1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2004/11/16 16:49:01 $
7 Version: $Revision: 1.130 $
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);
190 Filetype = ExplicitVR;
191 PrintLevel = 1; // 'Medium' print level by default
195 * \brief Canonical destructor.
197 Document::~Document ()
206 //-----------------------------------------------------------------------------
210 * \brief Prints The Dict Entries of THE public Dicom Dictionary
213 void Document::PrintPubDict(std::ostream & os)
215 RefPubDict->Print(os);
219 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
222 void Document::PrintShaDict(std::ostream & os)
224 RefShaDict->Print(os);
227 //-----------------------------------------------------------------------------
230 * \brief Get the public dictionary used
232 Dict* Document::GetPubDict()
238 * \brief Get the shadow dictionary used
240 Dict* Document::GetShaDict()
246 * \brief Set the shadow dictionary used
247 * \param dict dictionary to use in shadow
249 bool Document::SetShaDict(Dict *dict)
256 * \brief Set the shadow dictionary used
257 * \param dictName name of the dictionary to use in shadow
259 bool Document::SetShaDict(DictKey const & dictName)
261 RefShaDict = Global::GetDicts()->GetDict(dictName);
266 * \brief This predicate, based on hopefully reasonable heuristics,
267 * decides whether or not the current Document was properly parsed
268 * and contains the mandatory information for being considered as
269 * a well formed and usable Dicom/Acr File.
270 * @return true when Document is the one of a reasonable Dicom/Acr file,
273 bool Document::IsReadable()
275 if( Filetype == Unknown)
277 dbg.Verbose(0, "Document::IsReadable: wrong filetype");
283 dbg.Verbose(0, "Document::IsReadable: no tags in internal"
292 * \brief Accessor to the Transfer Syntax (when present) of the
293 * current document (it internally handles reading the
294 * value from disk when only parsing occured).
295 * @return The encountered Transfer Syntax of the current document.
297 TransferSyntaxType Document::GetTransferSyntax()
299 DocEntry *entry = GetDocEntryByNumber(0x0002, 0x0010);
305 // The entry might be present but not loaded (parsing and loading
306 // happen at different stages): try loading and proceed with check...
307 LoadDocEntrySafe(entry);
308 if (ValEntry* valEntry = dynamic_cast< ValEntry* >(entry) )
310 std::string transfer = valEntry->GetValue();
311 // The actual transfer (as read from disk) might be padded. We
312 // first need to remove the potential padding. We can make the
313 // weak assumption that padding was not executed with digits...
314 if ( transfer.length() == 0 )
316 // for brain damaged headers
319 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
321 transfer.erase(transfer.length()-1, 1);
323 for (int i = 0; TransferSyntaxStrings[i] != NULL; i++)
325 if ( TransferSyntaxStrings[i] == transfer )
327 return TransferSyntaxType(i);
334 bool Document::IsJPEGLossless()
336 TransferSyntaxType r = GetTransferSyntax();
337 return r == JPEGFullProgressionProcess10_12
338 || r == JPEGLosslessProcess14
339 || r == JPEGLosslessProcess14_1;
343 * \brief Determines if the Transfer Syntax was already encountered
344 * and if it corresponds to a JPEG2000 one
345 * @return True when JPEG2000 (Lossly or LossLess) found. False in all
348 bool Document::IsJPEG2000()
350 TransferSyntaxType r = GetTransferSyntax();
351 return r == JPEG2000Lossless || r == JPEG2000;
355 * \brief Determines if the Transfer Syntax corresponds to any form
356 * of Jpeg encoded Pixel data.
357 * @return True when any form of JPEG found. False otherwise.
359 bool Document::IsJPEG()
361 TransferSyntaxType r = GetTransferSyntax();
362 return r == JPEGBaselineProcess1
363 || r == JPEGExtendedProcess2_4
364 || r == JPEGExtendedProcess3_5
365 || r == JPEGSpectralSelectionProcess6_8
371 * \brief Determines if the Transfer Syntax corresponds to encapsulated
372 * of encoded Pixel Data (as opposed to native).
373 * @return True when encapsulated. False when native.
375 bool Document::IsEncapsulate()
377 TransferSyntaxType r = GetTransferSyntax();
378 return IsJPEG() || r == RLELossless;
382 * \brief Predicate for dicom version 3 file.
383 * @return True when the file is a dicom version 3.
385 bool Document::IsDicomV3()
387 // Checking if Transfert Syntax exists is enough
388 // Anyway, it's to late check if the 'Preamble' was found ...
389 // And ... would it be a rich idea to check ?
390 // (some 'no Preamble' DICOM images exist !)
391 return GetDocEntryByNumber(0x0002, 0x0010) != NULL;
395 * \brief returns the File Type
396 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
397 * @return the FileType code
399 FileType Document::GetFileType()
405 * \brief Tries to open the file \ref Document::Filename and
406 * checks the preamble when existing.
407 * @return The FILE pointer on success.
409 std::ifstream* Document::OpenFile()
411 if (Filename.length() == 0) return 0;
415 "Document::OpenFile is already opened when opening: ",
419 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
424 "Document::OpenFile cannot open file: ",
430 Fp->read((char*)&zero, (size_t)2 );
432 //ACR -- or DICOM with no Preamble --
433 if( zero == 0x0008 || zero == 0x0800 || zero == 0x0002 || zero == 0x0200 )
439 Fp->seekg(126L, std::ios_base::cur);
441 Fp->read(dicm, (size_t)4);
442 if( memcmp(dicm, "DICM", 4) == 0 )
449 "Document::OpenFile not DICOM/ACR (missing preamble)",
456 * \brief closes the file
457 * @return TRUE if the close was successfull
459 bool Document::CloseFile()
468 return true; //FIXME how do we detect a non-close ifstream ?
472 * \brief Writes in a file all the Header Entries (Dicom Elements)
473 * @param fp file pointer on an already open file
474 * @param filetype Type of the File to be written
475 * (ACR-NEMA, ExplicitVR, ImplicitVR)
476 * \return Always true.
478 void Document::Write(std::ofstream* fp, FileType filetype)
480 /// \todo move the following lines (and a lot of others, to be written)
481 /// to a future function CheckAndCorrectHeader
482 /// (necessary if user wants to write a DICOM V3 file
483 /// starting from an ACR-NEMA (V2) Header
485 if (filetype == ImplicitVR)
488 Util::DicomString( TransferSyntaxStrings[ImplicitVRLittleEndian] );
489 ReplaceOrCreateByNumber(ts, 0x0002, 0x0010);
491 /// \todo Refer to standards on page 21, chapter 6.2
492 /// "Value representation": values with a VR of UI shall be
493 /// padded with a single trailing null
494 /// in the following case we have to padd manually with a 0
496 SetEntryLengthByNumber(18, 0x0002, 0x0010);
499 if (filetype == ExplicitVR)
502 Util::DicomString( TransferSyntaxStrings[ExplicitVRLittleEndian] );
503 ReplaceOrCreateByNumber(ts, 0x0002, 0x0010); //LEAK
505 /// \todo Refer to standards on page 21, chapter 6.2
506 /// "Value representation": values with a VR of UI shall be
507 /// padded with a single trailing null
508 /// Dans le cas suivant on doit pader manuellement avec un 0
510 SetEntryLengthByNumber(20, 0x0002, 0x0010);
514 * \todo rewrite later, if really usefull
515 * - 'Group Length' element is optional in DICOM
516 * - but un-updated odd groups lengthes can causes pb
519 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
520 * UpdateGroupLength(false,filetype);
521 * if ( filetype == ACR)
522 * UpdateGroupLength(true,ACR);
525 ElementSet::Write(fp, filetype); // This one is recursive
530 * \brief Modifies the value of a given Header Entry (Dicom Element)
531 * when it exists. Create it with the given value when unexistant.
532 * @param value (string) Value to be set
533 * @param group Group number of the Entry
534 * @param elem Element number of the Entry
535 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
536 * \return pointer to the modified/created Header Entry (NULL when creation
539 ValEntry* Document::ReplaceOrCreateByNumber(
540 std::string const & value,
545 ValEntry* valEntry = 0;
546 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
550 // check if (group,element) DictEntry exists
551 // if it doesn't, create an entry in DictSet::VirtualEntry
554 // Find out if the tag we received is in the dictionaries:
555 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
556 DictEntry* dictEntry = pubDict->GetDictEntryByNumber(group, elem);
559 currentEntry = NewDocEntryByNumber(group, elem, vr);
563 currentEntry = NewDocEntryByNumber(group, elem);
568 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
569 " NewDocEntryByNumber failed.");
573 valEntry = new ValEntry(currentEntry);
576 if ( !AddEntry(valEntry))
579 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
580 " failed allthough this is a creation.");
586 valEntry = dynamic_cast< ValEntry* >(currentEntry);
587 if ( !valEntry ) // Euuuuh? It wasn't a ValEntry
588 // then we change it to a ValEntry ?
589 // Shouldn't it be considered as an error ?
591 // We need to promote the DocEntry to a ValEntry:
592 valEntry = new ValEntry(currentEntry);
593 if (!RemoveEntry(currentEntry))
596 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
597 " of previous DocEntry failed.");
600 if ( !AddEntry(valEntry))
603 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
604 " promoted ValEntry failed.");
610 SetEntryByNumber(value, group, elem);
616 * \brief Modifies the value of a given Header Entry (Dicom Element)
617 * when it exists. Create it with the given value when unexistant.
618 * @param binArea (binary) value to be set
619 * @param Group Group number of the Entry
620 * @param Elem Element number of the Entry
621 * \return pointer to the modified/created Header Entry (NULL when creation
624 BinEntry* Document::ReplaceOrCreateByNumber(
631 BinEntry* binEntry = 0;
632 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
636 // check if (group,element) DictEntry exists
637 // if it doesn't, create an entry in DictSet::VirtualEntry
640 // Find out if the tag we received is in the dictionaries:
641 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
642 DictEntry *dictEntry = pubDict->GetDictEntryByNumber(group, elem);
646 currentEntry = NewDocEntryByNumber(group, elem, vr);
650 currentEntry = NewDocEntryByNumber(group, elem);
654 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
655 " NewDocEntryByNumber failed.");
658 binEntry = new BinEntry(currentEntry);
659 if ( !AddEntry(binEntry))
661 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
662 " failed allthough this is a creation.");
667 binEntry = dynamic_cast< BinEntry* >(currentEntry);
668 if ( !binEntry ) // Euuuuh? It wasn't a BinEntry
669 // then we change it to a BinEntry ?
670 // Shouldn't it be considered as an error ?
672 // We need to promote the DocEntry to a BinEntry:
673 binEntry = new BinEntry(currentEntry);
674 if (!RemoveEntry(currentEntry))
676 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
677 " of previous DocEntry failed.");
680 if ( !AddEntry(binEntry))
682 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
683 " promoted BinEntry failed.");
689 SetEntryByNumber(binArea, lgth, group, elem);
696 * \brief Modifies the value of a given Header Entry (Dicom Element)
697 * when it exists. Create it when unexistant.
698 * @param Group Group number of the Entry
699 * @param Elem Element number of the Entry
700 * \return pointer to the modified/created SeqEntry (NULL when creation
703 SeqEntry* Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem)
706 DocEntry* a = GetDocEntryByNumber( group, elem);
709 a = NewSeqEntryByNumber(group, elem);
715 b = new SeqEntry(a, 1); // FIXME : 1 (Depth)
722 * \brief Set a new value if the invoked element exists
723 * Seems to be useless !!!
724 * @param value new element value
725 * @param group group number of the Entry
726 * @param elem element number of the Entry
729 bool Document::ReplaceIfExistByNumber(std::string const & value,
730 uint16_t group, uint16_t elem )
732 SetEntryByNumber(value, group, elem);
737 //-----------------------------------------------------------------------------
741 * \brief Checks if a given Dicom Element exists within the H table
742 * @param group Group number of the searched Dicom Element
743 * @param element Element number of the searched Dicom Element
744 * @return true is found
746 bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
748 const std::string &key = DictEntry::TranslateToKey(group, element );
749 return TagHT.count(key) != 0;
753 * \brief Searches within Header Entries (Dicom Elements) parsed with
754 * the public and private dictionaries
755 * for the element value of a given tag.
756 * \warning Don't use any longer : use GetPubEntryByName
757 * @param tagName name of the searched element.
758 * @return Corresponding element value when it exists,
759 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
761 std::string Document::GetEntryByName(TagName const & tagName)
763 DictEntry* dictEntry = RefPubDict->GetDictEntryByName(tagName);
769 return GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
773 * \brief Searches within Header Entries (Dicom Elements) parsed with
774 * the public and private dictionaries
775 * for the element value representation of a given tag.
777 * Obtaining the VR (Value Representation) might be needed by caller
778 * to convert the string typed content to caller's native type
779 * (think of C++ vs Python). The VR is actually of a higher level
780 * of semantics than just the native C++ type.
781 * @param tagName name of the searched element.
782 * @return Corresponding element value representation when it exists,
783 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
785 std::string Document::GetEntryVRByName(TagName const& tagName)
787 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
788 if( dictEntry == NULL)
793 DocEntry* elem = GetDocEntryByNumber(dictEntry->GetGroup(),
794 dictEntry->GetElement());
795 return elem->GetVR();
799 * \brief Searches within Header Entries (Dicom Elements) parsed with
800 * the public and private dictionaries
801 * for the element value representation of a given tag.
802 * @param group Group number of the searched tag.
803 * @param element Element number of the searched tag.
804 * @return Corresponding element value representation when it exists,
805 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
807 std::string Document::GetEntryByNumber(uint16_t group, uint16_t element)
809 TagKey key = DictEntry::TranslateToKey(group, element);
810 /// \todo use map methods, instead of multimap JPR
811 if ( !TagHT.count(key))
816 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
820 * \brief Searches within Header Entries (Dicom Elements) parsed with
821 * the public and private dictionaries
822 * for the element value representation of a given tag..
824 * Obtaining the VR (Value Representation) might be needed by caller
825 * to convert the string typed content to caller's native type
826 * (think of C++ vs Python). The VR is actually of a higher level
827 * of semantics than just the native C++ type.
828 * @param group Group number of the searched tag.
829 * @param element Element number of the searched tag.
830 * @return Corresponding element value representation when it exists,
831 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
833 std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element)
835 DocEntry* elem = GetDocEntryByNumber(group, element);
840 return elem->GetVR();
844 * \brief Searches within Header Entries (Dicom Elements) parsed with
845 * the public and private dictionaries
846 * for the value length of a given tag..
847 * @param group Group number of the searched tag.
848 * @param element Element number of the searched tag.
849 * @return Corresponding element length; -2 if not found
851 int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element)
853 DocEntry* elem = GetDocEntryByNumber(group, element);
856 return -2; //magic number
858 return elem->GetLength();
861 * \brief Sets the value (string) of the Header Entry (Dicom Element)
862 * @param content string value of the Dicom Element
863 * @param tagName name of the searched Dicom Element.
864 * @return true when found
866 bool Document::SetEntryByName( std::string const & content,
867 TagName const & tagName)
869 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
875 return SetEntryByNumber(content,dictEntry->GetGroup(),
876 dictEntry->GetElement());
880 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
881 * through it's (group, element) and modifies it's content with
883 * @param content new value (string) to substitute with
884 * @param group group number of the Dicom Element to modify
885 * @param element element number of the Dicom Element to modify
887 bool Document::SetEntryByNumber(std::string const& content,
888 uint16_t group, uint16_t element)
893 ValEntry* valEntry = GetValEntryByNumber(group, element);
896 dbg.Verbose(0, "Document::SetEntryByNumber: no corresponding",
897 " ValEntry (try promotion first).");
900 // Non even content must be padded with a space (020H)...
901 std::string finalContent = Util::DicomString( content.c_str() );
902 assert( !(finalContent.size() % 2) );
903 valEntry->SetValue(finalContent);
905 // Integers have a special treatement for their length:
907 l = finalContent.length();
908 if ( l != 0) // To avoid to be cheated by 'zero length' integers
910 VRKey vr = valEntry->GetVR();
911 if( vr == "US" || vr == "SS" )
913 // for multivaluated items
914 c = Util::CountSubstring(content, "\\") + 1;
917 else if( vr == "UL" || vr == "SL" )
919 // for multivaluated items
920 c = Util::CountSubstring(content, "\\") + 1;
924 valEntry->SetLength(l);
929 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
930 * through it's (group, element) and modifies it's content with
932 * @param content new value (void* -> uint8_t*) to substitute with
933 * @param lgth new value length
934 * @param group group number of the Dicom Element to modify
935 * @param element element number of the Dicom Element to modify
937 bool Document::SetEntryByNumber(uint8_t*content, int lgth,
938 uint16_t group, uint16_t element)
940 (void)lgth; //not used
941 TagKey key = DictEntry::TranslateToKey(group, element);
942 if ( !TagHT.count(key))
947 /* Hope Binary field length is *never* wrong
948 if(lgth%2) // Non even length are padded with a space (020H).
951 //content = content + '\0'; // fing a trick to enlarge a binary field?
954 BinEntry* entry = (BinEntry *)TagHT[key];
955 entry->SetBinArea(content);
956 entry->SetLength(lgth);
957 entry->SetValue(GDCM_BINLOADED);
963 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
964 * in the PubDocEntrySet of this instance
965 * through it's (group, element) and modifies it's length with
967 * \warning Use with extreme caution.
968 * @param l new length to substitute with
969 * @param group group number of the Entry to modify
970 * @param element element number of the Entry to modify
971 * @return true on success, false otherwise.
973 bool Document::SetEntryLengthByNumber(uint32_t l,
974 uint16_t group, uint16_t element)
976 /// \todo use map methods, instead of multimap JPR
977 TagKey key = DictEntry::TranslateToKey(group, element);
978 if ( !TagHT.count(key) )
984 l++; // length must be even
986 ( ((TagHT.equal_range(key)).first)->second )->SetLength(l);
992 * \brief Gets (from Header) the offset of a 'non string' element value
993 * (LoadElementValues has already be executed)
994 * @param group group number of the Entry
995 * @param elem element number of the Entry
996 * @return File Offset of the Element Value
998 size_t Document::GetEntryOffsetByNumber(uint16_t group, uint16_t elem)
1000 DocEntry* entry = GetDocEntryByNumber(group, elem);
1003 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry present.");
1006 return entry->GetOffset();
1010 * \brief Gets (from Header) a 'non string' element value
1011 * (LoadElementValues has already be executed)
1012 * @param group group number of the Entry
1013 * @param elem element number of the Entry
1014 * @return Pointer to the 'non string' area
1016 void* Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem)
1018 DocEntry* entry = GetDocEntryByNumber(group, elem);
1021 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry");
1024 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(entry) )
1026 return binEntry->GetBinArea();
1033 * \brief Loads (from disk) the element content
1034 * when a string is not suitable
1035 * @param group group number of the Entry
1036 * @param elem element number of the Entry
1038 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
1040 // Search the corresponding DocEntry
1041 DocEntry *docElement = GetDocEntryByNumber(group, elem);
1045 size_t o =(size_t)docElement->GetOffset();
1046 Fp->seekg( o, std::ios_base::beg);
1047 size_t l = docElement->GetLength();
1048 uint8_t* a = new uint8_t[l];
1051 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1056 Fp->read((char*)a, l);
1057 if( Fp->fail() || Fp->eof() )//Fp->gcount() == 1
1063 // Set the value to the DocEntry
1064 if( !SetEntryBinAreaByNumber( a, group, elem ) )
1067 dbg.Verbose(0, "Document::LoadEntryBinArea setting failed.");
1071 * \brief Loads (from disk) the element content
1072 * when a string is not suitable
1073 * @param element Entry whose binArea is going to be loaded
1075 void Document::LoadEntryBinArea(BinEntry* element)
1077 size_t o =(size_t)element->GetOffset();
1078 Fp->seekg(o, std::ios_base::beg);
1079 size_t l = element->GetLength();
1080 uint8_t* a = new uint8_t[l];
1083 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1087 /// \todo check the result
1088 Fp->read((char*)a, l);
1089 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
1095 element->SetBinArea(a);
1099 * \brief Sets a 'non string' value to a given Dicom Element
1100 * @param area area containing the 'non string' value
1101 * @param group Group number of the searched Dicom Element
1102 * @param element Element number of the searched Dicom Element
1105 bool Document::SetEntryBinAreaByNumber(uint8_t* area,
1106 uint16_t group, uint16_t element)
1108 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1109 if ( !currentEntry )
1114 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1116 binEntry->SetBinArea( area );
1124 * \brief Update the entries with the shadow dictionary.
1125 * Only non even entries are analyzed
1127 void Document::UpdateShaEntries()
1132 /// \todo TODO : still any use to explore recursively the whole structure?
1134 for(ListTag::iterator it=listEntries.begin();
1135 it!=listEntries.end();
1138 // Odd group => from public dictionary
1139 if((*it)->GetGroup()%2==0)
1142 // Peer group => search the corresponding dict entry
1144 entry=RefShaDict->GetDictEntryByNumber((*it)->GetGroup(),(*it)->GetElement());
1148 if((*it)->IsImplicitVR())
1153 (*it)->SetValue(GetDocEntryUnvalue(*it)); // to go on compiling
1155 // Set the new entry and the new value
1156 (*it)->SetDictEntry(entry);
1157 CheckDocEntryVR(*it,vr);
1159 (*it)->SetValue(GetDocEntryValue(*it)); // to go on compiling
1164 // Remove precedent value transformation
1165 (*it)->SetDictEntry(NewVirtualDictEntry((*it)->GetGroup(),(*it)->GetElement(),vr));
1172 * \brief Searches within the Header Entries for a Dicom Element of
1174 * @param tagName name of the searched Dicom Element.
1175 * @return Corresponding Dicom Element when it exists, and NULL
1178 DocEntry* Document::GetDocEntryByName(TagName const & tagName)
1180 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
1186 return GetDocEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
1190 * \brief retrieves a Dicom Element (the first one) using (group, element)
1191 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1192 * if you think it's NOT UNIQUE, check the count number
1193 * and use iterators to retrieve ALL the Dicoms Elements within
1194 * a given couple (group, element)
1195 * @param group Group number of the searched Dicom Element
1196 * @param element Element number of the searched Dicom Element
1199 DocEntry* Document::GetDocEntryByNumber(uint16_t group, uint16_t element)
1201 TagKey key = DictEntry::TranslateToKey(group, element);
1202 if ( !TagHT.count(key))
1206 return TagHT.find(key)->second;
1210 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1211 * returns a result when the corresponding entry is of type
1213 * @return When present, the corresponding ValEntry.
1215 ValEntry* Document::GetValEntryByNumber(uint16_t group, uint16_t element)
1217 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1218 if ( !currentEntry )
1222 if ( ValEntry* valEntry = dynamic_cast<ValEntry*>(currentEntry) )
1226 dbg.Verbose(0, "Document::GetValEntryByNumber: unfound ValEntry.");
1232 * \brief Loads the element while preserving the current
1233 * underlying file position indicator as opposed to
1234 * to LoadDocEntry that modifies it.
1235 * @param entry Header Entry whose value shall be loaded.
1238 void Document::LoadDocEntrySafe(DocEntry * entry)
1240 long PositionOnEntry = Fp->tellg();
1241 LoadDocEntry(entry);
1242 Fp->seekg(PositionOnEntry, std::ios_base::beg);
1246 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1248 * @return The properly swaped 32 bits integer.
1250 uint32_t Document::SwapLong(uint32_t a)
1257 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1258 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1262 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1266 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1269 //std::cout << "swapCode= " << SwapCode << std::endl;
1270 dbg.Error(" Document::SwapLong : unset swap code");
1277 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1279 * @return The properly unswaped 32 bits integer.
1281 uint32_t Document::UnswapLong(uint32_t a)
1287 * \brief Swaps the bytes so they agree with the processor order
1288 * @return The properly swaped 16 bits integer.
1290 uint16_t Document::SwapShort(uint16_t a)
1292 if ( SwapCode == 4321 || SwapCode == 2143 )
1294 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1300 * \brief Unswaps the bytes so they agree with the processor order
1301 * @return The properly unswaped 16 bits integer.
1303 uint16_t Document::UnswapShort(uint16_t a)
1305 return SwapShort(a);
1308 //-----------------------------------------------------------------------------
1312 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1313 * @return length of the parsed set.
1315 void Document::ParseDES(DocEntrySet *set, long offset,
1316 long l_max, bool delim_mode)
1318 DocEntry *newDocEntry = 0;
1322 if ( !delim_mode && (Fp->tellg()-offset) >= l_max)
1326 newDocEntry = ReadNextDocEntry( );
1332 VRKey vr = newDocEntry->GetVR();
1336 if ( Global::GetVR()->IsVROfGdcmStringRepresentable(vr) )
1338 /////////////////////// ValEntry
1339 ValEntry* newValEntry =
1340 new ValEntry( newDocEntry->GetDictEntry() ); //LEAK
1341 newValEntry->Copy( newDocEntry );
1343 // When "set" is a Document, then we are at the top of the
1344 // hierarchy and the Key is simply of the form ( group, elem )...
1345 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1348 newValEntry->SetKey( newValEntry->GetKey() );
1350 // ...but when "set" is a SQItem, we are inserting this new
1351 // valEntry in a sequence item. Hence the key has the
1352 // generalized form (refer to \ref BaseTagKey):
1353 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1355 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1356 + newValEntry->GetKey() );
1359 LoadDocEntry( newValEntry );
1360 bool delimitor=newValEntry->IsItemDelimitor();
1361 if( !set->AddEntry( newValEntry ) )
1363 // If here expect big troubles
1364 delete newValEntry; //otherwise mem leak
1371 if ( !delim_mode && (Fp->tellg()-offset) >= l_max)
1378 if ( ! Global::GetVR()->IsVROfGdcmBinaryRepresentable(vr) )
1380 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1381 dbg.Verbose(0, "Document::ParseDES: neither Valentry, "
1382 "nor BinEntry. Probably unknown VR.");
1385 //////////////////// BinEntry or UNKOWN VR:
1386 /* BinEntry* newBinEntry =
1387 new BinEntry( newDocEntry->GetDictEntry() ); //LEAK
1388 newBinEntry->Copy( newDocEntry );*/
1389 BinEntry* newBinEntry = new BinEntry( newDocEntry ); //LEAK
1391 // When "this" is a Document the Key is simply of the
1392 // form ( group, elem )...
1393 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1396 newBinEntry->SetKey( newBinEntry->GetKey() );
1398 // but when "this" is a SQItem, we are inserting this new
1399 // valEntry in a sequence item, and the kay has the
1400 // generalized form (refer to \ref BaseTagKey):
1401 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1403 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1404 + newBinEntry->GetKey() );
1407 LoadDocEntry( newBinEntry );
1408 if( !set->AddEntry( newBinEntry ) )
1410 //Expect big troubles if here
1415 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1416 && ( newDocEntry->GetElement() == 0x0010 ) )
1418 TransferSyntaxType ts = GetTransferSyntax();
1419 if ( ts == RLELossless )
1421 long PositionOnEntry = Fp->tellg();
1422 Fp->seekg( newDocEntry->GetOffset(), std::ios_base::beg );
1424 Fp->seekg( PositionOnEntry, std::ios_base::beg );
1426 else if ( IsJPEG() )
1428 long PositionOnEntry = Fp->tellg();
1429 Fp->seekg( newDocEntry->GetOffset(), std::ios_base::beg );
1430 ComputeJPEGFragmentInfo();
1431 Fp->seekg( PositionOnEntry, std::ios_base::beg );
1435 // Just to make sure we are at the beginning of next entry.
1436 SkipToNextDocEntry(newDocEntry);
1441 unsigned long l = newDocEntry->GetReadLength();
1442 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1444 if ( l == 0xffffffff )
1453 // no other way to create it ...
1454 SeqEntry* newSeqEntry =
1455 new SeqEntry( newDocEntry->GetDictEntry() );
1456 newSeqEntry->Copy( newDocEntry );
1457 newSeqEntry->SetDelimitorMode( delim_mode );
1459 // At the top of the hierarchy, stands a Document. When "set"
1460 // is a Document, then we are building the first depth level.
1461 // Hence the SeqEntry we are building simply has a depth
1463 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1466 newSeqEntry->SetDepthLevel( 1 );
1467 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1469 // But when "set" is allready a SQItem, we are building a nested
1470 // sequence, and hence the depth level of the new SeqEntry
1471 // we are building, is one level deeper:
1472 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1474 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1475 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1476 + newSeqEntry->GetKey() );
1480 { // Don't try to parse zero-length sequences
1481 ParseSQ( newSeqEntry,
1482 newDocEntry->GetOffset(),
1485 set->AddEntry( newSeqEntry );
1486 if ( !delim_mode && (Fp->tellg()-offset) >= l_max)
1496 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1497 * @return parsed length for this level
1499 void Document::ParseSQ( SeqEntry* seqEntry,
1500 long offset, long l_max, bool delim_mode)
1502 int SQItemNumber = 0;
1507 DocEntry* newDocEntry = ReadNextDocEntry();
1510 // FIXME Should warn user
1515 if ( newDocEntry->IsSequenceDelimitor() )
1517 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1521 if ( !delim_mode && (Fp->tellg()-offset) >= l_max)
1526 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1527 std::ostringstream newBase;
1528 newBase << seqEntry->GetKey()
1532 itemSQ->SetBaseTagKey( newBase.str() );
1533 unsigned int l = newDocEntry->GetReadLength();
1535 if ( l == 0xffffffff )
1544 ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1546 seqEntry->AddEntry( itemSQ, SQItemNumber );
1548 if ( !delim_mode && ( Fp->tellg() - offset ) >= l_max )
1556 * \brief Loads the element content if its length doesn't exceed
1557 * the value specified with Document::SetMaxSizeLoadEntry()
1558 * @param entry Header Entry (Dicom Element) to be dealt with
1560 void Document::LoadDocEntry(DocEntry* entry)
1562 uint16_t group = entry->GetGroup();
1563 std::string vr = entry->GetVR();
1564 uint32_t length = entry->GetLength();
1566 Fp->seekg((long)entry->GetOffset(), std::ios_base::beg);
1568 // A SeQuence "contains" a set of Elements.
1569 // (fffe e000) tells us an Element is beginning
1570 // (fffe e00d) tells us an Element just ended
1571 // (fffe e0dd) tells us the current SeQuence just ended
1572 if( group == 0xfffe )
1574 // NO more value field for SQ !
1578 // When the length is zero things are easy:
1581 ((ValEntry *)entry)->SetValue("");
1585 // The elements whose length is bigger than the specified upper bound
1586 // are not loaded. Instead we leave a short notice of the offset of
1587 // the element content and it's length.
1589 std::ostringstream s;
1590 if (length > MaxSizeLoadEntry)
1592 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1594 //s << "gdcm::NotLoaded (BinEntry)";
1595 s << GDCM_NOTLOADED;
1596 s << " Address:" << (long)entry->GetOffset();
1597 s << " Length:" << entry->GetLength();
1598 s << " x(" << std::hex << entry->GetLength() << ")";
1599 binEntryPtr->SetValue(s.str());
1601 // Be carefull : a BinEntry IS_A ValEntry ...
1602 else if (ValEntry* valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1604 // s << "gdcm::NotLoaded. (ValEntry)";
1605 s << GDCM_NOTLOADED;
1606 s << " Address:" << (long)entry->GetOffset();
1607 s << " Length:" << entry->GetLength();
1608 s << " x(" << std::hex << entry->GetLength() << ")";
1609 valEntryPtr->SetValue(s.str());
1614 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1615 << "nor a ValEntry ?! Should never print that !" << std::endl;
1618 // to be sure we are at the end of the value ...
1619 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1620 std::ios_base::beg);
1624 // When we find a BinEntry not very much can be done :
1625 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1627 s << GDCM_BINLOADED;
1628 binEntryPtr->SetValue(s.str());
1629 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1633 /// \todo Any compacter code suggested (?)
1634 if ( IsDocEntryAnInteger(entry) )
1638 // When short integer(s) are expected, read and convert the following
1639 // n *two characters properly i.e. consider them as short integers as
1640 // opposed to strings.
1641 // Elements with Value Multiplicity > 1
1642 // contain a set of integers (not a single one)
1643 if (vr == "US" || vr == "SS")
1646 NewInt = ReadInt16();
1650 for (int i=1; i < nbInt; i++)
1653 NewInt = ReadInt16();
1658 // See above comment on multiple integers (mutatis mutandis).
1659 else if (vr == "UL" || vr == "SL")
1662 NewInt = ReadInt32();
1666 for (int i=1; i < nbInt; i++)
1669 NewInt = ReadInt32();
1674 #ifdef GDCM_NO_ANSI_STRING_STREAM
1675 s << std::ends; // to avoid oddities on Solaris
1676 #endif //GDCM_NO_ANSI_STRING_STREAM
1678 ((ValEntry *)entry)->SetValue(s.str());
1682 // FIXME: We need an additional byte for storing \0 that is not on disk
1683 char *str = new char[length+1];
1684 Fp->read(str, (size_t)length);
1685 str[length] = '\0'; //this is only useful when length is odd
1686 // Special DicomString call to properly handle \0 and even length
1687 std::string newValue;
1690 newValue = Util::DicomString(str, length+1);
1691 //dbg.Verbose(0, "Warning: bad length: ", length );
1692 dbg.Verbose(0, "For string :", newValue.c_str());
1693 // Since we change the length of string update it length
1694 entry->SetReadLength(length+1);
1698 newValue = Util::DicomString(str, length);
1702 if ( ValEntry* valEntry = dynamic_cast<ValEntry* >(entry) )
1704 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1706 dbg.Verbose(1, "Document::LoadDocEntry",
1707 "unread element value");
1708 valEntry->SetValue(GDCM_UNREAD);
1714 // Because of correspondance with the VR dic
1715 valEntry->SetValue(newValue);
1719 valEntry->SetValue(newValue);
1724 dbg.Error(true, "Document::LoadDocEntry"
1725 "Should have a ValEntry, here !");
1731 * \brief Find the value Length of the passed Header Entry
1732 * @param entry Header Entry whose length of the value shall be loaded.
1734 void Document::FindDocEntryLength( DocEntry *entry )
1735 throw ( FormatError )
1737 uint16_t element = entry->GetElement();
1738 std::string vr = entry->GetVR();
1741 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1743 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1745 // The following reserved two bytes (see PS 3.5-2003, section
1746 // "7.1.2 Data element structure with explicit vr", p 27) must be
1747 // skipped before proceeding on reading the length on 4 bytes.
1748 Fp->seekg( 2L, std::ios_base::cur);
1749 uint32_t length32 = ReadInt32();
1751 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1756 /// \todo rename that to FindDocEntryLengthOBOrOW since
1757 /// the above test is on both OB and OW...
1758 lengthOB = FindDocEntryLengthOB();
1760 catch ( FormatUnexpected )
1762 // Computing the length failed (this happens with broken
1763 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1764 // chance to get the pixels by deciding the element goes
1765 // until the end of the file. Hence we artificially fix the
1766 // the length and proceed.
1767 long currentPosition = Fp->tellg();
1768 Fp->seekg(0L,std::ios_base::end);
1769 long lengthUntilEOF = Fp->tellg() - currentPosition;
1770 Fp->seekg(currentPosition, std::ios_base::beg);
1771 entry->SetLength(lengthUntilEOF);
1774 entry->SetLength(lengthOB);
1777 FixDocEntryFoundLength(entry, length32);
1781 // Length is encoded on 2 bytes.
1782 length16 = ReadInt16();
1784 // We can tell the current file is encoded in big endian (like
1785 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1786 // and it's value is the one of the encoding of a big endian file.
1787 // In order to deal with such big endian encoded files, we have
1788 // (at least) two strategies:
1789 // * when we load the "Transfer Syntax" tag with value of big endian
1790 // encoding, we raise the proper flags. Then we wait for the end
1791 // of the META group (0x0002) among which is "Transfer Syntax",
1792 // before switching the swap code to big endian. We have to postpone
1793 // the switching of the swap code since the META group is fully encoded
1794 // in little endian, and big endian coding only starts at the next
1795 // group. The corresponding code can be hard to analyse and adds
1796 // many additional unnecessary tests for regular tags.
1797 // * the second strategy consists in waiting for trouble, that shall
1798 // appear when we find the first group with big endian encoding. This
1799 // is easy to detect since the length of a "Group Length" tag (the
1800 // ones with zero as element number) has to be of 4 (0x0004). When we
1801 // encounter 1024 (0x0400) chances are the encoding changed and we
1802 // found a group with big endian encoding.
1803 // We shall use this second strategy. In order to make sure that we
1804 // can interpret the presence of an apparently big endian encoded
1805 // length of a "Group Length" without committing a big mistake, we
1806 // add an additional check: we look in the already parsed elements
1807 // for the presence of a "Transfer Syntax" whose value has to be "big
1808 // endian encoding". When this is the case, chances are we have got our
1809 // hands on a big endian encoded file: we switch the swap code to
1810 // big endian and proceed...
1811 if ( element == 0x0000 && length16 == 0x0400 )
1813 TransferSyntaxType ts = GetTransferSyntax();
1814 if ( ts != ExplicitVRBigEndian )
1816 throw FormatError( "Document::FindDocEntryLength()",
1817 " not explicit VR." );
1821 SwitchSwapToBigEndian();
1822 // Restore the unproperly loaded values i.e. the group, the element
1823 // and the dictionary entry depending on them.
1824 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1825 uint16_t correctElem = SwapShort( entry->GetElement() );
1826 DictEntry* newTag = GetDictEntryByNumber( correctGroup,
1830 // This correct tag is not in the dictionary. Create a new one.
1831 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1833 // FIXME this can create a memory leaks on the old entry that be
1834 // left unreferenced.
1835 entry->SetDictEntry( newTag );
1838 // Heuristic: well, some files are really ill-formed.
1839 if ( length16 == 0xffff)
1841 // 0xffff means that we deal with 'Unknown Length' Sequence
1844 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1849 // Either implicit VR or a non DICOM conformal (see note below) explicit
1850 // VR that ommited the VR of (at least) this element. Farts happen.
1851 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1852 // on Data elements "Implicit and Explicit VR Data Elements shall
1853 // not coexist in a Data Set and Data Sets nested within it".]
1854 // Length is on 4 bytes.
1856 FixDocEntryFoundLength( entry, ReadInt32() );
1862 * \brief Find the Value Representation of the current Dicom Element.
1865 void Document::FindDocEntryVR( DocEntry *entry )
1867 if ( Filetype != ExplicitVR )
1874 long positionOnEntry = Fp->tellg();
1875 // Warning: we believe this is explicit VR (Value Representation) because
1876 // we used a heuristic that found "UL" in the first tag. Alas this
1877 // doesn't guarantee that all the tags will be in explicit VR. In some
1878 // cases (see e-film filtered files) one finds implicit VR tags mixed
1879 // within an explicit VR file. Hence we make sure the present tag
1880 // is in explicit VR and try to fix things if it happens not to be
1883 Fp->read (vr, (size_t)2);
1886 if( !CheckDocEntryVR(entry, vr) )
1888 Fp->seekg(positionOnEntry, std::ios_base::beg);
1889 // When this element is known in the dictionary we shall use, e.g. for
1890 // the semantics (see the usage of IsAnInteger), the VR proposed by the
1891 // dictionary entry. Still we have to flag the element as implicit since
1892 // we know now our assumption on expliciteness is not furfilled.
1894 if ( entry->IsVRUnknown() )
1896 entry->SetVR("Implicit");
1898 entry->SetImplicitVR();
1903 * \brief Check the correspondance between the VR of the header entry
1904 * and the taken VR. If they are different, the header entry is
1905 * updated with the new VR.
1906 * @param entry Header Entry to check
1907 * @param vr Dicom Value Representation
1908 * @return false if the VR is incorrect of if the VR isn't referenced
1909 * otherwise, it returns true
1911 bool Document::CheckDocEntryVR(DocEntry *entry, VRKey vr)
1914 bool realExplicit = true;
1916 // Assume we are reading a falsely explicit VR file i.e. we reached
1917 // a tag where we expect reading a VR but are in fact we read the
1918 // first to bytes of the length. Then we will interogate (through find)
1919 // the dicom_vr dictionary with oddities like "\004\0" which crashes
1920 // both GCC and VC++ implementations of the STL map. Hence when the
1921 // expected VR read happens to be non-ascii characters we consider
1922 // we hit falsely explicit VR tag.
1924 if ( !isalpha((unsigned char)vr[0]) && !isalpha((unsigned char)vr[1]) )
1926 realExplicit = false;
1929 // CLEANME searching the dicom_vr at each occurence is expensive.
1930 // PostPone this test in an optional integrity check at the end
1931 // of parsing or only in debug mode.
1932 if ( realExplicit && !Global::GetVR()->Count(vr) )
1934 realExplicit = false;
1937 if ( !realExplicit )
1939 // We thought this was explicit VR, but we end up with an
1940 // implicit VR tag. Let's backtrack.
1941 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
1942 entry->GetGroup(), entry->GetElement());
1943 dbg.Verbose(1, "Document::FindVR: ", msg.c_str());
1945 if( entry->GetGroup() % 2 && entry->GetElement() == 0x0000)
1947 // Group length is UL !
1948 DictEntry* newEntry = NewVirtualDictEntry(
1949 entry->GetGroup(), entry->GetElement(),
1950 "UL", "FIXME", "Group Length");
1951 entry->SetDictEntry( newEntry );
1956 if ( entry->IsVRUnknown() )
1958 // When not a dictionary entry, we can safely overwrite the VR.
1959 if( entry->GetElement() == 0x0000 )
1961 // Group length is UL !
1969 else if ( entry->GetVR() != vr )
1971 // The VR present in the file and the dictionary disagree. We assume
1972 // the file writer knew best and use the VR of the file. Since it would
1973 // be unwise to overwrite the VR of a dictionary (since it would
1974 // compromise it's next user), we need to clone the actual DictEntry
1975 // and change the VR for the read one.
1976 DictEntry* newEntry = NewVirtualDictEntry(
1977 entry->GetGroup(), entry->GetElement(),
1978 vr, "FIXME", entry->GetName());
1979 entry->SetDictEntry(newEntry);
1986 * \brief Get the transformed value of the header entry. The VR value
1987 * is used to define the transformation to operate on the value
1988 * \warning NOT end user intended method !
1989 * @param entry entry to tranform
1990 * @return Transformed entry value
1992 std::string Document::GetDocEntryValue(DocEntry *entry)
1994 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1996 std::string val = ((ValEntry *)entry)->GetValue();
1997 std::string vr = entry->GetVR();
1998 uint32_t length = entry->GetLength();
1999 std::ostringstream s;
2002 // When short integer(s) are expected, read and convert the following
2003 // n * 2 bytes properly i.e. as a multivaluated strings
2004 // (each single value is separated fromthe next one by '\'
2005 // as usual for standard multivaluated filels
2006 // Elements with Value Multiplicity > 1
2007 // contain a set of short integers (not a single one)
2009 if( vr == "US" || vr == "SS" )
2014 for (int i=0; i < nbInt; i++)
2020 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
2021 newInt16 = SwapShort( newInt16 );
2026 // When integer(s) are expected, read and convert the following
2027 // n * 4 bytes properly i.e. as a multivaluated strings
2028 // (each single value is separated fromthe next one by '\'
2029 // as usual for standard multivaluated filels
2030 // Elements with Value Multiplicity > 1
2031 // contain a set of integers (not a single one)
2032 else if( vr == "UL" || vr == "SL" )
2037 for (int i=0; i < nbInt; i++)
2043 newInt32 = ( val[4*i+0] & 0xFF )
2044 + (( val[4*i+1] & 0xFF ) << 8 )
2045 + (( val[4*i+2] & 0xFF ) << 16 )
2046 + (( val[4*i+3] & 0xFF ) << 24 );
2047 newInt32 = SwapLong( newInt32 );
2051 #ifdef GDCM_NO_ANSI_STRING_STREAM
2052 s << std::ends; // to avoid oddities on Solaris
2053 #endif //GDCM_NO_ANSI_STRING_STREAM
2057 return ((ValEntry *)entry)->GetValue();
2061 * \brief Get the reverse transformed value of the header entry. The VR
2062 * value is used to define the reverse transformation to operate on
2064 * \warning NOT end user intended method !
2065 * @param entry Entry to reverse transform
2066 * @return Reverse transformed entry value
2068 std::string Document::GetDocEntryUnvalue(DocEntry* entry)
2070 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2072 std::string vr = entry->GetVR();
2073 std::vector<std::string> tokens;
2074 std::ostringstream s;
2076 if ( vr == "US" || vr == "SS" )
2080 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
2081 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2082 for (unsigned int i=0; i<tokens.size(); i++)
2084 newInt16 = atoi(tokens[i].c_str());
2085 s << ( newInt16 & 0xFF )
2086 << (( newInt16 >> 8 ) & 0xFF );
2090 if ( vr == "UL" || vr == "SL")
2094 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
2095 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2096 for (unsigned int i=0; i<tokens.size();i++)
2098 newInt32 = atoi(tokens[i].c_str());
2099 s << (char)( newInt32 & 0xFF )
2100 << (char)(( newInt32 >> 8 ) & 0xFF )
2101 << (char)(( newInt32 >> 16 ) & 0xFF )
2102 << (char)(( newInt32 >> 24 ) & 0xFF );
2107 #ifdef GDCM_NO_ANSI_STRING_STREAM
2108 s << std::ends; // to avoid oddities on Solaris
2109 #endif //GDCM_NO_ANSI_STRING_STREAM
2113 return ((ValEntry *)entry)->GetValue();
2117 * \brief Skip a given Header Entry
2118 * \warning NOT end user intended method !
2119 * @param entry entry to skip
2121 void Document::SkipDocEntry(DocEntry *entry)
2123 SkipBytes(entry->GetLength());
2127 * \brief Skips to the begining of the next Header Entry
2128 * \warning NOT end user intended method !
2129 * @param entry entry to skip
2131 void Document::SkipToNextDocEntry(DocEntry *entry)
2133 Fp->seekg((long)(entry->GetOffset()), std::ios_base::beg);
2134 Fp->seekg( (long)(entry->GetReadLength()), std::ios_base::cur);
2138 * \brief When the length of an element value is obviously wrong (because
2139 * the parser went Jabberwocky) one can hope improving things by
2140 * applying some heuristics.
2141 * @param entry entry to check
2142 * @param foundLength fist assumption about length
2144 void Document::FixDocEntryFoundLength(DocEntry *entry,
2145 uint32_t foundLength)
2147 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
2148 if ( foundLength == 0xffffffff)
2153 uint16_t gr = entry->GetGroup();
2154 uint16_t el = entry->GetElement();
2156 if ( foundLength % 2)
2158 std::ostringstream s;
2159 s << "Warning : Tag with uneven length "
2161 << " in x(" << std::hex << gr << "," << el <<")" << std::dec;
2162 dbg.Verbose(0, s.str().c_str());
2165 //////// Fix for some naughty General Electric images.
2166 // Allthough not recent many such GE corrupted images are still present
2167 // on Creatis hard disks. Hence this fix shall remain when such images
2168 // are no longer in user (we are talking a few years, here)...
2169 // Note: XMedCom probably uses such a trick since it is able to read
2170 // those pesky GE images ...
2171 if ( foundLength == 13)
2173 // Only happens for this length !
2174 if ( entry->GetGroup() != 0x0008
2175 || ( entry->GetElement() != 0x0070
2176 && entry->GetElement() != 0x0080 ) )
2179 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
2183 //////// Fix for some brain-dead 'Leonardo' Siemens images.
2184 // Occurence of such images is quite low (unless one leaves close to a
2185 // 'Leonardo' source. Hence, one might consider commenting out the
2186 // following fix on efficiency reasons.
2187 else if ( entry->GetGroup() == 0x0009
2188 && ( entry->GetElement() == 0x1113
2189 || entry->GetElement() == 0x1114 ) )
2192 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
2195 else if ( entry->GetVR() == "SQ" )
2197 foundLength = 0; // ReadLength is unchanged
2200 //////// We encountered a 'delimiter' element i.e. a tag of the form
2201 // "fffe|xxxx" which is just a marker. Delimiters length should not be
2202 // taken into account.
2203 else if( entry->GetGroup() == 0xfffe )
2205 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
2206 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
2207 // causes extra troubles...
2208 if( entry->GetElement() != 0x0000 )
2214 entry->SetUsableLength(foundLength);
2218 * \brief Apply some heuristics to predict whether the considered
2219 * element value contains/represents an integer or not.
2220 * @param entry The element value on which to apply the predicate.
2221 * @return The result of the heuristical predicate.
2223 bool Document::IsDocEntryAnInteger(DocEntry *entry)
2225 uint16_t element = entry->GetElement();
2226 uint16_t group = entry->GetGroup();
2227 const std::string & vr = entry->GetVR();
2228 uint32_t length = entry->GetLength();
2230 // When we have some semantics on the element we just read, and if we
2231 // a priori know we are dealing with an integer, then we shall be
2232 // able to swap it's element value properly.
2233 if ( element == 0 ) // This is the group length of the group
2241 // Allthough this should never happen, still some images have a
2242 // corrupted group length [e.g. have a glance at offset x(8336) of
2243 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2244 // Since for dicom compliant and well behaved headers, the present
2245 // test is useless (and might even look a bit paranoid), when we
2246 // encounter such an ill-formed image, we simply display a warning
2247 // message and proceed on parsing (while crossing fingers).
2248 std::ostringstream s;
2249 long filePosition = Fp->tellg();
2250 s << "Erroneous Group Length element length on : (" \
2251 << std::hex << group << " , " << element
2252 << ") -before- position x(" << filePosition << ")"
2253 << "lgt : " << length;
2254 dbg.Verbose(0, "Document::IsDocEntryAnInteger", s.str().c_str() );
2258 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2267 * \brief Find the Length till the next sequence delimiter
2268 * \warning NOT end user intended method !
2272 uint32_t Document::FindDocEntryLengthOB()
2273 throw( FormatUnexpected )
2275 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2276 long positionOnEntry = Fp->tellg();
2277 bool foundSequenceDelimiter = false;
2278 uint32_t totalLength = 0;
2280 while ( !foundSequenceDelimiter )
2286 group = ReadInt16();
2289 catch ( FormatError )
2291 throw FormatError("Document::FindDocEntryLengthOB()",
2292 " group or element not present.");
2295 // We have to decount the group and element we just read
2298 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2300 dbg.Verbose(1, "Document::FindDocEntryLengthOB: neither an Item "
2301 "tag nor a Sequence delimiter tag.");
2302 Fp->seekg(positionOnEntry, std::ios_base::beg);
2303 throw FormatUnexpected("Document::FindDocEntryLengthOB()",
2304 "Neither an Item tag nor a Sequence "
2308 if ( elem == 0xe0dd )
2310 foundSequenceDelimiter = true;
2313 uint32_t itemLength = ReadInt32();
2314 // We add 4 bytes since we just read the ItemLength with ReadInt32
2315 totalLength += itemLength + 4;
2316 SkipBytes(itemLength);
2318 if ( foundSequenceDelimiter )
2323 Fp->seekg( positionOnEntry, std::ios_base::beg);
2328 * \brief Reads a supposed to be 16 Bits integer
2329 * (swaps it depending on processor endianity)
2330 * @return read value
2332 uint16_t Document::ReadInt16()
2333 throw( FormatError )
2336 Fp->read ((char*)&g, (size_t)2);
2339 throw FormatError( "Document::ReadInt16()", " file error." );
2343 throw FormatError( "Document::ReadInt16()", "EOF." );
2350 * \brief Reads a supposed to be 32 Bits integer
2351 * (swaps it depending on processor endianity)
2352 * @return read value
2354 uint32_t Document::ReadInt32()
2355 throw( FormatError )
2358 Fp->read ((char*)&g, (size_t)4);
2361 throw FormatError( "Document::ReadInt32()", " file error." );
2365 throw FormatError( "Document::ReadInt32()", "EOF." );
2372 * \brief skips bytes inside the source file
2373 * \warning NOT end user intended method !
2376 void Document::SkipBytes(uint32_t nBytes)
2378 //FIXME don't dump the returned value
2379 Fp->seekg((long)nBytes, std::ios_base::cur);
2383 * \brief Loads all the needed Dictionaries
2384 * \warning NOT end user intended method !
2386 void Document::Initialise()
2388 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2390 RLEInfo = new RLEFramesInfo;
2391 JPEGInfo = new JPEGFragmentsInfo;
2395 * \brief Discover what the swap code is (among little endian, big endian,
2396 * bad little endian, bad big endian).
2398 * @return false when we are absolutely sure
2399 * it's neither ACR-NEMA nor DICOM
2400 * true when we hope ours assuptions are OK
2402 bool Document::CheckSwap()
2404 // The only guaranted way of finding the swap code is to find a
2405 // group tag since we know it's length has to be of four bytes i.e.
2406 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2407 // occurs when we can't find such group...
2409 uint32_t x = 4; // x : for ntohs
2410 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2414 char deb[256]; //HEADER_LENGTH_TO_READ];
2416 // First, compare HostByteOrder and NetworkByteOrder in order to
2417 // determine if we shall need to swap bytes (i.e. the Endian type).
2418 if ( x == ntohs(x) )
2427 // The easiest case is the one of a DICOM header, since it possesses a
2428 // file preamble where it suffice to look for the string "DICM".
2429 Fp->read(deb, HEADER_LENGTH_TO_READ);
2431 char *entCur = deb + 128;
2432 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2434 dbg.Verbose(1, "Document::CheckSwap:", "looks like DICOM Version3");
2436 // Next, determine the value representation (VR). Let's skip to the
2437 // first element (0002, 0000) and check there if we find "UL"
2438 // - or "OB" if the 1st one is (0002,0001) -,
2439 // in which case we (almost) know it is explicit VR.
2440 // WARNING: if it happens to be implicit VR then what we will read
2441 // is the length of the group. If this ascii representation of this
2442 // length happens to be "UL" then we shall believe it is explicit VR.
2443 // FIXME: in order to fix the above warning, we could read the next
2444 // element value (or a couple of elements values) in order to make
2445 // sure we are not commiting a big mistake.
2446 // We need to skip :
2447 // * the 128 bytes of File Preamble (often padded with zeroes),
2448 // * the 4 bytes of "DICM" string,
2449 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2450 // i.e. a total of 136 bytes.
2454 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2455 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2456 // *Implicit* VR. -and it is !-
2458 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2459 memcmp(entCur, "OB", (size_t)2) == 0 ||
2460 memcmp(entCur, "UI", (size_t)2) == 0 ||
2461 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2462 // when Write DCM *adds*
2464 // Use Document::dicom_vr to test all the possibilities
2465 // instead of just checking for UL, OB and UI !? group 0000
2467 Filetype = ExplicitVR;
2468 dbg.Verbose(1, "Document::CheckSwap:",
2469 "explicit Value Representation");
2473 Filetype = ImplicitVR;
2474 dbg.Verbose(1, "Document::CheckSwap:",
2475 "not an explicit Value Representation");
2481 dbg.Verbose(1, "Document::CheckSwap:",
2482 "HostByteOrder != NetworkByteOrder");
2487 dbg.Verbose(1, "Document::CheckSwap:",
2488 "HostByteOrder = NetworkByteOrder");
2491 // Position the file position indicator at first tag (i.e.
2492 // after the file preamble and the "DICM" string).
2493 Fp->seekg(0, std::ios_base::beg);
2494 Fp->seekg ( 132L, std::ios_base::beg);
2498 // Alas, this is not a DicomV3 file and whatever happens there is no file
2499 // preamble. We can reset the file position indicator to where the data
2500 // is (i.e. the beginning of the file).
2501 dbg.Verbose(1, "Document::CheckSwap:", "not a DICOM Version3 file");
2502 Fp->seekg(0, std::ios_base::beg);
2504 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2505 // By clean we mean that the length of the first tag is written down.
2506 // If this is the case and since the length of the first group HAS to be
2507 // four (bytes), then determining the proper swap code is straightforward.
2510 // We assume the array of char we are considering contains the binary
2511 // representation of a 32 bits integer. Hence the following dirty
2513 s32 = *((uint32_t *)(entCur));
2534 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2535 // It is time for despaired wild guesses.
2536 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2537 // i.e. the 'group length' element is not present :
2539 // check the supposed to be 'group number'
2540 // 0x0002 or 0x0004 or 0x0008
2541 // to determine ' SwapCode' value .
2542 // Only 0 or 4321 will be possible
2543 // (no oportunity to check for the formerly well known
2544 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2545 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -4, 8-)
2546 // the file IS NOT ACR-NEMA nor DICOM V3
2547 // Find a trick to tell it the caller...
2549 s16 = *((uint16_t *)(deb));
2566 dbg.Verbose(0, "Document::CheckSwap:",
2567 "ACR/NEMA unfound swap info (Really hopeless !)");
2571 // Then the only info we have is the net2host one.
2581 * \brief Restore the unproperly loaded values i.e. the group, the element
2582 * and the dictionary entry depending on them.
2584 void Document::SwitchSwapToBigEndian()
2586 dbg.Verbose(1, "Document::SwitchSwapToBigEndian",
2587 "Switching to BigEndian mode.");
2588 if ( SwapCode == 0 )
2592 else if ( SwapCode == 4321 )
2596 else if ( SwapCode == 3412 )
2600 else if ( SwapCode == 2143 )
2607 * \brief during parsing, Header Elements too long are not loaded in memory
2610 void Document::SetMaxSizeLoadEntry(long newSize)
2616 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2618 MaxSizeLoadEntry = 0xffffffff;
2621 MaxSizeLoadEntry = newSize;
2626 * \brief Header Elements too long will not be printed
2627 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2630 void Document::SetMaxSizePrintEntry(long newSize)
2632 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2637 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2639 MaxSizePrintEntry = 0xffffffff;
2642 MaxSizePrintEntry = newSize;
2648 * \brief Handle broken private tag from Philips NTSCAN
2649 * where the endianess is being switch to BigEndian for no
2653 void Document::HandleBrokenEndian(uint16_t group, uint16_t elem)
2655 // Endian reversion. Some files contain groups of tags with reversed endianess.
2656 static int reversedEndian = 0;
2657 // try to fix endian switching in the middle of headers
2658 if ((group == 0xfeff) && (elem == 0x00e0))
2660 // start endian swap mark for group found
2662 SwitchSwapToBigEndian();
2667 else if ((group == 0xfffe) && (elem == 0xe00d) && reversedEndian)
2669 // end of reversed endian group
2671 SwitchSwapToBigEndian();
2677 * \brief Read the next tag but WITHOUT loading it's value
2678 * (read the 'Group Number', the 'Element Number',
2679 * gets the Dict Entry
2680 * gets the VR, gets the length, gets the offset value)
2681 * @return On succes the newly created DocEntry, NULL on failure.
2683 DocEntry* Document::ReadNextDocEntry()
2690 group = ReadInt16();
2693 catch ( FormatError e )
2695 // We reached the EOF (or an error occured) therefore
2696 // header parsing has to be considered as finished.
2701 HandleBrokenEndian(group, elem);
2702 DocEntry *newEntry = NewDocEntryByNumber(group, elem);
2703 FindDocEntryVR(newEntry);
2707 FindDocEntryLength(newEntry);
2709 catch ( FormatError e )
2717 newEntry->SetOffset(Fp->tellg());
2724 * \brief Generate a free TagKey i.e. a TagKey that is not present
2725 * in the TagHt dictionary.
2726 * @param group The generated tag must belong to this group.
2727 * @return The element of tag with given group which is fee.
2729 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2731 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2733 TagKey key = DictEntry::TranslateToKey(group, elem);
2734 if (TagHT.count(key) == 0)
2743 * \brief Assuming the internal file pointer \ref Document::Fp
2744 * is placed at the beginning of a tag check whether this
2745 * tag is (TestGroup, TestElement).
2746 * \warning On success the internal file pointer \ref Document::Fp
2747 * is modified to point after the tag.
2748 * On failure (i.e. when the tag wasn't the expected tag
2749 * (TestGroup, TestElement) the internal file pointer
2750 * \ref Document::Fp is restored to it's original position.
2751 * @param testGroup The expected group of the tag.
2752 * @param testElement The expected Element of the tag.
2753 * @return True on success, false otherwise.
2755 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2757 long positionOnEntry = Fp->tellg();
2758 long currentPosition = Fp->tellg(); // On debugging purposes
2760 //// Read the Item Tag group and element, and make
2761 // sure they are what we expected:
2762 uint16_t itemTagGroup;
2763 uint16_t itemTagElement;
2766 itemTagGroup = ReadInt16();
2767 itemTagElement = ReadInt16();
2769 catch ( FormatError e )
2771 //std::cerr << e << std::endl;
2774 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2776 std::ostringstream s;
2777 s << " We should have found tag (";
2778 s << std::hex << testGroup << "," << testElement << ")" << std::endl;
2779 s << " but instead we encountered tag (";
2780 s << std::hex << itemTagGroup << "," << itemTagElement << ")"
2782 s << " at address: " << (unsigned)currentPosition << std::endl;
2783 dbg.Verbose(0, "Document::ReadItemTagLength: wrong Item Tag found:");
2784 dbg.Verbose(0, s.str().c_str());
2785 Fp->seekg(positionOnEntry, std::ios_base::beg);
2793 * \brief Assuming the internal file pointer \ref Document::Fp
2794 * is placed at the beginning of a tag (TestGroup, TestElement),
2795 * read the length associated to the Tag.
2796 * \warning On success the internal file pointer \ref Document::Fp
2797 * is modified to point after the tag and it's length.
2798 * On failure (i.e. when the tag wasn't the expected tag
2799 * (TestGroup, TestElement) the internal file pointer
2800 * \ref Document::Fp is restored to it's original position.
2801 * @param testGroup The expected group of the tag.
2802 * @param testElement The expected Element of the tag.
2803 * @return On success returns the length associated to the tag. On failure
2806 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2808 long positionOnEntry = Fp->tellg();
2809 (void)positionOnEntry;
2811 if ( !ReadTag(testGroup, testElement) )
2816 //// Then read the associated Item Length
2817 long currentPosition = Fp->tellg();
2818 uint32_t itemLength = ReadInt32();
2820 std::ostringstream s;
2821 s << "Basic Item Length is: "
2822 << itemLength << std::endl;
2823 s << " at address: " << (unsigned)currentPosition << std::endl;
2824 dbg.Verbose(0, "Document::ReadItemTagLength: ", s.str().c_str());
2830 * \brief When parsing the Pixel Data of an encapsulated file, read
2831 * the basic offset table (when present, and BTW dump it).
2833 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2835 //// Read the Basic Offset Table Item Tag length...
2836 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2838 // When present, read the basic offset table itself.
2839 // Notes: - since the presence of this basic offset table is optional
2840 // we can't rely on it for the implementation, and we will simply
2841 // trash it's content (when present).
2842 // - still, when present, we could add some further checks on the
2843 // lengths, but we won't bother with such fuses for the time being.
2844 if ( itemLength != 0 )
2846 char* basicOffsetTableItemValue = new char[itemLength + 1];
2847 Fp->read(basicOffsetTableItemValue, itemLength);
2850 for (unsigned int i=0; i < itemLength; i += 4 )
2852 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2854 std::ostringstream s;
2855 s << " Read one length: ";
2856 s << std::hex << individualLength << std::endl;
2858 "Document::ReadAndSkipEncapsulatedBasicOffsetTable: ",
2863 delete[] basicOffsetTableItemValue;
2868 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2869 * Compute the RLE extra information and store it in \ref RLEInfo
2870 * for later pixel retrieval usage.
2872 void Document::ComputeRLEInfo()
2874 TransferSyntaxType ts = GetTransferSyntax();
2875 if ( ts != RLELossless )
2880 // Encoded pixel data: for the time being we are only concerned with
2881 // Jpeg or RLE Pixel data encodings.
2882 // As stated in PS 3.5-2003, section 8.2 p44:
2883 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2884 // value representation OB is used".
2885 // Hence we expect an OB value representation. Concerning OB VR,
2886 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2887 // "For the Value Representations OB and OW, the encoding shall meet the
2888 // following specifications depending on the Data element tag:"
2890 // - the first item in the sequence of items before the encoded pixel
2891 // data stream shall be basic offset table item. The basic offset table
2892 // item value, however, is not required to be present"
2894 ReadAndSkipEncapsulatedBasicOffsetTable();
2896 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2897 // Loop on the individual frame[s] and store the information
2898 // on the RLE fragments in a RLEFramesInfo.
2899 // Note: - when only a single frame is present, this is a
2901 // - when more than one frame are present, then we are in
2902 // the case of a multi-frame image.
2904 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2906 // Parse the RLE Header and store the corresponding RLE Segment
2907 // Offset Table information on fragments of this current Frame.
2908 // Note that the fragment pixels themselves are not loaded
2909 // (but just skipped).
2910 long frameOffset = Fp->tellg();
2912 uint32_t nbRleSegments = ReadInt32();
2913 if ( nbRleSegments > 16 )
2915 // There should be at most 15 segments (refer to RLEFrame class)
2916 dbg.Verbose(0, "Document::ComputeRLEInfo: too many segments.");
2919 uint32_t rleSegmentOffsetTable[16];
2920 for( int k = 1; k <= 15; k++ )
2922 rleSegmentOffsetTable[k] = ReadInt32();
2925 // Deduce from both the RLE Header and the frameLength the
2926 // fragment length, and again store this info in a
2928 long rleSegmentLength[15];
2929 // skipping (not reading) RLE Segments
2930 if ( nbRleSegments > 1)
2932 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2934 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2935 - rleSegmentOffsetTable[k];
2936 SkipBytes(rleSegmentLength[k]);
2940 rleSegmentLength[nbRleSegments] = frameLength
2941 - rleSegmentOffsetTable[nbRleSegments];
2942 SkipBytes(rleSegmentLength[nbRleSegments]);
2944 // Store the collected info
2945 RLEFrame* newFrameInfo = new RLEFrame;
2946 newFrameInfo->NumberFragments = nbRleSegments;
2947 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2949 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2950 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2952 RLEInfo->Frames.push_back( newFrameInfo );
2955 // Make sure that at the end of the item we encounter a 'Sequence
2957 if ( !ReadTag(0xfffe, 0xe0dd) )
2959 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2960 dbg.Verbose(0, " item at end of RLE item sequence");
2965 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2966 * Compute the jpeg extra information (fragment[s] offset[s] and
2967 * length) and store it[them] in \ref JPEGInfo for later pixel
2970 void Document::ComputeJPEGFragmentInfo()
2972 // If you need to, look for comments of ComputeRLEInfo().
2978 ReadAndSkipEncapsulatedBasicOffsetTable();
2980 // Loop on the fragments[s] and store the parsed information in a
2982 long fragmentLength;
2983 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2985 long fragmentOffset = Fp->tellg();
2987 // Store the collected info
2988 JPEGFragment* newFragment = new JPEGFragment;
2989 newFragment->Offset = fragmentOffset;
2990 newFragment->Length = fragmentLength;
2991 JPEGInfo->Fragments.push_back( newFragment );
2993 SkipBytes( fragmentLength );
2996 // Make sure that at the end of the item we encounter a 'Sequence
2998 if ( !ReadTag(0xfffe, 0xe0dd) )
3000 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
3001 dbg.Verbose(0, " item at end of JPEG item sequence");
3006 * \brief Walk recursively the given \ref DocEntrySet, and feed
3007 * the given hash table (\ref TagDocEntryHT) with all the
3008 * \ref DocEntry (Dicom entries) encountered.
3009 * This method does the job for \ref BuildFlatHashTable.
3010 * @param builtHT Where to collect all the \ref DocEntry encountered
3011 * when recursively walking the given set.
3012 * @param set The structure to be traversed (recursively).
3014 void Document::BuildFlatHashTableRecurse( TagDocEntryHT& builtHT,
3017 if (ElementSet* elementSet = dynamic_cast< ElementSet* > ( set ) )
3019 TagDocEntryHT const & currentHT = elementSet->GetTagHT();
3020 for( TagDocEntryHT::const_iterator i = currentHT.begin();
3021 i != currentHT.end();
3024 DocEntry* entry = i->second;
3025 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3027 const ListSQItem& items = seqEntry->GetSQItems();
3028 for( ListSQItem::const_iterator item = items.begin();
3029 item != items.end();
3032 BuildFlatHashTableRecurse( builtHT, *item );
3036 builtHT[entry->GetKey()] = entry;
3041 if (SQItem* SQItemSet = dynamic_cast< SQItem* > ( set ) )
3043 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
3044 for (ListDocEntry::const_iterator i = currentList.begin();
3045 i != currentList.end();
3048 DocEntry* entry = *i;
3049 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3051 const ListSQItem& items = seqEntry->GetSQItems();
3052 for( ListSQItem::const_iterator item = items.begin();
3053 item != items.end();
3056 BuildFlatHashTableRecurse( builtHT, *item );
3060 builtHT[entry->GetKey()] = entry;
3067 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
3070 * The structure used by a Document (through \ref ElementSet),
3071 * in order to old the parsed entries of a Dicom header, is a recursive
3072 * one. This is due to the fact that the sequences (when present)
3073 * can be nested. Additionaly, the sequence items (represented in
3074 * gdcm as \ref SQItem) add an extra complexity to the data
3075 * structure. Hence, a gdcm user whishing to visit all the entries of
3076 * a Dicom header will need to dig in the gdcm internals (which
3077 * implies exposing all the internal data structures to the API).
3078 * In order to avoid this burden to the user, \ref BuildFlatHashTable
3079 * recursively builds a temporary hash table, which holds all the
3080 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
3082 * \warning Of course there is NO integrity constrain between the
3083 * returned \ref TagDocEntryHT and the \ref ElementSet used
3084 * to build it. Hence if the underlying \ref ElementSet is
3085 * altered, then it is the caller responsability to invoke
3086 * \ref BuildFlatHashTable again...
3087 * @return The flat std::map<> we juste build.
3089 TagDocEntryHT* Document::BuildFlatHashTable()
3091 TagDocEntryHT* FlatHT = new TagDocEntryHT;
3092 BuildFlatHashTableRecurse( *FlatHT, this );
3099 * \brief Compares two documents, according to \ref DicomDir rules
3100 * \warning Does NOT work with ACR-NEMA files
3101 * \todo Find a trick to solve the pb (use RET fields ?)
3103 * @return true if 'smaller'
3105 bool Document::operator<(Document &document)
3108 std::string s1 = GetEntryByNumber(0x0010,0x0010);
3109 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
3121 s1 = GetEntryByNumber(0x0010,0x0020);
3122 s2 = document.GetEntryByNumber(0x0010,0x0020);
3133 // Study Instance UID
3134 s1 = GetEntryByNumber(0x0020,0x000d);
3135 s2 = document.GetEntryByNumber(0x0020,0x000d);
3146 // Serie Instance UID
3147 s1 = GetEntryByNumber(0x0020,0x000e);
3148 s2 = document.GetEntryByNumber(0x0020,0x000e);
3163 } // end namespace gdcm
3165 //-----------------------------------------------------------------------------