1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/10 03:09:07 $
7 Version: $Revision: 1.169 $
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>
46 //-----------------------------------------------------------------------------
47 static const char *TransferSyntaxStrings[] = {
48 // Implicit VR Little Endian
50 // Implicit VR Big Endian DLX G.E?
52 // Explicit VR Little Endian
53 "1.2.840.10008.1.2.1",
54 // Deflated Explicit VR Little Endian
55 "1.2.840.10008.1.2.1.99",
56 // Explicit VR Big Endian
57 "1.2.840.10008.1.2.2",
58 // JPEG Baseline (Process 1)
59 "1.2.840.10008.1.2.4.50",
60 // JPEG Extended (Process 2 & 4)
61 "1.2.840.10008.1.2.4.51",
62 // JPEG Extended (Process 3 & 5)
63 "1.2.840.10008.1.2.4.52",
64 // JPEG Spectral Selection, Non-Hierarchical (Process 6 & 8)
65 "1.2.840.10008.1.2.4.53",
66 // JPEG Full Progression, Non-Hierarchical (Process 10 & 12)
67 "1.2.840.10008.1.2.4.55",
68 // JPEG Lossless, Non-Hierarchical (Process 14)
69 "1.2.840.10008.1.2.4.57",
70 // JPEG Lossless, Hierarchical, First-Order Prediction (Process 14, [Selection Value 1])
71 "1.2.840.10008.1.2.4.70",
73 "1.2.840.10008.1.2.4.90",
75 "1.2.840.10008.1.2.4.91",
77 "1.2.840.10008.1.2.5",
79 "Unknown Transfer Syntax"
82 //-----------------------------------------------------------------------------
83 // Refer to Document::CheckSwap()
84 //const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
86 // Refer to Document::SetMaxSizeLoadEntry()
87 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
88 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
90 //-----------------------------------------------------------------------------
91 // Constructor / Destructor
95 * @param filename file to be opened for parsing
97 Document::Document( std::string const &filename ) : ElementSet(-1)
99 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
109 gdcmVerboseMacro( "Starting parsing of file: " << Filename.c_str());
110 Fp->seekg( 0, std::ios::beg);
112 Fp->seekg(0, std::ios::end);
113 long lgt = Fp->tellg();
115 Fp->seekg( 0, std::ios::beg);
117 long beg = Fp->tellg();
120 ParseDES( this, beg, lgt, false); // le Load sera fait a la volee
122 Fp->seekg( 0, std::ios::beg);
124 // Load 'non string' values
126 std::string PhotometricInterpretation = GetEntry(0x0028,0x0004);
127 if( PhotometricInterpretation == "PALETTE COLOR " )
129 LoadEntryBinArea(0x0028,0x1200); // gray LUT
130 /// FIXME FIXME FIXME
131 /// The tags refered by the three following lines used to be CORRECTLY
132 /// defined as having an US Value Representation in the public
133 /// dictionnary. BUT the semantics implied by the three following
134 /// lines state that the corresponding tag contents are in fact
135 /// the ones of a BinEntry.
136 /// In order to fix things "Quick and Dirty" the dictionnary was
137 /// altered on PURPOUS but now contains a WRONG value.
138 /// In order to fix things and restore the dictionary to its
139 /// correct value, one needs to decided of the semantics by deciding
140 /// wether the following tags are either:
141 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
142 /// also used as BinEntry, which requires the proper conversion,
143 /// - OW, and hence loaded as BinEntry, but afterwards also used
144 /// as ValEntry, which requires the proper conversion.
145 LoadEntryBinArea(0x0028,0x1201); // R LUT
146 LoadEntryBinArea(0x0028,0x1202); // G LUT
147 LoadEntryBinArea(0x0028,0x1203); // B LUT
149 // Segmented Red Palette Color LUT Data
150 LoadEntryBinArea(0x0028,0x1221);
151 // Segmented Green Palette Color LUT Data
152 LoadEntryBinArea(0x0028,0x1222);
153 // Segmented Blue Palette Color LUT Data
154 LoadEntryBinArea(0x0028,0x1223);
156 //FIXME later : how to use it?
157 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
161 // --------------------------------------------------------------
162 // Specific code to allow gdcm to read ACR-LibIDO formated images
163 // Note: ACR-LibIDO is an extension of the ACR standard that was
164 // used at CREATIS. For the time being (say a couple years)
165 // we keep this kludge to allow a smooth move to gdcm for
166 // CREATIS developpers (sorry folks).
168 // if recognition code tells us we deal with a LibIDO image
169 // we switch lineNumber and columnNumber
172 RecCode = GetEntry(0x0008, 0x0010); // recognition code
173 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
174 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
175 // with "little-endian strings"
177 Filetype = ACR_LIBIDO;
178 std::string rows = GetEntry(0x0028, 0x0010);
179 std::string columns = GetEntry(0x0028, 0x0011);
180 SetEntry(columns, 0x0028, 0x0010);
181 SetEntry(rows , 0x0028, 0x0011);
183 // ----------------- End of ACR-LibIDO kludge ------------------
187 * \brief This default constructor doesn't parse the file. You should
188 * then invoke \ref Document::SetFileName and then the parsing.
190 Document::Document() : ElementSet(-1)
194 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
197 Filetype = ExplicitVR;
201 * \brief Canonical destructor.
203 Document::~Document ()
212 //-----------------------------------------------------------------------------
216 * \brief Prints The Dict Entries of THE public Dicom Dictionary
219 void Document::PrintPubDict(std::ostream &os)
221 RefPubDict->SetPrintLevel(PrintLevel);
222 RefPubDict->Print(os);
226 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
229 void Document::PrintShaDict(std::ostream &os)
231 RefShaDict->SetPrintLevel(PrintLevel);
232 RefShaDict->Print(os);
235 //-----------------------------------------------------------------------------
238 * \brief Get the public dictionary used
240 Dict *Document::GetPubDict()
246 * \brief Get the shadow dictionary used
248 Dict *Document::GetShaDict()
254 * \brief Set the shadow dictionary used
255 * @param dict dictionary to use in shadow
257 bool Document::SetShaDict(Dict *dict)
264 * \brief Set the shadow dictionary used
265 * @param dictName name of the dictionary to use in shadow
267 bool Document::SetShaDict(DictKey const &dictName)
269 RefShaDict = Global::GetDicts()->GetDict(dictName);
274 * \brief This predicate, based on hopefully reasonable heuristics,
275 * decides whether or not the current Document was properly parsed
276 * and contains the mandatory information for being considered as
277 * a well formed and usable Dicom/Acr File.
278 * @return true when Document is the one of a reasonable Dicom/Acr file,
281 bool Document::IsReadable()
283 if( Filetype == Unknown)
285 gdcmVerboseMacro( "Wrong filetype");
291 gdcmVerboseMacro( "No tags in internal hash table.");
299 * \brief Accessor to the Transfer Syntax (when present) of the
300 * current document (it internally handles reading the
301 * value from disk when only parsing occured).
302 * @return The encountered Transfer Syntax of the current document.
304 TransferSyntaxType Document::GetTransferSyntax()
306 DocEntry *entry = GetDocEntry(0x0002, 0x0010);
312 // The entry might be present but not loaded (parsing and loading
313 // happen at different stages): try loading and proceed with check...
314 LoadDocEntrySafe(entry);
315 if (ValEntry *valEntry = dynamic_cast< ValEntry* >(entry) )
317 std::string transfer = valEntry->GetValue();
318 // The actual transfer (as read from disk) might be padded. We
319 // first need to remove the potential padding. We can make the
320 // weak assumption that padding was not executed with digits...
321 if ( transfer.length() == 0 )
323 // for brain damaged headers
326 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
328 transfer.erase(transfer.length()-1, 1);
330 for (int i = 0; TransferSyntaxStrings[i] != NULL; i++)
332 if ( TransferSyntaxStrings[i] == transfer )
334 return TransferSyntaxType(i);
341 bool Document::IsJPEGLossless()
343 TransferSyntaxType r = GetTransferSyntax();
344 return r == JPEGFullProgressionProcess10_12
345 || r == JPEGLosslessProcess14
346 || r == JPEGLosslessProcess14_1;
350 * \brief Determines if the Transfer Syntax was already encountered
351 * and if it corresponds to a JPEG2000 one
352 * @return True when JPEG2000 (Lossly or LossLess) found. False in all
355 bool Document::IsJPEG2000()
357 TransferSyntaxType r = GetTransferSyntax();
358 return r == JPEG2000Lossless || r == JPEG2000;
362 * \brief Determines if the Transfer Syntax corresponds to any form
363 * of Jpeg encoded Pixel data.
364 * @return True when any form of JPEG found. False otherwise.
366 bool Document::IsJPEG()
368 TransferSyntaxType r = GetTransferSyntax();
369 return r == JPEGBaselineProcess1
370 || r == JPEGExtendedProcess2_4
371 || r == JPEGExtendedProcess3_5
372 || r == JPEGSpectralSelectionProcess6_8
378 * \brief Determines if the Transfer Syntax corresponds to encapsulated
379 * of encoded Pixel Data (as opposed to native).
380 * @return True when encapsulated. False when native.
382 bool Document::IsEncapsulate()
384 TransferSyntaxType r = GetTransferSyntax();
385 return IsJPEG() || r == RLELossless;
389 * \brief Predicate for dicom version 3 file.
390 * @return True when the file is a dicom version 3.
392 bool Document::IsDicomV3()
394 // Checking if Transfert Syntax exists is enough
395 // Anyway, it's to late check if the 'Preamble' was found ...
396 // And ... would it be a rich idea to check ?
397 // (some 'no Preamble' DICOM images exist !)
398 return GetDocEntry(0x0002, 0x0010) != NULL;
402 * \brief returns the File Type
403 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
404 * @return the FileType code
406 FileType Document::GetFileType()
412 * \brief Tries to open the file \ref Document::Filename and
413 * checks the preamble when existing.
414 * @return The FILE pointer on success.
416 std::ifstream *Document::OpenFile()
418 if (Filename.length() == 0)
425 gdcmVerboseMacro( "Is already opened when opening: " << Filename.c_str());
428 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
431 gdcmVerboseMacro( "Cannot open file: " << Filename.c_str());
438 Fp->read((char*)&zero, (size_t)2);
445 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
447 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
448 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
449 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
450 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
453 = Util::Format("ACR/DICOM with no preamble: (%04x)\n", zero);
454 gdcmVerboseMacro( msg.c_str() );
459 Fp->seekg(126L, std::ios::cur);
461 Fp->read(dicm, (size_t)4);
467 if( memcmp(dicm, "DICM", 4) == 0 )
473 gdcmVerboseMacro( "Not DICOM/ACR (missing preamble)" << Filename.c_str());
479 * \brief closes the file
480 * @return TRUE if the close was successfull
482 bool Document::CloseFile()
491 return true; //FIXME how do we detect a non-close ifstream ?
495 * \brief Writes in a file all the Header Entries (Dicom Elements)
496 * @param fp file pointer on an already open file
497 * @param filetype Type of the File to be written
498 * (ACR-NEMA, ExplicitVR, ImplicitVR)
499 * \return Always true.
501 void Document::WriteContent(std::ofstream *fp, FileType filetype)
503 /// \todo move the following lines (and a lot of others, to be written)
504 /// to a future function CheckAndCorrectHeader
505 /// (necessary if user wants to write a DICOM V3 file
506 /// starting from an ACR-NEMA (V2) Header
508 if ( filetype == ImplicitVR || filetype == ExplicitVR )
510 // writing Dicom File Preamble
511 char filePreamble[128];
512 memset(filePreamble, 0, 128);
513 fp->write(filePreamble, 128);
514 fp->write("DICM", 4);
518 * \todo rewrite later, if really usefull
519 * - 'Group Length' element is optional in DICOM
520 * - but un-updated odd groups lengthes can causes pb
523 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
524 * UpdateGroupLength(false,filetype);
525 * if ( filetype == ACR)
526 * UpdateGroupLength(true,ACR);
529 ElementSet::WriteContent(fp, filetype); // This one is recursive
533 * \brief Modifies the value of a given Doc Entry (Dicom Element)
534 * when it exists. Create it with the given value when unexistant.
535 * @param value (string) Value to be set
536 * @param group Group number of the Entry
537 * @param elem Element number of the Entry
538 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
539 * \return pointer to the modified/created Header Entry (NULL when creation
542 ValEntry *Document::ReplaceOrCreate(std::string const &value,
547 ValEntry *valEntry = 0;
548 DocEntry *currentEntry = GetDocEntry( group, elem);
552 valEntry = dynamic_cast< ValEntry* >(currentEntry);
556 if( valEntry->GetVR()!=vr )
559 // if currentEntry doesn't correspond to the requested valEntry
562 if (!RemoveEntry(currentEntry))
564 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
571 // Create a new valEntry if necessary
574 valEntry = NewValEntry(group, elem, vr);
576 if ( !AddEntry(valEntry))
578 gdcmVerboseMacro("AddEntry failed although this is a creation.");
585 // Set the binEntry value
586 SetEntry(value, valEntry);
591 * \brief Modifies the value of a given Header Entry (Dicom Element)
592 * when it exists. Create it with the given value when unexistant.
593 * A copy of the binArea is made to be kept in the Document.
594 * @param binArea (binary) value to be set
595 * @param Group Group number of the Entry
596 * @param Elem Element number of the Entry
597 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
598 * \return pointer to the modified/created Header Entry (NULL when creation
601 BinEntry *Document::ReplaceOrCreate(uint8_t *binArea,
607 BinEntry *binEntry = 0;
608 DocEntry *currentEntry = GetDocEntry( group, elem);
610 // Verify the currentEntry
613 binEntry = dynamic_cast< BinEntry* >(currentEntry);
617 if( binEntry->GetVR()!=vr )
620 // if currentEntry doesn't correspond to the requested valEntry
623 if (!RemoveEntry(currentEntry))
625 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
632 // Create a new binEntry if necessary
635 binEntry = NewBinEntry(group, elem, vr);
637 if ( !AddEntry(binEntry))
639 gdcmVerboseMacro( "AddEntry failed allthough this is a creation.");
646 // Set the binEntry value
648 if (lgth>0 && binArea)
650 tmpArea = new uint8_t[lgth];
651 memcpy(tmpArea,binArea,lgth);
657 if (!SetEntry(tmpArea,lgth,binEntry))
669 * \brief Modifies the value of a given Header Entry (Dicom Element)
670 * when it exists. Create it when unexistant.
671 * @param Group Group number of the Entry
672 * @param Elem Element number of the Entry
673 * \return pointer to the modified/created SeqEntry (NULL when creation
676 SeqEntry *Document::ReplaceOrCreate( uint16_t group, uint16_t elem)
678 SeqEntry *seqEntry = 0;
679 DocEntry *currentEntry = GetDocEntry( group, elem);
681 // Verify the currentEntry
684 seqEntry = dynamic_cast< SeqEntry* >(currentEntry);
688 if( seqEntry->GetVR()!="SQ" )
691 // if currentEntry doesn't correspond to the requested valEntry
694 if (!RemoveEntry(currentEntry))
696 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
703 // Create a new seqEntry if necessary
706 seqEntry = NewSeqEntry(group, elem);
708 if ( !AddEntry(seqEntry))
710 gdcmVerboseMacro( "AddEntry failed allthough this is a creation.");
721 * \brief Set a new value if the invoked element exists
722 * Seems to be useless !!!
723 * @param value new element value
724 * @param group group number of the Entry
725 * @param elem element number of the Entry
728 bool Document::ReplaceIfExist(std::string const &value,
729 uint16_t group, uint16_t elem )
731 SetEntry(value, group, elem);
736 std::string Document::GetTransferSyntaxValue(TransferSyntaxType type)
738 return TransferSyntaxStrings[type];
741 //-----------------------------------------------------------------------------
745 * \brief Checks if a given Dicom Element exists within the H table
746 * @param group Group number of the searched Dicom Element
747 * @param element Element number of the searched Dicom Element
748 * @return true is found
750 bool Document::CheckIfEntryExist(uint16_t group, uint16_t element )
752 const std::string &key = DictEntry::TranslateToKey(group, element );
753 return TagHT.count(key) != 0;
758 * \brief Searches within Header Entries (Dicom Elements) parsed with
759 * the public and private dictionaries
760 * for the element value representation of a given tag.
761 * @param group Group number of the searched tag.
762 * @param element Element number of the searched tag.
763 * @return Corresponding element value representation when it exists,
764 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
766 std::string Document::GetEntry(uint16_t group, uint16_t element)
768 TagKey key = DictEntry::TranslateToKey(group, element);
769 if ( !TagHT.count(key))
774 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
778 * \brief Searches within Header Entries (Dicom Elements) parsed with
779 * the public and private dictionaries
780 * for the element value representation of a given tag..
782 * Obtaining the VR (Value Representation) might be needed by caller
783 * to convert the string typed content to caller's native type
784 * (think of C++ vs Python). The VR is actually of a higher level
785 * of semantics than just the native C++ type.
786 * @param group Group number of the searched tag.
787 * @param element Element number of the searched tag.
788 * @return Corresponding element value representation when it exists,
789 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
791 std::string Document::GetEntryVR(uint16_t group, uint16_t element)
793 DocEntry *elem = GetDocEntry(group, element);
798 return elem->GetVR();
802 * \brief Searches within Header Entries (Dicom Elements) parsed with
803 * the public and private dictionaries
804 * for the value length of a given tag..
805 * @param group Group number of the searched tag.
806 * @param element Element number of the searched tag.
807 * @return Corresponding element length; -2 if not found
809 int Document::GetEntryLength(uint16_t group, uint16_t element)
811 DocEntry *elem = GetDocEntry(group, element);
814 return -2; //magic number
816 return elem->GetLength();
820 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
821 * through it's (group, element) and modifies it's content with
823 * @param content new value (string) to substitute with
824 * @param group group number of the Dicom Element to modify
825 * @param element element number of the Dicom Element to modify
827 bool Document::SetEntry(std::string const& content,
828 uint16_t group, uint16_t element)
830 ValEntry *entry = GetValEntry(group, element);
833 gdcmVerboseMacro( "No corresponding ValEntry (try promotion first).");
836 return SetEntry(content,entry);
840 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
841 * through it's (group, element) and modifies it's content with
843 * @param content new value (void* -> uint8_t*) to substitute with
844 * @param lgth new value length
845 * @param group group number of the Dicom Element to modify
846 * @param element element number of the Dicom Element to modify
848 bool Document::SetEntry(uint8_t*content, int lgth,
849 uint16_t group, uint16_t element)
851 BinEntry *entry = GetBinEntry(group, element);
854 gdcmVerboseMacro( "No corresponding ValEntry (try promotion first).");
858 return SetEntry(content,lgth,entry);
862 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
863 * and modifies it's content with the given value.
864 * @param content new value (string) to substitute with
865 * @param entry Entry to be modified
867 bool Document::SetEntry(std::string const &content,ValEntry *entry)
871 entry->SetValue(content);
878 * \brief Accesses an existing BinEntry (i.e. a Dicom Element)
879 * and modifies it's content with the given value.
880 * @param content new value (void* -> uint8_t*) to substitute with
881 * @param entry Entry to be modified
882 * @param lgth new value length
884 bool Document::SetEntry(uint8_t *content, int lgth, BinEntry *entry)
888 // Hope Binary field length is *never* wrong
889 /*if(lgth%2) // Non even length are padded with a space (020H).
892 //content = content + '\0'; // fing a trick to enlarge a binary field?
895 entry->SetBinArea(content);
896 entry->SetLength(lgth);
897 entry->SetValue(GDCM_BINLOADED);
904 * \brief Gets (from Header) a 'non string' element value
905 * (LoadElementValues has already be executed)
906 * @param group group number of the Entry
907 * @param elem element number of the Entry
908 * @return Pointer to the 'non string' area
910 void *Document::GetEntryBinArea(uint16_t group, uint16_t elem)
912 DocEntry *entry = GetDocEntry(group, elem);
915 gdcmVerboseMacro( "No entry");
918 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(entry) )
920 return binEntry->GetBinArea();
927 * \brief Loads (from disk) the element content
928 * when a string is not suitable
929 * @param group group number of the Entry
930 * @param elem element number of the Entry
932 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
934 // Search the corresponding DocEntry
935 DocEntry *docElement = GetDocEntry(group, elem);
939 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
943 LoadEntryBinArea(binElement);
947 * \brief Loads (from disk) the element content
948 * when a string is not suitable
949 * @param element Entry whose binArea is going to be loaded
951 void Document::LoadEntryBinArea(BinEntry *element)
953 if(element->GetBinArea())
960 size_t o =(size_t)element->GetOffset();
961 Fp->seekg(o, std::ios::beg);
963 size_t l = element->GetLength();
964 uint8_t *a = new uint8_t[l];
967 gdcmVerboseMacro( "Cannot allocate a");
971 /// \todo check the result
972 Fp->read((char*)a, l);
973 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
979 element->SetBinArea(a);
986 * \brief Sets a 'non string' value to a given Dicom Element
987 * @param area area containing the 'non string' value
988 * @param group Group number of the searched Dicom Element
989 * @param element Element number of the searched Dicom Element
992 /*bool Document::SetEntryBinArea(uint8_t *area,
993 uint16_t group, uint16_t element)
995 DocEntry *currentEntry = GetDocEntry(group, element);
1001 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1003 binEntry->SetBinArea( area );
1011 * \brief retrieves a Dicom Element (the first one) using (group, element)
1012 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1013 * if you think it's NOT UNIQUE, check the count number
1014 * and use iterators to retrieve ALL the Dicoms Elements within
1015 * a given couple (group, element)
1016 * @param group Group number of the searched Dicom Element
1017 * @param element Element number of the searched Dicom Element
1020 DocEntry *Document::GetDocEntry(uint16_t group, uint16_t element)
1022 TagKey key = DictEntry::TranslateToKey(group, element);
1023 if ( !TagHT.count(key))
1027 return TagHT.find(key)->second;
1031 * \brief Same as \ref Document::GetDocEntry except it only
1032 * returns a result when the corresponding entry is of type
1034 * @return When present, the corresponding ValEntry.
1036 ValEntry *Document::GetValEntry(uint16_t group, uint16_t element)
1038 DocEntry *currentEntry = GetDocEntry(group, element);
1039 if ( !currentEntry )
1043 if ( ValEntry *entry = dynamic_cast<ValEntry*>(currentEntry) )
1047 gdcmVerboseMacro( "Unfound ValEntry.");
1053 * \brief Same as \ref Document::GetDocEntry except it only
1054 * returns a result when the corresponding entry is of type
1056 * @return When present, the corresponding BinEntry.
1058 BinEntry *Document::GetBinEntry(uint16_t group, uint16_t element)
1060 DocEntry *currentEntry = GetDocEntry(group, element);
1061 if ( !currentEntry )
1065 if ( BinEntry *entry = dynamic_cast<BinEntry*>(currentEntry) )
1069 gdcmVerboseMacro( "Unfound BinEntry.");
1075 * \brief Loads the element while preserving the current
1076 * underlying file position indicator as opposed to
1077 * to LoadDocEntry that modifies it.
1078 * @param entry Header Entry whose value shall be loaded.
1081 void Document::LoadDocEntrySafe(DocEntry *entry)
1085 long PositionOnEntry = Fp->tellg();
1086 LoadDocEntry(entry);
1087 Fp->seekg(PositionOnEntry, std::ios::beg);
1092 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1094 * @return The properly swaped 32 bits integer.
1096 uint32_t Document::SwapLong(uint32_t a)
1103 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1104 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1108 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1112 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1115 //std::cout << "swapCode= " << SwapCode << std::endl;
1116 gdcmErrorMacro( "Unset swap code");
1123 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1125 * @return The properly unswaped 32 bits integer.
1127 uint32_t Document::UnswapLong(uint32_t a)
1133 * \brief Swaps the bytes so they agree with the processor order
1134 * @return The properly swaped 16 bits integer.
1136 uint16_t Document::SwapShort(uint16_t a)
1138 if ( SwapCode == 4321 || SwapCode == 2143 )
1140 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1146 * \brief Unswaps the bytes so they agree with the processor order
1147 * @return The properly unswaped 16 bits integer.
1149 uint16_t Document::UnswapShort(uint16_t a)
1151 return SwapShort(a);
1154 //-----------------------------------------------------------------------------
1158 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1159 * @return length of the parsed set.
1161 void Document::ParseDES(DocEntrySet *set, long offset,
1162 long l_max, bool delim_mode)
1164 DocEntry *newDocEntry = 0;
1165 ValEntry *newValEntry;
1166 BinEntry *newBinEntry;
1167 SeqEntry *newSeqEntry;
1173 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1179 newDocEntry = ReadNextDocEntry( );
1185 vr = newDocEntry->GetVR();
1186 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
1187 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
1188 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1190 if ( newValEntry || newBinEntry )
1194 if ( ! Global::GetVR()->IsVROfBinaryRepresentable(vr) )
1196 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1197 gdcmVerboseMacro( "Neither Valentry, nor BinEntry."
1198 "Probably unknown VR.");
1201 //////////////////// BinEntry or UNKOWN VR:
1202 // When "this" is a Document the Key is simply of the
1203 // form ( group, elem )...
1204 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1207 newBinEntry->SetKey( newBinEntry->GetKey() );
1209 // but when "this" is a SQItem, we are inserting this new
1210 // valEntry in a sequence item, and the kay has the
1211 // generalized form (refer to \ref BaseTagKey):
1212 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1214 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1215 + newBinEntry->GetKey() );
1218 LoadDocEntry( newBinEntry );
1219 if( !set->AddEntry( newBinEntry ) )
1221 //Expect big troubles if here
1222 //delete newBinEntry;
1228 /////////////////////// ValEntry
1229 // When "set" is a Document, then we are at the top of the
1230 // hierarchy and the Key is simply of the form ( group, elem )...
1231 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1234 newValEntry->SetKey( newValEntry->GetKey() );
1236 // ...but when "set" is a SQItem, we are inserting this new
1237 // valEntry in a sequence item. Hence the key has the
1238 // generalized form (refer to \ref BaseTagKey):
1239 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1241 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1242 + newValEntry->GetKey() );
1245 LoadDocEntry( newValEntry );
1246 bool delimitor=newValEntry->IsItemDelimitor();
1247 if( !set->AddEntry( newValEntry ) )
1249 // If here expect big troubles
1250 //delete newValEntry; //otherwise mem leak
1260 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1268 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1269 && ( newDocEntry->GetElement() == 0x0010 ) )
1271 TransferSyntaxType ts = GetTransferSyntax();
1272 if ( ts == RLELossless )
1274 long positionOnEntry = Fp->tellg();
1275 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1277 Fp->seekg( positionOnEntry, std::ios::beg );
1279 else if ( IsJPEG() )
1281 long positionOnEntry = Fp->tellg();
1282 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1283 ComputeJPEGFragmentInfo();
1284 Fp->seekg( positionOnEntry, std::ios::beg );
1288 // Just to make sure we are at the beginning of next entry.
1289 SkipToNextDocEntry(newDocEntry);
1294 unsigned long l = newDocEntry->GetReadLength();
1295 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1297 if ( l == 0xffffffff )
1306 // no other way to create it ...
1307 newSeqEntry->SetDelimitorMode( delim_mode );
1309 // At the top of the hierarchy, stands a Document. When "set"
1310 // is a Document, then we are building the first depth level.
1311 // Hence the SeqEntry we are building simply has a depth
1313 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1316 newSeqEntry->SetDepthLevel( 1 );
1317 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1319 // But when "set" is allready a SQItem, we are building a nested
1320 // sequence, and hence the depth level of the new SeqEntry
1321 // we are building, is one level deeper:
1322 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1324 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1325 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1326 + newSeqEntry->GetKey() );
1330 { // Don't try to parse zero-length sequences
1331 ParseSQ( newSeqEntry,
1332 newDocEntry->GetOffset(),
1335 set->AddEntry( newSeqEntry );
1336 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1348 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1349 * @return parsed length for this level
1351 void Document::ParseSQ( SeqEntry *seqEntry,
1352 long offset, long l_max, bool delim_mode)
1354 int SQItemNumber = 0;
1359 DocEntry *newDocEntry = ReadNextDocEntry();
1362 // FIXME Should warn user
1367 if ( newDocEntry->IsSequenceDelimitor() )
1369 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1373 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1379 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1380 std::ostringstream newBase;
1381 newBase << seqEntry->GetKey()
1385 itemSQ->SetBaseTagKey( newBase.str() );
1386 unsigned int l = newDocEntry->GetReadLength();
1388 if ( l == 0xffffffff )
1397 ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1400 seqEntry->AddEntry( itemSQ, SQItemNumber );
1402 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1410 * \brief Loads the element content if its length doesn't exceed
1411 * the value specified with Document::SetMaxSizeLoadEntry()
1412 * @param entry Header Entry (Dicom Element) to be dealt with
1414 void Document::LoadDocEntry(DocEntry *entry)
1416 uint16_t group = entry->GetGroup();
1417 std::string vr = entry->GetVR();
1418 uint32_t length = entry->GetLength();
1420 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1422 // A SeQuence "contains" a set of Elements.
1423 // (fffe e000) tells us an Element is beginning
1424 // (fffe e00d) tells us an Element just ended
1425 // (fffe e0dd) tells us the current SeQuence just ended
1426 if( group == 0xfffe )
1428 // NO more value field for SQ !
1432 // When the length is zero things are easy:
1435 ((ValEntry *)entry)->SetValue("");
1439 // The elements whose length is bigger than the specified upper bound
1440 // are not loaded. Instead we leave a short notice of the offset of
1441 // the element content and it's length.
1443 std::ostringstream s;
1444 if (length > MaxSizeLoadEntry)
1446 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1448 //s << "gdcm::NotLoaded (BinEntry)";
1449 s << GDCM_NOTLOADED;
1450 s << " Address:" << (long)entry->GetOffset();
1451 s << " Length:" << entry->GetLength();
1452 s << " x(" << std::hex << entry->GetLength() << ")";
1453 binEntryPtr->SetValue(s.str());
1455 // Be carefull : a BinEntry IS_A ValEntry ...
1456 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1458 // s << "gdcm::NotLoaded. (ValEntry)";
1459 s << GDCM_NOTLOADED;
1460 s << " Address:" << (long)entry->GetOffset();
1461 s << " Length:" << entry->GetLength();
1462 s << " x(" << std::hex << entry->GetLength() << ")";
1463 valEntryPtr->SetValue(s.str());
1468 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1469 << "nor a ValEntry ?! Should never print that !" << std::endl;
1472 // to be sure we are at the end of the value ...
1473 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1478 // When we find a BinEntry not very much can be done :
1479 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1481 s << GDCM_BINLOADED;
1482 binEntryPtr->SetValue(s.str());
1483 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1487 /// \todo Any compacter code suggested (?)
1488 if ( IsDocEntryAnInteger(entry) )
1492 // When short integer(s) are expected, read and convert the following
1493 // n *two characters properly i.e. consider them as short integers as
1494 // opposed to strings.
1495 // Elements with Value Multiplicity > 1
1496 // contain a set of integers (not a single one)
1497 if (vr == "US" || vr == "SS")
1500 NewInt = ReadInt16();
1504 for (int i=1; i < nbInt; i++)
1507 NewInt = ReadInt16();
1512 // See above comment on multiple integers (mutatis mutandis).
1513 else if (vr == "UL" || vr == "SL")
1516 NewInt = ReadInt32();
1520 for (int i=1; i < nbInt; i++)
1523 NewInt = ReadInt32();
1528 #ifdef GDCM_NO_ANSI_STRING_STREAM
1529 s << std::ends; // to avoid oddities on Solaris
1530 #endif //GDCM_NO_ANSI_STRING_STREAM
1532 ((ValEntry *)entry)->SetValue(s.str());
1536 // FIXME: We need an additional byte for storing \0 that is not on disk
1537 char *str = new char[length+1];
1538 Fp->read(str, (size_t)length);
1539 str[length] = '\0'; //this is only useful when length is odd
1540 // Special DicomString call to properly handle \0 and even length
1541 std::string newValue;
1544 newValue = Util::DicomString(str, length+1);
1545 gdcmVerboseMacro("Warning: bad length: " << length );
1546 gdcmVerboseMacro("For string :" << newValue.c_str());
1547 // Since we change the length of string update it length
1548 //entry->SetReadLength(length+1);
1552 newValue = Util::DicomString(str, length);
1556 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1558 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1560 gdcmVerboseMacro( "Unread element value");
1561 valEntry->SetValue(GDCM_UNREAD);
1567 // Because of correspondance with the VR dic
1568 valEntry->SetValue(newValue);
1572 valEntry->SetValue(newValue);
1577 gdcmErrorMacro( "Should have a ValEntry, here !");
1583 * \brief Find the value Length of the passed Header Entry
1584 * @param entry Header Entry whose length of the value shall be loaded.
1586 void Document::FindDocEntryLength( DocEntry *entry )
1587 throw ( FormatError )
1589 uint16_t element = entry->GetElement();
1590 std::string vr = entry->GetVR();
1593 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1595 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1597 // The following reserved two bytes (see PS 3.5-2003, section
1598 // "7.1.2 Data element structure with explicit vr", p 27) must be
1599 // skipped before proceeding on reading the length on 4 bytes.
1600 Fp->seekg( 2L, std::ios::cur);
1601 uint32_t length32 = ReadInt32();
1603 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1608 lengthOB = FindDocEntryLengthOBOrOW();
1610 catch ( FormatUnexpected )
1612 // Computing the length failed (this happens with broken
1613 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1614 // chance to get the pixels by deciding the element goes
1615 // until the end of the file. Hence we artificially fix the
1616 // the length and proceed.
1617 long currentPosition = Fp->tellg();
1618 Fp->seekg(0L,std::ios::end);
1620 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1621 Fp->seekg(currentPosition, std::ios::beg);
1623 entry->SetReadLength(lengthUntilEOF);
1624 entry->SetLength(lengthUntilEOF);
1627 entry->SetReadLength(lengthOB);
1628 entry->SetLength(lengthOB);
1631 FixDocEntryFoundLength(entry, length32);
1635 // Length is encoded on 2 bytes.
1636 length16 = ReadInt16();
1638 // We can tell the current file is encoded in big endian (like
1639 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1640 // and it's value is the one of the encoding of a big endian file.
1641 // In order to deal with such big endian encoded files, we have
1642 // (at least) two strategies:
1643 // * when we load the "Transfer Syntax" tag with value of big endian
1644 // encoding, we raise the proper flags. Then we wait for the end
1645 // of the META group (0x0002) among which is "Transfer Syntax",
1646 // before switching the swap code to big endian. We have to postpone
1647 // the switching of the swap code since the META group is fully encoded
1648 // in little endian, and big endian coding only starts at the next
1649 // group. The corresponding code can be hard to analyse and adds
1650 // many additional unnecessary tests for regular tags.
1651 // * the second strategy consists in waiting for trouble, that shall
1652 // appear when we find the first group with big endian encoding. This
1653 // is easy to detect since the length of a "Group Length" tag (the
1654 // ones with zero as element number) has to be of 4 (0x0004). When we
1655 // encounter 1024 (0x0400) chances are the encoding changed and we
1656 // found a group with big endian encoding.
1657 // We shall use this second strategy. In order to make sure that we
1658 // can interpret the presence of an apparently big endian encoded
1659 // length of a "Group Length" without committing a big mistake, we
1660 // add an additional check: we look in the already parsed elements
1661 // for the presence of a "Transfer Syntax" whose value has to be "big
1662 // endian encoding". When this is the case, chances are we have got our
1663 // hands on a big endian encoded file: we switch the swap code to
1664 // big endian and proceed...
1665 if ( element == 0x0000 && length16 == 0x0400 )
1667 TransferSyntaxType ts = GetTransferSyntax();
1668 if ( ts != ExplicitVRBigEndian )
1670 throw FormatError( "Document::FindDocEntryLength()",
1671 " not explicit VR." );
1675 SwitchSwapToBigEndian();
1677 // Restore the unproperly loaded values i.e. the group, the element
1678 // and the dictionary entry depending on them.
1679 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1680 uint16_t correctElem = SwapShort( entry->GetElement() );
1681 DictEntry *newTag = GetDictEntry( correctGroup,
1685 // This correct tag is not in the dictionary. Create a new one.
1686 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1688 // FIXME this can create a memory leaks on the old entry that be
1689 // left unreferenced.
1690 entry->SetDictEntry( newTag );
1693 // Heuristic: well, some files are really ill-formed.
1694 if ( length16 == 0xffff)
1696 // 0xffff means that we deal with 'Unknown Length' Sequence
1699 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1704 // Either implicit VR or a non DICOM conformal (see note below) explicit
1705 // VR that ommited the VR of (at least) this element. Farts happen.
1706 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1707 // on Data elements "Implicit and Explicit VR Data Elements shall
1708 // not coexist in a Data Set and Data Sets nested within it".]
1709 // Length is on 4 bytes.
1711 FixDocEntryFoundLength( entry, ReadInt32() );
1717 * \brief Find the Value Representation of the current Dicom Element.
1718 * @return Value Representation of the current Entry
1720 std::string Document::FindDocEntryVR()
1722 if ( Filetype != ExplicitVR )
1723 return(GDCM_UNKNOWN);
1725 long positionOnEntry = Fp->tellg();
1726 // Warning: we believe this is explicit VR (Value Representation) because
1727 // we used a heuristic that found "UL" in the first tag. Alas this
1728 // doesn't guarantee that all the tags will be in explicit VR. In some
1729 // cases (see e-film filtered files) one finds implicit VR tags mixed
1730 // within an explicit VR file. Hence we make sure the present tag
1731 // is in explicit VR and try to fix things if it happens not to be
1735 Fp->read (vr, (size_t)2);
1738 if( !CheckDocEntryVR(vr) )
1740 Fp->seekg(positionOnEntry, std::ios::beg);
1741 return(GDCM_UNKNOWN);
1747 * \brief Check the correspondance between the VR of the header entry
1748 * and the taken VR. If they are different, the header entry is
1749 * updated with the new VR.
1750 * @param vr Dicom Value Representation
1751 * @return false if the VR is incorrect of if the VR isn't referenced
1752 * otherwise, it returns true
1754 bool Document::CheckDocEntryVR(VRKey vr)
1756 // CLEANME searching the dicom_vr at each occurence is expensive.
1757 // PostPone this test in an optional integrity check at the end
1758 // of parsing or only in debug mode.
1759 if ( !Global::GetVR()->IsValidVR(vr) )
1766 * \brief Get the transformed value of the header entry. The VR value
1767 * is used to define the transformation to operate on the value
1768 * \warning NOT end user intended method !
1769 * @param entry entry to tranform
1770 * @return Transformed entry value
1772 std::string Document::GetDocEntryValue(DocEntry *entry)
1774 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1776 std::string val = ((ValEntry *)entry)->GetValue();
1777 std::string vr = entry->GetVR();
1778 uint32_t length = entry->GetLength();
1779 std::ostringstream s;
1782 // When short integer(s) are expected, read and convert the following
1783 // n * 2 bytes properly i.e. as a multivaluated strings
1784 // (each single value is separated fromthe next one by '\'
1785 // as usual for standard multivaluated filels
1786 // Elements with Value Multiplicity > 1
1787 // contain a set of short integers (not a single one)
1789 if( vr == "US" || vr == "SS" )
1794 for (int i=0; i < nbInt; i++)
1800 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1801 newInt16 = SwapShort( newInt16 );
1806 // When integer(s) are expected, read and convert the following
1807 // n * 4 bytes properly i.e. as a multivaluated strings
1808 // (each single value is separated fromthe next one by '\'
1809 // as usual for standard multivaluated filels
1810 // Elements with Value Multiplicity > 1
1811 // contain a set of integers (not a single one)
1812 else if( vr == "UL" || vr == "SL" )
1817 for (int i=0; i < nbInt; i++)
1823 newInt32 = ( val[4*i+0] & 0xFF )
1824 + (( val[4*i+1] & 0xFF ) << 8 )
1825 + (( val[4*i+2] & 0xFF ) << 16 )
1826 + (( val[4*i+3] & 0xFF ) << 24 );
1827 newInt32 = SwapLong( newInt32 );
1831 #ifdef GDCM_NO_ANSI_STRING_STREAM
1832 s << std::ends; // to avoid oddities on Solaris
1833 #endif //GDCM_NO_ANSI_STRING_STREAM
1837 return ((ValEntry *)entry)->GetValue();
1841 * \brief Get the reverse transformed value of the header entry. The VR
1842 * value is used to define the reverse transformation to operate on
1844 * \warning NOT end user intended method !
1845 * @param entry Entry to reverse transform
1846 * @return Reverse transformed entry value
1848 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1850 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1852 std::string vr = entry->GetVR();
1853 std::vector<std::string> tokens;
1854 std::ostringstream s;
1856 if ( vr == "US" || vr == "SS" )
1860 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1861 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1862 for (unsigned int i=0; i<tokens.size(); i++)
1864 newInt16 = atoi(tokens[i].c_str());
1865 s << ( newInt16 & 0xFF )
1866 << (( newInt16 >> 8 ) & 0xFF );
1870 if ( vr == "UL" || vr == "SL")
1874 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1875 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1876 for (unsigned int i=0; i<tokens.size();i++)
1878 newInt32 = atoi(tokens[i].c_str());
1879 s << (char)( newInt32 & 0xFF )
1880 << (char)(( newInt32 >> 8 ) & 0xFF )
1881 << (char)(( newInt32 >> 16 ) & 0xFF )
1882 << (char)(( newInt32 >> 24 ) & 0xFF );
1887 #ifdef GDCM_NO_ANSI_STRING_STREAM
1888 s << std::ends; // to avoid oddities on Solaris
1889 #endif //GDCM_NO_ANSI_STRING_STREAM
1893 return ((ValEntry *)entry)->GetValue();
1897 * \brief Skip a given Header Entry
1898 * \warning NOT end user intended method !
1899 * @param entry entry to skip
1901 void Document::SkipDocEntry(DocEntry *entry)
1903 SkipBytes(entry->GetLength());
1907 * \brief Skips to the begining of the next Header Entry
1908 * \warning NOT end user intended method !
1909 * @param offset start of skipping
1910 * @param readLgth length to skip
1913 void Document::SkipToNextDocEntry(DocEntry *newDocEntry)
1915 Fp->seekg((long)(newDocEntry->GetOffset()), std::ios::beg);
1916 Fp->seekg( (long)(newDocEntry->GetReadLength()),std::ios::cur);
1920 * \brief When the length of an element value is obviously wrong (because
1921 * the parser went Jabberwocky) one can hope improving things by
1922 * applying some heuristics.
1923 * @param entry entry to check
1924 * @param foundLength fist assumption about length
1926 void Document::FixDocEntryFoundLength(DocEntry *entry,
1927 uint32_t foundLength)
1929 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1930 if ( foundLength == 0xffffffff)
1935 uint16_t gr = entry->GetGroup();
1936 uint16_t elem = entry->GetElement();
1938 if ( foundLength % 2)
1940 gdcmVerboseMacro( "Warning : Tag with uneven length " << foundLength
1941 << " in x(" << std::hex << gr << "," << elem <<")");
1944 //////// Fix for some naughty General Electric images.
1945 // Allthough not recent many such GE corrupted images are still present
1946 // on Creatis hard disks. Hence this fix shall remain when such images
1947 // are no longer in use (we are talking a few years, here)...
1948 // Note: XMedCom probably uses such a trick since it is able to read
1949 // those pesky GE images ...
1950 if ( foundLength == 13)
1952 // Only happens for this length !
1953 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1956 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
1960 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1961 // Occurence of such images is quite low (unless one leaves close to a
1962 // 'Leonardo' source. Hence, one might consider commenting out the
1963 // following fix on efficiency reasons.
1964 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1967 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
1970 else if ( entry->GetVR() == "SQ" )
1972 foundLength = 0; // ReadLength is unchanged
1975 //////// We encountered a 'delimiter' element i.e. a tag of the form
1976 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1977 // taken into account.
1978 else if( gr == 0xfffe )
1980 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1981 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1982 // causes extra troubles...
1983 if( entry->GetElement() != 0x0000 )
1989 entry->SetLength(foundLength);
1993 * \brief Apply some heuristics to predict whether the considered
1994 * element value contains/represents an integer or not.
1995 * @param entry The element value on which to apply the predicate.
1996 * @return The result of the heuristical predicate.
1998 bool Document::IsDocEntryAnInteger(DocEntry *entry)
2000 uint16_t element = entry->GetElement();
2001 uint16_t group = entry->GetGroup();
2002 const std::string &vr = entry->GetVR();
2003 uint32_t length = entry->GetLength();
2005 // When we have some semantics on the element we just read, and if we
2006 // a priori know we are dealing with an integer, then we shall be
2007 // able to swap it's element value properly.
2008 if ( element == 0 ) // This is the group length of the group
2016 // Allthough this should never happen, still some images have a
2017 // corrupted group length [e.g. have a glance at offset x(8336) of
2018 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2019 // Since for dicom compliant and well behaved headers, the present
2020 // test is useless (and might even look a bit paranoid), when we
2021 // encounter such an ill-formed image, we simply display a warning
2022 // message and proceed on parsing (while crossing fingers).
2023 long filePosition = Fp->tellg();
2024 gdcmVerboseMacro( "Erroneous Group Length element length on : ("
2025 << std::hex << group << " , " << element
2026 << ") -before- position x(" << filePosition << ")"
2027 << "lgt : " << length );
2031 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2039 * \brief Find the Length till the next sequence delimiter
2040 * \warning NOT end user intended method !
2044 uint32_t Document::FindDocEntryLengthOBOrOW()
2045 throw( FormatUnexpected )
2047 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2048 long positionOnEntry = Fp->tellg();
2049 bool foundSequenceDelimiter = false;
2050 uint32_t totalLength = 0;
2052 while ( !foundSequenceDelimiter )
2058 group = ReadInt16();
2061 catch ( FormatError )
2063 throw FormatError("Document::FindDocEntryLengthOBOrOW()",
2064 " group or element not present.");
2067 // We have to decount the group and element we just read
2070 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2072 gdcmVerboseMacro( "Neither an Item tag nor a Sequence delimiter tag.");
2073 Fp->seekg(positionOnEntry, std::ios::beg);
2074 throw FormatUnexpected( "Neither an Item tag nor a Sequence delimiter tag.");
2077 if ( elem == 0xe0dd )
2079 foundSequenceDelimiter = true;
2082 uint32_t itemLength = ReadInt32();
2083 // We add 4 bytes since we just read the ItemLength with ReadInt32
2084 totalLength += itemLength + 4;
2085 SkipBytes(itemLength);
2087 if ( foundSequenceDelimiter )
2092 Fp->seekg( positionOnEntry, std::ios::beg);
2097 * \brief Reads a supposed to be 16 Bits integer
2098 * (swaps it depending on processor endianity)
2099 * @return read value
2101 uint16_t Document::ReadInt16()
2102 throw( FormatError )
2105 Fp->read ((char*)&g, (size_t)2);
2108 throw FormatError( "Document::ReadInt16()", " file error." );
2112 throw FormatError( "Document::ReadInt16()", "EOF." );
2119 * \brief Reads a supposed to be 32 Bits integer
2120 * (swaps it depending on processor endianity)
2121 * @return read value
2123 uint32_t Document::ReadInt32()
2124 throw( FormatError )
2127 Fp->read ((char*)&g, (size_t)4);
2130 throw FormatError( "Document::ReadInt32()", " file error." );
2134 throw FormatError( "Document::ReadInt32()", "EOF." );
2141 * \brief skips bytes inside the source file
2142 * \warning NOT end user intended method !
2145 void Document::SkipBytes(uint32_t nBytes)
2147 //FIXME don't dump the returned value
2148 Fp->seekg((long)nBytes, std::ios::cur);
2152 * \brief Loads all the needed Dictionaries
2153 * \warning NOT end user intended method !
2155 void Document::Initialise()
2157 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2159 RLEInfo = new RLEFramesInfo;
2160 JPEGInfo = new JPEGFragmentsInfo;
2165 * \brief Discover what the swap code is (among little endian, big endian,
2166 * bad little endian, bad big endian).
2168 * @return false when we are absolutely sure
2169 * it's neither ACR-NEMA nor DICOM
2170 * true when we hope ours assuptions are OK
2172 bool Document::CheckSwap()
2174 // The only guaranted way of finding the swap code is to find a
2175 // group tag since we know it's length has to be of four bytes i.e.
2176 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2177 // occurs when we can't find such group...
2179 uint32_t x = 4; // x : for ntohs
2180 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2186 // First, compare HostByteOrder and NetworkByteOrder in order to
2187 // determine if we shall need to swap bytes (i.e. the Endian type).
2188 if ( x == ntohs(x) )
2197 // The easiest case is the one of a DICOM header, since it possesses a
2198 // file preamble where it suffice to look for the string "DICM".
2201 char *entCur = deb + 128;
2202 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2204 gdcmVerboseMacro( "Looks like DICOM Version3" );
2206 // Next, determine the value representation (VR). Let's skip to the
2207 // first element (0002, 0000) and check there if we find "UL"
2208 // - or "OB" if the 1st one is (0002,0001) -,
2209 // in which case we (almost) know it is explicit VR.
2210 // WARNING: if it happens to be implicit VR then what we will read
2211 // is the length of the group. If this ascii representation of this
2212 // length happens to be "UL" then we shall believe it is explicit VR.
2213 // FIXME: in order to fix the above warning, we could read the next
2214 // element value (or a couple of elements values) in order to make
2215 // sure we are not commiting a big mistake.
2216 // We need to skip :
2217 // * the 128 bytes of File Preamble (often padded with zeroes),
2218 // * the 4 bytes of "DICM" string,
2219 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2220 // i.e. a total of 136 bytes.
2224 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2225 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2226 // *Implicit* VR. -and it is !-
2228 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2229 memcmp(entCur, "OB", (size_t)2) == 0 ||
2230 memcmp(entCur, "UI", (size_t)2) == 0 ||
2231 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2232 // when Write DCM *adds*
2234 // Use Document::dicom_vr to test all the possibilities
2235 // instead of just checking for UL, OB and UI !? group 0000
2237 Filetype = ExplicitVR;
2238 gdcmVerboseMacro( "Explicit Value Representation");
2242 Filetype = ImplicitVR;
2243 gdcmVerboseMacro( "Not an explicit Value Representation");
2249 gdcmVerboseMacro( "HostByteOrder != NetworkByteOrder");
2254 gdcmVerboseMacro( "HostByteOrder = NetworkByteOrder");
2257 // Position the file position indicator at first tag (i.e.
2258 // after the file preamble and the "DICM" string).
2259 Fp->seekg(0, std::ios::beg);
2260 Fp->seekg ( 132L, std::ios::beg);
2264 // Alas, this is not a DicomV3 file and whatever happens there is no file
2265 // preamble. We can reset the file position indicator to where the data
2266 // is (i.e. the beginning of the file).
2267 gdcmVerboseMacro( "Not a DICOM Version3 file");
2268 Fp->seekg(0, std::ios::beg);
2270 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2271 // By clean we mean that the length of the first tag is written down.
2272 // If this is the case and since the length of the first group HAS to be
2273 // four (bytes), then determining the proper swap code is straightforward.
2276 // We assume the array of char we are considering contains the binary
2277 // representation of a 32 bits integer. Hence the following dirty
2279 s32 = *((uint32_t *)(entCur));
2300 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2301 // It is time for despaired wild guesses.
2302 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2303 // i.e. the 'group length' element is not present :
2305 // check the supposed-to-be 'group number'
2306 // in ( 0x0001 .. 0x0008 )
2307 // to determine ' SwapCode' value .
2308 // Only 0 or 4321 will be possible
2309 // (no oportunity to check for the formerly well known
2310 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2311 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2312 // the file IS NOT ACR-NEMA nor DICOM V3
2313 // Find a trick to tell it the caller...
2315 s16 = *((uint16_t *)(deb));
2342 gdcmVerboseMacro( "ACR/NEMA unfound swap info (Really hopeless !)");
2346 // Then the only info we have is the net2host one.
2358 * \brief Restore the unproperly loaded values i.e. the group, the element
2359 * and the dictionary entry depending on them.
2361 void Document::SwitchSwapToBigEndian()
2363 gdcmVerboseMacro( "Switching to BigEndian mode.");
2364 if ( SwapCode == 0 )
2368 else if ( SwapCode == 4321 )
2372 else if ( SwapCode == 3412 )
2376 else if ( SwapCode == 2143 )
2383 * \brief during parsing, Header Elements too long are not loaded in memory
2386 void Document::SetMaxSizeLoadEntry(long newSize)
2392 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2394 MaxSizeLoadEntry = 0xffffffff;
2397 MaxSizeLoadEntry = newSize;
2402 * \brief Header Elements too long will not be printed
2403 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2406 void Document::SetMaxSizePrintEntry(long newSize)
2408 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2413 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2415 MaxSizePrintEntry = 0xffffffff;
2418 MaxSizePrintEntry = newSize;
2424 * \brief Handle broken private tag from Philips NTSCAN
2425 * where the endianess is being switch to BigEndian for no
2429 void Document::HandleBrokenEndian(uint16_t group, uint16_t elem)
2431 // Endian reversion. Some files contain groups of tags with reversed endianess.
2432 static int reversedEndian = 0;
2433 // try to fix endian switching in the middle of headers
2434 if ((group == 0xfeff) && (elem == 0x00e0))
2436 // start endian swap mark for group found
2438 SwitchSwapToBigEndian();
2443 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2445 // end of reversed endian group
2447 SwitchSwapToBigEndian();
2452 * \brief Read the next tag but WITHOUT loading it's value
2453 * (read the 'Group Number', the 'Element Number',
2454 * gets the Dict Entry
2455 * gets the VR, gets the length, gets the offset value)
2456 * @return On succes the newly created DocEntry, NULL on failure.
2458 DocEntry *Document::ReadNextDocEntry()
2465 group = ReadInt16();
2468 catch ( FormatError e )
2470 // We reached the EOF (or an error occured) therefore
2471 // header parsing has to be considered as finished.
2476 HandleBrokenEndian(group, elem);
2477 std::string vr = FindDocEntryVR();
2478 std::string realVR = vr;
2480 if( vr == GDCM_UNKNOWN)
2482 DictEntry *dictEntry = GetDictEntry(group,elem);
2484 realVR = dictEntry->GetVR();
2488 if( Global::GetVR()->IsVROfSequence(realVR) )
2489 newEntry = NewSeqEntry(group, elem);
2490 else if( Global::GetVR()->IsVROfStringRepresentable(realVR) )
2491 newEntry = NewValEntry(group, elem,vr);
2493 newEntry = NewBinEntry(group, elem,vr);
2495 if( vr == GDCM_UNKNOWN )
2497 if( Filetype == ExplicitVR )
2499 // We thought this was explicit VR, but we end up with an
2500 // implicit VR tag. Let's backtrack.
2502 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
2503 newEntry->GetGroup(), newEntry->GetElement());
2504 gdcmVerboseMacro( msg.c_str() );
2506 newEntry->SetImplicitVR();
2511 FindDocEntryLength(newEntry);
2513 catch ( FormatError e )
2521 newEntry->SetOffset(Fp->tellg());
2528 * \brief Generate a free TagKey i.e. a TagKey that is not present
2529 * in the TagHt dictionary.
2530 * @param group The generated tag must belong to this group.
2531 * @return The element of tag with given group which is fee.
2533 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2535 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2537 TagKey key = DictEntry::TranslateToKey(group, elem);
2538 if (TagHT.count(key) == 0)
2547 * \brief Assuming the internal file pointer \ref Document::Fp
2548 * is placed at the beginning of a tag check whether this
2549 * tag is (TestGroup, TestElement).
2550 * \warning On success the internal file pointer \ref Document::Fp
2551 * is modified to point after the tag.
2552 * On failure (i.e. when the tag wasn't the expected tag
2553 * (TestGroup, TestElement) the internal file pointer
2554 * \ref Document::Fp is restored to it's original position.
2555 * @param testGroup The expected group of the tag.
2556 * @param testElement The expected Element of the tag.
2557 * @return True on success, false otherwise.
2559 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2561 long positionOnEntry = Fp->tellg();
2562 long currentPosition = Fp->tellg(); // On debugging purposes
2564 //// Read the Item Tag group and element, and make
2565 // sure they are what we expected:
2566 uint16_t itemTagGroup;
2567 uint16_t itemTagElement;
2570 itemTagGroup = ReadInt16();
2571 itemTagElement = ReadInt16();
2573 catch ( FormatError e )
2575 //std::cerr << e << std::endl;
2578 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2580 gdcmVerboseMacro( "Wrong Item Tag found:"
2581 << " We should have found tag ("
2582 << std::hex << testGroup << "," << testElement << ")" << std::endl
2583 << " but instead we encountered tag ("
2584 << std::hex << itemTagGroup << "," << itemTagElement << ")"
2586 << " at address: " << (unsigned int)currentPosition );
2587 Fp->seekg(positionOnEntry, std::ios::beg);
2595 * \brief Assuming the internal file pointer \ref Document::Fp
2596 * is placed at the beginning of a tag (TestGroup, TestElement),
2597 * read the length associated to the Tag.
2598 * \warning On success the internal file pointer \ref Document::Fp
2599 * is modified to point after the tag and it's length.
2600 * On failure (i.e. when the tag wasn't the expected tag
2601 * (TestGroup, TestElement) the internal file pointer
2602 * \ref Document::Fp is restored to it's original position.
2603 * @param testGroup The expected group of the tag.
2604 * @param testElement The expected Element of the tag.
2605 * @return On success returns the length associated to the tag. On failure
2608 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2610 long positionOnEntry = Fp->tellg();
2611 (void)positionOnEntry;
2613 if ( !ReadTag(testGroup, testElement) )
2618 //// Then read the associated Item Length
2619 long currentPosition = Fp->tellg();
2620 uint32_t itemLength = ReadInt32();
2622 gdcmVerboseMacro( "Basic Item Length is: "
2623 << itemLength << std::endl
2624 << " at address: " << (unsigned int)currentPosition);
2630 * \brief When parsing the Pixel Data of an encapsulated file, read
2631 * the basic offset table (when present, and BTW dump it).
2633 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2635 //// Read the Basic Offset Table Item Tag length...
2636 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2638 // When present, read the basic offset table itself.
2639 // Notes: - since the presence of this basic offset table is optional
2640 // we can't rely on it for the implementation, and we will simply
2641 // trash it's content (when present).
2642 // - still, when present, we could add some further checks on the
2643 // lengths, but we won't bother with such fuses for the time being.
2644 if ( itemLength != 0 )
2646 char *basicOffsetTableItemValue = new char[itemLength + 1];
2647 Fp->read(basicOffsetTableItemValue, itemLength);
2650 for (unsigned int i=0; i < itemLength; i += 4 )
2652 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2654 gdcmVerboseMacro( "Read one length: " <<
2655 std::hex << individualLength );
2659 delete[] basicOffsetTableItemValue;
2664 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2665 * Compute the RLE extra information and store it in \ref RLEInfo
2666 * for later pixel retrieval usage.
2668 void Document::ComputeRLEInfo()
2670 TransferSyntaxType ts = GetTransferSyntax();
2671 if ( ts != RLELossless )
2676 // Encoded pixel data: for the time being we are only concerned with
2677 // Jpeg or RLE Pixel data encodings.
2678 // As stated in PS 3.5-2003, section 8.2 p44:
2679 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2680 // value representation OB is used".
2681 // Hence we expect an OB value representation. Concerning OB VR,
2682 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2683 // "For the Value Representations OB and OW, the encoding shall meet the
2684 // following specifications depending on the Data element tag:"
2686 // - the first item in the sequence of items before the encoded pixel
2687 // data stream shall be basic offset table item. The basic offset table
2688 // item value, however, is not required to be present"
2690 ReadAndSkipEncapsulatedBasicOffsetTable();
2692 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2693 // Loop on the individual frame[s] and store the information
2694 // on the RLE fragments in a RLEFramesInfo.
2695 // Note: - when only a single frame is present, this is a
2697 // - when more than one frame are present, then we are in
2698 // the case of a multi-frame image.
2700 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2702 // Parse the RLE Header and store the corresponding RLE Segment
2703 // Offset Table information on fragments of this current Frame.
2704 // Note that the fragment pixels themselves are not loaded
2705 // (but just skipped).
2706 long frameOffset = Fp->tellg();
2708 uint32_t nbRleSegments = ReadInt32();
2709 if ( nbRleSegments > 16 )
2711 // There should be at most 15 segments (refer to RLEFrame class)
2712 gdcmVerboseMacro( "Too many segments.");
2715 uint32_t rleSegmentOffsetTable[16];
2716 for( int k = 1; k <= 15; k++ )
2718 rleSegmentOffsetTable[k] = ReadInt32();
2721 // Deduce from both the RLE Header and the frameLength the
2722 // fragment length, and again store this info in a
2724 long rleSegmentLength[15];
2725 // skipping (not reading) RLE Segments
2726 if ( nbRleSegments > 1)
2728 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2730 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2731 - rleSegmentOffsetTable[k];
2732 SkipBytes(rleSegmentLength[k]);
2736 rleSegmentLength[nbRleSegments] = frameLength
2737 - rleSegmentOffsetTable[nbRleSegments];
2738 SkipBytes(rleSegmentLength[nbRleSegments]);
2740 // Store the collected info
2741 RLEFrame *newFrameInfo = new RLEFrame;
2742 newFrameInfo->NumberFragments = nbRleSegments;
2743 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2745 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2746 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2748 RLEInfo->Frames.push_back( newFrameInfo );
2751 // Make sure that at the end of the item we encounter a 'Sequence
2753 if ( !ReadTag(0xfffe, 0xe0dd) )
2755 gdcmVerboseMacro( "No sequence delimiter item at end of RLE item sequence");
2760 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2761 * Compute the jpeg extra information (fragment[s] offset[s] and
2762 * length) and store it[them] in \ref JPEGInfo for later pixel
2765 void Document::ComputeJPEGFragmentInfo()
2767 // If you need to, look for comments of ComputeRLEInfo().
2773 ReadAndSkipEncapsulatedBasicOffsetTable();
2775 // Loop on the fragments[s] and store the parsed information in a
2777 long fragmentLength;
2778 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2780 long fragmentOffset = Fp->tellg();
2782 // Store the collected info
2783 JPEGFragment *newFragment = new JPEGFragment;
2784 newFragment->Offset = fragmentOffset;
2785 newFragment->Length = fragmentLength;
2786 JPEGInfo->Fragments.push_back( newFragment );
2788 SkipBytes( fragmentLength );
2791 // Make sure that at the end of the item we encounter a 'Sequence
2793 if ( !ReadTag(0xfffe, 0xe0dd) )
2795 gdcmVerboseMacro( "No sequence delimiter item at end of JPEG item sequence");
2800 * \brief Walk recursively the given \ref DocEntrySet, and feed
2801 * the given hash table (\ref TagDocEntryHT) with all the
2802 * \ref DocEntry (Dicom entries) encountered.
2803 * This method does the job for \ref BuildFlatHashTable.
2804 * @param builtHT Where to collect all the \ref DocEntry encountered
2805 * when recursively walking the given set.
2806 * @param set The structure to be traversed (recursively).
2808 void Document::BuildFlatHashTableRecurse( TagDocEntryHT &builtHT,
2811 if (ElementSet *elementSet = dynamic_cast< ElementSet* > ( set ) )
2813 TagDocEntryHT const ¤tHT = elementSet->GetTagHT();
2814 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2815 i != currentHT.end();
2818 DocEntry *entry = i->second;
2819 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2821 const ListSQItem& items = seqEntry->GetSQItems();
2822 for( ListSQItem::const_iterator item = items.begin();
2823 item != items.end();
2826 BuildFlatHashTableRecurse( builtHT, *item );
2830 builtHT[entry->GetKey()] = entry;
2835 if (SQItem *SQItemSet = dynamic_cast< SQItem* > ( set ) )
2837 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
2838 for (ListDocEntry::const_iterator i = currentList.begin();
2839 i != currentList.end();
2842 DocEntry *entry = *i;
2843 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2845 const ListSQItem& items = seqEntry->GetSQItems();
2846 for( ListSQItem::const_iterator item = items.begin();
2847 item != items.end();
2850 BuildFlatHashTableRecurse( builtHT, *item );
2854 builtHT[entry->GetKey()] = entry;
2861 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2864 * The structure used by a Document (through \ref ElementSet),
2865 * in order to hold the parsed entries of a Dicom header, is a recursive
2866 * one. This is due to the fact that the sequences (when present)
2867 * can be nested. Additionaly, the sequence items (represented in
2868 * gdcm as \ref SQItem) add an extra complexity to the data
2869 * structure. Hence, a gdcm user whishing to visit all the entries of
2870 * a Dicom header will need to dig in the gdcm internals (which
2871 * implies exposing all the internal data structures to the API).
2872 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2873 * recursively builds a temporary hash table, which holds all the
2874 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
2876 * \warning Of course there is NO integrity constrain between the
2877 * returned \ref TagDocEntryHT and the \ref ElementSet used
2878 * to build it. Hence if the underlying \ref ElementSet is
2879 * altered, then it is the caller responsability to invoke
2880 * \ref BuildFlatHashTable again...
2881 * @return The flat std::map<> we juste build.
2883 TagDocEntryHT *Document::BuildFlatHashTable()
2885 TagDocEntryHT *FlatHT = new TagDocEntryHT;
2886 BuildFlatHashTableRecurse( *FlatHT, this );
2893 * \brief Compares two documents, according to \ref DicomDir rules
2894 * \warning Does NOT work with ACR-NEMA files
2895 * \todo Find a trick to solve the pb (use RET fields ?)
2897 * @return true if 'smaller'
2899 bool Document::operator<(Document &document)
2902 std::string s1 = GetEntry(0x0010,0x0010);
2903 std::string s2 = document.GetEntry(0x0010,0x0010);
2915 s1 = GetEntry(0x0010,0x0020);
2916 s2 = document.GetEntry(0x0010,0x0020);
2927 // Study Instance UID
2928 s1 = GetEntry(0x0020,0x000d);
2929 s2 = document.GetEntry(0x0020,0x000d);
2940 // Serie Instance UID
2941 s1 = GetEntry(0x0020,0x000e);
2942 s2 = document.GetEntry(0x0020,0x000e);
2959 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
2960 * @param filetype Type of the File to be written
2962 int Document::ComputeGroup0002Length( FileType filetype )
2967 int groupLength = 0;
2968 bool found0002 = false;
2970 // for each zero-level Tag in the DCM Header
2974 entry = GetNextEntry();
2977 gr = entry->GetGroup();
2983 el = entry->GetElement();
2984 vr = entry->GetVR();
2986 if (filetype == ExplicitVR)
2988 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
2990 groupLength += 4; // explicit VR AND OB, OW, SQ : 4 more bytes
2993 groupLength += 2 + 2 + 4 + entry->GetLength();
2995 else if (found0002 )
2998 entry = GetNextEntry();
3003 } // end namespace gdcm
3005 //-----------------------------------------------------------------------------