1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2004/11/25 15:46:11 $
7 Version: $Revision: 1.140 $
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 //-----------------------------------------------------------------------------
40 // Refer to Document::CheckSwap()
41 const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
43 // Refer to Document::SetMaxSizeLoadEntry()
44 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
45 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
47 //-----------------------------------------------------------------------------
48 // Constructor / Destructor
52 * @param filename file to be opened for parsing
54 Document::Document( std::string const & filename ) : ElementSet(-1)
56 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
66 dbg.Verbose(0, "Document::Document: starting parsing of file: ",
68 Fp->seekg( 0, std::ios_base::beg);
70 Fp->seekg(0, std::ios_base::end);
71 long lgt = Fp->tellg();
73 Fp->seekg( 0, std::ios_base::beg);
75 long beg = Fp->tellg();
78 ParseDES( this, beg, lgt, false); // le Load sera fait a la volee
80 Fp->seekg( 0, std::ios_base::beg);
82 // Load 'non string' values
84 std::string PhotometricInterpretation = GetEntryByNumber(0x0028,0x0004);
85 if( PhotometricInterpretation == "PALETTE COLOR " )
87 LoadEntryBinArea(0x0028,0x1200); // gray LUT
89 /// The tags refered by the three following lines used to be CORRECTLY
90 /// defined as having an US Value Representation in the public
91 /// dictionnary. BUT the semantics implied by the three following
92 /// lines state that the corresponding tag contents are in fact
93 /// the ones of a BinEntry.
94 /// In order to fix things "Quick and Dirty" the dictionnary was
95 /// altered on PURPOUS but now contains a WRONG value.
96 /// In order to fix things and restore the dictionary to its
97 /// correct value, one needs to decided of the semantics by deciding
98 /// wether the following tags are either:
99 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
100 /// also used as BinEntry, which requires the proper conversion,
101 /// - OW, and hence loaded as BinEntry, but afterwards also used
102 /// as ValEntry, which requires the proper conversion.
103 LoadEntryBinArea(0x0028,0x1201); // R LUT
104 LoadEntryBinArea(0x0028,0x1202); // G LUT
105 LoadEntryBinArea(0x0028,0x1203); // B LUT
107 // Segmented Red Palette Color LUT Data
108 LoadEntryBinArea(0x0028,0x1221);
109 // Segmented Green Palette Color LUT Data
110 LoadEntryBinArea(0x0028,0x1222);
111 // Segmented Blue Palette Color LUT Data
112 LoadEntryBinArea(0x0028,0x1223);
114 //FIXME later : how to use it?
115 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
119 // --------------------------------------------------------------
120 // Specific code to allow gdcm to read ACR-LibIDO formated images
121 // Note: ACR-LibIDO is an extension of the ACR standard that was
122 // used at CREATIS. For the time being (say a couple years)
123 // we keep this kludge to allow a smooth move to gdcm for
124 // CREATIS developpers (sorry folks).
126 // if recognition code tells us we deal with a LibIDO image
127 // we switch lineNumber and columnNumber
130 RecCode = GetEntryByNumber(0x0008, 0x0010); // recognition code
131 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
132 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
133 // with "little-endian strings"
135 Filetype = ACR_LIBIDO;
136 std::string rows = GetEntryByNumber(0x0028, 0x0010);
137 std::string columns = GetEntryByNumber(0x0028, 0x0011);
138 SetEntryByNumber(columns, 0x0028, 0x0010);
139 SetEntryByNumber(rows , 0x0028, 0x0011);
141 // ----------------- End of ACR-LibIDO kludge ------------------
143 PrintLevel = 1; // 'Medium' print level by default
147 * \brief This default constructor doesn't parse the file. You should
148 * then invoke \ref Document::SetFileName and then the parsing.
150 Document::Document() : ElementSet(-1)
154 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
157 Filetype = ExplicitVR;
158 PrintLevel = 1; // 'Medium' print level by default
162 * \brief Canonical destructor.
164 Document::~Document ()
173 //-----------------------------------------------------------------------------
177 * \brief Prints The Dict Entries of THE public Dicom Dictionary
180 void Document::PrintPubDict(std::ostream & os)
182 RefPubDict->Print(os);
186 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
189 void Document::PrintShaDict(std::ostream & os)
191 RefShaDict->Print(os);
194 //-----------------------------------------------------------------------------
197 * \brief Get the public dictionary used
199 Dict* Document::GetPubDict()
205 * \brief Get the shadow dictionary used
207 Dict* Document::GetShaDict()
213 * \brief Set the shadow dictionary used
214 * \param dict dictionary to use in shadow
216 bool Document::SetShaDict(Dict *dict)
223 * \brief Set the shadow dictionary used
224 * \param dictName name of the dictionary to use in shadow
226 bool Document::SetShaDict(DictKey const & dictName)
228 RefShaDict = Global::GetDicts()->GetDict(dictName);
233 * \brief This predicate, based on hopefully reasonable heuristics,
234 * decides whether or not the current Document was properly parsed
235 * and contains the mandatory information for being considered as
236 * a well formed and usable Dicom/Acr File.
237 * @return true when Document is the one of a reasonable Dicom/Acr file,
240 bool Document::IsReadable()
242 if( Filetype == Unknown)
244 dbg.Verbose(0, "Document::IsReadable: wrong filetype");
250 dbg.Verbose(0, "Document::IsReadable: no tags in internal"
259 * \brief Accessor to the Transfer Syntax (when present) of the
260 * current document (it internally handles reading the
261 * value from disk when only parsing occured).
262 * @return The encountered Transfer Syntax of the current document.
264 TransferSyntaxType Document::GetTransferSyntax()
266 DocEntry *entry = GetDocEntryByNumber(0x0002, 0x0010);
272 // The entry might be present but not loaded (parsing and loading
273 // happen at different stages): try loading and proceed with check...
274 LoadDocEntrySafe(entry);
275 if (ValEntry* valEntry = dynamic_cast< ValEntry* >(entry) )
277 std::string transfer = valEntry->GetValue();
278 // The actual transfer (as read from disk) might be padded. We
279 // first need to remove the potential padding. We can make the
280 // weak assumption that padding was not executed with digits...
281 if ( transfer.length() == 0 )
283 // for brain damaged headers
286 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
288 transfer.erase(transfer.length()-1, 1);
290 for (int i = 0; TransferSyntaxStrings[i] != NULL; i++)
292 if ( TransferSyntaxStrings[i] == transfer )
294 return TransferSyntaxType(i);
301 bool Document::IsJPEGLossless()
303 TransferSyntaxType r = GetTransferSyntax();
304 return r == JPEGFullProgressionProcess10_12
305 || r == JPEGLosslessProcess14
306 || r == JPEGLosslessProcess14_1;
310 * \brief Determines if the Transfer Syntax was already encountered
311 * and if it corresponds to a JPEG2000 one
312 * @return True when JPEG2000 (Lossly or LossLess) found. False in all
315 bool Document::IsJPEG2000()
317 TransferSyntaxType r = GetTransferSyntax();
318 return r == JPEG2000Lossless || r == JPEG2000;
322 * \brief Determines if the Transfer Syntax corresponds to any form
323 * of Jpeg encoded Pixel data.
324 * @return True when any form of JPEG found. False otherwise.
326 bool Document::IsJPEG()
328 TransferSyntaxType r = GetTransferSyntax();
329 return r == JPEGBaselineProcess1
330 || r == JPEGExtendedProcess2_4
331 || r == JPEGExtendedProcess3_5
332 || r == JPEGSpectralSelectionProcess6_8
338 * \brief Determines if the Transfer Syntax corresponds to encapsulated
339 * of encoded Pixel Data (as opposed to native).
340 * @return True when encapsulated. False when native.
342 bool Document::IsEncapsulate()
344 TransferSyntaxType r = GetTransferSyntax();
345 return IsJPEG() || r == RLELossless;
349 * \brief Predicate for dicom version 3 file.
350 * @return True when the file is a dicom version 3.
352 bool Document::IsDicomV3()
354 // Checking if Transfert Syntax exists is enough
355 // Anyway, it's to late check if the 'Preamble' was found ...
356 // And ... would it be a rich idea to check ?
357 // (some 'no Preamble' DICOM images exist !)
358 return GetDocEntryByNumber(0x0002, 0x0010) != NULL;
362 * \brief returns the File Type
363 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
364 * @return the FileType code
366 FileType Document::GetFileType()
372 * \brief Tries to open the file \ref Document::Filename and
373 * checks the preamble when existing.
374 * @return The FILE pointer on success.
376 std::ifstream* Document::OpenFile()
378 if (Filename.length() == 0) return 0;
382 "Document::OpenFile is already opened when opening: ",
386 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
391 "Document::OpenFile cannot open file: ",
397 Fp->read((char*)&zero, (size_t)2 );
399 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
401 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
402 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
403 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
404 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
410 Fp->seekg(126L, std::ios_base::cur);
412 Fp->read(dicm, (size_t)4);
413 if( memcmp(dicm, "DICM", 4) == 0 )
420 "Document::OpenFile not DICOM/ACR (missing preamble)",
427 * \brief closes the file
428 * @return TRUE if the close was successfull
430 bool Document::CloseFile()
439 return true; //FIXME how do we detect a non-close ifstream ?
443 * \brief Writes in a file all the Header Entries (Dicom Elements)
444 * @param fp file pointer on an already open file
445 * @param filetype Type of the File to be written
446 * (ACR-NEMA, ExplicitVR, ImplicitVR)
447 * \return Always true.
449 void Document::WriteContent(std::ofstream* fp, FileType filetype)
451 /// \todo move the following lines (and a lot of others, to be written)
452 /// to a future function CheckAndCorrectHeader
453 /// (necessary if user wants to write a DICOM V3 file
454 /// starting from an ACR-NEMA (V2) Header
456 if ( filetype == ImplicitVR || filetype == ExplicitVR )
458 // writing Dicom File Preamble
459 char filePreamble[128];
460 memset(filePreamble, 0, 128);
461 fp->write(filePreamble, 128);
462 fp->write("DICM", 4);
466 * \todo rewrite later, if really usefull
467 * - 'Group Length' element is optional in DICOM
468 * - but un-updated odd groups lengthes can causes pb
471 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
472 * UpdateGroupLength(false,filetype);
473 * if ( filetype == ACR)
474 * UpdateGroupLength(true,ACR);
477 ElementSet::WriteContent(fp, filetype); // This one is recursive
481 * \brief Modifies the value of a given Header Entry (Dicom Element)
482 * when it exists. Create it with the given value when unexistant.
483 * @param value (string) Value to be set
484 * @param group Group number of the Entry
485 * @param elem Element number of the Entry
486 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
487 * \return pointer to the modified/created Header Entry (NULL when creation
490 ValEntry* Document::ReplaceOrCreateByNumber(
491 std::string const & value,
496 ValEntry* valEntry = 0;
497 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
501 // check if (group,element) DictEntry exists
502 // if it doesn't, create an entry in DictSet::VirtualEntry
505 // Find out if the tag we received is in the dictionaries:
506 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
507 DictEntry* dictEntry = pubDict->GetDictEntryByNumber(group, elem);
510 currentEntry = NewDocEntryByNumber(group, elem, vr);
514 currentEntry = NewDocEntryByNumber(group, elem);
519 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
520 " NewDocEntryByNumber failed.");
524 valEntry = new ValEntry(currentEntry);
527 if ( !AddEntry(valEntry))
530 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
531 " failed allthough this is a creation.");
537 valEntry = dynamic_cast< ValEntry* >(currentEntry);
538 if ( !valEntry ) // Euuuuh? It wasn't a ValEntry
539 // then we change it to a ValEntry ?
540 // Shouldn't it be considered as an error ?
542 // We need to promote the DocEntry to a ValEntry:
543 valEntry = new ValEntry(currentEntry);
544 if (!RemoveEntry(currentEntry))
547 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
548 " of previous DocEntry failed.");
551 if ( !AddEntry(valEntry))
554 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
555 " promoted ValEntry failed.");
561 SetEntryByNumber(value, group, elem);
567 * \brief Modifies the value of a given Header Entry (Dicom Element)
568 * when it exists. Create it with the given value when unexistant.
569 * A copy of the binArea is made to be kept in the Document.
570 * @param binArea (binary) value to be set
571 * @param Group Group number of the Entry
572 * @param Elem Element number of the Entry
573 * \return pointer to the modified/created Header Entry (NULL when creation
576 BinEntry* Document::ReplaceOrCreateByNumber(
583 BinEntry* binEntry = 0;
584 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
588 // check if (group,element) DictEntry exists
589 // if it doesn't, create an entry in DictSet::VirtualEntry
592 // Find out if the tag we received is in the dictionaries:
593 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
594 DictEntry *dictEntry = pubDict->GetDictEntryByNumber(group, elem);
598 currentEntry = NewDocEntryByNumber(group, elem, vr);
602 currentEntry = NewDocEntryByNumber(group, elem);
606 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
607 " NewDocEntryByNumber failed.");
610 binEntry = new BinEntry(currentEntry);
611 if ( !AddEntry(binEntry))
613 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
614 " failed allthough this is a creation.");
619 binEntry = dynamic_cast< BinEntry* >(currentEntry);
620 if ( !binEntry ) // Euuuuh? It wasn't a BinEntry
621 // then we change it to a BinEntry ?
622 // Shouldn't it be considered as an error ?
624 // We need to promote the DocEntry to a BinEntry:
625 binEntry = new BinEntry(currentEntry);
626 if (!RemoveEntry(currentEntry))
628 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
629 " of previous DocEntry failed.");
632 if ( !AddEntry(binEntry))
634 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
635 " promoted BinEntry failed.");
642 if (lgth>0 && binArea)
644 tmpArea = new uint8_t[lgth];
645 memcpy(tmpArea,binArea,lgth);
651 if (!SetEntryByNumber(tmpArea, lgth, group, elem))
664 * \brief Modifies the value of a given Header Entry (Dicom Element)
665 * when it exists. Create it when unexistant.
666 * @param Group Group number of the Entry
667 * @param Elem Element number of the Entry
668 * \return pointer to the modified/created SeqEntry (NULL when creation
671 SeqEntry* Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem)
674 DocEntry* a = GetDocEntryByNumber( group, elem);
677 a = NewSeqEntryByNumber(group, elem);
683 b = new SeqEntry(a, 1); // FIXME : 1 (Depth)
690 * \brief Set a new value if the invoked element exists
691 * Seems to be useless !!!
692 * @param value new element value
693 * @param group group number of the Entry
694 * @param elem element number of the Entry
697 bool Document::ReplaceIfExistByNumber(std::string const & value,
698 uint16_t group, uint16_t elem )
700 SetEntryByNumber(value, group, elem);
705 //-----------------------------------------------------------------------------
709 * \brief Checks if a given Dicom Element exists within the H table
710 * @param group Group number of the searched Dicom Element
711 * @param element Element number of the searched Dicom Element
712 * @return true is found
714 bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
716 const std::string &key = DictEntry::TranslateToKey(group, element );
717 return TagHT.count(key) != 0;
721 * \brief Searches within Header Entries (Dicom Elements) parsed with
722 * the public and private dictionaries
723 * for the element value of a given tag.
724 * \warning Don't use any longer : use GetPubEntryByName
725 * @param tagName name of the searched element.
726 * @return Corresponding element value when it exists,
727 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
729 std::string Document::GetEntryByName(TagName const & tagName)
731 DictEntry* dictEntry = RefPubDict->GetDictEntryByName(tagName);
737 return GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
741 * \brief Searches within Header Entries (Dicom Elements) parsed with
742 * the public and private dictionaries
743 * for the element value representation of a given tag.
745 * Obtaining the VR (Value Representation) might be needed by caller
746 * to convert the string typed content to caller's native type
747 * (think of C++ vs Python). The VR is actually of a higher level
748 * of semantics than just the native C++ type.
749 * @param tagName name of the searched element.
750 * @return Corresponding element value representation when it exists,
751 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
753 std::string Document::GetEntryVRByName(TagName const& tagName)
755 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
756 if( dictEntry == NULL)
761 DocEntry* elem = GetDocEntryByNumber(dictEntry->GetGroup(),
762 dictEntry->GetElement());
763 return elem->GetVR();
767 * \brief Searches within Header Entries (Dicom Elements) parsed with
768 * the public and private dictionaries
769 * for the element value representation of a given tag.
770 * @param group Group number of the searched tag.
771 * @param element Element number of the searched tag.
772 * @return Corresponding element value representation when it exists,
773 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
775 std::string Document::GetEntryByNumber(uint16_t group, uint16_t element)
777 TagKey key = DictEntry::TranslateToKey(group, element);
778 /// \todo use map methods, instead of multimap JPR
779 if ( !TagHT.count(key))
784 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
788 * \brief Searches within Header Entries (Dicom Elements) parsed with
789 * the public and private dictionaries
790 * for the element value representation of a given tag..
792 * Obtaining the VR (Value Representation) might be needed by caller
793 * to convert the string typed content to caller's native type
794 * (think of C++ vs Python). The VR is actually of a higher level
795 * of semantics than just the native C++ type.
796 * @param group Group number of the searched tag.
797 * @param element Element number of the searched tag.
798 * @return Corresponding element value representation when it exists,
799 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
801 std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element)
803 DocEntry* elem = GetDocEntryByNumber(group, element);
808 return elem->GetVR();
812 * \brief Searches within Header Entries (Dicom Elements) parsed with
813 * the public and private dictionaries
814 * for the value length of a given tag..
815 * @param group Group number of the searched tag.
816 * @param element Element number of the searched tag.
817 * @return Corresponding element length; -2 if not found
819 int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element)
821 DocEntry* elem = GetDocEntryByNumber(group, element);
824 return -2; //magic number
826 return elem->GetLength();
829 * \brief Sets the value (string) of the Header Entry (Dicom Element)
830 * @param content string value of the Dicom Element
831 * @param tagName name of the searched Dicom Element.
832 * @return true when found
834 bool Document::SetEntryByName( std::string const & content,
835 TagName const & tagName)
837 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
843 return SetEntryByNumber(content,dictEntry->GetGroup(),
844 dictEntry->GetElement());
848 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
849 * through it's (group, element) and modifies it's content with
851 * @param content new value (string) to substitute with
852 * @param group group number of the Dicom Element to modify
853 * @param element element number of the Dicom Element to modify
855 bool Document::SetEntryByNumber(std::string const& content,
856 uint16_t group, uint16_t element)
861 ValEntry* valEntry = GetValEntryByNumber(group, element);
864 dbg.Verbose(0, "Document::SetEntryByNumber: no corresponding",
865 " ValEntry (try promotion first).");
868 // Non even content must be padded with a space (020H)...
869 std::string finalContent = Util::DicomString( content.c_str() );
870 assert( !(finalContent.size() % 2) );
871 valEntry->SetValue(finalContent);
873 // Integers have a special treatement for their length:
875 l = finalContent.length();
876 if ( l != 0) // To avoid to be cheated by 'zero length' integers
878 VRKey vr = valEntry->GetVR();
879 if( vr == "US" || vr == "SS" )
881 // for multivaluated items
882 c = Util::CountSubstring(content, "\\") + 1;
885 else if( vr == "UL" || vr == "SL" )
887 // for multivaluated items
888 c = Util::CountSubstring(content, "\\") + 1;
892 valEntry->SetLength(l);
897 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
898 * through it's (group, element) and modifies it's content with
900 * @param content new value (void* -> uint8_t*) to substitute with
901 * @param lgth new value length
902 * @param group group number of the Dicom Element to modify
903 * @param element element number of the Dicom Element to modify
905 bool Document::SetEntryByNumber(uint8_t*content, int lgth,
906 uint16_t group, uint16_t element)
908 (void)lgth; //not used
909 TagKey key = DictEntry::TranslateToKey(group, element);
910 if ( !TagHT.count(key))
915 /* Hope Binary field length is *never* wrong
916 if(lgth%2) // Non even length are padded with a space (020H).
919 //content = content + '\0'; // fing a trick to enlarge a binary field?
922 BinEntry* entry = (BinEntry *)TagHT[key];
923 entry->SetBinArea(content);
924 entry->SetLength(lgth);
925 entry->SetValue(GDCM_BINLOADED);
931 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
932 * in the PubDocEntrySet of this instance
933 * through it's (group, element) and modifies it's length with
935 * \warning Use with extreme caution.
936 * @param l new length to substitute with
937 * @param group group number of the Entry to modify
938 * @param element element number of the Entry to modify
939 * @return true on success, false otherwise.
941 bool Document::SetEntryLengthByNumber(uint32_t l,
942 uint16_t group, uint16_t element)
944 /// \todo use map methods, instead of multimap JPR
945 TagKey key = DictEntry::TranslateToKey(group, element);
946 if ( !TagHT.count(key) )
952 l++; // length must be even
954 ( ((TagHT.equal_range(key)).first)->second )->SetLength(l);
960 * \brief Gets (from Header) the offset of a 'non string' element value
961 * (LoadElementValues has already be executed)
962 * @param group group number of the Entry
963 * @param elem element number of the Entry
964 * @return File Offset of the Element Value
966 size_t Document::GetEntryOffsetByNumber(uint16_t group, uint16_t elem)
968 DocEntry* entry = GetDocEntryByNumber(group, elem);
971 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry present.");
974 return entry->GetOffset();
978 * \brief Gets (from Header) a 'non string' element value
979 * (LoadElementValues has already be executed)
980 * @param group group number of the Entry
981 * @param elem element number of the Entry
982 * @return Pointer to the 'non string' area
984 void* Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem)
986 DocEntry* entry = GetDocEntryByNumber(group, elem);
989 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry");
992 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(entry) )
994 return binEntry->GetBinArea();
1001 * \brief Loads (from disk) the element content
1002 * when a string is not suitable
1003 * @param group group number of the Entry
1004 * @param elem element number of the Entry
1006 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
1008 // Search the corresponding DocEntry
1009 DocEntry *docElement = GetDocEntryByNumber(group, elem);
1013 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
1017 LoadEntryBinArea(binElement);
1018 /* size_t o =(size_t)docElement->GetOffset();
1019 Fp->seekg( o, std::ios_base::beg);
1020 size_t l = docElement->GetLength();
1021 uint8_t* a = new uint8_t[l];
1024 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1029 Fp->read((char*)a, l);
1030 if( Fp->fail() || Fp->eof() )//Fp->gcount() == 1
1036 // Set the value to the DocEntry
1037 if( !SetEntryBinAreaByNumber( a, group, elem ) )
1040 dbg.Verbose(0, "Document::LoadEntryBinArea setting failed.");
1045 * \brief Loads (from disk) the element content
1046 * when a string is not suitable
1047 * @param element Entry whose binArea is going to be loaded
1049 void Document::LoadEntryBinArea(BinEntry* element)
1051 if(element->GetBinArea())
1054 bool openFile = !Fp;
1058 size_t o =(size_t)element->GetOffset();
1059 Fp->seekg(o, std::ios_base::beg);
1061 size_t l = element->GetLength();
1062 uint8_t* a = new uint8_t[l];
1065 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1069 /// \todo check the result
1070 Fp->read((char*)a, l);
1071 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
1077 element->SetBinArea(a);
1084 * \brief Sets a 'non string' value to a given Dicom Element
1085 * @param area area containing the 'non string' value
1086 * @param group Group number of the searched Dicom Element
1087 * @param element Element number of the searched Dicom Element
1090 bool Document::SetEntryBinAreaByNumber(uint8_t* area,
1091 uint16_t group, uint16_t element)
1093 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1094 if ( !currentEntry )
1099 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1101 binEntry->SetBinArea( area );
1109 * \brief Update the entries with the shadow dictionary.
1110 * Only non even entries are analyzed
1112 void Document::UpdateShaEntries()
1117 /// \todo TODO : still any use to explore recursively the whole structure?
1119 for(ListTag::iterator it=listEntries.begin();
1120 it!=listEntries.end();
1123 // Odd group => from public dictionary
1124 if((*it)->GetGroup()%2==0)
1127 // Peer group => search the corresponding dict entry
1129 entry=RefShaDict->GetDictEntryByNumber((*it)->GetGroup(),(*it)->GetElement());
1133 if((*it)->IsImplicitVR())
1138 (*it)->SetValue(GetDocEntryUnvalue(*it)); // to go on compiling
1140 // Set the new entry and the new value
1141 (*it)->SetDictEntry(entry);
1142 CheckDocEntryVR(*it,vr);
1144 (*it)->SetValue(GetDocEntryValue(*it)); // to go on compiling
1149 // Remove precedent value transformation
1150 (*it)->SetDictEntry(NewVirtualDictEntry((*it)->GetGroup(),(*it)->GetElement(),vr));
1157 * \brief Searches within the Header Entries for a Dicom Element of
1159 * @param tagName name of the searched Dicom Element.
1160 * @return Corresponding Dicom Element when it exists, and NULL
1163 DocEntry* Document::GetDocEntryByName(TagName const & tagName)
1165 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
1171 return GetDocEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
1175 * \brief retrieves a Dicom Element (the first one) using (group, element)
1176 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1177 * if you think it's NOT UNIQUE, check the count number
1178 * and use iterators to retrieve ALL the Dicoms Elements within
1179 * a given couple (group, element)
1180 * @param group Group number of the searched Dicom Element
1181 * @param element Element number of the searched Dicom Element
1184 DocEntry* Document::GetDocEntryByNumber(uint16_t group, uint16_t element)
1186 TagKey key = DictEntry::TranslateToKey(group, element);
1187 if ( !TagHT.count(key))
1191 return TagHT.find(key)->second;
1195 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1196 * returns a result when the corresponding entry is of type
1198 * @return When present, the corresponding ValEntry.
1200 ValEntry* Document::GetValEntryByNumber(uint16_t group, uint16_t element)
1202 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1203 if ( !currentEntry )
1207 if ( ValEntry* valEntry = dynamic_cast<ValEntry*>(currentEntry) )
1211 dbg.Verbose(0, "Document::GetValEntryByNumber: unfound ValEntry.");
1217 * \brief Loads the element while preserving the current
1218 * underlying file position indicator as opposed to
1219 * to LoadDocEntry that modifies it.
1220 * @param entry Header Entry whose value shall be loaded.
1223 void Document::LoadDocEntrySafe(DocEntry * entry)
1227 long PositionOnEntry = Fp->tellg();
1228 LoadDocEntry(entry);
1229 Fp->seekg(PositionOnEntry, std::ios_base::beg);
1234 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1236 * @return The properly swaped 32 bits integer.
1238 uint32_t Document::SwapLong(uint32_t a)
1245 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1246 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1250 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1254 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1257 //std::cout << "swapCode= " << SwapCode << std::endl;
1258 dbg.Error(" Document::SwapLong : unset swap code");
1265 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1267 * @return The properly unswaped 32 bits integer.
1269 uint32_t Document::UnswapLong(uint32_t a)
1275 * \brief Swaps the bytes so they agree with the processor order
1276 * @return The properly swaped 16 bits integer.
1278 uint16_t Document::SwapShort(uint16_t a)
1280 if ( SwapCode == 4321 || SwapCode == 2143 )
1282 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1288 * \brief Unswaps the bytes so they agree with the processor order
1289 * @return The properly unswaped 16 bits integer.
1291 uint16_t Document::UnswapShort(uint16_t a)
1293 return SwapShort(a);
1296 //-----------------------------------------------------------------------------
1300 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1301 * @return length of the parsed set.
1303 void Document::ParseDES(DocEntrySet *set, long offset,
1304 long l_max, bool delim_mode)
1306 DocEntry *newDocEntry = 0;
1310 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1314 newDocEntry = ReadNextDocEntry( );
1320 VRKey vr = newDocEntry->GetVR();
1324 if ( Global::GetVR()->IsVROfGdcmStringRepresentable(vr) )
1326 /////////////////////// ValEntry
1327 ValEntry* newValEntry =
1328 new ValEntry( newDocEntry->GetDictEntry() ); //LEAK
1329 newValEntry->Copy( newDocEntry );
1331 // When "set" is a Document, then we are at the top of the
1332 // hierarchy and the Key is simply of the form ( group, elem )...
1333 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1336 newValEntry->SetKey( newValEntry->GetKey() );
1338 // ...but when "set" is a SQItem, we are inserting this new
1339 // valEntry in a sequence item. Hence the key has the
1340 // generalized form (refer to \ref BaseTagKey):
1341 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1343 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1344 + newValEntry->GetKey() );
1347 LoadDocEntry( newValEntry );
1348 bool delimitor=newValEntry->IsItemDelimitor();
1349 if( !set->AddEntry( newValEntry ) )
1351 // If here expect big troubles
1352 delete newValEntry; //otherwise mem leak
1360 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1368 if ( ! Global::GetVR()->IsVROfGdcmBinaryRepresentable(vr) )
1370 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1371 dbg.Verbose(0, "Document::ParseDES: neither Valentry, "
1372 "nor BinEntry. Probably unknown VR.");
1375 //////////////////// BinEntry or UNKOWN VR:
1376 BinEntry* newBinEntry = new BinEntry( newDocEntry ); //LEAK
1378 // When "this" is a Document the Key is simply of the
1379 // form ( group, elem )...
1380 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1383 newBinEntry->SetKey( newBinEntry->GetKey() );
1385 // but when "this" is a SQItem, we are inserting this new
1386 // valEntry in a sequence item, and the kay has the
1387 // generalized form (refer to \ref BaseTagKey):
1388 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1390 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1391 + newBinEntry->GetKey() );
1394 LoadDocEntry( newBinEntry );
1395 if( !set->AddEntry( newBinEntry ) )
1397 //Expect big troubles if here
1402 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1403 && ( newDocEntry->GetElement() == 0x0010 ) )
1405 TransferSyntaxType ts = GetTransferSyntax();
1406 if ( ts == RLELossless )
1408 long PositionOnEntry = Fp->tellg();
1409 Fp->seekg( newDocEntry->GetOffset(), std::ios_base::beg );
1411 Fp->seekg( PositionOnEntry, std::ios_base::beg );
1413 else if ( IsJPEG() )
1415 long PositionOnEntry = Fp->tellg();
1416 Fp->seekg( newDocEntry->GetOffset(), std::ios_base::beg );
1417 ComputeJPEGFragmentInfo();
1418 Fp->seekg( PositionOnEntry, std::ios_base::beg );
1422 // Just to make sure we are at the beginning of next entry.
1423 SkipToNextDocEntry(newDocEntry);
1424 //delete newDocEntry;
1429 unsigned long l = newDocEntry->GetReadLength();
1430 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1432 if ( l == 0xffffffff )
1441 // no other way to create it ...
1442 SeqEntry* newSeqEntry =
1443 new SeqEntry( newDocEntry->GetDictEntry() );
1444 newSeqEntry->Copy( newDocEntry );
1445 newSeqEntry->SetDelimitorMode( delim_mode );
1447 // At the top of the hierarchy, stands a Document. When "set"
1448 // is a Document, then we are building the first depth level.
1449 // Hence the SeqEntry we are building simply has a depth
1451 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1454 newSeqEntry->SetDepthLevel( 1 );
1455 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1457 // But when "set" is allready a SQItem, we are building a nested
1458 // sequence, and hence the depth level of the new SeqEntry
1459 // we are building, is one level deeper:
1460 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1462 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1463 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1464 + newSeqEntry->GetKey() );
1468 { // Don't try to parse zero-length sequences
1469 ParseSQ( newSeqEntry,
1470 newDocEntry->GetOffset(),
1473 set->AddEntry( newSeqEntry );
1474 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1485 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1486 * @return parsed length for this level
1488 void Document::ParseSQ( SeqEntry* seqEntry,
1489 long offset, long l_max, bool delim_mode)
1491 int SQItemNumber = 0;
1496 DocEntry* newDocEntry = ReadNextDocEntry();
1499 // FIXME Should warn user
1504 if ( newDocEntry->IsSequenceDelimitor() )
1506 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1510 if ( !delim_mode && ((long)(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);
1537 seqEntry->AddEntry( itemSQ, SQItemNumber );
1539 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1547 * \brief Loads the element content if its length doesn't exceed
1548 * the value specified with Document::SetMaxSizeLoadEntry()
1549 * @param entry Header Entry (Dicom Element) to be dealt with
1551 void Document::LoadDocEntry(DocEntry* entry)
1553 uint16_t group = entry->GetGroup();
1554 std::string vr = entry->GetVR();
1555 uint32_t length = entry->GetLength();
1557 Fp->seekg((long)entry->GetOffset(), std::ios_base::beg);
1559 // A SeQuence "contains" a set of Elements.
1560 // (fffe e000) tells us an Element is beginning
1561 // (fffe e00d) tells us an Element just ended
1562 // (fffe e0dd) tells us the current SeQuence just ended
1563 if( group == 0xfffe )
1565 // NO more value field for SQ !
1569 // When the length is zero things are easy:
1572 ((ValEntry *)entry)->SetValue("");
1576 // The elements whose length is bigger than the specified upper bound
1577 // are not loaded. Instead we leave a short notice of the offset of
1578 // the element content and it's length.
1580 std::ostringstream s;
1581 if (length > MaxSizeLoadEntry)
1583 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1585 //s << "gdcm::NotLoaded (BinEntry)";
1586 s << GDCM_NOTLOADED;
1587 s << " Address:" << (long)entry->GetOffset();
1588 s << " Length:" << entry->GetLength();
1589 s << " x(" << std::hex << entry->GetLength() << ")";
1590 binEntryPtr->SetValue(s.str());
1592 // Be carefull : a BinEntry IS_A ValEntry ...
1593 else if (ValEntry* valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1595 // s << "gdcm::NotLoaded. (ValEntry)";
1596 s << GDCM_NOTLOADED;
1597 s << " Address:" << (long)entry->GetOffset();
1598 s << " Length:" << entry->GetLength();
1599 s << " x(" << std::hex << entry->GetLength() << ")";
1600 valEntryPtr->SetValue(s.str());
1605 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1606 << "nor a ValEntry ?! Should never print that !" << std::endl;
1609 // to be sure we are at the end of the value ...
1610 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1611 std::ios_base::beg);
1615 // When we find a BinEntry not very much can be done :
1616 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1618 s << GDCM_BINLOADED;
1619 binEntryPtr->SetValue(s.str());
1620 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1624 /// \todo Any compacter code suggested (?)
1625 if ( IsDocEntryAnInteger(entry) )
1629 // When short integer(s) are expected, read and convert the following
1630 // n *two characters properly i.e. consider them as short integers as
1631 // opposed to strings.
1632 // Elements with Value Multiplicity > 1
1633 // contain a set of integers (not a single one)
1634 if (vr == "US" || vr == "SS")
1637 NewInt = ReadInt16();
1641 for (int i=1; i < nbInt; i++)
1644 NewInt = ReadInt16();
1649 // See above comment on multiple integers (mutatis mutandis).
1650 else if (vr == "UL" || vr == "SL")
1653 NewInt = ReadInt32();
1657 for (int i=1; i < nbInt; i++)
1660 NewInt = ReadInt32();
1665 #ifdef GDCM_NO_ANSI_STRING_STREAM
1666 s << std::ends; // to avoid oddities on Solaris
1667 #endif //GDCM_NO_ANSI_STRING_STREAM
1669 ((ValEntry *)entry)->SetValue(s.str());
1673 // FIXME: We need an additional byte for storing \0 that is not on disk
1674 char *str = new char[length+1];
1675 Fp->read(str, (size_t)length);
1676 str[length] = '\0'; //this is only useful when length is odd
1677 // Special DicomString call to properly handle \0 and even length
1678 std::string newValue;
1681 newValue = Util::DicomString(str, length+1);
1682 //dbg.Verbose(0, "Warning: bad length: ", length );
1683 dbg.Verbose(0, "For string :", newValue.c_str());
1684 // Since we change the length of string update it length
1685 entry->SetReadLength(length+1);
1689 newValue = Util::DicomString(str, length);
1693 if ( ValEntry* valEntry = dynamic_cast<ValEntry* >(entry) )
1695 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1697 dbg.Verbose(1, "Document::LoadDocEntry",
1698 "unread element value");
1699 valEntry->SetValue(GDCM_UNREAD);
1705 // Because of correspondance with the VR dic
1706 valEntry->SetValue(newValue);
1710 valEntry->SetValue(newValue);
1715 dbg.Error(true, "Document::LoadDocEntry"
1716 "Should have a ValEntry, here !");
1722 * \brief Find the value Length of the passed Header Entry
1723 * @param entry Header Entry whose length of the value shall be loaded.
1725 void Document::FindDocEntryLength( DocEntry *entry )
1726 throw ( FormatError )
1728 uint16_t element = entry->GetElement();
1729 std::string vr = entry->GetVR();
1732 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1734 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1736 // The following reserved two bytes (see PS 3.5-2003, section
1737 // "7.1.2 Data element structure with explicit vr", p 27) must be
1738 // skipped before proceeding on reading the length on 4 bytes.
1739 Fp->seekg( 2L, std::ios_base::cur);
1740 uint32_t length32 = ReadInt32();
1742 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1747 /// \todo rename that to FindDocEntryLengthOBOrOW since
1748 /// the above test is on both OB and OW...
1749 lengthOB = FindDocEntryLengthOB();
1751 catch ( FormatUnexpected )
1753 // Computing the length failed (this happens with broken
1754 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1755 // chance to get the pixels by deciding the element goes
1756 // until the end of the file. Hence we artificially fix the
1757 // the length and proceed.
1758 long currentPosition = Fp->tellg();
1759 Fp->seekg(0L,std::ios_base::end);
1760 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1761 Fp->seekg(currentPosition, std::ios_base::beg);
1762 entry->SetLength(lengthUntilEOF);
1765 entry->SetLength(lengthOB);
1768 FixDocEntryFoundLength(entry, length32);
1772 // Length is encoded on 2 bytes.
1773 length16 = ReadInt16();
1775 // We can tell the current file is encoded in big endian (like
1776 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1777 // and it's value is the one of the encoding of a big endian file.
1778 // In order to deal with such big endian encoded files, we have
1779 // (at least) two strategies:
1780 // * when we load the "Transfer Syntax" tag with value of big endian
1781 // encoding, we raise the proper flags. Then we wait for the end
1782 // of the META group (0x0002) among which is "Transfer Syntax",
1783 // before switching the swap code to big endian. We have to postpone
1784 // the switching of the swap code since the META group is fully encoded
1785 // in little endian, and big endian coding only starts at the next
1786 // group. The corresponding code can be hard to analyse and adds
1787 // many additional unnecessary tests for regular tags.
1788 // * the second strategy consists in waiting for trouble, that shall
1789 // appear when we find the first group with big endian encoding. This
1790 // is easy to detect since the length of a "Group Length" tag (the
1791 // ones with zero as element number) has to be of 4 (0x0004). When we
1792 // encounter 1024 (0x0400) chances are the encoding changed and we
1793 // found a group with big endian encoding.
1794 // We shall use this second strategy. In order to make sure that we
1795 // can interpret the presence of an apparently big endian encoded
1796 // length of a "Group Length" without committing a big mistake, we
1797 // add an additional check: we look in the already parsed elements
1798 // for the presence of a "Transfer Syntax" whose value has to be "big
1799 // endian encoding". When this is the case, chances are we have got our
1800 // hands on a big endian encoded file: we switch the swap code to
1801 // big endian and proceed...
1802 if ( element == 0x0000 && length16 == 0x0400 )
1804 TransferSyntaxType ts = GetTransferSyntax();
1805 if ( ts != ExplicitVRBigEndian )
1807 throw FormatError( "Document::FindDocEntryLength()",
1808 " not explicit VR." );
1812 SwitchSwapToBigEndian();
1813 // Restore the unproperly loaded values i.e. the group, the element
1814 // and the dictionary entry depending on them.
1815 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1816 uint16_t correctElem = SwapShort( entry->GetElement() );
1817 DictEntry* newTag = GetDictEntryByNumber( correctGroup,
1821 // This correct tag is not in the dictionary. Create a new one.
1822 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1824 // FIXME this can create a memory leaks on the old entry that be
1825 // left unreferenced.
1826 entry->SetDictEntry( newTag );
1829 // Heuristic: well, some files are really ill-formed.
1830 if ( length16 == 0xffff)
1832 // 0xffff means that we deal with 'Unknown Length' Sequence
1835 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1840 // Either implicit VR or a non DICOM conformal (see note below) explicit
1841 // VR that ommited the VR of (at least) this element. Farts happen.
1842 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1843 // on Data elements "Implicit and Explicit VR Data Elements shall
1844 // not coexist in a Data Set and Data Sets nested within it".]
1845 // Length is on 4 bytes.
1847 FixDocEntryFoundLength( entry, ReadInt32() );
1853 * \brief Find the Value Representation of the current Dicom Element.
1856 void Document::FindDocEntryVR( DocEntry *entry )
1858 if ( Filetype != ExplicitVR )
1865 long positionOnEntry = Fp->tellg();
1866 // Warning: we believe this is explicit VR (Value Representation) because
1867 // we used a heuristic that found "UL" in the first tag. Alas this
1868 // doesn't guarantee that all the tags will be in explicit VR. In some
1869 // cases (see e-film filtered files) one finds implicit VR tags mixed
1870 // within an explicit VR file. Hence we make sure the present tag
1871 // is in explicit VR and try to fix things if it happens not to be
1874 Fp->read (vr, (size_t)2);
1877 if( !CheckDocEntryVR(entry, vr) )
1879 Fp->seekg(positionOnEntry, std::ios_base::beg);
1880 // When this element is known in the dictionary we shall use, e.g. for
1881 // the semantics (see the usage of IsAnInteger), the VR proposed by the
1882 // dictionary entry. Still we have to flag the element as implicit since
1883 // we know now our assumption on expliciteness is not furfilled.
1885 if ( entry->IsVRUnknown() )
1887 entry->SetVR("Implicit");
1889 entry->SetImplicitVR();
1894 * \brief Check the correspondance between the VR of the header entry
1895 * and the taken VR. If they are different, the header entry is
1896 * updated with the new VR.
1897 * @param entry Header Entry to check
1898 * @param vr Dicom Value Representation
1899 * @return false if the VR is incorrect of if the VR isn't referenced
1900 * otherwise, it returns true
1902 bool Document::CheckDocEntryVR(DocEntry *entry, VRKey vr)
1905 bool realExplicit = true;
1907 // Assume we are reading a falsely explicit VR file i.e. we reached
1908 // a tag where we expect reading a VR but are in fact we read the
1909 // first to bytes of the length. Then we will interogate (through find)
1910 // the dicom_vr dictionary with oddities like "\004\0" which crashes
1911 // both GCC and VC++ implementations of the STL map. Hence when the
1912 // expected VR read happens to be non-ascii characters we consider
1913 // we hit falsely explicit VR tag.
1915 if ( !isalpha((unsigned char)vr[0]) && !isalpha((unsigned char)vr[1]) )
1917 realExplicit = false;
1920 // CLEANME searching the dicom_vr at each occurence is expensive.
1921 // PostPone this test in an optional integrity check at the end
1922 // of parsing or only in debug mode.
1923 if ( realExplicit && !Global::GetVR()->Count(vr) )
1925 realExplicit = false;
1928 if ( !realExplicit )
1930 // We thought this was explicit VR, but we end up with an
1931 // implicit VR tag. Let's backtrack.
1932 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
1933 entry->GetGroup(), entry->GetElement());
1934 dbg.Verbose(1, "Document::FindVR: ", msg.c_str());
1936 if( entry->GetGroup() % 2 && entry->GetElement() == 0x0000)
1938 // Group length is UL !
1939 DictEntry* newEntry = NewVirtualDictEntry(
1940 entry->GetGroup(), entry->GetElement(),
1941 "UL", "FIXME", "Group Length");
1942 entry->SetDictEntry( newEntry );
1947 if ( entry->IsVRUnknown() )
1949 // When not a dictionary entry, we can safely overwrite the VR.
1950 if( entry->GetElement() == 0x0000 )
1952 // Group length is UL !
1960 else if ( entry->GetVR() != vr )
1962 // The VR present in the file and the dictionary disagree. We assume
1963 // the file writer knew best and use the VR of the file. Since it would
1964 // be unwise to overwrite the VR of a dictionary (since it would
1965 // compromise it's next user), we need to clone the actual DictEntry
1966 // and change the VR for the read one.
1967 DictEntry* newEntry = NewVirtualDictEntry(
1968 entry->GetGroup(), entry->GetElement(),
1969 vr, "FIXME", entry->GetName());
1970 entry->SetDictEntry(newEntry);
1977 * \brief Get the transformed value of the header entry. The VR value
1978 * is used to define the transformation to operate on the value
1979 * \warning NOT end user intended method !
1980 * @param entry entry to tranform
1981 * @return Transformed entry value
1983 std::string Document::GetDocEntryValue(DocEntry *entry)
1985 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1987 std::string val = ((ValEntry *)entry)->GetValue();
1988 std::string vr = entry->GetVR();
1989 uint32_t length = entry->GetLength();
1990 std::ostringstream s;
1993 // When short integer(s) are expected, read and convert the following
1994 // n * 2 bytes properly i.e. as a multivaluated strings
1995 // (each single value is separated fromthe next one by '\'
1996 // as usual for standard multivaluated filels
1997 // Elements with Value Multiplicity > 1
1998 // contain a set of short integers (not a single one)
2000 if( vr == "US" || vr == "SS" )
2005 for (int i=0; i < nbInt; i++)
2011 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
2012 newInt16 = SwapShort( newInt16 );
2017 // When integer(s) are expected, read and convert the following
2018 // n * 4 bytes properly i.e. as a multivaluated strings
2019 // (each single value is separated fromthe next one by '\'
2020 // as usual for standard multivaluated filels
2021 // Elements with Value Multiplicity > 1
2022 // contain a set of integers (not a single one)
2023 else if( vr == "UL" || vr == "SL" )
2028 for (int i=0; i < nbInt; i++)
2034 newInt32 = ( val[4*i+0] & 0xFF )
2035 + (( val[4*i+1] & 0xFF ) << 8 )
2036 + (( val[4*i+2] & 0xFF ) << 16 )
2037 + (( val[4*i+3] & 0xFF ) << 24 );
2038 newInt32 = SwapLong( newInt32 );
2042 #ifdef GDCM_NO_ANSI_STRING_STREAM
2043 s << std::ends; // to avoid oddities on Solaris
2044 #endif //GDCM_NO_ANSI_STRING_STREAM
2048 return ((ValEntry *)entry)->GetValue();
2052 * \brief Get the reverse transformed value of the header entry. The VR
2053 * value is used to define the reverse transformation to operate on
2055 * \warning NOT end user intended method !
2056 * @param entry Entry to reverse transform
2057 * @return Reverse transformed entry value
2059 std::string Document::GetDocEntryUnvalue(DocEntry* entry)
2061 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2063 std::string vr = entry->GetVR();
2064 std::vector<std::string> tokens;
2065 std::ostringstream s;
2067 if ( vr == "US" || vr == "SS" )
2071 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
2072 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2073 for (unsigned int i=0; i<tokens.size(); i++)
2075 newInt16 = atoi(tokens[i].c_str());
2076 s << ( newInt16 & 0xFF )
2077 << (( newInt16 >> 8 ) & 0xFF );
2081 if ( vr == "UL" || vr == "SL")
2085 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
2086 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2087 for (unsigned int i=0; i<tokens.size();i++)
2089 newInt32 = atoi(tokens[i].c_str());
2090 s << (char)( newInt32 & 0xFF )
2091 << (char)(( newInt32 >> 8 ) & 0xFF )
2092 << (char)(( newInt32 >> 16 ) & 0xFF )
2093 << (char)(( newInt32 >> 24 ) & 0xFF );
2098 #ifdef GDCM_NO_ANSI_STRING_STREAM
2099 s << std::ends; // to avoid oddities on Solaris
2100 #endif //GDCM_NO_ANSI_STRING_STREAM
2104 return ((ValEntry *)entry)->GetValue();
2108 * \brief Skip a given Header Entry
2109 * \warning NOT end user intended method !
2110 * @param entry entry to skip
2112 void Document::SkipDocEntry(DocEntry *entry)
2114 SkipBytes(entry->GetLength());
2118 * \brief Skips to the begining of the next Header Entry
2119 * \warning NOT end user intended method !
2120 * @param entry entry to skip
2122 void Document::SkipToNextDocEntry(DocEntry *entry)
2124 Fp->seekg((long)(entry->GetOffset()), std::ios_base::beg);
2125 Fp->seekg( (long)(entry->GetReadLength()), std::ios_base::cur);
2129 * \brief When the length of an element value is obviously wrong (because
2130 * the parser went Jabberwocky) one can hope improving things by
2131 * applying some heuristics.
2132 * @param entry entry to check
2133 * @param foundLength fist assumption about length
2135 void Document::FixDocEntryFoundLength(DocEntry *entry,
2136 uint32_t foundLength)
2138 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
2139 if ( foundLength == 0xffffffff)
2144 uint16_t gr = entry->GetGroup();
2145 uint16_t el = entry->GetElement();
2147 if ( foundLength % 2)
2149 std::ostringstream s;
2150 s << "Warning : Tag with uneven length "
2152 << " in x(" << std::hex << gr << "," << el <<")" << std::dec;
2153 dbg.Verbose(0, s.str().c_str());
2156 //////// Fix for some naughty General Electric images.
2157 // Allthough not recent many such GE corrupted images are still present
2158 // on Creatis hard disks. Hence this fix shall remain when such images
2159 // are no longer in user (we are talking a few years, here)...
2160 // Note: XMedCom probably uses such a trick since it is able to read
2161 // those pesky GE images ...
2162 if ( foundLength == 13)
2164 // Only happens for this length !
2165 if ( entry->GetGroup() != 0x0008
2166 || ( entry->GetElement() != 0x0070
2167 && entry->GetElement() != 0x0080 ) )
2170 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
2174 //////// Fix for some brain-dead 'Leonardo' Siemens images.
2175 // Occurence of such images is quite low (unless one leaves close to a
2176 // 'Leonardo' source. Hence, one might consider commenting out the
2177 // following fix on efficiency reasons.
2178 else if ( entry->GetGroup() == 0x0009
2179 && ( entry->GetElement() == 0x1113
2180 || entry->GetElement() == 0x1114 ) )
2183 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
2186 else if ( entry->GetVR() == "SQ" )
2188 foundLength = 0; // ReadLength is unchanged
2191 //////// We encountered a 'delimiter' element i.e. a tag of the form
2192 // "fffe|xxxx" which is just a marker. Delimiters length should not be
2193 // taken into account.
2194 else if( entry->GetGroup() == 0xfffe )
2196 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
2197 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
2198 // causes extra troubles...
2199 if( entry->GetElement() != 0x0000 )
2205 entry->SetUsableLength(foundLength);
2209 * \brief Apply some heuristics to predict whether the considered
2210 * element value contains/represents an integer or not.
2211 * @param entry The element value on which to apply the predicate.
2212 * @return The result of the heuristical predicate.
2214 bool Document::IsDocEntryAnInteger(DocEntry *entry)
2216 uint16_t element = entry->GetElement();
2217 uint16_t group = entry->GetGroup();
2218 const std::string & vr = entry->GetVR();
2219 uint32_t length = entry->GetLength();
2221 // When we have some semantics on the element we just read, and if we
2222 // a priori know we are dealing with an integer, then we shall be
2223 // able to swap it's element value properly.
2224 if ( element == 0 ) // This is the group length of the group
2232 // Allthough this should never happen, still some images have a
2233 // corrupted group length [e.g. have a glance at offset x(8336) of
2234 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2235 // Since for dicom compliant and well behaved headers, the present
2236 // test is useless (and might even look a bit paranoid), when we
2237 // encounter such an ill-formed image, we simply display a warning
2238 // message and proceed on parsing (while crossing fingers).
2239 std::ostringstream s;
2240 long filePosition = Fp->tellg();
2241 s << "Erroneous Group Length element length on : (" \
2242 << std::hex << group << " , " << element
2243 << ") -before- position x(" << filePosition << ")"
2244 << "lgt : " << length;
2245 dbg.Verbose(0, "Document::IsDocEntryAnInteger", s.str().c_str() );
2249 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2258 * \brief Find the Length till the next sequence delimiter
2259 * \warning NOT end user intended method !
2263 uint32_t Document::FindDocEntryLengthOB()
2264 throw( FormatUnexpected )
2266 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2267 long positionOnEntry = Fp->tellg();
2268 bool foundSequenceDelimiter = false;
2269 uint32_t totalLength = 0;
2271 while ( !foundSequenceDelimiter )
2277 group = ReadInt16();
2280 catch ( FormatError )
2282 throw FormatError("Document::FindDocEntryLengthOB()",
2283 " group or element not present.");
2286 // We have to decount the group and element we just read
2289 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2291 dbg.Verbose(1, "Document::FindDocEntryLengthOB: neither an Item "
2292 "tag nor a Sequence delimiter tag.");
2293 Fp->seekg(positionOnEntry, std::ios_base::beg);
2294 throw FormatUnexpected("Document::FindDocEntryLengthOB()",
2295 "Neither an Item tag nor a Sequence "
2299 if ( elem == 0xe0dd )
2301 foundSequenceDelimiter = true;
2304 uint32_t itemLength = ReadInt32();
2305 // We add 4 bytes since we just read the ItemLength with ReadInt32
2306 totalLength += itemLength + 4;
2307 SkipBytes(itemLength);
2309 if ( foundSequenceDelimiter )
2314 Fp->seekg( positionOnEntry, std::ios_base::beg);
2319 * \brief Reads a supposed to be 16 Bits integer
2320 * (swaps it depending on processor endianity)
2321 * @return read value
2323 uint16_t Document::ReadInt16()
2324 throw( FormatError )
2327 Fp->read ((char*)&g, (size_t)2);
2330 throw FormatError( "Document::ReadInt16()", " file error." );
2334 throw FormatError( "Document::ReadInt16()", "EOF." );
2341 * \brief Reads a supposed to be 32 Bits integer
2342 * (swaps it depending on processor endianity)
2343 * @return read value
2345 uint32_t Document::ReadInt32()
2346 throw( FormatError )
2349 Fp->read ((char*)&g, (size_t)4);
2352 throw FormatError( "Document::ReadInt32()", " file error." );
2356 throw FormatError( "Document::ReadInt32()", "EOF." );
2363 * \brief skips bytes inside the source file
2364 * \warning NOT end user intended method !
2367 void Document::SkipBytes(uint32_t nBytes)
2369 //FIXME don't dump the returned value
2370 Fp->seekg((long)nBytes, std::ios_base::cur);
2374 * \brief Loads all the needed Dictionaries
2375 * \warning NOT end user intended method !
2377 void Document::Initialise()
2379 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2381 RLEInfo = new RLEFramesInfo;
2382 JPEGInfo = new JPEGFragmentsInfo;
2386 * \brief Discover what the swap code is (among little endian, big endian,
2387 * bad little endian, bad big endian).
2389 * @return false when we are absolutely sure
2390 * it's neither ACR-NEMA nor DICOM
2391 * true when we hope ours assuptions are OK
2393 bool Document::CheckSwap()
2395 // The only guaranted way of finding the swap code is to find a
2396 // group tag since we know it's length has to be of four bytes i.e.
2397 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2398 // occurs when we can't find such group...
2400 uint32_t x = 4; // x : for ntohs
2401 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2405 char deb[256]; //HEADER_LENGTH_TO_READ];
2407 // First, compare HostByteOrder and NetworkByteOrder in order to
2408 // determine if we shall need to swap bytes (i.e. the Endian type).
2409 if ( x == ntohs(x) )
2418 // The easiest case is the one of a DICOM header, since it possesses a
2419 // file preamble where it suffice to look for the string "DICM".
2420 Fp->read(deb, HEADER_LENGTH_TO_READ);
2422 char *entCur = deb + 128;
2423 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2425 dbg.Verbose(1, "Document::CheckSwap:", "looks like DICOM Version3");
2427 // Next, determine the value representation (VR). Let's skip to the
2428 // first element (0002, 0000) and check there if we find "UL"
2429 // - or "OB" if the 1st one is (0002,0001) -,
2430 // in which case we (almost) know it is explicit VR.
2431 // WARNING: if it happens to be implicit VR then what we will read
2432 // is the length of the group. If this ascii representation of this
2433 // length happens to be "UL" then we shall believe it is explicit VR.
2434 // FIXME: in order to fix the above warning, we could read the next
2435 // element value (or a couple of elements values) in order to make
2436 // sure we are not commiting a big mistake.
2437 // We need to skip :
2438 // * the 128 bytes of File Preamble (often padded with zeroes),
2439 // * the 4 bytes of "DICM" string,
2440 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2441 // i.e. a total of 136 bytes.
2445 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2446 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2447 // *Implicit* VR. -and it is !-
2449 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2450 memcmp(entCur, "OB", (size_t)2) == 0 ||
2451 memcmp(entCur, "UI", (size_t)2) == 0 ||
2452 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2453 // when Write DCM *adds*
2455 // Use Document::dicom_vr to test all the possibilities
2456 // instead of just checking for UL, OB and UI !? group 0000
2458 Filetype = ExplicitVR;
2459 dbg.Verbose(1, "Document::CheckSwap:",
2460 "explicit Value Representation");
2464 Filetype = ImplicitVR;
2465 dbg.Verbose(1, "Document::CheckSwap:",
2466 "not an explicit Value Representation");
2472 dbg.Verbose(1, "Document::CheckSwap:",
2473 "HostByteOrder != NetworkByteOrder");
2478 dbg.Verbose(1, "Document::CheckSwap:",
2479 "HostByteOrder = NetworkByteOrder");
2482 // Position the file position indicator at first tag (i.e.
2483 // after the file preamble and the "DICM" string).
2484 Fp->seekg(0, std::ios_base::beg);
2485 Fp->seekg ( 132L, std::ios_base::beg);
2489 // Alas, this is not a DicomV3 file and whatever happens there is no file
2490 // preamble. We can reset the file position indicator to where the data
2491 // is (i.e. the beginning of the file).
2492 dbg.Verbose(1, "Document::CheckSwap:", "not a DICOM Version3 file");
2493 Fp->seekg(0, std::ios_base::beg);
2495 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2496 // By clean we mean that the length of the first tag is written down.
2497 // If this is the case and since the length of the first group HAS to be
2498 // four (bytes), then determining the proper swap code is straightforward.
2501 // We assume the array of char we are considering contains the binary
2502 // representation of a 32 bits integer. Hence the following dirty
2504 s32 = *((uint32_t *)(entCur));
2525 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2526 // It is time for despaired wild guesses.
2527 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2528 // i.e. the 'group length' element is not present :
2530 // check the supposed-to-be 'group number'
2531 // in ( 0x0001 .. 0x0008 )
2532 // to determine ' SwapCode' value .
2533 // Only 0 or 4321 will be possible
2534 // (no oportunity to check for the formerly well known
2535 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2536 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2537 // the file IS NOT ACR-NEMA nor DICOM V3
2538 // Find a trick to tell it the caller...
2540 s16 = *((uint16_t *)(deb));
2567 dbg.Verbose(0, "Document::CheckSwap:",
2568 "ACR/NEMA unfound swap info (Really hopeless !)");
2572 // Then the only info we have is the net2host one.
2584 * \brief Restore the unproperly loaded values i.e. the group, the element
2585 * and the dictionary entry depending on them.
2587 void Document::SwitchSwapToBigEndian()
2589 dbg.Verbose(1, "Document::SwitchSwapToBigEndian",
2590 "Switching to BigEndian mode.");
2591 if ( SwapCode == 0 )
2595 else if ( SwapCode == 4321 )
2599 else if ( SwapCode == 3412 )
2603 else if ( SwapCode == 2143 )
2610 * \brief during parsing, Header Elements too long are not loaded in memory
2613 void Document::SetMaxSizeLoadEntry(long newSize)
2619 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2621 MaxSizeLoadEntry = 0xffffffff;
2624 MaxSizeLoadEntry = newSize;
2629 * \brief Header Elements too long will not be printed
2630 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2633 void Document::SetMaxSizePrintEntry(long newSize)
2635 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2640 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2642 MaxSizePrintEntry = 0xffffffff;
2645 MaxSizePrintEntry = newSize;
2651 * \brief Handle broken private tag from Philips NTSCAN
2652 * where the endianess is being switch to BigEndian for no
2656 void Document::HandleBrokenEndian(uint16_t group, uint16_t elem)
2658 // Endian reversion. Some files contain groups of tags with reversed endianess.
2659 static int reversedEndian = 0;
2660 // try to fix endian switching in the middle of headers
2661 if ((group == 0xfeff) && (elem == 0x00e0))
2663 // start endian swap mark for group found
2665 SwitchSwapToBigEndian();
2670 else if ((group == 0xfffe) && (elem == 0xe00d) && reversedEndian)
2672 // end of reversed endian group
2674 SwitchSwapToBigEndian();
2680 * \brief Read the next tag but WITHOUT loading it's value
2681 * (read the 'Group Number', the 'Element Number',
2682 * gets the Dict Entry
2683 * gets the VR, gets the length, gets the offset value)
2684 * @return On succes the newly created DocEntry, NULL on failure.
2686 DocEntry* Document::ReadNextDocEntry()
2693 group = ReadInt16();
2696 catch ( FormatError e )
2698 // We reached the EOF (or an error occured) therefore
2699 // header parsing has to be considered as finished.
2704 HandleBrokenEndian(group, elem);
2705 DocEntry *newEntry = NewDocEntryByNumber(group, elem);
2706 FindDocEntryVR(newEntry);
2710 FindDocEntryLength(newEntry);
2712 catch ( FormatError e )
2720 newEntry->SetOffset(Fp->tellg());
2727 * \brief Generate a free TagKey i.e. a TagKey that is not present
2728 * in the TagHt dictionary.
2729 * @param group The generated tag must belong to this group.
2730 * @return The element of tag with given group which is fee.
2732 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2734 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2736 TagKey key = DictEntry::TranslateToKey(group, elem);
2737 if (TagHT.count(key) == 0)
2746 * \brief Assuming the internal file pointer \ref Document::Fp
2747 * is placed at the beginning of a tag check whether this
2748 * tag is (TestGroup, TestElement).
2749 * \warning On success the internal file pointer \ref Document::Fp
2750 * is modified to point after the tag.
2751 * On failure (i.e. when the tag wasn't the expected tag
2752 * (TestGroup, TestElement) the internal file pointer
2753 * \ref Document::Fp is restored to it's original position.
2754 * @param testGroup The expected group of the tag.
2755 * @param testElement The expected Element of the tag.
2756 * @return True on success, false otherwise.
2758 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2760 long positionOnEntry = Fp->tellg();
2761 long currentPosition = Fp->tellg(); // On debugging purposes
2763 //// Read the Item Tag group and element, and make
2764 // sure they are what we expected:
2765 uint16_t itemTagGroup;
2766 uint16_t itemTagElement;
2769 itemTagGroup = ReadInt16();
2770 itemTagElement = ReadInt16();
2772 catch ( FormatError e )
2774 //std::cerr << e << std::endl;
2777 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2779 std::ostringstream s;
2780 s << " We should have found tag (";
2781 s << std::hex << testGroup << "," << testElement << ")" << std::endl;
2782 s << " but instead we encountered tag (";
2783 s << std::hex << itemTagGroup << "," << itemTagElement << ")"
2785 s << " at address: " << (unsigned)currentPosition << std::endl;
2786 dbg.Verbose(0, "Document::ReadItemTagLength: wrong Item Tag found:");
2787 dbg.Verbose(0, s.str().c_str());
2788 Fp->seekg(positionOnEntry, std::ios_base::beg);
2796 * \brief Assuming the internal file pointer \ref Document::Fp
2797 * is placed at the beginning of a tag (TestGroup, TestElement),
2798 * read the length associated to the Tag.
2799 * \warning On success the internal file pointer \ref Document::Fp
2800 * is modified to point after the tag and it's length.
2801 * On failure (i.e. when the tag wasn't the expected tag
2802 * (TestGroup, TestElement) the internal file pointer
2803 * \ref Document::Fp is restored to it's original position.
2804 * @param testGroup The expected group of the tag.
2805 * @param testElement The expected Element of the tag.
2806 * @return On success returns the length associated to the tag. On failure
2809 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2811 long positionOnEntry = Fp->tellg();
2812 (void)positionOnEntry;
2814 if ( !ReadTag(testGroup, testElement) )
2819 //// Then read the associated Item Length
2820 long currentPosition = Fp->tellg();
2821 uint32_t itemLength = ReadInt32();
2823 std::ostringstream s;
2824 s << "Basic Item Length is: "
2825 << itemLength << std::endl;
2826 s << " at address: " << (unsigned)currentPosition << std::endl;
2827 dbg.Verbose(0, "Document::ReadItemTagLength: ", s.str().c_str());
2833 * \brief When parsing the Pixel Data of an encapsulated file, read
2834 * the basic offset table (when present, and BTW dump it).
2836 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2838 //// Read the Basic Offset Table Item Tag length...
2839 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2841 // When present, read the basic offset table itself.
2842 // Notes: - since the presence of this basic offset table is optional
2843 // we can't rely on it for the implementation, and we will simply
2844 // trash it's content (when present).
2845 // - still, when present, we could add some further checks on the
2846 // lengths, but we won't bother with such fuses for the time being.
2847 if ( itemLength != 0 )
2849 char* basicOffsetTableItemValue = new char[itemLength + 1];
2850 Fp->read(basicOffsetTableItemValue, itemLength);
2853 for (unsigned int i=0; i < itemLength; i += 4 )
2855 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2857 std::ostringstream s;
2858 s << " Read one length: ";
2859 s << std::hex << individualLength << std::endl;
2861 "Document::ReadAndSkipEncapsulatedBasicOffsetTable: ",
2866 delete[] basicOffsetTableItemValue;
2871 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2872 * Compute the RLE extra information and store it in \ref RLEInfo
2873 * for later pixel retrieval usage.
2875 void Document::ComputeRLEInfo()
2877 TransferSyntaxType ts = GetTransferSyntax();
2878 if ( ts != RLELossless )
2883 // Encoded pixel data: for the time being we are only concerned with
2884 // Jpeg or RLE Pixel data encodings.
2885 // As stated in PS 3.5-2003, section 8.2 p44:
2886 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2887 // value representation OB is used".
2888 // Hence we expect an OB value representation. Concerning OB VR,
2889 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2890 // "For the Value Representations OB and OW, the encoding shall meet the
2891 // following specifications depending on the Data element tag:"
2893 // - the first item in the sequence of items before the encoded pixel
2894 // data stream shall be basic offset table item. The basic offset table
2895 // item value, however, is not required to be present"
2897 ReadAndSkipEncapsulatedBasicOffsetTable();
2899 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2900 // Loop on the individual frame[s] and store the information
2901 // on the RLE fragments in a RLEFramesInfo.
2902 // Note: - when only a single frame is present, this is a
2904 // - when more than one frame are present, then we are in
2905 // the case of a multi-frame image.
2907 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2909 // Parse the RLE Header and store the corresponding RLE Segment
2910 // Offset Table information on fragments of this current Frame.
2911 // Note that the fragment pixels themselves are not loaded
2912 // (but just skipped).
2913 long frameOffset = Fp->tellg();
2915 uint32_t nbRleSegments = ReadInt32();
2916 if ( nbRleSegments > 16 )
2918 // There should be at most 15 segments (refer to RLEFrame class)
2919 dbg.Verbose(0, "Document::ComputeRLEInfo: too many segments.");
2922 uint32_t rleSegmentOffsetTable[16];
2923 for( int k = 1; k <= 15; k++ )
2925 rleSegmentOffsetTable[k] = ReadInt32();
2928 // Deduce from both the RLE Header and the frameLength the
2929 // fragment length, and again store this info in a
2931 long rleSegmentLength[15];
2932 // skipping (not reading) RLE Segments
2933 if ( nbRleSegments > 1)
2935 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2937 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2938 - rleSegmentOffsetTable[k];
2939 SkipBytes(rleSegmentLength[k]);
2943 rleSegmentLength[nbRleSegments] = frameLength
2944 - rleSegmentOffsetTable[nbRleSegments];
2945 SkipBytes(rleSegmentLength[nbRleSegments]);
2947 // Store the collected info
2948 RLEFrame* newFrameInfo = new RLEFrame;
2949 newFrameInfo->NumberFragments = nbRleSegments;
2950 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2952 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2953 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2955 RLEInfo->Frames.push_back( newFrameInfo );
2958 // Make sure that at the end of the item we encounter a 'Sequence
2960 if ( !ReadTag(0xfffe, 0xe0dd) )
2962 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2963 dbg.Verbose(0, " item at end of RLE item sequence");
2968 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2969 * Compute the jpeg extra information (fragment[s] offset[s] and
2970 * length) and store it[them] in \ref JPEGInfo for later pixel
2973 void Document::ComputeJPEGFragmentInfo()
2975 // If you need to, look for comments of ComputeRLEInfo().
2981 ReadAndSkipEncapsulatedBasicOffsetTable();
2983 // Loop on the fragments[s] and store the parsed information in a
2985 long fragmentLength;
2986 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2988 long fragmentOffset = Fp->tellg();
2990 // Store the collected info
2991 JPEGFragment* newFragment = new JPEGFragment;
2992 newFragment->Offset = fragmentOffset;
2993 newFragment->Length = fragmentLength;
2994 JPEGInfo->Fragments.push_back( newFragment );
2996 SkipBytes( fragmentLength );
2999 // Make sure that at the end of the item we encounter a 'Sequence
3001 if ( !ReadTag(0xfffe, 0xe0dd) )
3003 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
3004 dbg.Verbose(0, " item at end of JPEG item sequence");
3009 * \brief Walk recursively the given \ref DocEntrySet, and feed
3010 * the given hash table (\ref TagDocEntryHT) with all the
3011 * \ref DocEntry (Dicom entries) encountered.
3012 * This method does the job for \ref BuildFlatHashTable.
3013 * @param builtHT Where to collect all the \ref DocEntry encountered
3014 * when recursively walking the given set.
3015 * @param set The structure to be traversed (recursively).
3017 void Document::BuildFlatHashTableRecurse( TagDocEntryHT& builtHT,
3020 if (ElementSet* elementSet = dynamic_cast< ElementSet* > ( set ) )
3022 TagDocEntryHT const & currentHT = elementSet->GetTagHT();
3023 for( TagDocEntryHT::const_iterator i = currentHT.begin();
3024 i != currentHT.end();
3027 DocEntry* entry = i->second;
3028 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3030 const ListSQItem& items = seqEntry->GetSQItems();
3031 for( ListSQItem::const_iterator item = items.begin();
3032 item != items.end();
3035 BuildFlatHashTableRecurse( builtHT, *item );
3039 builtHT[entry->GetKey()] = entry;
3044 if (SQItem* SQItemSet = dynamic_cast< SQItem* > ( set ) )
3046 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
3047 for (ListDocEntry::const_iterator i = currentList.begin();
3048 i != currentList.end();
3051 DocEntry* entry = *i;
3052 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3054 const ListSQItem& items = seqEntry->GetSQItems();
3055 for( ListSQItem::const_iterator item = items.begin();
3056 item != items.end();
3059 BuildFlatHashTableRecurse( builtHT, *item );
3063 builtHT[entry->GetKey()] = entry;
3070 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
3073 * The structure used by a Document (through \ref ElementSet),
3074 * in order to hold the parsed entries of a Dicom header, is a recursive
3075 * one. This is due to the fact that the sequences (when present)
3076 * can be nested. Additionaly, the sequence items (represented in
3077 * gdcm as \ref SQItem) add an extra complexity to the data
3078 * structure. Hence, a gdcm user whishing to visit all the entries of
3079 * a Dicom header will need to dig in the gdcm internals (which
3080 * implies exposing all the internal data structures to the API).
3081 * In order to avoid this burden to the user, \ref BuildFlatHashTable
3082 * recursively builds a temporary hash table, which holds all the
3083 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
3085 * \warning Of course there is NO integrity constrain between the
3086 * returned \ref TagDocEntryHT and the \ref ElementSet used
3087 * to build it. Hence if the underlying \ref ElementSet is
3088 * altered, then it is the caller responsability to invoke
3089 * \ref BuildFlatHashTable again...
3090 * @return The flat std::map<> we juste build.
3092 TagDocEntryHT* Document::BuildFlatHashTable()
3094 TagDocEntryHT* FlatHT = new TagDocEntryHT;
3095 BuildFlatHashTableRecurse( *FlatHT, this );
3102 * \brief Compares two documents, according to \ref DicomDir rules
3103 * \warning Does NOT work with ACR-NEMA files
3104 * \todo Find a trick to solve the pb (use RET fields ?)
3106 * @return true if 'smaller'
3108 bool Document::operator<(Document &document)
3111 std::string s1 = GetEntryByNumber(0x0010,0x0010);
3112 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
3124 s1 = GetEntryByNumber(0x0010,0x0020);
3125 s2 = document.GetEntryByNumber(0x0010,0x0020);
3136 // Study Instance UID
3137 s1 = GetEntryByNumber(0x0020,0x000d);
3138 s2 = document.GetEntryByNumber(0x0020,0x000d);
3149 // Serie Instance UID
3150 s1 = GetEntryByNumber(0x0020,0x000e);
3151 s2 = document.GetEntryByNumber(0x0020,0x000e);
3166 } // end namespace gdcm
3168 //-----------------------------------------------------------------------------