1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2004/11/10 16:13:18 $
7 Version: $Revision: 1.120 $
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"
31 #if defined(_MSC_VER) || defined(__BORLANDC__)
34 #include <netinet/in.h>
39 static const char *TransferSyntaxStrings[] = {
40 // Implicit VR Little Endian
42 // Explicit VR Little Endian
43 "1.2.840.10008.1.2.1",
44 // Deflated Explicit VR Little Endian
45 "1.2.840.10008.1.2.1.99",
46 // Explicit VR Big Endian
47 "1.2.840.10008.1.2.2",
48 // JPEG Baseline (Process 1)
49 "1.2.840.10008.1.2.4.50",
50 // JPEG Extended (Process 2 & 4)
51 "1.2.840.10008.1.2.4.51",
52 // JPEG Extended (Process 3 & 5)
53 "1.2.840.10008.1.2.4.52",
54 // JPEG Spectral Selection, Non-Hierarchical (Process 6 & 8)
55 "1.2.840.10008.1.2.4.53",
56 // JPEG Full Progression, Non-Hierarchical (Process 10 & 12)
57 "1.2.840.10008.1.2.4.55",
58 // JPEG Lossless, Non-Hierarchical (Process 14)
59 "1.2.840.10008.1.2.4.57",
60 // JPEG Lossless, Hierarchical, First-Order Prediction (Process 14, [Selection Value 1])
61 "1.2.840.10008.1.2.4.70",
63 "1.2.840.10008.1.2.4.90",
65 "1.2.840.10008.1.2.4.91",
67 "1.2.840.10008.1.2.5",
69 "Unknown Transfer Syntax"
72 //-----------------------------------------------------------------------------
73 // Refer to Document::CheckSwap()
74 const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
76 // Refer to Document::SetMaxSizeLoadEntry()
77 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
78 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
80 //-----------------------------------------------------------------------------
81 // Constructor / Destructor
85 * @param filename file to be opened for parsing
87 Document::Document( std::string const & filename ) : ElementSet(-1)
89 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
99 dbg.Verbose(0, "Document::Document: starting parsing of file: ",
101 Fp->seekg( 0, std::ios_base::beg);
103 Fp->seekg(0, std::ios_base::end);
104 long lgt = Fp->tellg();
106 Fp->seekg( 0, std::ios_base::beg);
108 long beg = Fp->tellg();
111 ParseDES( this, beg, lgt, false); // le Load sera fait a la volee
113 Fp->seekg( 0, std::ios_base::beg);
115 // Load 'non string' values
117 std::string PhotometricInterpretation = GetEntryByNumber(0x0028,0x0004);
118 if( PhotometricInterpretation == "PALETTE COLOR " )
120 LoadEntryBinArea(0x0028,0x1200); // gray LUT
121 /// FIXME FIXME FIXME
122 /// The tags refered by the three following lines used to be CORRECTLY
123 /// defined as having an US Value Representation in the public
124 /// dictionnary. BUT the semantics implied by the three following
125 /// lines state that the corresponding tag contents are in fact
126 /// the ones of a BinEntry.
127 /// In order to fix things "Quick and Dirty" the dictionnary was
128 /// altered on PURPOUS but now contains a WRONG value.
129 /// In order to fix things and restore the dictionary to its
130 /// correct value, one needs to decided of the semantics by deciding
131 /// wether the following tags are either:
132 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
133 /// also used as BinEntry, which requires the proper conversion,
134 /// - OW, and hence loaded as BinEntry, but afterwards also used
135 /// as ValEntry, which requires the proper conversion.
136 LoadEntryBinArea(0x0028,0x1201); // R LUT
137 LoadEntryBinArea(0x0028,0x1202); // G LUT
138 LoadEntryBinArea(0x0028,0x1203); // B LUT
140 // Segmented Red Palette Color LUT Data
141 LoadEntryBinArea(0x0028,0x1221);
142 // Segmented Green Palette Color LUT Data
143 LoadEntryBinArea(0x0028,0x1222);
144 // Segmented Blue Palette Color LUT Data
145 LoadEntryBinArea(0x0028,0x1223);
147 //FIXME later : how to use it?
148 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
152 // --------------------------------------------------------------
153 // Specific code to allow gdcm to read ACR-LibIDO formated images
154 // Note: ACR-LibIDO is an extension of the ACR standard that was
155 // used at CREATIS. For the time being (say a couple years)
156 // we keep this kludge to allow a smooth move to gdcm for
157 // CREATIS developpers (sorry folks).
159 // if recognition code tells us we deal with a LibIDO image
160 // we switch lineNumber and columnNumber
163 RecCode = GetEntryByNumber(0x0008, 0x0010); // recognition code
164 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
165 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
166 // with "little-endian strings"
168 Filetype = ACR_LIBIDO;
169 std::string rows = GetEntryByNumber(0x0028, 0x0010);
170 std::string columns = GetEntryByNumber(0x0028, 0x0011);
171 SetEntryByNumber(columns, 0x0028, 0x0010);
172 SetEntryByNumber(rows , 0x0028, 0x0011);
174 // ----------------- End of ACR-LibIDO kludge ------------------
176 PrintLevel = 1; // 'Medium' print level by default
180 * \brief This default constructor doesn't parse the file. You should
181 * then invoke \ref Document::SetFileName and then the parsing.
183 Document::Document() : ElementSet(-1)
187 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
189 PrintLevel = 1; // 'Medium' print level by default
193 * \brief Canonical destructor.
195 Document::~Document ()
200 // Recursive clean up of sequences
201 for (TagDocEntryHT::const_iterator it = TagHT.begin();
202 it != TagHT.end(); ++it )
204 //delete it->second; //temp remove
211 //-----------------------------------------------------------------------------
215 * \brief Prints The Dict Entries of THE public Dicom Dictionary
218 void Document::PrintPubDict(std::ostream & os)
220 RefPubDict->Print(os);
224 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
227 void Document::PrintShaDict(std::ostream & os)
229 RefShaDict->Print(os);
232 //-----------------------------------------------------------------------------
235 * \brief Get the public dictionary used
237 Dict* Document::GetPubDict()
243 * \brief Get the shadow dictionary used
245 Dict* Document::GetShaDict()
251 * \brief Set the shadow dictionary used
252 * \param dict dictionary to use in shadow
254 bool Document::SetShaDict(Dict *dict)
261 * \brief Set the shadow dictionary used
262 * \param dictName name of the dictionary to use in shadow
264 bool Document::SetShaDict(DictKey const & dictName)
266 RefShaDict = Global::GetDicts()->GetDict(dictName);
271 * \brief This predicate, based on hopefully reasonable heuristics,
272 * decides whether or not the current Document was properly parsed
273 * and contains the mandatory information for being considered as
274 * a well formed and usable Dicom/Acr File.
275 * @return true when Document is the one of a reasonable Dicom/Acr file,
278 bool Document::IsReadable()
280 if( Filetype == Unknown)
282 dbg.Verbose(0, "Document::IsReadable: wrong filetype");
288 dbg.Verbose(0, "Document::IsReadable: no tags in internal"
298 * \brief Internal function that checks whether the Transfer Syntax given
299 * as argument is the one present in the current document.
300 * @param syntaxToCheck The transfert syntax we need to check against.
301 * @return True when SyntaxToCheck corresponds to the Transfer Syntax of
302 * the current document. False either when the document contains
303 * no Transfer Syntax, or when the Tranfer Syntaxes doesn't match.
305 TransferSyntaxType Document::GetTransferSyntax()
307 DocEntry *entry = GetDocEntryByNumber(0x0002, 0x0010);
313 // The entry might be present but not loaded (parsing and loading
314 // happen at different stages): try loading and proceed with check...
315 LoadDocEntrySafe(entry);
316 if (ValEntry* valEntry = dynamic_cast< ValEntry* >(entry) )
318 std::string transfer = valEntry->GetValue();
319 // The actual transfer (as read from disk) might be padded. We
320 // first need to remove the potential padding. We can make the
321 // weak assumption that padding was not executed with digits...
322 if ( transfer.length() == 0 )
324 // for brain damaged headers
327 while ( !isdigit(transfer[transfer.length()-1]) )
329 transfer.erase(transfer.length()-1, 1);
331 for (int i = 0; TransferSyntaxStrings[i] != NULL; i++)
333 if ( TransferSyntaxStrings[i] == transfer )
335 return TransferSyntaxType(i);
342 bool Document::IsJPEGLossless()
344 TransferSyntaxType r = GetTransferSyntax();
345 return r == JPEGFullProgressionProcess10_12
346 || r == JPEGLosslessProcess14
347 || r == JPEGLosslessProcess14_1;
351 * \brief Determines if the Transfer Syntax was already encountered
352 * and if it corresponds to a JPEG2000 one
353 * @return True when JPEG2000 (Lossly or LossLess) found. False in all
356 bool Document::IsJPEG2000()
358 TransferSyntaxType r = GetTransferSyntax();
359 return r == JPEG2000Lossless || r == JPEG2000;
363 * \brief Determines if the Transfer Syntax corresponds to any form
364 * of Jpeg encoded Pixel data.
365 * @return True when any form of JPEG found. False otherwise.
367 bool Document::IsJPEG()
369 TransferSyntaxType r = GetTransferSyntax();
370 return r == JPEGBaselineProcess1
371 || r == JPEGExtendedProcess2_4
372 || r == JPEGExtendedProcess3_5
373 || r == JPEGSpectralSelectionProcess6_8
379 * \brief Determines if the Transfer Syntax corresponds to encapsulated
380 * of encoded Pixel Data (as opposed to native).
381 * @return True when encapsulated. False when native.
383 bool Document::IsEncapsulate()
385 TransferSyntaxType r = GetTransferSyntax();
386 return IsJPEG() || r == RLELossless;
390 * \brief Predicate for dicom version 3 file.
391 * @return True when the file is a dicom version 3.
393 bool Document::IsDicomV3()
395 // Checking if Transfert Syntax exists is enough
396 // Anyway, it's to late check if the 'Preamble' was found ...
397 // And ... would it be a rich idea to check ?
398 // (some 'no Preamble' DICOM images exist !)
399 return GetDocEntryByNumber(0x0002, 0x0010) != NULL;
403 * \brief returns the File Type
404 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
405 * @return the FileType code
407 FileType Document::GetFileType()
413 * \brief Tries to open the file \ref Document::Filename and
414 * checks the preamble when existing.
415 * @return The FILE pointer on success.
417 std::ifstream* Document::OpenFile()
422 "Document::OpenFile is already opened when opening: ",
426 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
431 "Document::OpenFile cannot open file: ",
437 Fp->read((char*)&zero, (size_t)2 );
439 //ACR -- or DICOM with no Preamble --
440 if( zero == 0x0008 || zero == 0x0800 || zero == 0x0002 || zero == 0x0200 )
446 Fp->seekg(126L, std::ios_base::cur);
448 Fp->read(dicm, (size_t)4);
449 if( memcmp(dicm, "DICM", 4) == 0 )
456 "Document::OpenFile not DICOM/ACR (missing preamble)",
463 * \brief closes the file
464 * @return TRUE if the close was successfull
466 bool Document::CloseFile()
475 return true; //FIXME how do we detect a non-close ifstream ?
479 * \brief Writes in a file all the Header Entries (Dicom Elements)
480 * @param fp file pointer on an already open file
481 * @param filetype Type of the File to be written
482 * (ACR-NEMA, ExplicitVR, ImplicitVR)
483 * \return Always true.
485 void Document::Write(std::ofstream* fp, FileType filetype)
487 /// \todo move the following lines (and a lot of others, to be written)
488 /// to a future function CheckAndCorrectHeader
489 /// (necessary if user wants to write a DICOM V3 file
490 /// starting from an ACR-NEMA (V2) Header
492 if (filetype == ImplicitVR)
495 Util::DicomString( TransferSyntaxStrings[ImplicitVRLittleEndian] );
496 ReplaceOrCreateByNumber(ts, 0x0002, 0x0010);
498 /// \todo Refer to standards on page 21, chapter 6.2
499 /// "Value representation": values with a VR of UI shall be
500 /// padded with a single trailing null
501 /// in the following case we have to padd manually with a 0
503 SetEntryLengthByNumber(18, 0x0002, 0x0010);
506 if (filetype == ExplicitVR)
509 Util::DicomString( TransferSyntaxStrings[ExplicitVRLittleEndian] );
510 ReplaceOrCreateByNumber(ts, 0x0002, 0x0010); //LEAK
512 /// \todo Refer to standards on page 21, chapter 6.2
513 /// "Value representation": values with a VR of UI shall be
514 /// padded with a single trailing null
515 /// Dans le cas suivant on doit pader manuellement avec un 0
517 SetEntryLengthByNumber(20, 0x0002, 0x0010);
521 * \todo rewrite later, if really usefull
522 * - 'Group Length' element is optional in DICOM
523 * - but un-updated odd groups lengthes can causes pb
526 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
527 * UpdateGroupLength(false,filetype);
528 * if ( filetype == ACR)
529 * UpdateGroupLength(true,ACR);
532 ElementSet::Write(fp, filetype); // This one is recursive
537 * \brief Modifies the value of a given Header Entry (Dicom Element)
538 * when it exists. Create it with the given value when unexistant.
539 * @param value (string) Value to be set
540 * @param group Group number of the Entry
541 * @param elem Element number of the Entry
542 * @param VR V(alue) R(epresentation) of the Entry -if private Entry-
543 * \return pointer to the modified/created Header Entry (NULL when creation
546 ValEntry* Document::ReplaceOrCreateByNumber(
547 std::string const & value,
552 ValEntry* valEntry = 0;
553 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
557 // check if (group,element) DictEntry exists
558 // if it doesn't, create an entry in DictSet::VirtualEntry
561 // Find out if the tag we received is in the dictionaries:
562 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
563 DictEntry* dictEntry = pubDict->GetDictEntryByNumber(group, elem);
566 currentEntry = NewDocEntryByNumber(group, elem, vr);
570 currentEntry = NewDocEntryByNumber(group, elem);
575 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
576 " NewDocEntryByNumber failed.");
579 valEntry = new ValEntry(currentEntry);
580 if ( !AddEntry(valEntry))
583 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
584 " failed allthough this is a creation.");
586 // This is the reason why NewDocEntryByNumber are a real
587 // bad habit !!! FIXME
592 valEntry = dynamic_cast< ValEntry* >(currentEntry);
593 if ( !valEntry ) // Euuuuh? It wasn't a ValEntry
594 // then we change it to a ValEntry ?
595 // Shouldn't it be considered as an error ?
597 // We need to promote the DocEntry to a ValEntry:
598 valEntry = new ValEntry(currentEntry);
599 if (!RemoveEntry(currentEntry))
602 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
603 " of previous DocEntry failed.");
606 if ( !AddEntry(valEntry))
609 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
610 " promoted ValEntry failed.");
616 SetEntryByNumber(value, group, elem);
622 * \brief Modifies the value of a given Header Entry (Dicom Element)
623 * when it exists. Create it with the given value when unexistant.
624 * @param binArea (binary) value to be set
625 * @param Group Group number of the Entry
626 * @param Elem Element number of the Entry
627 * \return pointer to the modified/created Header Entry (NULL when creation
630 BinEntry* Document::ReplaceOrCreateByNumber(
637 BinEntry* binEntry = 0;
638 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
642 // check if (group,element) DictEntry exists
643 // if it doesn't, create an entry in DictSet::VirtualEntry
646 // Find out if the tag we received is in the dictionaries:
647 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
648 DictEntry *dictEntry = pubDict->GetDictEntryByNumber(group, elem);
652 currentEntry = NewDocEntryByNumber(group, elem, vr);
656 currentEntry = NewDocEntryByNumber(group, elem);
660 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
661 " NewDocEntryByNumber failed.");
664 binEntry = new BinEntry(currentEntry);
665 if ( !AddEntry(binEntry))
667 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
668 " failed allthough this is a creation.");
673 binEntry = dynamic_cast< BinEntry* >(currentEntry);
674 if ( !binEntry ) // Euuuuh? It wasn't a BinEntry
675 // then we change it to a BinEntry ?
676 // Shouldn't it be considered as an error ?
678 // We need to promote the DocEntry to a BinEntry:
679 binEntry = new BinEntry(currentEntry);
680 if (!RemoveEntry(currentEntry))
682 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
683 " of previous DocEntry failed.");
686 if ( !AddEntry(binEntry))
688 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
689 " promoted BinEntry failed.");
695 SetEntryByNumber(binArea, lgth, group, elem);
702 * \brief Modifies the value of a given Header Entry (Dicom Element)
703 * when it exists. Create it when unexistant.
704 * @param Group Group number of the Entry
705 * @param Elem Element number of the Entry
706 * \return pointer to the modified/created SeqEntry (NULL when creation
709 SeqEntry* Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem)
712 DocEntry* a = GetDocEntryByNumber( group, elem);
715 a = NewSeqEntryByNumber(group, elem);
721 b = new SeqEntry(a, 1); // FIXME : 1 (Depth)
728 * \brief Set a new value if the invoked element exists
729 * Seems to be useless !!!
730 * @param value new element value
731 * @param group group number of the Entry
732 * @param elem element number of the Entry
735 bool Document::ReplaceIfExistByNumber(std::string const & value,
736 uint16_t group, uint16_t elem )
738 SetEntryByNumber(value, group, elem);
743 //-----------------------------------------------------------------------------
747 * \brief Checks if a given Dicom Element exists within the H table
748 * @param group Group number of the searched Dicom Element
749 * @param element Element number of the searched Dicom Element
750 * @return true is found
752 bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
754 const std::string &key = DictEntry::TranslateToKey(group, element );
755 return TagHT.count(key) != 0;
759 * \brief Searches within Header Entries (Dicom Elements) parsed with
760 * the public and private dictionaries
761 * for the element value of a given tag.
762 * \warning Don't use any longer : use GetPubEntryByName
763 * @param tagName name of the searched element.
764 * @return Corresponding element value when it exists,
765 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
767 std::string Document::GetEntryByName(TagName const & tagName)
769 DictEntry* dictEntry = RefPubDict->GetDictEntryByName(tagName);
775 return GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
779 * \brief Searches within Header Entries (Dicom Elements) parsed with
780 * the public and private dictionaries
781 * for the element value representation of a given tag.
783 * Obtaining the VR (Value Representation) might be needed by caller
784 * to convert the string typed content to caller's native type
785 * (think of C++ vs Python). The VR is actually of a higher level
786 * of semantics than just the native C++ type.
787 * @param tagName name of the searched element.
788 * @return Corresponding element value representation when it exists,
789 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
791 std::string Document::GetEntryVRByName(TagName const& tagName)
793 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
794 if( dictEntry == NULL)
799 DocEntry* elem = GetDocEntryByNumber(dictEntry->GetGroup(),
800 dictEntry->GetElement());
801 return elem->GetVR();
805 * \brief Searches within Header Entries (Dicom Elements) parsed with
806 * the public and private dictionaries
807 * for the element value representation of a given tag.
808 * @param group Group number of the searched tag.
809 * @param element Element number of the searched tag.
810 * @return Corresponding element value representation when it exists,
811 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
813 std::string Document::GetEntryByNumber(uint16_t group, uint16_t element)
815 TagKey key = DictEntry::TranslateToKey(group, element);
816 /// \todo use map methods, instead of multimap JPR
817 if ( !TagHT.count(key))
822 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
826 * \brief Searches within Header Entries (Dicom Elements) parsed with
827 * the public and private dictionaries
828 * for the element value representation of a given tag..
830 * Obtaining the VR (Value Representation) might be needed by caller
831 * to convert the string typed content to caller's native type
832 * (think of C++ vs Python). The VR is actually of a higher level
833 * of semantics than just the native C++ type.
834 * @param group Group number of the searched tag.
835 * @param element Element number of the searched tag.
836 * @return Corresponding element value representation when it exists,
837 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
839 std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element)
841 DocEntry* elem = GetDocEntryByNumber(group, element);
846 return elem->GetVR();
850 * \brief Searches within Header Entries (Dicom Elements) parsed with
851 * the public and private dictionaries
852 * for the value length of a given tag..
853 * @param group Group number of the searched tag.
854 * @param element Element number of the searched tag.
855 * @return Corresponding element length; -2 if not found
857 int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element)
859 DocEntry* elem = GetDocEntryByNumber(group, element);
862 return -2; //magic number
864 return elem->GetLength();
867 * \brief Sets the value (string) of the Header Entry (Dicom Element)
868 * @param content string value of the Dicom Element
869 * @param tagName name of the searched Dicom Element.
870 * @return true when found
872 bool Document::SetEntryByName(std::string const & content,
873 TagName const & tagName)
875 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
881 return SetEntryByNumber(content,dictEntry->GetGroup(),
882 dictEntry->GetElement());
886 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
887 * through it's (group, element) and modifies it's content with
889 * @param content new value (string) to substitute with
890 * @param group group number of the Dicom Element to modify
891 * @param element element number of the Dicom Element to modify
893 bool Document::SetEntryByNumber(std::string const& content,
894 uint16_t group, uint16_t element)
899 ValEntry* valEntry = GetValEntryByNumber(group, element);
902 dbg.Verbose(0, "Document::SetEntryByNumber: no corresponding",
903 " ValEntry (try promotion first).");
906 // Non even content must be padded with a space (020H)...
907 std::string finalContent = Util::DicomString( content.c_str() );
908 assert( !(finalContent.size() % 2) );
909 valEntry->SetValue(finalContent);
911 // Integers have a special treatement for their length:
913 l = finalContent.length();
914 if ( l != 0) // To avoid to be cheated by 'zero length' integers
916 VRKey vr = valEntry->GetVR();
917 if( vr == "US" || vr == "SS" )
919 // for multivaluated items
920 c = Util::CountSubstring(content, "\\") + 1;
923 else if( vr == "UL" || vr == "SL" )
925 // for multivaluated items
926 c = Util::CountSubstring(content, "\\") + 1;
930 valEntry->SetLength(l);
935 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
936 * through it's (group, element) and modifies it's content with
938 * @param content new value (void* -> uint8_t*) to substitute with
939 * @param lgth new value length
940 * @param group group number of the Dicom Element to modify
941 * @param element element number of the Dicom Element to modify
943 bool Document::SetEntryByNumber(uint8_t*content, int lgth,
944 uint16_t group, uint16_t element)
946 (void)lgth; //not used
947 TagKey key = DictEntry::TranslateToKey(group, element);
948 if ( !TagHT.count(key))
953 /* Hope Binary field length is *never* wrong
954 if(lgth%2) // Non even length are padded with a space (020H).
957 //content = content + '\0'; // fing a trick to enlarge a binary field?
960 BinEntry* a = (BinEntry *)TagHT[key];
961 a->SetBinArea(content);
963 a->SetValue(GDCM_BINLOADED);
969 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
970 * in the PubDocEntrySet of this instance
971 * through it's (group, element) and modifies it's length with
973 * \warning Use with extreme caution.
974 * @param l new length to substitute with
975 * @param group group number of the Entry to modify
976 * @param element element number of the Entry to modify
977 * @return true on success, false otherwise.
979 bool Document::SetEntryLengthByNumber(uint32_t l,
980 uint16_t group, uint16_t element)
982 /// \todo use map methods, instead of multimap JPR
983 TagKey key = DictEntry::TranslateToKey(group, element);
984 if ( !TagHT.count(key) )
990 l++; // length must be even
992 ( ((TagHT.equal_range(key)).first)->second )->SetLength(l);
998 * \brief Gets (from Header) the offset of a 'non string' element value
999 * (LoadElementValues has already be executed)
1000 * @param group group number of the Entry
1001 * @param elem element number of the Entry
1002 * @return File Offset of the Element Value
1004 size_t Document::GetEntryOffsetByNumber(uint16_t group, uint16_t elem)
1006 DocEntry* entry = GetDocEntryByNumber(group, elem);
1009 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry present.");
1012 return entry->GetOffset();
1016 * \brief Gets (from Header) a 'non string' element value
1017 * (LoadElementValues has already be executed)
1018 * @param group group number of the Entry
1019 * @param elem element number of the Entry
1020 * @return Pointer to the 'non string' area
1022 void* Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem)
1024 DocEntry* entry = GetDocEntryByNumber(group, elem);
1027 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry");
1030 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(entry) )
1032 return binEntry->GetBinArea();
1039 * \brief Loads (from disk) the element content
1040 * when a string is not suitable
1041 * @param group group number of the Entry
1042 * @param elem element number of the Entry
1044 void* Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
1046 DocEntry *docElement = GetDocEntryByNumber(group, elem);
1051 size_t o =(size_t)docElement->GetOffset();
1052 Fp->seekg( o, std::ios_base::beg);
1053 size_t l = docElement->GetLength();
1054 uint8_t* a = new uint8_t[l];
1057 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1060 Fp->read((char*)a, l);
1061 if( Fp->fail() || Fp->eof() )//Fp->gcount() == 1
1066 /// \todo Drop any already existing void area! JPR
1067 if( !SetEntryBinAreaByNumber( a, group, elem ) )
1069 dbg.Verbose(0, "Document::LoadEntryBinArea setting failed.");
1074 * \brief Loads (from disk) the element content
1075 * when a string is not suitable
1076 * @param element Entry whose binArea is going to be loaded
1078 void* Document::LoadEntryBinArea(BinEntry* element)
1080 size_t o =(size_t)element->GetOffset();
1081 Fp->seekg(o, std::ios_base::beg);
1082 size_t l = element->GetLength();
1083 uint8_t* a = new uint8_t[l];
1086 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1089 element->SetBinArea((uint8_t*)a);
1090 /// \todo check the result
1091 Fp->read((char*)a, l);
1092 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
1102 * \brief Sets a 'non string' value to a given Dicom Element
1103 * @param area area containing the 'non string' value
1104 * @param group Group number of the searched Dicom Element
1105 * @param element Element number of the searched Dicom Element
1108 bool Document::SetEntryBinAreaByNumber(uint8_t* area,
1109 uint16_t group, uint16_t element)
1111 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1112 if ( !currentEntry )
1116 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1118 binEntry->SetBinArea( area );
1125 * \brief Update the entries with the shadow dictionary.
1126 * Only non even entries are analyzed
1128 void Document::UpdateShaEntries()
1133 /// \todo TODO : still any use to explore recursively the whole structure?
1135 for(ListTag::iterator it=listEntries.begin();
1136 it!=listEntries.end();
1139 // Odd group => from public dictionary
1140 if((*it)->GetGroup()%2==0)
1143 // Peer group => search the corresponding dict entry
1145 entry=RefShaDict->GetDictEntryByNumber((*it)->GetGroup(),(*it)->GetElement());
1149 if((*it)->IsImplicitVR())
1154 (*it)->SetValue(GetDocEntryUnvalue(*it)); // to go on compiling
1156 // Set the new entry and the new value
1157 (*it)->SetDictEntry(entry);
1158 CheckDocEntryVR(*it,vr);
1160 (*it)->SetValue(GetDocEntryValue(*it)); // to go on compiling
1165 // Remove precedent value transformation
1166 (*it)->SetDictEntry(NewVirtualDictEntry((*it)->GetGroup(),(*it)->GetElement(),vr));
1173 * \brief Searches within the Header Entries for a Dicom Element of
1175 * @param tagName name of the searched Dicom Element.
1176 * @return Corresponding Dicom Element when it exists, and NULL
1179 DocEntry* Document::GetDocEntryByName(TagName const & tagName)
1181 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
1187 return GetDocEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
1191 * \brief retrieves a Dicom Element (the first one) using (group, element)
1192 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1193 * if you think it's NOT UNIQUE, check the count number
1194 * and use iterators to retrieve ALL the Dicoms Elements within
1195 * a given couple (group, element)
1196 * @param group Group number of the searched Dicom Element
1197 * @param element Element number of the searched Dicom Element
1200 DocEntry* Document::GetDocEntryByNumber(uint16_t group, uint16_t element)
1202 TagKey key = DictEntry::TranslateToKey(group, element);
1203 if ( !TagHT.count(key))
1207 return TagHT.find(key)->second;
1211 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1212 * returns a result when the corresponding entry is of type
1214 * @return When present, the corresponding ValEntry.
1216 ValEntry* Document::GetValEntryByNumber(uint16_t group, uint16_t element)
1218 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1219 if ( !currentEntry )
1223 if ( ValEntry* valEntry = dynamic_cast<ValEntry*>(currentEntry) )
1227 dbg.Verbose(0, "Document::GetValEntryByNumber: unfound ValEntry.");
1233 * \brief Loads the element while preserving the current
1234 * underlying file position indicator as opposed to
1235 * to LoadDocEntry that modifies it.
1236 * @param entry Header Entry whose value shall be loaded.
1239 void Document::LoadDocEntrySafe(DocEntry * entry)
1241 long PositionOnEntry = Fp->tellg();
1242 LoadDocEntry(entry);
1243 Fp->seekg(PositionOnEntry, std::ios_base::beg);
1247 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1249 * @return The properly swaped 32 bits integer.
1251 uint32_t Document::SwapLong(uint32_t a)
1258 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1259 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1263 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1267 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1270 //std::cout << "swapCode= " << SwapCode << std::endl;
1271 dbg.Error(" Document::SwapLong : unset swap code");
1278 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1280 * @return The properly unswaped 32 bits integer.
1282 uint32_t Document::UnswapLong(uint32_t a)
1288 * \brief Swaps the bytes so they agree with the processor order
1289 * @return The properly swaped 16 bits integer.
1291 uint16_t Document::SwapShort(uint16_t a)
1293 if ( SwapCode == 4321 || SwapCode == 2143 )
1295 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1301 * \brief Unswaps the bytes so they agree with the processor order
1302 * @return The properly unswaped 16 bits integer.
1304 uint16_t Document::UnswapShort(uint16_t a)
1306 return SwapShort(a);
1309 //-----------------------------------------------------------------------------
1313 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1314 * @return length of the parsed set.
1316 void Document::ParseDES(DocEntrySet *set, long offset,
1317 long l_max, bool delim_mode)
1319 DocEntry *newDocEntry = 0;
1323 if ( !delim_mode && (Fp->tellg()-offset) >= l_max)
1327 newDocEntry = ReadNextDocEntry( );
1333 VRKey vr = newDocEntry->GetVR();
1337 if ( Global::GetVR()->IsVROfGdcmStringRepresentable(vr) )
1339 /////////////////////// ValEntry
1340 ValEntry* newValEntry =
1341 new ValEntry( newDocEntry->GetDictEntry() );
1342 newValEntry->Copy( newDocEntry );
1344 // When "set" is a Document, then we are at the top of the
1345 // hierarchy and the Key is simply of the form ( group, elem )...
1346 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1349 newValEntry->SetKey( newValEntry->GetKey() );
1351 // ...but when "set" is a SQItem, we are inserting this new
1352 // valEntry in a sequence item. Hence the key has the
1353 // generalized form (refer to \ref BaseTagKey):
1354 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1356 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1357 + newValEntry->GetKey() );
1360 set->AddEntry( newValEntry );
1361 LoadDocEntry( newValEntry );
1362 if (newValEntry->IsItemDelimitor())
1366 if ( !delim_mode && (Fp->tellg()-offset) >= l_max)
1373 if ( ! Global::GetVR()->IsVROfGdcmBinaryRepresentable(vr) )
1375 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1376 dbg.Verbose(0, "Document::ParseDES: neither Valentry, "
1377 "nor BinEntry. Probably unknown VR.");
1380 //////////////////// BinEntry or UNKOWN VR:
1381 BinEntry* newBinEntry =
1382 new BinEntry( newDocEntry->GetDictEntry() ); //LEAK
1383 newBinEntry->Copy( newDocEntry );
1385 // When "this" is a Document the Key is simply of the
1386 // form ( group, elem )...
1387 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1390 newBinEntry->SetKey( newBinEntry->GetKey() );
1392 // but when "this" is a SQItem, we are inserting this new
1393 // valEntry in a sequence item, and the kay has the
1394 // generalized form (refer to \ref BaseTagKey):
1395 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1397 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1398 + newBinEntry->GetKey() );
1401 set->AddEntry( newBinEntry );
1402 LoadDocEntry( newBinEntry );
1405 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1406 && ( newDocEntry->GetElement() == 0x0010 ) )
1408 TransferSyntaxType ts = GetTransferSyntax();
1409 if ( ts == RLELossless )
1411 long PositionOnEntry = Fp->tellg();
1412 Fp->seekg( newDocEntry->GetOffset(), std::ios_base::beg );
1414 Fp->seekg( PositionOnEntry, std::ios_base::beg );
1416 else if ( IsJPEG() )
1418 long PositionOnEntry = Fp->tellg();
1419 Fp->seekg( newDocEntry->GetOffset(), std::ios_base::beg );
1420 ComputeJPEGFragmentInfo();
1421 Fp->seekg( PositionOnEntry, std::ios_base::beg );
1425 // Just to make sure we are at the beginning of next entry.
1426 SkipToNextDocEntry(newDocEntry);
1431 unsigned long l = newDocEntry->GetReadLength();
1432 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1434 if ( l == 0xffffffff )
1443 // no other way to create it ...
1444 SeqEntry* newSeqEntry =
1445 new SeqEntry( newDocEntry->GetDictEntry() );
1446 newSeqEntry->Copy( newDocEntry );
1447 newSeqEntry->SetDelimitorMode( delim_mode );
1449 // At the top of the hierarchy, stands a Document. When "set"
1450 // is a Document, then we are building the first depth level.
1451 // Hence the SeqEntry we are building simply has a depth
1453 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1456 newSeqEntry->SetDepthLevel( 1 );
1457 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1459 // But when "set" is allready a SQItem, we are building a nested
1460 // sequence, and hence the depth level of the new SeqEntry
1461 // we are building, is one level deeper:
1462 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1464 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1465 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1466 + newSeqEntry->GetKey() );
1470 { // Don't try to parse zero-length sequences
1471 ParseSQ( newSeqEntry,
1472 newDocEntry->GetOffset(),
1475 set->AddEntry( newSeqEntry );
1476 if ( !delim_mode && (Fp->tellg()-offset) >= l_max)
1486 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1487 * @return parsed length for this level
1489 void Document::ParseSQ( SeqEntry* seqEntry,
1490 long offset, long l_max, bool delim_mode)
1492 int SQItemNumber = 0;
1497 DocEntry* newDocEntry = ReadNextDocEntry();
1500 // FIXME Should warn user
1505 if ( newDocEntry->IsSequenceDelimitor() )
1507 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1511 if ( !delim_mode && (Fp->tellg()-offset) >= l_max)
1516 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1517 std::ostringstream newBase;
1518 newBase << seqEntry->GetKey()
1522 itemSQ->SetBaseTagKey( newBase.str() );
1523 unsigned int l = newDocEntry->GetReadLength();
1525 if ( l == 0xffffffff )
1534 ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1536 seqEntry->AddEntry( itemSQ, SQItemNumber );
1538 if ( !delim_mode && ( Fp->tellg() - offset ) >= l_max )
1546 * \brief Loads the element content if its length doesn't exceed
1547 * the value specified with Document::SetMaxSizeLoadEntry()
1548 * @param entry Header Entry (Dicom Element) to be dealt with
1550 void Document::LoadDocEntry(DocEntry* entry)
1552 uint16_t group = entry->GetGroup();
1553 std::string vr = entry->GetVR();
1554 uint32_t length = entry->GetLength();
1556 Fp->seekg((long)entry->GetOffset(), std::ios_base::beg);
1558 // A SeQuence "contains" a set of Elements.
1559 // (fffe e000) tells us an Element is beginning
1560 // (fffe e00d) tells us an Element just ended
1561 // (fffe e0dd) tells us the current SeQuence just ended
1562 if( group == 0xfffe )
1564 // NO more value field for SQ !
1568 // When the length is zero things are easy:
1571 ((ValEntry *)entry)->SetValue("");
1575 // The elements whose length is bigger than the specified upper bound
1576 // are not loaded. Instead we leave a short notice of the offset of
1577 // the element content and it's length.
1579 std::ostringstream s;
1580 if (length > MaxSizeLoadEntry)
1582 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1584 //s << "gdcm::NotLoaded (BinEntry)";
1585 s << GDCM_NOTLOADED;
1586 s << " Address:" << (long)entry->GetOffset();
1587 s << " Length:" << entry->GetLength();
1588 s << " x(" << std::hex << entry->GetLength() << ")";
1589 binEntryPtr->SetValue(s.str());
1591 // Be carefull : a BinEntry IS_A ValEntry ...
1592 else if (ValEntry* valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1594 // s << "gdcm::NotLoaded. (ValEntry)";
1595 s << GDCM_NOTLOADED;
1596 s << " Address:" << (long)entry->GetOffset();
1597 s << " Length:" << entry->GetLength();
1598 s << " x(" << std::hex << entry->GetLength() << ")";
1599 valEntryPtr->SetValue(s.str());
1604 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1605 << "nor a ValEntry ?! Should never print that !" << std::endl;
1608 // to be sure we are at the end of the value ...
1609 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1610 std::ios_base::beg);
1614 // When we find a BinEntry not very much can be done :
1615 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1617 s << GDCM_BINLOADED;
1618 binEntryPtr->SetValue(s.str());
1619 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1623 /// \todo Any compacter code suggested (?)
1624 if ( IsDocEntryAnInteger(entry) )
1628 // When short integer(s) are expected, read and convert the following
1629 // n *two characters properly i.e. consider them as short integers as
1630 // opposed to strings.
1631 // Elements with Value Multiplicity > 1
1632 // contain a set of integers (not a single one)
1633 if (vr == "US" || vr == "SS")
1636 NewInt = ReadInt16();
1640 for (int i=1; i < nbInt; i++)
1643 NewInt = ReadInt16();
1648 // See above comment on multiple integers (mutatis mutandis).
1649 else if (vr == "UL" || vr == "SL")
1652 NewInt = ReadInt32();
1656 for (int i=1; i < nbInt; i++)
1659 NewInt = ReadInt32();
1664 #ifdef GDCM_NO_ANSI_STRING_STREAM
1665 s << std::ends; // to avoid oddities on Solaris
1666 #endif //GDCM_NO_ANSI_STRING_STREAM
1668 ((ValEntry *)entry)->SetValue(s.str());
1672 // FIXME: We need an additional byte for storing \0 that is not on disk
1673 char *str = new char[length+1];
1674 Fp->read(str, (size_t)length);
1675 str[length] = '\0'; //this is only useful when length is odd
1676 // Special DicomString call to properly handle \0 and even length
1677 std::string newValue;
1680 newValue = Util::DicomString(str, length+1);
1681 //dbg.Verbose(0, "Warning: bad length: ", length );
1682 dbg.Verbose(0, "For string :", newValue.c_str());
1683 // Since we change the length of string update it length
1684 entry->SetReadLength(length+1);
1688 newValue = Util::DicomString(str, length);
1692 if ( ValEntry* valEntry = dynamic_cast<ValEntry* >(entry) )
1694 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1696 dbg.Verbose(1, "Document::LoadDocEntry",
1697 "unread element value");
1698 valEntry->SetValue(GDCM_UNREAD);
1704 // Because of correspondance with the VR dic
1705 valEntry->SetValue(newValue);
1709 valEntry->SetValue(newValue);
1714 dbg.Error(true, "Document::LoadDocEntry"
1715 "Should have a ValEntry, here !");
1721 * \brief Find the value Length of the passed Header Entry
1722 * @param entry Header Entry whose length of the value shall be loaded.
1724 void Document::FindDocEntryLength( DocEntry *entry )
1725 throw ( FormatError )
1727 uint16_t element = entry->GetElement();
1728 std::string vr = entry->GetVR();
1731 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1733 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1735 // The following reserved two bytes (see PS 3.5-2003, section
1736 // "7.1.2 Data element structure with explicit vr", p 27) must be
1737 // skipped before proceeding on reading the length on 4 bytes.
1738 Fp->seekg( 2L, std::ios_base::cur);
1739 uint32_t length32 = ReadInt32();
1741 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1746 /// \todo rename that to FindDocEntryLengthOBOrOW since
1747 /// the above test is on both OB and OW...
1748 lengthOB = FindDocEntryLengthOB();
1750 catch ( FormatUnexpected )
1752 // Computing the length failed (this happens with broken
1753 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1754 // chance to get the pixels by deciding the element goes
1755 // until the end of the file. Hence we artificially fix the
1756 // the length and proceed.
1757 long currentPosition = Fp->tellg();
1758 Fp->seekg(0L,std::ios_base::end);
1759 long lengthUntilEOF = Fp->tellg() - currentPosition;
1760 Fp->seekg(currentPosition, std::ios_base::beg);
1761 entry->SetLength(lengthUntilEOF);
1764 entry->SetLength(lengthOB);
1767 FixDocEntryFoundLength(entry, length32);
1771 // Length is encoded on 2 bytes.
1772 length16 = ReadInt16();
1774 // We can tell the current file is encoded in big endian (like
1775 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1776 // and it's value is the one of the encoding of a big endian file.
1777 // In order to deal with such big endian encoded files, we have
1778 // (at least) two strategies:
1779 // * when we load the "Transfer Syntax" tag with value of big endian
1780 // encoding, we raise the proper flags. Then we wait for the end
1781 // of the META group (0x0002) among which is "Transfer Syntax",
1782 // before switching the swap code to big endian. We have to postpone
1783 // the switching of the swap code since the META group is fully encoded
1784 // in little endian, and big endian coding only starts at the next
1785 // group. The corresponding code can be hard to analyse and adds
1786 // many additional unnecessary tests for regular tags.
1787 // * the second strategy consists in waiting for trouble, that shall
1788 // appear when we find the first group with big endian encoding. This
1789 // is easy to detect since the length of a "Group Length" tag (the
1790 // ones with zero as element number) has to be of 4 (0x0004). When we
1791 // encounter 1024 (0x0400) chances are the encoding changed and we
1792 // found a group with big endian encoding.
1793 // We shall use this second strategy. In order to make sure that we
1794 // can interpret the presence of an apparently big endian encoded
1795 // length of a "Group Length" without committing a big mistake, we
1796 // add an additional check: we look in the already parsed elements
1797 // for the presence of a "Transfer Syntax" whose value has to be "big
1798 // endian encoding". When this is the case, chances are we have got our
1799 // hands on a big endian encoded file: we switch the swap code to
1800 // big endian and proceed...
1801 if ( element == 0x0000 && length16 == 0x0400 )
1803 TransferSyntaxType ts = GetTransferSyntax();
1804 if ( ts != ExplicitVRBigEndian )
1806 throw FormatError( "Document::FindDocEntryLength()",
1807 " not explicit VR." );
1811 SwitchSwapToBigEndian();
1812 // Restore the unproperly loaded values i.e. the group, the element
1813 // and the dictionary entry depending on them.
1814 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1815 uint16_t correctElem = SwapShort( entry->GetElement() );
1816 DictEntry* newTag = GetDictEntryByNumber( correctGroup,
1820 // This correct tag is not in the dictionary. Create a new one.
1821 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1823 // FIXME this can create a memory leaks on the old entry that be
1824 // left unreferenced.
1825 entry->SetDictEntry( newTag );
1828 // Heuristic: well, some files are really ill-formed.
1829 if ( length16 == 0xffff)
1831 // 0xffff means that we deal with 'Unknown Length' Sequence
1834 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1839 // Either implicit VR or a non DICOM conformal (see note below) explicit
1840 // VR that ommited the VR of (at least) this element. Farts happen.
1841 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1842 // on Data elements "Implicit and Explicit VR Data Elements shall
1843 // not coexist in a Data Set and Data Sets nested within it".]
1844 // Length is on 4 bytes.
1846 FixDocEntryFoundLength( entry, ReadInt32() );
1852 * \brief Find the Value Representation of the current Dicom Element.
1855 void Document::FindDocEntryVR( DocEntry *entry )
1857 if ( Filetype != ExplicitVR )
1864 long positionOnEntry = Fp->tellg();
1865 // Warning: we believe this is explicit VR (Value Representation) because
1866 // we used a heuristic that found "UL" in the first tag. Alas this
1867 // doesn't guarantee that all the tags will be in explicit VR. In some
1868 // cases (see e-film filtered files) one finds implicit VR tags mixed
1869 // within an explicit VR file. Hence we make sure the present tag
1870 // is in explicit VR and try to fix things if it happens not to be
1873 Fp->read (vr, (size_t)2);
1876 if( !CheckDocEntryVR(entry, vr) )
1878 Fp->seekg(positionOnEntry, std::ios_base::beg);
1879 // When this element is known in the dictionary we shall use, e.g. for
1880 // the semantics (see the usage of IsAnInteger), the VR proposed by the
1881 // dictionary entry. Still we have to flag the element as implicit since
1882 // we know now our assumption on expliciteness is not furfilled.
1884 if ( entry->IsVRUnknown() )
1886 entry->SetVR("Implicit");
1888 entry->SetImplicitVR();
1893 * \brief Check the correspondance between the VR of the header entry
1894 * and the taken VR. If they are different, the header entry is
1895 * updated with the new VR.
1896 * @param entry Header Entry to check
1897 * @param vr Dicom Value Representation
1898 * @return false if the VR is incorrect of if the VR isn't referenced
1899 * otherwise, it returns true
1901 bool Document::CheckDocEntryVR(DocEntry *entry, VRKey vr)
1904 bool realExplicit = true;
1906 // Assume we are reading a falsely explicit VR file i.e. we reached
1907 // a tag where we expect reading a VR but are in fact we read the
1908 // first to bytes of the length. Then we will interogate (through find)
1909 // the dicom_vr dictionary with oddities like "\004\0" which crashes
1910 // both GCC and VC++ implementations of the STL map. Hence when the
1911 // expected VR read happens to be non-ascii characters we consider
1912 // we hit falsely explicit VR tag.
1914 if ( !isalpha(vr[0]) && !isalpha(vr[1]) )
1916 realExplicit = false;
1919 // CLEANME searching the dicom_vr at each occurence is expensive.
1920 // PostPone this test in an optional integrity check at the end
1921 // of parsing or only in debug mode.
1922 if ( realExplicit && !Global::GetVR()->Count(vr) )
1924 realExplicit = false;
1927 if ( !realExplicit )
1929 // We thought this was explicit VR, but we end up with an
1930 // implicit VR tag. Let's backtrack.
1931 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
1932 entry->GetGroup(), entry->GetElement());
1933 dbg.Verbose(1, "Document::FindVR: ", msg.c_str());
1935 if( entry->GetGroup() % 2 && entry->GetElement() == 0x0000)
1937 // Group length is UL !
1938 DictEntry* newEntry = NewVirtualDictEntry(
1939 entry->GetGroup(), entry->GetElement(),
1940 "UL", "FIXME", "Group Length");
1941 entry->SetDictEntry( newEntry );
1946 if ( entry->IsVRUnknown() )
1948 // When not a dictionary entry, we can safely overwrite the VR.
1949 if( entry->GetElement() == 0x0000 )
1951 // Group length is UL !
1959 else if ( entry->GetVR() != vr )
1961 // The VR present in the file and the dictionary disagree. We assume
1962 // the file writer knew best and use the VR of the file. Since it would
1963 // be unwise to overwrite the VR of a dictionary (since it would
1964 // compromise it's next user), we need to clone the actual DictEntry
1965 // and change the VR for the read one.
1966 DictEntry* newEntry = NewVirtualDictEntry(
1967 entry->GetGroup(), entry->GetElement(),
1968 vr, "FIXME", entry->GetName());
1969 entry->SetDictEntry(newEntry);
1976 * \brief Get the transformed value of the header entry. The VR value
1977 * is used to define the transformation to operate on the value
1978 * \warning NOT end user intended method !
1979 * @param entry entry to tranform
1980 * @return Transformed entry value
1982 std::string Document::GetDocEntryValue(DocEntry *entry)
1984 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1986 std::string val = ((ValEntry *)entry)->GetValue();
1987 std::string vr = entry->GetVR();
1988 uint32_t length = entry->GetLength();
1989 std::ostringstream s;
1992 // When short integer(s) are expected, read and convert the following
1993 // n * 2 bytes properly i.e. as a multivaluated strings
1994 // (each single value is separated fromthe next one by '\'
1995 // as usual for standard multivaluated filels
1996 // Elements with Value Multiplicity > 1
1997 // contain a set of short integers (not a single one)
1999 if( vr == "US" || vr == "SS" )
2004 for (int i=0; i < nbInt; i++)
2010 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
2011 newInt16 = SwapShort( newInt16 );
2016 // When integer(s) are expected, read and convert the following
2017 // n * 4 bytes properly i.e. as a multivaluated strings
2018 // (each single value is separated fromthe next one by '\'
2019 // as usual for standard multivaluated filels
2020 // Elements with Value Multiplicity > 1
2021 // contain a set of integers (not a single one)
2022 else if( vr == "UL" || vr == "SL" )
2027 for (int i=0; i < nbInt; i++)
2033 newInt32 = ( val[4*i+0] & 0xFF )
2034 + (( val[4*i+1] & 0xFF ) << 8 )
2035 + (( val[4*i+2] & 0xFF ) << 16 )
2036 + (( val[4*i+3] & 0xFF ) << 24 );
2037 newInt32 = SwapLong( newInt32 );
2041 #ifdef GDCM_NO_ANSI_STRING_STREAM
2042 s << std::ends; // to avoid oddities on Solaris
2043 #endif //GDCM_NO_ANSI_STRING_STREAM
2047 return ((ValEntry *)entry)->GetValue();
2051 * \brief Get the reverse transformed value of the header entry. The VR
2052 * value is used to define the reverse transformation to operate on
2054 * \warning NOT end user intended method !
2055 * @param entry Entry to reverse transform
2056 * @return Reverse transformed entry value
2058 std::string Document::GetDocEntryUnvalue(DocEntry* entry)
2060 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2062 std::string vr = entry->GetVR();
2063 std::vector<std::string> tokens;
2064 std::ostringstream s;
2066 if ( vr == "US" || vr == "SS" )
2070 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
2071 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2072 for (unsigned int i=0; i<tokens.size(); i++)
2074 newInt16 = atoi(tokens[i].c_str());
2075 s << ( newInt16 & 0xFF )
2076 << (( newInt16 >> 8 ) & 0xFF );
2080 if ( vr == "UL" || vr == "SL")
2084 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
2085 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2086 for (unsigned int i=0; i<tokens.size();i++)
2088 newInt32 = atoi(tokens[i].c_str());
2089 s << (char)( newInt32 & 0xFF )
2090 << (char)(( newInt32 >> 8 ) & 0xFF )
2091 << (char)(( newInt32 >> 16 ) & 0xFF )
2092 << (char)(( newInt32 >> 24 ) & 0xFF );
2097 #ifdef GDCM_NO_ANSI_STRING_STREAM
2098 s << std::ends; // to avoid oddities on Solaris
2099 #endif //GDCM_NO_ANSI_STRING_STREAM
2103 return ((ValEntry *)entry)->GetValue();
2107 * \brief Skip a given Header Entry
2108 * \warning NOT end user intended method !
2109 * @param entry entry to skip
2111 void Document::SkipDocEntry(DocEntry *entry)
2113 SkipBytes(entry->GetLength());
2117 * \brief Skips to the begining of the next Header Entry
2118 * \warning NOT end user intended method !
2119 * @param entry entry to skip
2121 void Document::SkipToNextDocEntry(DocEntry *entry)
2123 Fp->seekg((long)(entry->GetOffset()), std::ios_base::beg);
2124 Fp->seekg( (long)(entry->GetReadLength()), std::ios_base::cur);
2128 * \brief When the length of an element value is obviously wrong (because
2129 * the parser went Jabberwocky) one can hope improving things by
2130 * applying some heuristics.
2131 * @param entry entry to check
2132 * @param foundLength fist assumption about length
2134 void Document::FixDocEntryFoundLength(DocEntry *entry,
2135 uint32_t foundLength)
2137 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
2138 if ( foundLength == 0xffffffff)
2143 uint16_t gr = entry->GetGroup();
2144 uint16_t el = entry->GetElement();
2146 if ( foundLength % 2)
2148 std::ostringstream s;
2149 s << "Warning : Tag with uneven length "
2151 << " in x(" << std::hex << gr << "," << el <<")" << std::dec;
2152 dbg.Verbose(0, s.str().c_str());
2155 //////// Fix for some naughty General Electric images.
2156 // Allthough not recent many such GE corrupted images are still present
2157 // on Creatis hard disks. Hence this fix shall remain when such images
2158 // are no longer in user (we are talking a few years, here)...
2159 // Note: XMedCom probably uses such a trick since it is able to read
2160 // those pesky GE images ...
2161 if ( foundLength == 13)
2163 // Only happens for this length !
2164 if ( entry->GetGroup() != 0x0008
2165 || ( entry->GetElement() != 0x0070
2166 && entry->GetElement() != 0x0080 ) )
2169 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
2173 //////// Fix for some brain-dead 'Leonardo' Siemens images.
2174 // Occurence of such images is quite low (unless one leaves close to a
2175 // 'Leonardo' source. Hence, one might consider commenting out the
2176 // following fix on efficiency reasons.
2177 else if ( entry->GetGroup() == 0x0009
2178 && ( entry->GetElement() == 0x1113
2179 || entry->GetElement() == 0x1114 ) )
2182 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
2185 else if ( entry->GetVR() == "SQ" )
2187 foundLength = 0; // ReadLength is unchanged
2190 //////// We encountered a 'delimiter' element i.e. a tag of the form
2191 // "fffe|xxxx" which is just a marker. Delimiters length should not be
2192 // taken into account.
2193 else if( entry->GetGroup() == 0xfffe )
2195 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
2196 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
2197 // causes extra troubles...
2198 if( entry->GetElement() != 0x0000 )
2204 entry->SetUsableLength(foundLength);
2208 * \brief Apply some heuristics to predict whether the considered
2209 * element value contains/represents an integer or not.
2210 * @param entry The element value on which to apply the predicate.
2211 * @return The result of the heuristical predicate.
2213 bool Document::IsDocEntryAnInteger(DocEntry *entry)
2215 uint16_t element = entry->GetElement();
2216 uint16_t group = entry->GetGroup();
2217 const std::string & vr = entry->GetVR();
2218 uint32_t length = entry->GetLength();
2220 // When we have some semantics on the element we just read, and if we
2221 // a priori know we are dealing with an integer, then we shall be
2222 // able to swap it's element value properly.
2223 if ( element == 0 ) // This is the group length of the group
2231 // Allthough this should never happen, still some images have a
2232 // corrupted group length [e.g. have a glance at offset x(8336) of
2233 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2234 // Since for dicom compliant and well behaved headers, the present
2235 // test is useless (and might even look a bit paranoid), when we
2236 // encounter such an ill-formed image, we simply display a warning
2237 // message and proceed on parsing (while crossing fingers).
2238 std::ostringstream s;
2239 long filePosition = Fp->tellg();
2240 s << "Erroneous Group Length element length on : (" \
2241 << std::hex << group << " , " << element
2242 << ") -before- position x(" << filePosition << ")"
2243 << "lgt : " << length;
2244 dbg.Verbose(0, "Document::IsDocEntryAnInteger", s.str().c_str() );
2248 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2257 * \brief Find the Length till the next sequence delimiter
2258 * \warning NOT end user intended method !
2262 uint32_t Document::FindDocEntryLengthOB()
2263 throw( FormatUnexpected )
2265 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2266 long positionOnEntry = Fp->tellg();
2267 bool foundSequenceDelimiter = false;
2268 uint32_t totalLength = 0;
2270 while ( !foundSequenceDelimiter )
2276 group = ReadInt16();
2279 catch ( FormatError )
2281 throw FormatError("Document::FindDocEntryLengthOB()",
2282 " group or element not present.");
2285 // We have to decount the group and element we just read
2288 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2290 dbg.Verbose(1, "Document::FindDocEntryLengthOB: neither an Item "
2291 "tag nor a Sequence delimiter tag.");
2292 Fp->seekg(positionOnEntry, std::ios_base::beg);
2293 throw FormatUnexpected("Document::FindDocEntryLengthOB()",
2294 "Neither an Item tag nor a Sequence "
2298 if ( elem == 0xe0dd )
2300 foundSequenceDelimiter = true;
2303 uint32_t itemLength = ReadInt32();
2304 // We add 4 bytes since we just read the ItemLength with ReadInt32
2305 totalLength += itemLength + 4;
2306 SkipBytes(itemLength);
2308 if ( foundSequenceDelimiter )
2313 Fp->seekg( positionOnEntry, std::ios_base::beg);
2318 * \brief Reads a supposed to be 16 Bits integer
2319 * (swaps it depending on processor endianity)
2320 * @return read value
2322 uint16_t Document::ReadInt16()
2323 throw( FormatError )
2326 Fp->read ((char*)&g, (size_t)2);
2329 throw FormatError( "Document::ReadInt16()", " file error." );
2333 throw FormatError( "Document::ReadInt16()", "EOF." );
2340 * \brief Reads a supposed to be 32 Bits integer
2341 * (swaps it depending on processor endianity)
2342 * @return read value
2344 uint32_t Document::ReadInt32()
2345 throw( FormatError )
2348 Fp->read ((char*)&g, (size_t)4);
2351 throw FormatError( "Document::ReadInt32()", " file error." );
2355 throw FormatError( "Document::ReadInt32()", "EOF." );
2362 * \brief skips bytes inside the source file
2363 * \warning NOT end user intended method !
2366 void Document::SkipBytes(uint32_t nBytes)
2368 //FIXME don't dump the returned value
2369 Fp->seekg((long)nBytes, std::ios_base::cur);
2373 * \brief Loads all the needed Dictionaries
2374 * \warning NOT end user intended method !
2376 void Document::Initialise()
2378 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2380 RLEInfo = new RLEFramesInfo;
2381 JPEGInfo = new JPEGFragmentsInfo;
2385 * \brief Discover what the swap code is (among little endian, big endian,
2386 * bad little endian, bad big endian).
2388 * @return false when we are absolutely sure
2389 * it's neither ACR-NEMA nor DICOM
2390 * true when we hope ours assuptions are OK
2392 bool Document::CheckSwap()
2394 // The only guaranted way of finding the swap code is to find a
2395 // group tag since we know it's length has to be of four bytes i.e.
2396 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2397 // occurs when we can't find such group...
2399 uint32_t x = 4; // x : for ntohs
2400 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2404 char deb[256]; //HEADER_LENGTH_TO_READ];
2406 // First, compare HostByteOrder and NetworkByteOrder in order to
2407 // determine if we shall need to swap bytes (i.e. the Endian type).
2408 if ( x == ntohs(x) )
2417 // The easiest case is the one of a DICOM header, since it possesses a
2418 // file preamble where it suffice to look for the string "DICM".
2419 Fp->read(deb, HEADER_LENGTH_TO_READ);
2421 char *entCur = deb + 128;
2422 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2424 dbg.Verbose(1, "Document::CheckSwap:", "looks like DICOM Version3");
2426 // Next, determine the value representation (VR). Let's skip to the
2427 // first element (0002, 0000) and check there if we find "UL"
2428 // - or "OB" if the 1st one is (0002,0001) -,
2429 // in which case we (almost) know it is explicit VR.
2430 // WARNING: if it happens to be implicit VR then what we will read
2431 // is the length of the group. If this ascii representation of this
2432 // length happens to be "UL" then we shall believe it is explicit VR.
2433 // FIXME: in order to fix the above warning, we could read the next
2434 // element value (or a couple of elements values) in order to make
2435 // sure we are not commiting a big mistake.
2436 // We need to skip :
2437 // * the 128 bytes of File Preamble (often padded with zeroes),
2438 // * the 4 bytes of "DICM" string,
2439 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2440 // i.e. a total of 136 bytes.
2444 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2445 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2446 // *Implicit* VR. -and it is !-
2448 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2449 memcmp(entCur, "OB", (size_t)2) == 0 ||
2450 memcmp(entCur, "UI", (size_t)2) == 0 ||
2451 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2452 // when Write DCM *adds*
2454 // Use Document::dicom_vr to test all the possibilities
2455 // instead of just checking for UL, OB and UI !? group 0000
2457 Filetype = ExplicitVR;
2458 dbg.Verbose(1, "Document::CheckSwap:",
2459 "explicit Value Representation");
2463 Filetype = ImplicitVR;
2464 dbg.Verbose(1, "Document::CheckSwap:",
2465 "not an explicit Value Representation");
2471 dbg.Verbose(1, "Document::CheckSwap:",
2472 "HostByteOrder != NetworkByteOrder");
2477 dbg.Verbose(1, "Document::CheckSwap:",
2478 "HostByteOrder = NetworkByteOrder");
2481 // Position the file position indicator at first tag (i.e.
2482 // after the file preamble and the "DICM" string).
2483 Fp->seekg(0, std::ios_base::beg);
2484 Fp->seekg ( 132L, std::ios_base::beg);
2488 // Alas, this is not a DicomV3 file and whatever happens there is no file
2489 // preamble. We can reset the file position indicator to where the data
2490 // is (i.e. the beginning of the file).
2491 dbg.Verbose(1, "Document::CheckSwap:", "not a DICOM Version3 file");
2492 Fp->seekg(0, std::ios_base::beg);
2494 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2495 // By clean we mean that the length of the first tag is written down.
2496 // If this is the case and since the length of the first group HAS to be
2497 // four (bytes), then determining the proper swap code is straightforward.
2500 // We assume the array of char we are considering contains the binary
2501 // representation of a 32 bits integer. Hence the following dirty
2503 s32 = *((uint32_t *)(entCur));
2524 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2525 // It is time for despaired wild guesses.
2526 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2527 // i.e. the 'group length' element is not present :
2529 // check the supposed to be 'group number'
2530 // 0x0002 or 0x0004 or 0x0008
2531 // to determine ' SwapCode' value .
2532 // Only 0 or 4321 will be possible
2533 // (no oportunity to check for the formerly well known
2534 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2535 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -4, 8-)
2536 // the file IS NOT ACR-NEMA nor DICOM V3
2537 // Find a trick to tell it the caller...
2539 s16 = *((uint16_t *)(deb));
2556 dbg.Verbose(0, "Document::CheckSwap:",
2557 "ACR/NEMA unfound swap info (Really hopeless !)");
2561 // Then the only info we have is the net2host one.
2571 * \brief Restore the unproperly loaded values i.e. the group, the element
2572 * and the dictionary entry depending on them.
2574 void Document::SwitchSwapToBigEndian()
2576 dbg.Verbose(1, "Document::SwitchSwapToBigEndian",
2577 "Switching to BigEndian mode.");
2578 if ( SwapCode == 0 )
2582 else if ( SwapCode == 4321 )
2586 else if ( SwapCode == 3412 )
2590 else if ( SwapCode == 2143 )
2597 * \brief during parsing, Header Elements too long are not loaded in memory
2600 void Document::SetMaxSizeLoadEntry(long newSize)
2606 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2608 MaxSizeLoadEntry = 0xffffffff;
2611 MaxSizeLoadEntry = newSize;
2616 * \brief Header Elements too long will not be printed
2617 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2620 void Document::SetMaxSizePrintEntry(long newSize)
2622 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2627 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2629 MaxSizePrintEntry = 0xffffffff;
2632 MaxSizePrintEntry = newSize;
2638 * \brief Read the next tag but WITHOUT loading it's value
2639 * (read the 'Group Number', the 'Element Number',
2640 * gets the Dict Entry
2641 * gets the VR, gets the length, gets the offset value)
2642 * @return On succes the newly created DocEntry, NULL on failure.
2644 DocEntry* Document::ReadNextDocEntry()
2651 group = ReadInt16();
2654 catch ( FormatError e )
2656 // We reached the EOF (or an error occured) therefore
2657 // header parsing has to be considered as finished.
2662 DocEntry *newEntry = NewDocEntryByNumber(group, elem);
2663 FindDocEntryVR(newEntry);
2667 FindDocEntryLength(newEntry);
2669 catch ( FormatError e )
2677 newEntry->SetOffset(Fp->tellg());
2684 * \brief Generate a free TagKey i.e. a TagKey that is not present
2685 * in the TagHt dictionary.
2686 * @param group The generated tag must belong to this group.
2687 * @return The element of tag with given group which is fee.
2689 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2691 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2693 TagKey key = DictEntry::TranslateToKey(group, elem);
2694 if (TagHT.count(key) == 0)
2703 * \brief Assuming the internal file pointer \ref Document::Fp
2704 * is placed at the beginning of a tag check whether this
2705 * tag is (TestGroup, TestElement).
2706 * \warning On success the internal file pointer \ref Document::Fp
2707 * is modified to point after the tag.
2708 * On failure (i.e. when the tag wasn't the expected tag
2709 * (TestGroup, TestElement) the internal file pointer
2710 * \ref Document::Fp is restored to it's original position.
2711 * @param testGroup The expected group of the tag.
2712 * @param testElement The expected Element of the tag.
2713 * @return True on success, false otherwise.
2715 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2717 long positionOnEntry = Fp->tellg();
2718 long currentPosition = Fp->tellg(); // On debugging purposes
2720 //// Read the Item Tag group and element, and make
2721 // sure they are what we expected:
2722 uint16_t itemTagGroup;
2723 uint16_t itemTagElement;
2726 itemTagGroup = ReadInt16();
2727 itemTagElement = ReadInt16();
2729 catch ( FormatError e )
2731 //std::cerr << e << std::endl;
2734 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2736 std::ostringstream s;
2737 s << " We should have found tag (";
2738 s << std::hex << testGroup << "," << testElement << ")" << std::endl;
2739 s << " but instead we encountered tag (";
2740 s << std::hex << itemTagGroup << "," << itemTagElement << ")"
2742 s << " at address: " << (unsigned)currentPosition << std::endl;
2743 dbg.Verbose(0, "Document::ReadItemTagLength: wrong Item Tag found:");
2744 dbg.Verbose(0, s.str().c_str());
2745 Fp->seekg(positionOnEntry, std::ios_base::beg);
2753 * \brief Assuming the internal file pointer \ref Document::Fp
2754 * is placed at the beginning of a tag (TestGroup, TestElement),
2755 * read the length associated to the Tag.
2756 * \warning On success the internal file pointer \ref Document::Fp
2757 * is modified to point after the tag and it's length.
2758 * On failure (i.e. when the tag wasn't the expected tag
2759 * (TestGroup, TestElement) the internal file pointer
2760 * \ref Document::Fp is restored to it's original position.
2761 * @param testGroup The expected group of the tag.
2762 * @param testElement The expected Element of the tag.
2763 * @return On success returns the length associated to the tag. On failure
2766 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2768 long positionOnEntry = Fp->tellg();
2769 (void)positionOnEntry;
2771 if ( !ReadTag(testGroup, testElement) )
2776 //// Then read the associated Item Length
2777 long currentPosition = Fp->tellg();
2778 uint32_t itemLength = ReadInt32();
2780 std::ostringstream s;
2781 s << "Basic Item Length is: "
2782 << itemLength << std::endl;
2783 s << " at address: " << (unsigned)currentPosition << std::endl;
2784 dbg.Verbose(0, "Document::ReadItemTagLength: ", s.str().c_str());
2790 * \brief When parsing the Pixel Data of an encapsulated file, read
2791 * the basic offset table (when present, and BTW dump it).
2793 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2795 //// Read the Basic Offset Table Item Tag length...
2796 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2798 // When present, read the basic offset table itself.
2799 // Notes: - since the presence of this basic offset table is optional
2800 // we can't rely on it for the implementation, and we will simply
2801 // trash it's content (when present).
2802 // - still, when present, we could add some further checks on the
2803 // lengths, but we won't bother with such fuses for the time being.
2804 if ( itemLength != 0 )
2806 char* basicOffsetTableItemValue = new char[itemLength + 1];
2807 Fp->read(basicOffsetTableItemValue, itemLength);
2810 for (unsigned int i=0; i < itemLength; i += 4 )
2812 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2814 std::ostringstream s;
2815 s << " Read one length: ";
2816 s << std::hex << individualLength << std::endl;
2818 "Document::ReadAndSkipEncapsulatedBasicOffsetTable: ",
2823 delete[] basicOffsetTableItemValue;
2828 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2829 * Compute the RLE extra information and store it in \ref RLEInfo
2830 * for later pixel retrieval usage.
2832 void Document::ComputeRLEInfo()
2834 TransferSyntaxType ts = GetTransferSyntax();
2835 if ( ts != RLELossless )
2840 // Encoded pixel data: for the time being we are only concerned with
2841 // Jpeg or RLE Pixel data encodings.
2842 // As stated in PS 3.5-2003, section 8.2 p44:
2843 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2844 // value representation OB is used".
2845 // Hence we expect an OB value representation. Concerning OB VR,
2846 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2847 // "For the Value Representations OB and OW, the encoding shall meet the
2848 // following specifications depending on the Data element tag:"
2850 // - the first item in the sequence of items before the encoded pixel
2851 // data stream shall be basic offset table item. The basic offset table
2852 // item value, however, is not required to be present"
2854 ReadAndSkipEncapsulatedBasicOffsetTable();
2856 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2857 // Loop on the individual frame[s] and store the information
2858 // on the RLE fragments in a RLEFramesInfo.
2859 // Note: - when only a single frame is present, this is a
2861 // - when more than one frame are present, then we are in
2862 // the case of a multi-frame image.
2864 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2866 // Parse the RLE Header and store the corresponding RLE Segment
2867 // Offset Table information on fragments of this current Frame.
2868 // Note that the fragment pixels themselves are not loaded
2869 // (but just skipped).
2870 long frameOffset = Fp->tellg();
2872 uint32_t nbRleSegments = ReadInt32();
2873 if ( nbRleSegments > 16 )
2875 // There should be at most 15 segments (refer to RLEFrame class)
2876 dbg.Verbose(0, "Document::ComputeRLEInfo: too many segments.");
2879 uint32_t rleSegmentOffsetTable[15];
2880 for( int k = 1; k <= 15; k++ )
2882 rleSegmentOffsetTable[k] = ReadInt32();
2885 // Deduce from both the RLE Header and the frameLength the
2886 // fragment length, and again store this info in a
2888 long rleSegmentLength[15];
2889 // skipping (not reading) RLE Segments
2890 if ( nbRleSegments > 1)
2892 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2894 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2895 - rleSegmentOffsetTable[k];
2896 SkipBytes(rleSegmentLength[k]);
2900 rleSegmentLength[nbRleSegments] = frameLength
2901 - rleSegmentOffsetTable[nbRleSegments];
2902 SkipBytes(rleSegmentLength[nbRleSegments]);
2904 // Store the collected info
2905 RLEFrame* newFrameInfo = new RLEFrame;
2906 newFrameInfo->NumberFragments = nbRleSegments;
2907 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2909 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2910 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2912 RLEInfo->Frames.push_back( newFrameInfo );
2915 // Make sure that at the end of the item we encounter a 'Sequence
2917 if ( !ReadTag(0xfffe, 0xe0dd) )
2919 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2920 dbg.Verbose(0, " item at end of RLE item sequence");
2925 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2926 * Compute the jpeg extra information (fragment[s] offset[s] and
2927 * length) and store it[them] in \ref JPEGInfo for later pixel
2930 void Document::ComputeJPEGFragmentInfo()
2932 // If you need to, look for comments of ComputeRLEInfo().
2938 ReadAndSkipEncapsulatedBasicOffsetTable();
2940 // Loop on the fragments[s] and store the parsed information in a
2942 long fragmentLength;
2943 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2945 long fragmentOffset = Fp->tellg();
2947 // Store the collected info
2948 JPEGFragment* newFragment = new JPEGFragment;
2949 newFragment->Offset = fragmentOffset;
2950 newFragment->Length = fragmentLength;
2951 JPEGInfo->Fragments.push_back( newFragment );
2953 SkipBytes( fragmentLength );
2956 // Make sure that at the end of the item we encounter a 'Sequence
2958 if ( !ReadTag(0xfffe, 0xe0dd) )
2960 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2961 dbg.Verbose(0, " item at end of JPEG item sequence");
2966 * \brief Walk recursively the given \ref DocEntrySet, and feed
2967 * the given hash table (\ref TagDocEntryHT) with all the
2968 * \ref DocEntry (Dicom entries) encountered.
2969 * This method does the job for \ref BuildFlatHashTable.
2970 * @param builtHT Where to collect all the \ref DocEntry encountered
2971 * when recursively walking the given set.
2972 * @param set The structure to be traversed (recursively).
2974 void Document::BuildFlatHashTableRecurse( TagDocEntryHT& builtHT,
2977 if (ElementSet* elementSet = dynamic_cast< ElementSet* > ( set ) )
2979 TagDocEntryHT const & currentHT = elementSet->GetTagHT();
2980 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2981 i != currentHT.end();
2984 DocEntry* entry = i->second;
2985 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
2987 const ListSQItem& items = seqEntry->GetSQItems();
2988 for( ListSQItem::const_iterator item = items.begin();
2989 item != items.end();
2992 BuildFlatHashTableRecurse( builtHT, *item );
2996 builtHT[entry->GetKey()] = entry;
3001 if (SQItem* SQItemSet = dynamic_cast< SQItem* > ( set ) )
3003 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
3004 for (ListDocEntry::const_iterator i = currentList.begin();
3005 i != currentList.end();
3008 DocEntry* entry = *i;
3009 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3011 const ListSQItem& items = seqEntry->GetSQItems();
3012 for( ListSQItem::const_iterator item = items.begin();
3013 item != items.end();
3016 BuildFlatHashTableRecurse( builtHT, *item );
3020 builtHT[entry->GetKey()] = entry;
3027 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
3030 * The structure used by a Document (through \ref ElementSet),
3031 * in order to old the parsed entries of a Dicom header, is a recursive
3032 * one. This is due to the fact that the sequences (when present)
3033 * can be nested. Additionaly, the sequence items (represented in
3034 * gdcm as \ref SQItem) add an extra complexity to the data
3035 * structure. Hence, a gdcm user whishing to visit all the entries of
3036 * a Dicom header will need to dig in the gdcm internals (which
3037 * implies exposing all the internal data structures to the API).
3038 * In order to avoid this burden to the user, \ref BuildFlatHashTable
3039 * recursively builds a temporary hash table, which holds all the
3040 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
3042 * \warning Of course there is NO integrity constrain between the
3043 * returned \ref TagDocEntryHT and the \ref ElementSet used
3044 * to build it. Hence if the underlying \ref ElementSet is
3045 * altered, then it is the caller responsability to invoke
3046 * \ref BuildFlatHashTable again...
3047 * @return The flat std::map<> we juste build.
3049 TagDocEntryHT* Document::BuildFlatHashTable()
3051 TagDocEntryHT* FlatHT = new TagDocEntryHT;
3052 BuildFlatHashTableRecurse( *FlatHT, this );
3059 * \brief Compares two documents, according to \ref DicomDir rules
3060 * \warning Does NOT work with ACR-NEMA files
3061 * \todo Find a trick to solve the pb (use RET fields ?)
3063 * @return true if 'smaller'
3065 bool Document::operator<(Document &document)
3068 std::string s1 = GetEntryByNumber(0x0010,0x0010);
3069 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
3081 s1 = GetEntryByNumber(0x0010,0x0020);
3082 s2 = document.GetEntryByNumber(0x0010,0x0020);
3093 // Study Instance UID
3094 s1 = GetEntryByNumber(0x0020,0x000d);
3095 s2 = document.GetEntryByNumber(0x0020,0x000d);
3106 // Serie Instance UID
3107 s1 = GetEntryByNumber(0x0020,0x000e);
3108 s2 = document.GetEntryByNumber(0x0020,0x000e);
3123 } // end namespace gdcm
3125 //-----------------------------------------------------------------------------