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1744 newInt16 = SwapShort( newInt16 );
1749 // When integer(s) are expected, read and convert the following
1750 // n * 4 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 integers (not a single one)
1755 else if( vr == "UL" || vr == "SL" )
1760 for (int i=0; i < nbInt; i++)
1766 newInt32 = ( val[4*i+0] & 0xFF )
1767 + (( val[4*i+1] & 0xFF ) << 8 )
1768 + (( val[4*i+2] & 0xFF ) << 16 )
1769 + (( val[4*i+3] & 0xFF ) << 24 );
1770 newInt32 = SwapLong( newInt32 );
1774 #ifdef GDCM_NO_ANSI_STRING_STREAM
1775 s << std::ends; // to avoid oddities on Solaris
1776 #endif //GDCM_NO_ANSI_STRING_STREAM
1780 return ((ValEntry *)entry)->GetValue();
1784 * \brief Get the reverse transformed value of the header entry. The VR
1785 * value is used to define the reverse transformation to operate on
1787 * \warning NOT end user intended method !
1788 * @param entry Entry to reverse transform
1789 * @return Reverse transformed entry value
1791 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1793 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1795 std::string vr = entry->GetVR();
1796 std::vector<std::string> tokens;
1797 std::ostringstream s;
1799 if ( vr == "US" || vr == "SS" )
1803 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1804 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1805 for (unsigned int i=0; i<tokens.size(); i++)
1807 newInt16 = atoi(tokens[i].c_str());
1808 s << ( newInt16 & 0xFF )
1809 << (( newInt16 >> 8 ) & 0xFF );
1813 if ( vr == "UL" || vr == "SL")
1817 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1818 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1819 for (unsigned int i=0; i<tokens.size();i++)
1821 newInt32 = atoi(tokens[i].c_str());
1822 s << (char)( newInt32 & 0xFF )
1823 << (char)(( newInt32 >> 8 ) & 0xFF )
1824 << (char)(( newInt32 >> 16 ) & 0xFF )
1825 << (char)(( newInt32 >> 24 ) & 0xFF );
1830 #ifdef GDCM_NO_ANSI_STRING_STREAM
1831 s << std::ends; // to avoid oddities on Solaris
1832 #endif //GDCM_NO_ANSI_STRING_STREAM
1836 return ((ValEntry *)entry)->GetValue();
1840 * \brief Skip a given Header Entry
1841 * \warning NOT end user intended method !
1842 * @param entry entry to skip
1844 void Document::SkipDocEntry(DocEntry *entry)
1846 SkipBytes(entry->GetLength());
1850 * \brief Skips to the begining of the next Header Entry
1851 * \warning NOT end user intended method !
1852 * @param currentDocEntry entry to skip
1854 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1856 Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg);
1857 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1858 Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur);
1862 * \brief When the length of an element value is obviously wrong (because
1863 * the parser went Jabberwocky) one can hope improving things by
1864 * applying some heuristics.
1865 * @param entry entry to check
1866 * @param foundLength first assumption about length
1868 void Document::FixDocEntryFoundLength(DocEntry *entry,
1869 uint32_t foundLength)
1871 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1872 if ( foundLength == 0xffffffff)
1877 uint16_t gr = entry->GetGroup();
1878 uint16_t elem = entry->GetElement();
1880 if ( foundLength % 2)
1882 gdcmVerboseMacro( "Warning : Tag with uneven length " << foundLength
1883 << " in x(" << std::hex << gr << "," << elem <<")");
1886 //////// Fix for some naughty General Electric images.
1887 // Allthough not recent many such GE corrupted images are still present
1888 // on Creatis hard disks. Hence this fix shall remain when such images
1889 // are no longer in use (we are talking a few years, here)...
1890 // Note: XMedCom probably uses such a trick since it is able to read
1891 // those pesky GE images ...
1892 if ( foundLength == 13)
1894 // Only happens for this length !
1895 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1898 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
1902 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1903 // Occurence of such images is quite low (unless one leaves close to a
1904 // 'Leonardo' source. Hence, one might consider commenting out the
1905 // following fix on efficiency reasons.
1906 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1909 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
1912 else if ( entry->GetVR() == "SQ" )
1914 foundLength = 0; // ReadLength is unchanged
1917 //////// We encountered a 'delimiter' element i.e. a tag of the form
1918 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1919 // taken into account.
1920 else if( gr == 0xfffe )
1922 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1923 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1924 // causes extra troubles...
1925 if( entry->GetElement() != 0x0000 )
1931 entry->SetLength(foundLength);
1935 * \brief Apply some heuristics to predict whether the considered
1936 * element value contains/represents an integer or not.
1937 * @param entry The element value on which to apply the predicate.
1938 * @return The result of the heuristical predicate.
1940 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1942 uint16_t elem = entry->GetElement();
1943 uint16_t group = entry->GetGroup();
1944 const std::string &vr = entry->GetVR();
1945 uint32_t length = entry->GetLength();
1947 // When we have some semantics on the element we just read, and if we
1948 // a priori know we are dealing with an integer, then we shall be
1949 // able to swap it's element value properly.
1950 if ( elem == 0 ) // This is the group length of the group
1958 // Allthough this should never happen, still some images have a
1959 // corrupted group length [e.g. have a glance at offset x(8336) of
1960 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
1961 // Since for dicom compliant and well behaved headers, the present
1962 // test is useless (and might even look a bit paranoid), when we
1963 // encounter such an ill-formed image, we simply display a warning
1964 // message and proceed on parsing (while crossing fingers).
1965 long filePosition = Fp->tellg();
1966 gdcmVerboseMacro( "Erroneous Group Length element length on : ("
1967 << std::hex << group << " , " << elem
1968 << ") -before- position x(" << filePosition << ")"
1969 << "lgt : " << length );
1973 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1981 * \brief Find the Length till the next sequence delimiter
1982 * \warning NOT end user intended method !
1986 uint32_t Document::FindDocEntryLengthOBOrOW()
1987 throw( FormatUnexpected )
1989 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1990 long positionOnEntry = Fp->tellg();
1991 bool foundSequenceDelimiter = false;
1992 uint32_t totalLength = 0;
1994 while ( !foundSequenceDelimiter )
2000 group = ReadInt16();
2003 catch ( FormatError )
2005 throw FormatError("Unexpected end of file encountered during ",
2006 "Document::FindDocEntryLengthOBOrOW()");
2009 // We have to decount the group and element we just read
2012 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2014 long filePosition = Fp->tellg();
2015 gdcmVerboseMacro( "Neither an Item tag nor a Sequence delimiter tag on :"
2016 << std::hex << group << " , " << elem
2017 << ") -before- position x(" << filePosition << ")" );
2019 Fp->seekg(positionOnEntry, std::ios::beg);
2020 throw FormatUnexpected( "Neither an Item tag nor a Sequence delimiter tag.");
2023 if ( elem == 0xe0dd )
2025 foundSequenceDelimiter = true;
2028 uint32_t itemLength = ReadInt32();
2029 // We add 4 bytes since we just read the ItemLength with ReadInt32
2030 totalLength += itemLength + 4;
2031 SkipBytes(itemLength);
2033 if ( foundSequenceDelimiter )
2038 Fp->seekg( positionOnEntry, std::ios::beg);
2043 * \brief Reads a supposed to be 16 Bits integer
2044 * (swaps it depending on processor endianity)
2045 * @return read value
2047 uint16_t Document::ReadInt16()
2048 throw( FormatError )
2051 Fp->read ((char*)&g, (size_t)2);
2054 throw FormatError( "Document::ReadInt16()", " file error." );
2058 throw FormatError( "Document::ReadInt16()", "EOF." );
2065 * \brief Reads a supposed to be 32 Bits integer
2066 * (swaps it depending on processor endianity)
2067 * @return read value
2069 uint32_t Document::ReadInt32()
2070 throw( FormatError )
2073 Fp->read ((char*)&g, (size_t)4);
2076 throw FormatError( "Document::ReadInt32()", " file error." );
2080 throw FormatError( "Document::ReadInt32()", "EOF." );
2087 * \brief skips bytes inside the source file
2088 * \warning NOT end user intended method !
2091 void Document::SkipBytes(uint32_t nBytes)
2093 //FIXME don't dump the returned value
2094 Fp->seekg((long)nBytes, std::ios::cur);
2098 * \brief Loads all the needed Dictionaries
2099 * \warning NOT end user intended method !
2101 void Document::Initialise()
2103 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2105 RLEInfo = new RLEFramesInfo;
2106 JPEGInfo = new JPEGFragmentsInfo;
2111 * \brief Discover what the swap code is (among little endian, big endian,
2112 * bad little endian, bad big endian).
2114 * @return false when we are absolutely sure
2115 * it's neither ACR-NEMA nor DICOM
2116 * true when we hope ours assuptions are OK
2118 bool Document::CheckSwap()
2120 // The only guaranted way of finding the swap code is to find a
2121 // group tag since we know it's length has to be of four bytes i.e.
2122 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2123 // occurs when we can't find such group...
2125 uint32_t x = 4; // x : for ntohs
2126 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2132 // First, compare HostByteOrder and NetworkByteOrder in order to
2133 // determine if we shall need to swap bytes (i.e. the Endian type).
2134 if ( x == ntohs(x) )
2143 // The easiest case is the one of a 'true' DICOM header, we just have
2144 // to look for the string "DICM" inside the file preamble.
2147 char *entCur = deb + 128;
2148 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2150 gdcmVerboseMacro( "Looks like DICOM Version3 (preamble + DCM)" );
2152 // Group 0002 should always be VR, and the first element 0000
2153 // Let's be carefull (so many wrong headers ...)
2154 // and determine the value representation (VR) :
2155 // Let's skip to the first element (0002,0000) and check there if we find
2156 // "UL" - or "OB" if the 1st one is (0002,0001) -,
2157 // in which case we (almost) know it is explicit VR.
2158 // WARNING: if it happens to be implicit VR then what we will read
2159 // is the length of the group. If this ascii representation of this
2160 // length happens to be "UL" then we shall believe it is explicit VR.
2161 // We need to skip :
2162 // * the 128 bytes of File Preamble (often padded with zeroes),
2163 // * the 4 bytes of "DICM" string,
2164 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2165 // i.e. a total of 136 bytes.
2168 // group 0x0002 *is always* Explicit VR Sometimes ,
2169 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
2170 // *Implicit* VR (see former 'gdcmData/icone.dcm')
2172 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2173 memcmp(entCur, "OB", (size_t)2) == 0 ||
2174 memcmp(entCur, "UI", (size_t)2) == 0 ||
2175 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2176 // when Write DCM *adds*
2178 // Use Document::dicom_vr to test all the possibilities
2179 // instead of just checking for UL, OB and UI !? group 0000
2181 Filetype = ExplicitVR;
2182 gdcmVerboseMacro( "Group 0002 : Explicit Value Representation");
2186 Filetype = ImplicitVR;
2187 gdcmVerboseMacro( "Group 0002 :Not an explicit Value Representation;"
2188 << "Looks like a bugged Header!");
2194 gdcmVerboseMacro( "HostByteOrder != NetworkByteOrder");
2199 gdcmVerboseMacro( "HostByteOrder = NetworkByteOrder");
2202 // Position the file position indicator at first tag
2203 // (i.e. after the file preamble and the "DICM" string).
2204 Fp->seekg(0, std::ios::beg);
2205 Fp->seekg ( 132L, std::ios::beg);
2209 // Alas, this is not a DicomV3 file and whatever happens there is no file
2210 // preamble. We can reset the file position indicator to where the data
2211 // is (i.e. the beginning of the file).
2212 gdcmVerboseMacro( "Not a DICOM Version3 file");
2213 Fp->seekg(0, std::ios::beg);
2215 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2216 // By clean we mean that the length of the first tag is written down.
2217 // If this is the case and since the length of the first group HAS to be
2218 // four (bytes), then determining the proper swap code is straightforward.
2221 // We assume the array of char we are considering contains the binary
2222 // representation of a 32 bits integer. Hence the following dirty
2224 s32 = *((uint32_t *)(entCur));
2245 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2246 // It is time for despaired wild guesses.
2247 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2248 // i.e. the 'group length' element is not present :
2250 // check the supposed-to-be 'group number'
2251 // in ( 0x0001 .. 0x0008 )
2252 // to determine ' SwapCode' value .
2253 // Only 0 or 4321 will be possible
2254 // (no oportunity to check for the formerly well known
2255 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2256 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2257 // the file IS NOT ACR-NEMA nor DICOM V3
2258 // Find a trick to tell it the caller...
2260 s16 = *((uint16_t *)(deb));
2287 gdcmVerboseMacro( "ACR/NEMA unfound swap info (Really hopeless !)");
2297 * \brief Change the Byte Swap code.
2299 void Document::SwitchByteSwapCode()
2301 gdcmVerboseMacro( "Switching Byte Swap code from "<< SwapCode);
2302 if ( SwapCode == 1234 )
2306 else if ( SwapCode == 4321 )
2310 else if ( SwapCode == 3412 )
2314 else if ( SwapCode == 2143 )
2321 * \brief during parsing, Header Elements too long are not loaded in memory
2324 void Document::SetMaxSizeLoadEntry(long newSize)
2330 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2332 MaxSizeLoadEntry = 0xffffffff;
2335 MaxSizeLoadEntry = newSize;
2340 * \brief Header Elements too long will not be printed
2341 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2344 void Document::SetMaxSizePrintEntry(long newSize)
2346 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
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
2392 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
2394 std::string Document::GetTransferSyntaxName()
2396 // use the TS (TS : Transfer Syntax)
2397 std::string transferSyntax = GetEntry(0x0002,0x0010);
2399 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
2401 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
2402 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
2403 return "Uncompressed ACR-NEMA";
2405 if ( transferSyntax == GDCM_UNFOUND )
2407 gdcmVerboseMacro( "Unfound Transfer Syntax (0002,0010)");
2408 return "Uncompressed ACR-NEMA";
2411 // we do it only when we need it
2412 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
2414 // Global::GetTS() is a global static you shall never try to delete it!
2419 * \brief Group 0002 is always coded Little Endian
2420 * whatever Transfer Syntax is
2423 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2425 // Endian reversion. Some files contain groups of tags with reversed endianess.
2426 if ( !Group0002Parsed && group != 0x0002)
2428 Group0002Parsed = true;
2429 // we just came out of group 0002
2430 // if Transfer syntax is Big Endian we have to change CheckSwap
2432 std::string ts = GetTransferSyntax();
2433 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2435 gdcmVerboseMacro("True DICOM File, with NO Tansfer Syntax: " << ts );
2439 // FIXME Strangely, this works with
2440 //'Implicit VR Transfer Syntax (GE Private)
2441 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ExplicitVRBigEndian )
2443 gdcmVerboseMacro("Transfer Syntax Name = ["
2444 << GetTransferSyntaxName() << "]" );
2445 SwitchByteSwapCode();
2446 group = SwapShort(group);
2447 elem = SwapShort(elem);
2453 * \brief Read the next tag but WITHOUT loading it's value
2454 * (read the 'Group Number', the 'Element Number',
2455 * gets the Dict Entry
2456 * gets the VR, gets the length, gets the offset value)
2457 * @return On succes the newly created DocEntry, NULL on failure.
2459 DocEntry *Document::ReadNextDocEntry()
2466 group = ReadInt16();
2469 catch ( FormatError e )
2471 // We reached the EOF (or an error occured) therefore
2472 // header parsing has to be considered as finished.
2477 // Sometimes file contains groups of tags with reversed endianess.
2478 HandleBrokenEndian(group, elem);
2480 // In 'true DICOM' files Group 0002 is always little endian
2481 if ( HasDCMPreamble )
2482 HandleOutOfGroup0002(group, elem);
2484 std::string vr = FindDocEntryVR();
2485 std::string realVR = vr;
2487 if( vr == GDCM_UNKNOWN)
2489 DictEntry *dictEntry = GetDictEntry(group,elem);
2491 realVR = dictEntry->GetVR();
2495 if( Global::GetVR()->IsVROfSequence(realVR) )
2496 newEntry = NewSeqEntry(group, elem);
2497 else if( Global::GetVR()->IsVROfStringRepresentable(realVR) )
2498 newEntry = NewValEntry(group, elem,vr);
2500 newEntry = NewBinEntry(group, elem,vr);
2502 if( vr == GDCM_UNKNOWN )
2504 if( Filetype == ExplicitVR )
2506 // We thought this was explicit VR, but we end up with an
2507 // implicit VR tag. Let's backtrack.
2508 if ( newEntry->GetGroup() != 0xfffe )
2511 msg = Util::Format("Entry (%04x,%04x) should be Explicit VR\n",
2512 newEntry->GetGroup(), newEntry->GetElement());
2513 gdcmVerboseMacro( msg.c_str() );
2516 newEntry->SetImplicitVR();
2521 FindDocEntryLength(newEntry);
2523 catch ( FormatError e )
2531 newEntry->SetOffset(Fp->tellg());
2538 * \brief Generate a free TagKey i.e. a TagKey that is not present
2539 * in the TagHt dictionary.
2540 * @param group The generated tag must belong to this group.
2541 * @return The element of tag with given group which is fee.
2543 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2545 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2547 TagKey key = DictEntry::TranslateToKey(group, elem);
2548 if (TagHT.count(key) == 0)
2557 * \brief Assuming the internal file pointer \ref Document::Fp
2558 * is placed at the beginning of a tag check whether this
2559 * tag is (TestGroup, TestElement).
2560 * \warning On success the internal file pointer \ref Document::Fp
2561 * is modified to point after the tag.
2562 * On failure (i.e. when the tag wasn't the expected tag
2563 * (TestGroup, TestElement) the internal file pointer
2564 * \ref Document::Fp is restored to it's original position.
2565 * @param testGroup The expected group of the tag.
2566 * @param testElement The expected Element of the tag.
2567 * @return True on success, false otherwise.
2569 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2571 long positionOnEntry = Fp->tellg();
2572 long currentPosition = Fp->tellg(); // On debugging purposes
2574 //// Read the Item Tag group and element, and make
2575 // sure they are what we expected:
2576 uint16_t itemTagGroup;
2577 uint16_t itemTagElement;
2580 itemTagGroup = ReadInt16();
2581 itemTagElement = ReadInt16();
2583 catch ( FormatError e )
2585 //std::cerr << e << std::endl;
2588 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2590 gdcmVerboseMacro( "Wrong Item Tag found:"
2591 << " We should have found tag ("
2592 << std::hex << testGroup << "," << testElement << ")" << std::endl
2593 << " but instead we encountered tag ("
2594 << std::hex << itemTagGroup << "," << itemTagElement << ")"
2595 << " at address: " << " 0x(" << (unsigned int)currentPosition << ")"
2597 Fp->seekg(positionOnEntry, std::ios::beg);
2605 * \brief Assuming the internal file pointer \ref Document::Fp
2606 * is placed at the beginning of a tag (TestGroup, TestElement),
2607 * read the length associated to the Tag.
2608 * \warning On success the internal file pointer \ref Document::Fp
2609 * is modified to point after the tag and it's length.
2610 * On failure (i.e. when the tag wasn't the expected tag
2611 * (TestGroup, TestElement) the internal file pointer
2612 * \ref Document::Fp is restored to it's original position.
2613 * @param testGroup The expected group of the tag.
2614 * @param testElement The expected Element of the tag.
2615 * @return On success returns the length associated to the tag. On failure
2618 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2621 if ( !ReadTag(testGroup, testElement) )
2626 //// Then read the associated Item Length
2627 long currentPosition = Fp->tellg();
2628 uint32_t itemLength = ReadInt32();
2630 gdcmVerboseMacro( "Basic Item Length is: "
2631 << itemLength << std::endl
2632 << " at address: " << std::hex << (unsigned int)currentPosition);
2638 * \brief When parsing the Pixel Data of an encapsulated file, read
2639 * the basic offset table (when present, and BTW dump it).
2641 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2643 //// Read the Basic Offset Table Item Tag length...
2644 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2646 // When present, read the basic offset table itself.
2647 // Notes: - since the presence of this basic offset table is optional
2648 // we can't rely on it for the implementation, and we will simply
2649 // trash it's content (when present).
2650 // - still, when present, we could add some further checks on the
2651 // lengths, but we won't bother with such fuses for the time being.
2652 if ( itemLength != 0 )
2654 char *basicOffsetTableItemValue = new char[itemLength + 1];
2655 Fp->read(basicOffsetTableItemValue, itemLength);
2658 for (unsigned int i=0; i < itemLength; i += 4 )
2660 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2662 gdcmVerboseMacro( "Read one length: " <<
2663 std::hex << individualLength );
2667 delete[] basicOffsetTableItemValue;
2672 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2673 * Compute the RLE extra information and store it in \ref RLEInfo
2674 * for later pixel retrieval usage.
2676 void Document::ComputeRLEInfo()
2678 std::string ts = GetTransferSyntax();
2679 if ( !Global::GetTS()->IsRLELossless(ts) )
2684 // Encoded pixel data: for the time being we are only concerned with
2685 // Jpeg or RLE Pixel data encodings.
2686 // As stated in PS 3.5-2003, section 8.2 p44:
2687 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2688 // value representation OB is used".
2689 // Hence we expect an OB value representation. Concerning OB VR,
2690 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2691 // "For the Value Representations OB and OW, the encoding shall meet the
2692 // following specifications depending on the Data element tag:"
2694 // - the first item in the sequence of items before the encoded pixel
2695 // data stream shall be basic offset table item. The basic offset table
2696 // item value, however, is not required to be present"
2698 ReadAndSkipEncapsulatedBasicOffsetTable();
2700 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2701 // Loop on the individual frame[s] and store the information
2702 // on the RLE fragments in a RLEFramesInfo.
2703 // Note: - when only a single frame is present, this is a
2705 // - when more than one frame are present, then we are in
2706 // the case of a multi-frame image.
2708 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2710 // Parse the RLE Header and store the corresponding RLE Segment
2711 // Offset Table information on fragments of this current Frame.
2712 // Note that the fragment pixels themselves are not loaded
2713 // (but just skipped).
2714 long frameOffset = Fp->tellg();
2716 uint32_t nbRleSegments = ReadInt32();
2717 if ( nbRleSegments > 16 )
2719 // There should be at most 15 segments (refer to RLEFrame class)
2720 gdcmVerboseMacro( "Too many segments.");
2723 uint32_t rleSegmentOffsetTable[16];
2724 for( int k = 1; k <= 15; k++ )
2726 rleSegmentOffsetTable[k] = ReadInt32();
2729 // Deduce from both the RLE Header and the frameLength the
2730 // fragment length, and again store this info in a
2732 long rleSegmentLength[15];
2733 // skipping (not reading) RLE Segments
2734 if ( nbRleSegments > 1)
2736 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2738 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2739 - rleSegmentOffsetTable[k];
2740 SkipBytes(rleSegmentLength[k]);
2744 rleSegmentLength[nbRleSegments] = frameLength
2745 - rleSegmentOffsetTable[nbRleSegments];
2746 SkipBytes(rleSegmentLength[nbRleSegments]);
2748 // Store the collected info
2749 RLEFrame *newFrameInfo = new RLEFrame;
2750 newFrameInfo->NumberFragments = nbRleSegments;
2751 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2753 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2754 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2756 RLEInfo->Frames.push_back( newFrameInfo );
2759 // Make sure that at the end of the item we encounter a 'Sequence
2761 if ( !ReadTag(0xfffe, 0xe0dd) )
2763 gdcmVerboseMacro( "No sequence delimiter item at end of RLE item sequence");
2768 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2769 * Compute the jpeg extra information (fragment[s] offset[s] and
2770 * length) and store it[them] in \ref JPEGInfo for later pixel
2773 void Document::ComputeJPEGFragmentInfo()
2775 // If you need to, look for comments of ComputeRLEInfo().
2776 std::string ts = GetTransferSyntax();
2777 if ( ! Global::GetTS()->IsJPEG(ts) )
2782 ReadAndSkipEncapsulatedBasicOffsetTable();
2784 // Loop on the fragments[s] and store the parsed information in a
2786 long fragmentLength;
2787 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2789 long fragmentOffset = Fp->tellg();
2791 // Store the collected info
2792 JPEGFragment *newFragment = new JPEGFragment;
2793 newFragment->Offset = fragmentOffset;
2794 newFragment->Length = fragmentLength;
2795 JPEGInfo->Fragments.push_back( newFragment );
2797 SkipBytes( fragmentLength );
2800 // Make sure that at the end of the item we encounter a 'Sequence
2802 if ( !ReadTag(0xfffe, 0xe0dd) )
2804 gdcmVerboseMacro( "No sequence delimiter item at end of JPEG item sequence");
2809 * \brief Walk recursively the given \ref DocEntrySet, and feed
2810 * the given hash table (\ref TagDocEntryHT) with all the
2811 * \ref DocEntry (Dicom entries) encountered.
2812 * This method does the job for \ref BuildFlatHashTable.
2813 * @param builtHT Where to collect all the \ref DocEntry encountered
2814 * when recursively walking the given set.
2815 * @param set The structure to be traversed (recursively).
2817 /*void Document::BuildFlatHashTableRecurse( TagDocEntryHT &builtHT,
2820 if (ElementSet *elementSet = dynamic_cast< ElementSet* > ( set ) )
2822 TagDocEntryHT const ¤tHT = elementSet->GetTagHT();
2823 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2824 i != currentHT.end();
2827 DocEntry *entry = i->second;
2828 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2830 const ListSQItem& items = seqEntry->GetSQItems();
2831 for( ListSQItem::const_iterator item = items.begin();
2832 item != items.end();
2835 BuildFlatHashTableRecurse( builtHT, *item );
2839 builtHT[entry->GetKey()] = entry;
2844 if (SQItem *SQItemSet = dynamic_cast< SQItem* > ( set ) )
2846 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
2847 for (ListDocEntry::const_iterator i = currentList.begin();
2848 i != currentList.end();
2851 DocEntry *entry = *i;
2852 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2854 const ListSQItem& items = seqEntry->GetSQItems();
2855 for( ListSQItem::const_iterator item = items.begin();
2856 item != items.end();
2859 BuildFlatHashTableRecurse( builtHT, *item );
2863 builtHT[entry->GetKey()] = entry;
2870 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2873 * The structure used by a Document (through \ref ElementSet),
2874 * in order to hold the parsed entries of a Dicom header, is a recursive
2875 * one. This is due to the fact that the sequences (when present)
2876 * can be nested. Additionaly, the sequence items (represented in
2877 * gdcm as \ref SQItem) add an extra complexity to the data
2878 * structure. Hence, a gdcm user whishing to visit all the entries of
2879 * a Dicom header will need to dig in the gdcm internals (which
2880 * implies exposing all the internal data structures to the API).
2881 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2882 * recursively builds a temporary hash table, which holds all the
2883 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
2885 * \warning Of course there is NO integrity constrain between the
2886 * returned \ref TagDocEntryHT and the \ref ElementSet used
2887 * to build it. Hence if the underlying \ref ElementSet is
2888 * altered, then it is the caller responsability to invoke
2889 * \ref BuildFlatHashTable again...
2890 * @return The flat std::map<> we juste build.
2892 /*TagDocEntryHT *Document::BuildFlatHashTable()
2894 TagDocEntryHT *FlatHT = new TagDocEntryHT;
2895 BuildFlatHashTableRecurse( *FlatHT, this );
2902 * \brief Compares two documents, according to \ref DicomDir rules
2903 * \warning Does NOT work with ACR-NEMA files
2904 * \todo Find a trick to solve the pb (use RET fields ?)
2906 * @return true if 'smaller'
2908 bool Document::operator<(Document &document)
2911 std::string s1 = GetEntry(0x0010,0x0010);
2912 std::string s2 = document.GetEntry(0x0010,0x0010);
2924 s1 = GetEntry(0x0010,0x0020);
2925 s2 = document.GetEntry(0x0010,0x0020);
2936 // Study Instance UID
2937 s1 = GetEntry(0x0020,0x000d);
2938 s2 = document.GetEntry(0x0020,0x000d);
2949 // Serie Instance UID
2950 s1 = GetEntry(0x0020,0x000e);
2951 s2 = document.GetEntry(0x0020,0x000e);
2968 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
2969 * @param filetype Type of the File to be written
2971 int Document::ComputeGroup0002Length( FileType filetype )
2976 int groupLength = 0;
2977 bool found0002 = false;
2979 // for each zero-level Tag in the DCM Header
2983 entry = GetNextEntry();
2986 gr = entry->GetGroup();
2992 el = entry->GetElement();
2993 vr = entry->GetVR();
2995 if (filetype == ExplicitVR)
2997 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
2999 groupLength += 4; // explicit VR AND OB, OW, SQ : 4 more bytes
3002 groupLength += 2 + 2 + 4 + entry->GetLength();
3004 else if (found0002 )
3007 entry = GetNextEntry();
3012 } // end namespace gdcm
3014 //-----------------------------------------------------------------------------