1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/07 22:03:30 $
7 Version: $Revision: 1.166 $
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 = GetEntryByNumber(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 = GetEntryByNumber(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 = GetEntryByNumber(0x0028, 0x0010);
179 std::string columns = GetEntryByNumber(0x0028, 0x0011);
180 SetEntryByNumber(columns, 0x0028, 0x0010);
181 SetEntryByNumber(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 = GetDocEntryByNumber(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 GetDocEntryByNumber(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 )
456 Fp->seekg(126L, std::ios::cur);
458 Fp->read(dicm, (size_t)4);
464 if( memcmp(dicm, "DICM", 4) == 0 )
470 gdcmVerboseMacro( "Not DICOM/ACR (missing preamble)" << Filename.c_str());
476 * \brief closes the file
477 * @return TRUE if the close was successfull
479 bool Document::CloseFile()
488 return true; //FIXME how do we detect a non-close ifstream ?
492 * \brief Writes in a file all the Header Entries (Dicom Elements)
493 * @param fp file pointer on an already open file
494 * @param filetype Type of the File to be written
495 * (ACR-NEMA, ExplicitVR, ImplicitVR)
496 * \return Always true.
498 void Document::WriteContent(std::ofstream *fp, FileType filetype)
500 /// \todo move the following lines (and a lot of others, to be written)
501 /// to a future function CheckAndCorrectHeader
502 /// (necessary if user wants to write a DICOM V3 file
503 /// starting from an ACR-NEMA (V2) Header
505 if ( filetype == ImplicitVR || filetype == ExplicitVR )
507 // writing Dicom File Preamble
508 char filePreamble[128];
509 memset(filePreamble, 0, 128);
510 fp->write(filePreamble, 128);
511 fp->write("DICM", 4);
515 * \todo rewrite later, if really usefull
516 * - 'Group Length' element is optional in DICOM
517 * - but un-updated odd groups lengthes can causes pb
520 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
521 * UpdateGroupLength(false,filetype);
522 * if ( filetype == ACR)
523 * UpdateGroupLength(true,ACR);
526 ElementSet::WriteContent(fp, filetype); // This one is recursive
530 * \brief Modifies the value of a given Doc Entry (Dicom Element)
531 * when it exists. Create it with the given value when unexistant.
532 * @param value (string) Value to be set
533 * @param group Group number of the Entry
534 * @param elem Element number of the Entry
535 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
536 * \return pointer to the modified/created Header Entry (NULL when creation
539 ValEntry *Document::ReplaceOrCreateByNumber(std::string const &value,
544 ValEntry *valEntry = 0;
545 DocEntry *currentEntry = GetDocEntryByNumber( group, elem);
549 valEntry = dynamic_cast< ValEntry* >(currentEntry);
553 if( valEntry->GetVR()!=vr )
556 // if currentEntry doesn't correspond to the requested valEntry
559 if (!RemoveEntry(currentEntry))
561 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
568 // Create a new valEntry if necessary
571 valEntry = NewValEntryByNumber(group, elem, vr);
573 if ( !AddEntry(valEntry))
575 gdcmVerboseMacro("AddEntry failed although this is a creation.");
582 // Set the binEntry value
583 SetEntry(value, valEntry);
588 * \brief Modifies the value of a given Header Entry (Dicom Element)
589 * when it exists. Create it with the given value when unexistant.
590 * A copy of the binArea is made to be kept in the Document.
591 * @param binArea (binary) value to be set
592 * @param Group Group number of the Entry
593 * @param Elem Element number of the Entry
594 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
595 * \return pointer to the modified/created Header Entry (NULL when creation
598 BinEntry *Document::ReplaceOrCreateByNumber(uint8_t *binArea,
604 BinEntry *binEntry = 0;
605 DocEntry *currentEntry = GetDocEntryByNumber( group, elem);
607 // Verify the currentEntry
610 binEntry = dynamic_cast< BinEntry* >(currentEntry);
614 if( binEntry->GetVR()!=vr )
617 // if currentEntry doesn't correspond to the requested valEntry
620 if (!RemoveEntry(currentEntry))
622 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
629 // Create a new binEntry if necessary
632 binEntry = NewBinEntryByNumber(group, elem, vr);
634 if ( !AddEntry(binEntry))
636 gdcmVerboseMacro( "AddEntry failed allthough this is a creation.");
643 // Set the binEntry value
645 if (lgth>0 && binArea)
647 tmpArea = new uint8_t[lgth];
648 memcpy(tmpArea,binArea,lgth);
654 if (!SetEntry(tmpArea,lgth,binEntry))
666 * \brief Modifies the value of a given Header Entry (Dicom Element)
667 * when it exists. Create it when unexistant.
668 * @param Group Group number of the Entry
669 * @param Elem Element number of the Entry
670 * \return pointer to the modified/created SeqEntry (NULL when creation
673 SeqEntry *Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem)
675 SeqEntry *seqEntry = 0;
676 DocEntry *currentEntry = GetDocEntryByNumber( group, elem);
678 // Verify the currentEntry
681 seqEntry = dynamic_cast< SeqEntry* >(currentEntry);
685 if( seqEntry->GetVR()!="SQ" )
688 // if currentEntry doesn't correspond to the requested valEntry
691 if (!RemoveEntry(currentEntry))
693 gdcmVerboseMacro( "Removal of previous DocEntry failed.");
700 // Create a new seqEntry if necessary
703 seqEntry = NewSeqEntryByNumber(group, elem);
705 if ( !AddEntry(seqEntry))
707 gdcmVerboseMacro( "AddEntry failed allthough this is a creation.");
718 * \brief Set a new value if the invoked element exists
719 * Seems to be useless !!!
720 * @param value new element value
721 * @param group group number of the Entry
722 * @param elem element number of the Entry
725 bool Document::ReplaceIfExistByNumber(std::string const &value,
726 uint16_t group, uint16_t elem )
728 SetEntryByNumber(value, group, elem);
733 std::string Document::GetTransferSyntaxValue(TransferSyntaxType type)
735 return TransferSyntaxStrings[type];
738 //-----------------------------------------------------------------------------
742 * \brief Checks if a given Dicom Element exists within the H table
743 * @param group Group number of the searched Dicom Element
744 * @param element Element number of the searched Dicom Element
745 * @return true is found
747 bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
749 const std::string &key = DictEntry::TranslateToKey(group, element );
750 return TagHT.count(key) != 0;
755 * \brief Searches within Header Entries (Dicom Elements) parsed with
756 * the public and private dictionaries
757 * for the element value representation of a given tag.
758 * @param group Group number of the searched tag.
759 * @param element Element number of the searched tag.
760 * @return Corresponding element value representation when it exists,
761 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
763 std::string Document::GetEntryByNumber(uint16_t group, uint16_t element)
765 TagKey key = DictEntry::TranslateToKey(group, element);
766 if ( !TagHT.count(key))
771 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
775 * \brief Searches within Header Entries (Dicom Elements) parsed with
776 * the public and private dictionaries
777 * for the element value representation of a given tag..
779 * Obtaining the VR (Value Representation) might be needed by caller
780 * to convert the string typed content to caller's native type
781 * (think of C++ vs Python). The VR is actually of a higher level
782 * of semantics than just the native C++ type.
783 * @param group Group number of the searched tag.
784 * @param element Element number of the searched tag.
785 * @return Corresponding element value representation when it exists,
786 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
788 std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element)
790 DocEntry *elem = GetDocEntryByNumber(group, element);
795 return elem->GetVR();
799 * \brief Searches within Header Entries (Dicom Elements) parsed with
800 * the public and private dictionaries
801 * for the value length of a given tag..
802 * @param group Group number of the searched tag.
803 * @param element Element number of the searched tag.
804 * @return Corresponding element length; -2 if not found
806 int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element)
808 DocEntry *elem = GetDocEntryByNumber(group, element);
811 return -2; //magic number
813 return elem->GetLength();
817 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
818 * through it's (group, element) and modifies it's content with
820 * @param content new value (string) to substitute with
821 * @param group group number of the Dicom Element to modify
822 * @param element element number of the Dicom Element to modify
824 bool Document::SetEntryByNumber(std::string const& content,
825 uint16_t group, uint16_t element)
827 ValEntry *entry = GetValEntryByNumber(group, element);
830 gdcmVerboseMacro( "No corresponding ValEntry (try promotion first).");
833 return SetEntry(content,entry);
837 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
838 * through it's (group, element) and modifies it's content with
840 * @param content new value (void* -> uint8_t*) to substitute with
841 * @param lgth new value length
842 * @param group group number of the Dicom Element to modify
843 * @param element element number of the Dicom Element to modify
845 bool Document::SetEntryByNumber(uint8_t*content, int lgth,
846 uint16_t group, uint16_t element)
848 BinEntry *entry = GetBinEntryByNumber(group, element);
851 gdcmVerboseMacro( "No corresponding ValEntry (try promotion first).");
855 return SetEntry(content,lgth,entry);
859 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
860 * and modifies it's content with the given value.
861 * @param content new value (string) to substitute with
862 * @param entry Entry to be modified
864 bool Document::SetEntry(std::string const &content,ValEntry *entry)
868 entry->SetValue(content);
875 * \brief Accesses an existing BinEntry (i.e. a Dicom Element)
876 * and modifies it's content with the given value.
877 * @param content new value (void* -> uint8_t*) to substitute with
878 * @param entry Entry to be modified
879 * @param lgth new value length
881 bool Document::SetEntry(uint8_t *content, int lgth, BinEntry *entry)
885 // Hope Binary field length is *never* wrong
886 /*if(lgth%2) // Non even length are padded with a space (020H).
889 //content = content + '\0'; // fing a trick to enlarge a binary field?
892 entry->SetBinArea(content);
893 entry->SetLength(lgth);
894 entry->SetValue(GDCM_BINLOADED);
901 * \brief Gets (from Header) a 'non string' element value
902 * (LoadElementValues has already be executed)
903 * @param group group number of the Entry
904 * @param elem element number of the Entry
905 * @return Pointer to the 'non string' area
907 void *Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem)
909 DocEntry *entry = GetDocEntryByNumber(group, elem);
912 gdcmVerboseMacro( "No entry");
915 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(entry) )
917 return binEntry->GetBinArea();
924 * \brief Loads (from disk) the element content
925 * when a string is not suitable
926 * @param group group number of the Entry
927 * @param elem element number of the Entry
929 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
931 // Search the corresponding DocEntry
932 DocEntry *docElement = GetDocEntryByNumber(group, elem);
936 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
940 LoadEntryBinArea(binElement);
944 * \brief Loads (from disk) the element content
945 * when a string is not suitable
946 * @param element Entry whose binArea is going to be loaded
948 void Document::LoadEntryBinArea(BinEntry *element)
950 if(element->GetBinArea())
957 size_t o =(size_t)element->GetOffset();
958 Fp->seekg(o, std::ios::beg);
960 size_t l = element->GetLength();
961 uint8_t *a = new uint8_t[l];
964 gdcmVerboseMacro( "Cannot allocate a");
968 /// \todo check the result
969 Fp->read((char*)a, l);
970 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
976 element->SetBinArea(a);
983 * \brief Sets a 'non string' value to a given Dicom Element
984 * @param area area containing the 'non string' value
985 * @param group Group number of the searched Dicom Element
986 * @param element Element number of the searched Dicom Element
989 /*bool Document::SetEntryBinAreaByNumber(uint8_t *area,
990 uint16_t group, uint16_t element)
992 DocEntry *currentEntry = GetDocEntryByNumber(group, element);
998 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1000 binEntry->SetBinArea( area );
1008 * \brief retrieves a Dicom Element (the first one) using (group, element)
1009 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1010 * if you think it's NOT UNIQUE, check the count number
1011 * and use iterators to retrieve ALL the Dicoms Elements within
1012 * a given couple (group, element)
1013 * @param group Group number of the searched Dicom Element
1014 * @param element Element number of the searched Dicom Element
1017 DocEntry *Document::GetDocEntryByNumber(uint16_t group, uint16_t element)
1019 TagKey key = DictEntry::TranslateToKey(group, element);
1020 if ( !TagHT.count(key))
1024 return TagHT.find(key)->second;
1028 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1029 * returns a result when the corresponding entry is of type
1031 * @return When present, the corresponding ValEntry.
1033 ValEntry *Document::GetValEntryByNumber(uint16_t group, uint16_t element)
1035 DocEntry *currentEntry = GetDocEntryByNumber(group, element);
1036 if ( !currentEntry )
1040 if ( ValEntry *entry = dynamic_cast<ValEntry*>(currentEntry) )
1044 gdcmVerboseMacro( "Unfound ValEntry.");
1050 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1051 * returns a result when the corresponding entry is of type
1053 * @return When present, the corresponding BinEntry.
1055 BinEntry *Document::GetBinEntryByNumber(uint16_t group, uint16_t element)
1057 DocEntry *currentEntry = GetDocEntryByNumber(group, element);
1058 if ( !currentEntry )
1062 if ( BinEntry *entry = dynamic_cast<BinEntry*>(currentEntry) )
1066 gdcmVerboseMacro( "Unfound BinEntry.");
1072 * \brief Loads the element while preserving the current
1073 * underlying file position indicator as opposed to
1074 * to LoadDocEntry that modifies it.
1075 * @param entry Header Entry whose value shall be loaded.
1078 void Document::LoadDocEntrySafe(DocEntry *entry)
1082 long PositionOnEntry = Fp->tellg();
1083 LoadDocEntry(entry);
1084 Fp->seekg(PositionOnEntry, std::ios::beg);
1089 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1091 * @return The properly swaped 32 bits integer.
1093 uint32_t Document::SwapLong(uint32_t a)
1100 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1101 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1105 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1109 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1112 //std::cout << "swapCode= " << SwapCode << std::endl;
1113 gdcmErrorMacro(" Document::SwapLong : unset swap code");
1120 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1122 * @return The properly unswaped 32 bits integer.
1124 uint32_t Document::UnswapLong(uint32_t a)
1130 * \brief Swaps the bytes so they agree with the processor order
1131 * @return The properly swaped 16 bits integer.
1133 uint16_t Document::SwapShort(uint16_t a)
1135 if ( SwapCode == 4321 || SwapCode == 2143 )
1137 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1143 * \brief Unswaps the bytes so they agree with the processor order
1144 * @return The properly unswaped 16 bits integer.
1146 uint16_t Document::UnswapShort(uint16_t a)
1148 return SwapShort(a);
1151 //-----------------------------------------------------------------------------
1155 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1156 * @return length of the parsed set.
1158 void Document::ParseDES(DocEntrySet *set, long offset,
1159 long l_max, bool delim_mode)
1161 DocEntry *newDocEntry = 0;
1162 ValEntry *newValEntry;
1163 BinEntry *newBinEntry;
1164 SeqEntry *newSeqEntry;
1170 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1176 newDocEntry = ReadNextDocEntry( );
1182 vr = newDocEntry->GetVR();
1183 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
1184 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
1185 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1187 if ( newValEntry || newBinEntry )
1191 if ( ! Global::GetVR()->IsVROfBinaryRepresentable(vr) )
1193 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1194 gdcmVerboseMacro( "Neither Valentry, nor BinEntry."
1195 "Probably unknown VR.");
1198 //////////////////// BinEntry or UNKOWN VR:
1199 // When "this" is a Document the Key is simply of the
1200 // form ( group, elem )...
1201 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1204 newBinEntry->SetKey( newBinEntry->GetKey() );
1206 // but when "this" is a SQItem, we are inserting this new
1207 // valEntry in a sequence item, and the kay has the
1208 // generalized form (refer to \ref BaseTagKey):
1209 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1211 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1212 + newBinEntry->GetKey() );
1215 LoadDocEntry( newBinEntry );
1216 if( !set->AddEntry( newBinEntry ) )
1218 //Expect big troubles if here
1219 //delete newBinEntry;
1225 /////////////////////// ValEntry
1226 // When "set" is a Document, then we are at the top of the
1227 // hierarchy and the Key is simply of the form ( group, elem )...
1228 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1231 newValEntry->SetKey( newValEntry->GetKey() );
1233 // ...but when "set" is a SQItem, we are inserting this new
1234 // valEntry in a sequence item. Hence the key has the
1235 // generalized form (refer to \ref BaseTagKey):
1236 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1238 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1239 + newValEntry->GetKey() );
1242 LoadDocEntry( newValEntry );
1243 bool delimitor=newValEntry->IsItemDelimitor();
1244 if( !set->AddEntry( newValEntry ) )
1246 // If here expect big troubles
1247 //delete newValEntry; //otherwise mem leak
1257 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1265 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1266 && ( newDocEntry->GetElement() == 0x0010 ) )
1268 TransferSyntaxType ts = GetTransferSyntax();
1269 if ( ts == RLELossless )
1271 long positionOnEntry = Fp->tellg();
1272 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1274 Fp->seekg( positionOnEntry, std::ios::beg );
1276 else if ( IsJPEG() )
1278 long positionOnEntry = Fp->tellg();
1279 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1280 ComputeJPEGFragmentInfo();
1281 Fp->seekg( positionOnEntry, std::ios::beg );
1285 // Just to make sure we are at the beginning of next entry.
1286 SkipToNextDocEntry(newDocEntry);
1291 unsigned long l = newDocEntry->GetReadLength();
1292 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1294 if ( l == 0xffffffff )
1303 // no other way to create it ...
1304 newSeqEntry->SetDelimitorMode( delim_mode );
1306 // At the top of the hierarchy, stands a Document. When "set"
1307 // is a Document, then we are building the first depth level.
1308 // Hence the SeqEntry we are building simply has a depth
1310 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1313 newSeqEntry->SetDepthLevel( 1 );
1314 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1316 // But when "set" is allready a SQItem, we are building a nested
1317 // sequence, and hence the depth level of the new SeqEntry
1318 // we are building, is one level deeper:
1319 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1321 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1322 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1323 + newSeqEntry->GetKey() );
1327 { // Don't try to parse zero-length sequences
1328 ParseSQ( newSeqEntry,
1329 newDocEntry->GetOffset(),
1332 set->AddEntry( newSeqEntry );
1333 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1345 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1346 * @return parsed length for this level
1348 void Document::ParseSQ( SeqEntry *seqEntry,
1349 long offset, long l_max, bool delim_mode)
1351 int SQItemNumber = 0;
1356 DocEntry *newDocEntry = ReadNextDocEntry();
1359 // FIXME Should warn user
1364 if ( newDocEntry->IsSequenceDelimitor() )
1366 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1370 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1376 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1377 std::ostringstream newBase;
1378 newBase << seqEntry->GetKey()
1382 itemSQ->SetBaseTagKey( newBase.str() );
1383 unsigned int l = newDocEntry->GetReadLength();
1385 if ( l == 0xffffffff )
1394 ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1397 seqEntry->AddEntry( itemSQ, SQItemNumber );
1399 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1407 * \brief Loads the element content if its length doesn't exceed
1408 * the value specified with Document::SetMaxSizeLoadEntry()
1409 * @param entry Header Entry (Dicom Element) to be dealt with
1411 void Document::LoadDocEntry(DocEntry *entry)
1413 uint16_t group = entry->GetGroup();
1414 std::string vr = entry->GetVR();
1415 uint32_t length = entry->GetLength();
1417 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1419 // A SeQuence "contains" a set of Elements.
1420 // (fffe e000) tells us an Element is beginning
1421 // (fffe e00d) tells us an Element just ended
1422 // (fffe e0dd) tells us the current SeQuence just ended
1423 if( group == 0xfffe )
1425 // NO more value field for SQ !
1429 // When the length is zero things are easy:
1432 ((ValEntry *)entry)->SetValue("");
1436 // The elements whose length is bigger than the specified upper bound
1437 // are not loaded. Instead we leave a short notice of the offset of
1438 // the element content and it's length.
1440 std::ostringstream s;
1441 if (length > MaxSizeLoadEntry)
1443 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1445 //s << "gdcm::NotLoaded (BinEntry)";
1446 s << GDCM_NOTLOADED;
1447 s << " Address:" << (long)entry->GetOffset();
1448 s << " Length:" << entry->GetLength();
1449 s << " x(" << std::hex << entry->GetLength() << ")";
1450 binEntryPtr->SetValue(s.str());
1452 // Be carefull : a BinEntry IS_A ValEntry ...
1453 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1455 // s << "gdcm::NotLoaded. (ValEntry)";
1456 s << GDCM_NOTLOADED;
1457 s << " Address:" << (long)entry->GetOffset();
1458 s << " Length:" << entry->GetLength();
1459 s << " x(" << std::hex << entry->GetLength() << ")";
1460 valEntryPtr->SetValue(s.str());
1465 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1466 << "nor a ValEntry ?! Should never print that !" << std::endl;
1469 // to be sure we are at the end of the value ...
1470 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1475 // When we find a BinEntry not very much can be done :
1476 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1478 s << GDCM_BINLOADED;
1479 binEntryPtr->SetValue(s.str());
1480 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1484 /// \todo Any compacter code suggested (?)
1485 if ( IsDocEntryAnInteger(entry) )
1489 // When short integer(s) are expected, read and convert the following
1490 // n *two characters properly i.e. consider them as short integers as
1491 // opposed to strings.
1492 // Elements with Value Multiplicity > 1
1493 // contain a set of integers (not a single one)
1494 if (vr == "US" || vr == "SS")
1497 NewInt = ReadInt16();
1501 for (int i=1; i < nbInt; i++)
1504 NewInt = ReadInt16();
1509 // See above comment on multiple integers (mutatis mutandis).
1510 else if (vr == "UL" || vr == "SL")
1513 NewInt = ReadInt32();
1517 for (int i=1; i < nbInt; i++)
1520 NewInt = ReadInt32();
1525 #ifdef GDCM_NO_ANSI_STRING_STREAM
1526 s << std::ends; // to avoid oddities on Solaris
1527 #endif //GDCM_NO_ANSI_STRING_STREAM
1529 ((ValEntry *)entry)->SetValue(s.str());
1533 // FIXME: We need an additional byte for storing \0 that is not on disk
1534 char *str = new char[length+1];
1535 Fp->read(str, (size_t)length);
1536 str[length] = '\0'; //this is only useful when length is odd
1537 // Special DicomString call to properly handle \0 and even length
1538 std::string newValue;
1541 newValue = Util::DicomString(str, length+1);
1542 gdcmVerboseMacro("Warning: bad length: " << length );
1543 gdcmVerboseMacro("For string :" << newValue.c_str());
1544 // Since we change the length of string update it length
1545 //entry->SetReadLength(length+1);
1549 newValue = Util::DicomString(str, length);
1553 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1555 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1557 gdcmVerboseMacro( "Unread element value");
1558 valEntry->SetValue(GDCM_UNREAD);
1564 // Because of correspondance with the VR dic
1565 valEntry->SetValue(newValue);
1569 valEntry->SetValue(newValue);
1574 gdcmErrorMacro("Document::LoadDocEntry"
1575 "Should have a ValEntry, here !");
1581 * \brief Find the value Length of the passed Header Entry
1582 * @param entry Header Entry whose length of the value shall be loaded.
1584 void Document::FindDocEntryLength( DocEntry *entry )
1585 throw ( FormatError )
1587 uint16_t element = entry->GetElement();
1588 std::string vr = entry->GetVR();
1591 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1593 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1595 // The following reserved two bytes (see PS 3.5-2003, section
1596 // "7.1.2 Data element structure with explicit vr", p 27) must be
1597 // skipped before proceeding on reading the length on 4 bytes.
1598 Fp->seekg( 2L, std::ios::cur);
1599 uint32_t length32 = ReadInt32();
1601 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1606 lengthOB = FindDocEntryLengthOBOrOW();
1608 catch ( FormatUnexpected )
1610 // Computing the length failed (this happens with broken
1611 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1612 // chance to get the pixels by deciding the element goes
1613 // until the end of the file. Hence we artificially fix the
1614 // the length and proceed.
1615 long currentPosition = Fp->tellg();
1616 Fp->seekg(0L,std::ios::end);
1618 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1619 Fp->seekg(currentPosition, std::ios::beg);
1621 entry->SetReadLength(lengthUntilEOF);
1622 entry->SetLength(lengthUntilEOF);
1625 entry->SetReadLength(lengthOB);
1626 entry->SetLength(lengthOB);
1629 FixDocEntryFoundLength(entry, length32);
1633 // Length is encoded on 2 bytes.
1634 length16 = ReadInt16();
1636 // We can tell the current file is encoded in big endian (like
1637 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1638 // and it's value is the one of the encoding of a big endian file.
1639 // In order to deal with such big endian encoded files, we have
1640 // (at least) two strategies:
1641 // * when we load the "Transfer Syntax" tag with value of big endian
1642 // encoding, we raise the proper flags. Then we wait for the end
1643 // of the META group (0x0002) among which is "Transfer Syntax",
1644 // before switching the swap code to big endian. We have to postpone
1645 // the switching of the swap code since the META group is fully encoded
1646 // in little endian, and big endian coding only starts at the next
1647 // group. The corresponding code can be hard to analyse and adds
1648 // many additional unnecessary tests for regular tags.
1649 // * the second strategy consists in waiting for trouble, that shall
1650 // appear when we find the first group with big endian encoding. This
1651 // is easy to detect since the length of a "Group Length" tag (the
1652 // ones with zero as element number) has to be of 4 (0x0004). When we
1653 // encounter 1024 (0x0400) chances are the encoding changed and we
1654 // found a group with big endian encoding.
1655 // We shall use this second strategy. In order to make sure that we
1656 // can interpret the presence of an apparently big endian encoded
1657 // length of a "Group Length" without committing a big mistake, we
1658 // add an additional check: we look in the already parsed elements
1659 // for the presence of a "Transfer Syntax" whose value has to be "big
1660 // endian encoding". When this is the case, chances are we have got our
1661 // hands on a big endian encoded file: we switch the swap code to
1662 // big endian and proceed...
1663 if ( element == 0x0000 && length16 == 0x0400 )
1665 TransferSyntaxType ts = GetTransferSyntax();
1666 if ( ts != ExplicitVRBigEndian )
1668 throw FormatError( "Document::FindDocEntryLength()",
1669 " not explicit VR." );
1673 SwitchSwapToBigEndian();
1675 // Restore the unproperly loaded values i.e. the group, the element
1676 // and the dictionary entry depending on them.
1677 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1678 uint16_t correctElem = SwapShort( entry->GetElement() );
1679 DictEntry *newTag = GetDictEntryByNumber( correctGroup,
1683 // This correct tag is not in the dictionary. Create a new one.
1684 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1686 // FIXME this can create a memory leaks on the old entry that be
1687 // left unreferenced.
1688 entry->SetDictEntry( newTag );
1691 // Heuristic: well, some files are really ill-formed.
1692 if ( length16 == 0xffff)
1694 // 0xffff means that we deal with 'Unknown Length' Sequence
1697 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1702 // Either implicit VR or a non DICOM conformal (see note below) explicit
1703 // VR that ommited the VR of (at least) this element. Farts happen.
1704 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1705 // on Data elements "Implicit and Explicit VR Data Elements shall
1706 // not coexist in a Data Set and Data Sets nested within it".]
1707 // Length is on 4 bytes.
1709 FixDocEntryFoundLength( entry, ReadInt32() );
1715 * \brief Find the Value Representation of the current Dicom Element.
1716 * @return Value Representation of the current Entry
1718 std::string Document::FindDocEntryVR()
1720 if ( Filetype != ExplicitVR )
1721 return(GDCM_UNKNOWN);
1723 long positionOnEntry = Fp->tellg();
1724 // Warning: we believe this is explicit VR (Value Representation) because
1725 // we used a heuristic that found "UL" in the first tag. Alas this
1726 // doesn't guarantee that all the tags will be in explicit VR. In some
1727 // cases (see e-film filtered files) one finds implicit VR tags mixed
1728 // within an explicit VR file. Hence we make sure the present tag
1729 // is in explicit VR and try to fix things if it happens not to be
1733 Fp->read (vr, (size_t)2);
1736 if( !CheckDocEntryVR(vr) )
1738 Fp->seekg(positionOnEntry, std::ios::beg);
1739 return(GDCM_UNKNOWN);
1745 * \brief Check the correspondance between the VR of the header entry
1746 * and the taken VR. If they are different, the header entry is
1747 * updated with the new VR.
1748 * @param vr Dicom Value Representation
1749 * @return false if the VR is incorrect of if the VR isn't referenced
1750 * otherwise, it returns true
1752 bool Document::CheckDocEntryVR(VRKey vr)
1754 // CLEANME searching the dicom_vr at each occurence is expensive.
1755 // PostPone this test in an optional integrity check at the end
1756 // of parsing or only in debug mode.
1757 if ( !Global::GetVR()->IsValidVR(vr) )
1764 * \brief Get the transformed value of the header entry. The VR value
1765 * is used to define the transformation to operate on the value
1766 * \warning NOT end user intended method !
1767 * @param entry entry to tranform
1768 * @return Transformed entry value
1770 std::string Document::GetDocEntryValue(DocEntry *entry)
1772 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1774 std::string val = ((ValEntry *)entry)->GetValue();
1775 std::string vr = entry->GetVR();
1776 uint32_t length = entry->GetLength();
1777 std::ostringstream s;
1780 // When short integer(s) are expected, read and convert the following
1781 // n * 2 bytes properly i.e. as a multivaluated strings
1782 // (each single value is separated fromthe next one by '\'
1783 // as usual for standard multivaluated filels
1784 // Elements with Value Multiplicity > 1
1785 // contain a set of short integers (not a single one)
1787 if( vr == "US" || vr == "SS" )
1792 for (int i=0; i < nbInt; i++)
1798 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1799 newInt16 = SwapShort( newInt16 );
1804 // When integer(s) are expected, read and convert the following
1805 // n * 4 bytes properly i.e. as a multivaluated strings
1806 // (each single value is separated fromthe next one by '\'
1807 // as usual for standard multivaluated filels
1808 // Elements with Value Multiplicity > 1
1809 // contain a set of integers (not a single one)
1810 else if( vr == "UL" || vr == "SL" )
1815 for (int i=0; i < nbInt; i++)
1821 newInt32 = ( val[4*i+0] & 0xFF )
1822 + (( val[4*i+1] & 0xFF ) << 8 )
1823 + (( val[4*i+2] & 0xFF ) << 16 )
1824 + (( val[4*i+3] & 0xFF ) << 24 );
1825 newInt32 = SwapLong( newInt32 );
1829 #ifdef GDCM_NO_ANSI_STRING_STREAM
1830 s << std::ends; // to avoid oddities on Solaris
1831 #endif //GDCM_NO_ANSI_STRING_STREAM
1835 return ((ValEntry *)entry)->GetValue();
1839 * \brief Get the reverse transformed value of the header entry. The VR
1840 * value is used to define the reverse transformation to operate on
1842 * \warning NOT end user intended method !
1843 * @param entry Entry to reverse transform
1844 * @return Reverse transformed entry value
1846 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1848 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1850 std::string vr = entry->GetVR();
1851 std::vector<std::string> tokens;
1852 std::ostringstream s;
1854 if ( vr == "US" || vr == "SS" )
1858 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1859 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1860 for (unsigned int i=0; i<tokens.size(); i++)
1862 newInt16 = atoi(tokens[i].c_str());
1863 s << ( newInt16 & 0xFF )
1864 << (( newInt16 >> 8 ) & 0xFF );
1868 if ( vr == "UL" || vr == "SL")
1872 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1873 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1874 for (unsigned int i=0; i<tokens.size();i++)
1876 newInt32 = atoi(tokens[i].c_str());
1877 s << (char)( newInt32 & 0xFF )
1878 << (char)(( newInt32 >> 8 ) & 0xFF )
1879 << (char)(( newInt32 >> 16 ) & 0xFF )
1880 << (char)(( newInt32 >> 24 ) & 0xFF );
1885 #ifdef GDCM_NO_ANSI_STRING_STREAM
1886 s << std::ends; // to avoid oddities on Solaris
1887 #endif //GDCM_NO_ANSI_STRING_STREAM
1891 return ((ValEntry *)entry)->GetValue();
1895 * \brief Skip a given Header Entry
1896 * \warning NOT end user intended method !
1897 * @param entry entry to skip
1899 void Document::SkipDocEntry(DocEntry *entry)
1901 SkipBytes(entry->GetLength());
1905 * \brief Skips to the begining of the next Header Entry
1906 * \warning NOT end user intended method !
1907 * @param offset start of skipping
1908 * @param readLgth length to skip
1911 void Document::SkipToNextDocEntry(DocEntry *newDocEntry)
1913 Fp->seekg((long)(newDocEntry->GetOffset()), std::ios::beg);
1914 Fp->seekg( (long)(newDocEntry->GetReadLength()),std::ios::cur);
1918 * \brief When the length of an element value is obviously wrong (because
1919 * the parser went Jabberwocky) one can hope improving things by
1920 * applying some heuristics.
1921 * @param entry entry to check
1922 * @param foundLength fist assumption about length
1924 void Document::FixDocEntryFoundLength(DocEntry *entry,
1925 uint32_t foundLength)
1927 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1928 if ( foundLength == 0xffffffff)
1933 uint16_t gr = entry->GetGroup();
1934 uint16_t elem = entry->GetElement();
1936 if ( foundLength % 2)
1938 gdcmVerboseMacro( "Warning : Tag with uneven length " << foundLength
1939 << " in x(" << std::hex << gr << "," << elem <<")");
1942 //////// Fix for some naughty General Electric images.
1943 // Allthough not recent many such GE corrupted images are still present
1944 // on Creatis hard disks. Hence this fix shall remain when such images
1945 // are no longer in use (we are talking a few years, here)...
1946 // Note: XMedCom probably uses such a trick since it is able to read
1947 // those pesky GE images ...
1948 if ( foundLength == 13)
1950 // Only happens for this length !
1951 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1954 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
1958 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1959 // Occurence of such images is quite low (unless one leaves close to a
1960 // 'Leonardo' source. Hence, one might consider commenting out the
1961 // following fix on efficiency reasons.
1962 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1965 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
1968 else if ( entry->GetVR() == "SQ" )
1970 foundLength = 0; // ReadLength is unchanged
1973 //////// We encountered a 'delimiter' element i.e. a tag of the form
1974 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1975 // taken into account.
1976 else if( gr == 0xfffe )
1978 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1979 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1980 // causes extra troubles...
1981 if( entry->GetElement() != 0x0000 )
1987 entry->SetLength(foundLength);
1991 * \brief Apply some heuristics to predict whether the considered
1992 * element value contains/represents an integer or not.
1993 * @param entry The element value on which to apply the predicate.
1994 * @return The result of the heuristical predicate.
1996 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1998 uint16_t element = entry->GetElement();
1999 uint16_t group = entry->GetGroup();
2000 const std::string &vr = entry->GetVR();
2001 uint32_t length = entry->GetLength();
2003 // When we have some semantics on the element we just read, and if we
2004 // a priori know we are dealing with an integer, then we shall be
2005 // able to swap it's element value properly.
2006 if ( element == 0 ) // This is the group length of the group
2014 // Allthough this should never happen, still some images have a
2015 // corrupted group length [e.g. have a glance at offset x(8336) of
2016 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2017 // Since for dicom compliant and well behaved headers, the present
2018 // test is useless (and might even look a bit paranoid), when we
2019 // encounter such an ill-formed image, we simply display a warning
2020 // message and proceed on parsing (while crossing fingers).
2021 long filePosition = Fp->tellg();
2022 gdcmVerboseMacro( "Erroneous Group Length element length on : ("
2023 << std::hex << group << " , " << element
2024 << ") -before- position x(" << filePosition << ")"
2025 << "lgt : " << length );
2029 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2037 * \brief Find the Length till the next sequence delimiter
2038 * \warning NOT end user intended method !
2042 uint32_t Document::FindDocEntryLengthOBOrOW()
2043 throw( FormatUnexpected )
2045 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2046 long positionOnEntry = Fp->tellg();
2047 bool foundSequenceDelimiter = false;
2048 uint32_t totalLength = 0;
2050 while ( !foundSequenceDelimiter )
2056 group = ReadInt16();
2059 catch ( FormatError )
2061 throw FormatError("Document::FindDocEntryLengthOBOrOW()",
2062 " group or element not present.");
2065 // We have to decount the group and element we just read
2068 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2070 gdcmVerboseMacro( "Neither an Item tag nor a Sequence delimiter tag.");
2071 Fp->seekg(positionOnEntry, std::ios::beg);
2072 throw FormatUnexpected( "Neither an Item tag nor a Sequence delimiter tag.");
2075 if ( elem == 0xe0dd )
2077 foundSequenceDelimiter = true;
2080 uint32_t itemLength = ReadInt32();
2081 // We add 4 bytes since we just read the ItemLength with ReadInt32
2082 totalLength += itemLength + 4;
2083 SkipBytes(itemLength);
2085 if ( foundSequenceDelimiter )
2090 Fp->seekg( positionOnEntry, std::ios::beg);
2095 * \brief Reads a supposed to be 16 Bits integer
2096 * (swaps it depending on processor endianity)
2097 * @return read value
2099 uint16_t Document::ReadInt16()
2100 throw( FormatError )
2103 Fp->read ((char*)&g, (size_t)2);
2106 throw FormatError( "Document::ReadInt16()", " file error." );
2110 throw FormatError( "Document::ReadInt16()", "EOF." );
2117 * \brief Reads a supposed to be 32 Bits integer
2118 * (swaps it depending on processor endianity)
2119 * @return read value
2121 uint32_t Document::ReadInt32()
2122 throw( FormatError )
2125 Fp->read ((char*)&g, (size_t)4);
2128 throw FormatError( "Document::ReadInt32()", " file error." );
2132 throw FormatError( "Document::ReadInt32()", "EOF." );
2139 * \brief skips bytes inside the source file
2140 * \warning NOT end user intended method !
2143 void Document::SkipBytes(uint32_t nBytes)
2145 //FIXME don't dump the returned value
2146 Fp->seekg((long)nBytes, std::ios::cur);
2150 * \brief Loads all the needed Dictionaries
2151 * \warning NOT end user intended method !
2153 void Document::Initialise()
2155 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2157 RLEInfo = new RLEFramesInfo;
2158 JPEGInfo = new JPEGFragmentsInfo;
2163 * \brief Discover what the swap code is (among little endian, big endian,
2164 * bad little endian, bad big endian).
2166 * @return false when we are absolutely sure
2167 * it's neither ACR-NEMA nor DICOM
2168 * true when we hope ours assuptions are OK
2170 bool Document::CheckSwap()
2172 // The only guaranted way of finding the swap code is to find a
2173 // group tag since we know it's length has to be of four bytes i.e.
2174 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2175 // occurs when we can't find such group...
2177 uint32_t x = 4; // x : for ntohs
2178 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2184 // First, compare HostByteOrder and NetworkByteOrder in order to
2185 // determine if we shall need to swap bytes (i.e. the Endian type).
2186 if ( x == ntohs(x) )
2195 // The easiest case is the one of a DICOM header, since it possesses a
2196 // file preamble where it suffice to look for the string "DICM".
2199 char *entCur = deb + 128;
2200 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2202 gdcmVerboseMacro( "Looks like DICOM Version3" );
2204 // Next, determine the value representation (VR). Let's skip to the
2205 // first element (0002, 0000) and check there if we find "UL"
2206 // - or "OB" if the 1st one is (0002,0001) -,
2207 // in which case we (almost) know it is explicit VR.
2208 // WARNING: if it happens to be implicit VR then what we will read
2209 // is the length of the group. If this ascii representation of this
2210 // length happens to be "UL" then we shall believe it is explicit VR.
2211 // FIXME: in order to fix the above warning, we could read the next
2212 // element value (or a couple of elements values) in order to make
2213 // sure we are not commiting a big mistake.
2214 // We need to skip :
2215 // * the 128 bytes of File Preamble (often padded with zeroes),
2216 // * the 4 bytes of "DICM" string,
2217 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2218 // i.e. a total of 136 bytes.
2222 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2223 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2224 // *Implicit* VR. -and it is !-
2226 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2227 memcmp(entCur, "OB", (size_t)2) == 0 ||
2228 memcmp(entCur, "UI", (size_t)2) == 0 ||
2229 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2230 // when Write DCM *adds*
2232 // Use Document::dicom_vr to test all the possibilities
2233 // instead of just checking for UL, OB and UI !? group 0000
2235 Filetype = ExplicitVR;
2236 gdcmVerboseMacro( "Explicit Value Representation");
2240 Filetype = ImplicitVR;
2241 gdcmVerboseMacro( "Not an explicit Value Representation");
2247 gdcmVerboseMacro( "HostByteOrder != NetworkByteOrder");
2252 gdcmVerboseMacro( "HostByteOrder = NetworkByteOrder");
2255 // Position the file position indicator at first tag (i.e.
2256 // after the file preamble and the "DICM" string).
2257 Fp->seekg(0, std::ios::beg);
2258 Fp->seekg ( 132L, std::ios::beg);
2262 // Alas, this is not a DicomV3 file and whatever happens there is no file
2263 // preamble. We can reset the file position indicator to where the data
2264 // is (i.e. the beginning of the file).
2265 gdcmVerboseMacro( "Not a DICOM Version3 file");
2266 Fp->seekg(0, std::ios::beg);
2268 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2269 // By clean we mean that the length of the first tag is written down.
2270 // If this is the case and since the length of the first group HAS to be
2271 // four (bytes), then determining the proper swap code is straightforward.
2274 // We assume the array of char we are considering contains the binary
2275 // representation of a 32 bits integer. Hence the following dirty
2277 s32 = *((uint32_t *)(entCur));
2298 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2299 // It is time for despaired wild guesses.
2300 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2301 // i.e. the 'group length' element is not present :
2303 // check the supposed-to-be 'group number'
2304 // in ( 0x0001 .. 0x0008 )
2305 // to determine ' SwapCode' value .
2306 // Only 0 or 4321 will be possible
2307 // (no oportunity to check for the formerly well known
2308 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2309 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2310 // the file IS NOT ACR-NEMA nor DICOM V3
2311 // Find a trick to tell it the caller...
2313 s16 = *((uint16_t *)(deb));
2340 gdcmVerboseMacro( "ACR/NEMA unfound swap info (Really hopeless !)");
2344 // Then the only info we have is the net2host one.
2356 * \brief Restore the unproperly loaded values i.e. the group, the element
2357 * and the dictionary entry depending on them.
2359 void Document::SwitchSwapToBigEndian()
2361 gdcmVerboseMacro( "Switching to BigEndian mode.");
2362 if ( SwapCode == 0 )
2366 else if ( SwapCode == 4321 )
2370 else if ( SwapCode == 3412 )
2374 else if ( SwapCode == 2143 )
2381 * \brief during parsing, Header Elements too long are not loaded in memory
2384 void Document::SetMaxSizeLoadEntry(long newSize)
2390 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2392 MaxSizeLoadEntry = 0xffffffff;
2395 MaxSizeLoadEntry = newSize;
2400 * \brief Header Elements too long will not be printed
2401 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2404 void Document::SetMaxSizePrintEntry(long newSize)
2406 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2411 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2413 MaxSizePrintEntry = 0xffffffff;
2416 MaxSizePrintEntry = newSize;
2422 * \brief Handle broken private tag from Philips NTSCAN
2423 * where the endianess is being switch to BigEndian for no
2427 void Document::HandleBrokenEndian(uint16_t group, uint16_t elem)
2429 // Endian reversion. Some files contain groups of tags with reversed endianess.
2430 static int reversedEndian = 0;
2431 // try to fix endian switching in the middle of headers
2432 if ((group == 0xfeff) && (elem == 0x00e0))
2434 // start endian swap mark for group found
2436 SwitchSwapToBigEndian();
2441 else if ((group == 0xfffe) && (elem == 0xe00d) && reversedEndian)
2443 // end of reversed endian group
2445 SwitchSwapToBigEndian();
2450 * \brief Read the next tag but WITHOUT loading it's value
2451 * (read the 'Group Number', the 'Element Number',
2452 * gets the Dict Entry
2453 * gets the VR, gets the length, gets the offset value)
2454 * @return On succes the newly created DocEntry, NULL on failure.
2456 DocEntry *Document::ReadNextDocEntry()
2463 group = ReadInt16();
2466 catch ( FormatError e )
2468 // We reached the EOF (or an error occured) therefore
2469 // header parsing has to be considered as finished.
2474 HandleBrokenEndian(group, elem);
2475 std::string vr = FindDocEntryVR();
2476 std::string realVR = vr;
2478 if( vr == GDCM_UNKNOWN)
2480 DictEntry *dictEntry = GetDictEntryByNumber(group,elem);
2482 realVR = dictEntry->GetVR();
2486 if( Global::GetVR()->IsVROfSequence(realVR) )
2487 newEntry = NewSeqEntryByNumber(group, elem);
2488 else if( Global::GetVR()->IsVROfStringRepresentable(realVR) )
2489 newEntry = NewValEntryByNumber(group, elem,vr);
2491 newEntry = NewBinEntryByNumber(group, elem,vr);
2493 if( vr == GDCM_UNKNOWN )
2495 if( Filetype == ExplicitVR )
2497 // We thought this was explicit VR, but we end up with an
2498 // implicit VR tag. Let's backtrack.
2500 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
2501 newEntry->GetGroup(), newEntry->GetElement());
2502 gdcmVerboseMacro( msg.c_str() );
2504 newEntry->SetImplicitVR();
2509 FindDocEntryLength(newEntry);
2511 catch ( FormatError e )
2519 newEntry->SetOffset(Fp->tellg());
2526 * \brief Generate a free TagKey i.e. a TagKey that is not present
2527 * in the TagHt dictionary.
2528 * @param group The generated tag must belong to this group.
2529 * @return The element of tag with given group which is fee.
2531 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2533 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2535 TagKey key = DictEntry::TranslateToKey(group, elem);
2536 if (TagHT.count(key) == 0)
2545 * \brief Assuming the internal file pointer \ref Document::Fp
2546 * is placed at the beginning of a tag check whether this
2547 * tag is (TestGroup, TestElement).
2548 * \warning On success the internal file pointer \ref Document::Fp
2549 * is modified to point after the tag.
2550 * On failure (i.e. when the tag wasn't the expected tag
2551 * (TestGroup, TestElement) the internal file pointer
2552 * \ref Document::Fp is restored to it's original position.
2553 * @param testGroup The expected group of the tag.
2554 * @param testElement The expected Element of the tag.
2555 * @return True on success, false otherwise.
2557 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2559 long positionOnEntry = Fp->tellg();
2560 long currentPosition = Fp->tellg(); // On debugging purposes
2562 //// Read the Item Tag group and element, and make
2563 // sure they are what we expected:
2564 uint16_t itemTagGroup;
2565 uint16_t itemTagElement;
2568 itemTagGroup = ReadInt16();
2569 itemTagElement = ReadInt16();
2571 catch ( FormatError e )
2573 //std::cerr << e << std::endl;
2576 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2578 gdcmVerboseMacro( "Wrong Item Tag found:"
2579 << " We should have found tag ("
2580 << std::hex << testGroup << "," << testElement << ")" << std::endl
2581 << " but instead we encountered tag ("
2582 << std::hex << itemTagGroup << "," << itemTagElement << ")"
2584 << " at address: " << (unsigned int)currentPosition );
2585 Fp->seekg(positionOnEntry, std::ios::beg);
2593 * \brief Assuming the internal file pointer \ref Document::Fp
2594 * is placed at the beginning of a tag (TestGroup, TestElement),
2595 * read the length associated to the Tag.
2596 * \warning On success the internal file pointer \ref Document::Fp
2597 * is modified to point after the tag and it's length.
2598 * On failure (i.e. when the tag wasn't the expected tag
2599 * (TestGroup, TestElement) the internal file pointer
2600 * \ref Document::Fp is restored to it's original position.
2601 * @param testGroup The expected group of the tag.
2602 * @param testElement The expected Element of the tag.
2603 * @return On success returns the length associated to the tag. On failure
2606 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2608 long positionOnEntry = Fp->tellg();
2609 (void)positionOnEntry;
2611 if ( !ReadTag(testGroup, testElement) )
2616 //// Then read the associated Item Length
2617 long currentPosition = Fp->tellg();
2618 uint32_t itemLength = ReadInt32();
2620 gdcmVerboseMacro( "Basic Item Length is: "
2621 << itemLength << std::endl
2622 << " at address: " << (unsigned int)currentPosition);
2628 * \brief When parsing the Pixel Data of an encapsulated file, read
2629 * the basic offset table (when present, and BTW dump it).
2631 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2633 //// Read the Basic Offset Table Item Tag length...
2634 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2636 // When present, read the basic offset table itself.
2637 // Notes: - since the presence of this basic offset table is optional
2638 // we can't rely on it for the implementation, and we will simply
2639 // trash it's content (when present).
2640 // - still, when present, we could add some further checks on the
2641 // lengths, but we won't bother with such fuses for the time being.
2642 if ( itemLength != 0 )
2644 char *basicOffsetTableItemValue = new char[itemLength + 1];
2645 Fp->read(basicOffsetTableItemValue, itemLength);
2648 for (unsigned int i=0; i < itemLength; i += 4 )
2650 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2652 gdcmVerboseMacro( "Read one length: " <<
2653 std::hex << individualLength );
2657 delete[] basicOffsetTableItemValue;
2662 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2663 * Compute the RLE extra information and store it in \ref RLEInfo
2664 * for later pixel retrieval usage.
2666 void Document::ComputeRLEInfo()
2668 TransferSyntaxType ts = GetTransferSyntax();
2669 if ( ts != RLELossless )
2674 // Encoded pixel data: for the time being we are only concerned with
2675 // Jpeg or RLE Pixel data encodings.
2676 // As stated in PS 3.5-2003, section 8.2 p44:
2677 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2678 // value representation OB is used".
2679 // Hence we expect an OB value representation. Concerning OB VR,
2680 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2681 // "For the Value Representations OB and OW, the encoding shall meet the
2682 // following specifications depending on the Data element tag:"
2684 // - the first item in the sequence of items before the encoded pixel
2685 // data stream shall be basic offset table item. The basic offset table
2686 // item value, however, is not required to be present"
2688 ReadAndSkipEncapsulatedBasicOffsetTable();
2690 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2691 // Loop on the individual frame[s] and store the information
2692 // on the RLE fragments in a RLEFramesInfo.
2693 // Note: - when only a single frame is present, this is a
2695 // - when more than one frame are present, then we are in
2696 // the case of a multi-frame image.
2698 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2700 // Parse the RLE Header and store the corresponding RLE Segment
2701 // Offset Table information on fragments of this current Frame.
2702 // Note that the fragment pixels themselves are not loaded
2703 // (but just skipped).
2704 long frameOffset = Fp->tellg();
2706 uint32_t nbRleSegments = ReadInt32();
2707 if ( nbRleSegments > 16 )
2709 // There should be at most 15 segments (refer to RLEFrame class)
2710 gdcmVerboseMacro( "Too many segments.");
2713 uint32_t rleSegmentOffsetTable[16];
2714 for( int k = 1; k <= 15; k++ )
2716 rleSegmentOffsetTable[k] = ReadInt32();
2719 // Deduce from both the RLE Header and the frameLength the
2720 // fragment length, and again store this info in a
2722 long rleSegmentLength[15];
2723 // skipping (not reading) RLE Segments
2724 if ( nbRleSegments > 1)
2726 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2728 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2729 - rleSegmentOffsetTable[k];
2730 SkipBytes(rleSegmentLength[k]);
2734 rleSegmentLength[nbRleSegments] = frameLength
2735 - rleSegmentOffsetTable[nbRleSegments];
2736 SkipBytes(rleSegmentLength[nbRleSegments]);
2738 // Store the collected info
2739 RLEFrame *newFrameInfo = new RLEFrame;
2740 newFrameInfo->NumberFragments = nbRleSegments;
2741 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2743 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2744 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2746 RLEInfo->Frames.push_back( newFrameInfo );
2749 // Make sure that at the end of the item we encounter a 'Sequence
2751 if ( !ReadTag(0xfffe, 0xe0dd) )
2753 gdcmVerboseMacro( "No sequence delimiter item at end of RLE item sequence");
2758 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2759 * Compute the jpeg extra information (fragment[s] offset[s] and
2760 * length) and store it[them] in \ref JPEGInfo for later pixel
2763 void Document::ComputeJPEGFragmentInfo()
2765 // If you need to, look for comments of ComputeRLEInfo().
2771 ReadAndSkipEncapsulatedBasicOffsetTable();
2773 // Loop on the fragments[s] and store the parsed information in a
2775 long fragmentLength;
2776 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2778 long fragmentOffset = Fp->tellg();
2780 // Store the collected info
2781 JPEGFragment *newFragment = new JPEGFragment;
2782 newFragment->Offset = fragmentOffset;
2783 newFragment->Length = fragmentLength;
2784 JPEGInfo->Fragments.push_back( newFragment );
2786 SkipBytes( fragmentLength );
2789 // Make sure that at the end of the item we encounter a 'Sequence
2791 if ( !ReadTag(0xfffe, 0xe0dd) )
2793 gdcmVerboseMacro( "No sequence delimiter item at end of JPEG item sequence");
2798 * \brief Walk recursively the given \ref DocEntrySet, and feed
2799 * the given hash table (\ref TagDocEntryHT) with all the
2800 * \ref DocEntry (Dicom entries) encountered.
2801 * This method does the job for \ref BuildFlatHashTable.
2802 * @param builtHT Where to collect all the \ref DocEntry encountered
2803 * when recursively walking the given set.
2804 * @param set The structure to be traversed (recursively).
2806 void Document::BuildFlatHashTableRecurse( TagDocEntryHT &builtHT,
2809 if (ElementSet *elementSet = dynamic_cast< ElementSet* > ( set ) )
2811 TagDocEntryHT const ¤tHT = elementSet->GetTagHT();
2812 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2813 i != currentHT.end();
2816 DocEntry *entry = i->second;
2817 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2819 const ListSQItem& items = seqEntry->GetSQItems();
2820 for( ListSQItem::const_iterator item = items.begin();
2821 item != items.end();
2824 BuildFlatHashTableRecurse( builtHT, *item );
2828 builtHT[entry->GetKey()] = entry;
2833 if (SQItem *SQItemSet = dynamic_cast< SQItem* > ( set ) )
2835 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
2836 for (ListDocEntry::const_iterator i = currentList.begin();
2837 i != currentList.end();
2840 DocEntry *entry = *i;
2841 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2843 const ListSQItem& items = seqEntry->GetSQItems();
2844 for( ListSQItem::const_iterator item = items.begin();
2845 item != items.end();
2848 BuildFlatHashTableRecurse( builtHT, *item );
2852 builtHT[entry->GetKey()] = entry;
2859 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2862 * The structure used by a Document (through \ref ElementSet),
2863 * in order to hold the parsed entries of a Dicom header, is a recursive
2864 * one. This is due to the fact that the sequences (when present)
2865 * can be nested. Additionaly, the sequence items (represented in
2866 * gdcm as \ref SQItem) add an extra complexity to the data
2867 * structure. Hence, a gdcm user whishing to visit all the entries of
2868 * a Dicom header will need to dig in the gdcm internals (which
2869 * implies exposing all the internal data structures to the API).
2870 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2871 * recursively builds a temporary hash table, which holds all the
2872 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
2874 * \warning Of course there is NO integrity constrain between the
2875 * returned \ref TagDocEntryHT and the \ref ElementSet used
2876 * to build it. Hence if the underlying \ref ElementSet is
2877 * altered, then it is the caller responsability to invoke
2878 * \ref BuildFlatHashTable again...
2879 * @return The flat std::map<> we juste build.
2881 TagDocEntryHT *Document::BuildFlatHashTable()
2883 TagDocEntryHT *FlatHT = new TagDocEntryHT;
2884 BuildFlatHashTableRecurse( *FlatHT, this );
2891 * \brief Compares two documents, according to \ref DicomDir rules
2892 * \warning Does NOT work with ACR-NEMA files
2893 * \todo Find a trick to solve the pb (use RET fields ?)
2895 * @return true if 'smaller'
2897 bool Document::operator<(Document &document)
2900 std::string s1 = GetEntryByNumber(0x0010,0x0010);
2901 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
2913 s1 = GetEntryByNumber(0x0010,0x0020);
2914 s2 = document.GetEntryByNumber(0x0010,0x0020);
2925 // Study Instance UID
2926 s1 = GetEntryByNumber(0x0020,0x000d);
2927 s2 = document.GetEntryByNumber(0x0020,0x000d);
2938 // Serie Instance UID
2939 s1 = GetEntryByNumber(0x0020,0x000e);
2940 s2 = document.GetEntryByNumber(0x0020,0x000e);
2957 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
2958 * @param filetype Type of the File to be written
2960 int Document::ComputeGroup0002Length( FileType filetype )
2965 int groupLength = 0;
2966 bool found0002 = false;
2968 // for each zero-level Tag in the DCM Header
2972 entry = GetNextEntry();
2975 gr = entry->GetGroup();
2981 el = entry->GetElement();
2982 vr = entry->GetVR();
2984 if (filetype == ExplicitVR)
2986 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
2988 groupLength += 4; // explicit VR AND OB, OW, SQ : 4 more bytes
2991 groupLength += 2 + 2 + 4 + entry->GetLength();
2993 else if (found0002 )
2996 entry = GetNextEntry();
3001 } // end namespace gdcm
3003 //-----------------------------------------------------------------------------