1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2004/11/17 19:49:13 $
7 Version: $Revision: 1.134 $
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 // Implicit VR Little Endian DLX G.E?
44 // Explicit VR Little Endian
45 "1.2.840.10008.1.2.1",
46 // Deflated Explicit VR Little Endian
47 "1.2.840.10008.1.2.1.99",
48 // Explicit VR Big Endian
49 "1.2.840.10008.1.2.2",
50 // JPEG Baseline (Process 1)
51 "1.2.840.10008.1.2.4.50",
52 // JPEG Extended (Process 2 & 4)
53 "1.2.840.10008.1.2.4.51",
54 // JPEG Extended (Process 3 & 5)
55 "1.2.840.10008.1.2.4.52",
56 // JPEG Spectral Selection, Non-Hierarchical (Process 6 & 8)
57 "1.2.840.10008.1.2.4.53",
58 // JPEG Full Progression, Non-Hierarchical (Process 10 & 12)
59 "1.2.840.10008.1.2.4.55",
60 // JPEG Lossless, Non-Hierarchical (Process 14)
61 "1.2.840.10008.1.2.4.57",
62 // JPEG Lossless, Hierarchical, First-Order Prediction (Process 14, [Selection Value 1])
63 "1.2.840.10008.1.2.4.70",
65 "1.2.840.10008.1.2.4.90",
67 "1.2.840.10008.1.2.4.91",
69 "1.2.840.10008.1.2.5",
71 "Unknown Transfer Syntax"
74 //-----------------------------------------------------------------------------
75 // Refer to Document::CheckSwap()
76 const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
78 // Refer to Document::SetMaxSizeLoadEntry()
79 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
80 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
82 //-----------------------------------------------------------------------------
83 // Constructor / Destructor
87 * @param filename file to be opened for parsing
89 Document::Document( std::string const & filename ) : ElementSet(-1)
91 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
101 dbg.Verbose(0, "Document::Document: starting parsing of file: ",
103 Fp->seekg( 0, std::ios_base::beg);
105 Fp->seekg(0, std::ios_base::end);
106 long lgt = Fp->tellg();
108 Fp->seekg( 0, std::ios_base::beg);
110 long beg = Fp->tellg();
113 ParseDES( this, beg, lgt, false); // le Load sera fait a la volee
115 Fp->seekg( 0, std::ios_base::beg);
117 // Load 'non string' values
119 std::string PhotometricInterpretation = GetEntryByNumber(0x0028,0x0004);
120 if( PhotometricInterpretation == "PALETTE COLOR " )
122 LoadEntryBinArea(0x0028,0x1200); // gray LUT
123 /// FIXME FIXME FIXME
124 /// The tags refered by the three following lines used to be CORRECTLY
125 /// defined as having an US Value Representation in the public
126 /// dictionnary. BUT the semantics implied by the three following
127 /// lines state that the corresponding tag contents are in fact
128 /// the ones of a BinEntry.
129 /// In order to fix things "Quick and Dirty" the dictionnary was
130 /// altered on PURPOUS but now contains a WRONG value.
131 /// In order to fix things and restore the dictionary to its
132 /// correct value, one needs to decided of the semantics by deciding
133 /// wether the following tags are either:
134 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
135 /// also used as BinEntry, which requires the proper conversion,
136 /// - OW, and hence loaded as BinEntry, but afterwards also used
137 /// as ValEntry, which requires the proper conversion.
138 LoadEntryBinArea(0x0028,0x1201); // R LUT
139 LoadEntryBinArea(0x0028,0x1202); // G LUT
140 LoadEntryBinArea(0x0028,0x1203); // B LUT
142 // Segmented Red Palette Color LUT Data
143 LoadEntryBinArea(0x0028,0x1221);
144 // Segmented Green Palette Color LUT Data
145 LoadEntryBinArea(0x0028,0x1222);
146 // Segmented Blue Palette Color LUT Data
147 LoadEntryBinArea(0x0028,0x1223);
149 //FIXME later : how to use it?
150 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
154 // --------------------------------------------------------------
155 // Specific code to allow gdcm to read ACR-LibIDO formated images
156 // Note: ACR-LibIDO is an extension of the ACR standard that was
157 // used at CREATIS. For the time being (say a couple years)
158 // we keep this kludge to allow a smooth move to gdcm for
159 // CREATIS developpers (sorry folks).
161 // if recognition code tells us we deal with a LibIDO image
162 // we switch lineNumber and columnNumber
165 RecCode = GetEntryByNumber(0x0008, 0x0010); // recognition code
166 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
167 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
168 // with "little-endian strings"
170 Filetype = ACR_LIBIDO;
171 std::string rows = GetEntryByNumber(0x0028, 0x0010);
172 std::string columns = GetEntryByNumber(0x0028, 0x0011);
173 SetEntryByNumber(columns, 0x0028, 0x0010);
174 SetEntryByNumber(rows , 0x0028, 0x0011);
176 // ----------------- End of ACR-LibIDO kludge ------------------
178 PrintLevel = 1; // 'Medium' print level by default
182 * \brief This default constructor doesn't parse the file. You should
183 * then invoke \ref Document::SetFileName and then the parsing.
185 Document::Document() : ElementSet(-1)
189 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
192 Filetype = ExplicitVR;
193 PrintLevel = 1; // 'Medium' print level by default
197 * \brief Canonical destructor.
199 Document::~Document ()
208 //-----------------------------------------------------------------------------
212 * \brief Prints The Dict Entries of THE public Dicom Dictionary
215 void Document::PrintPubDict(std::ostream & os)
217 RefPubDict->Print(os);
221 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
224 void Document::PrintShaDict(std::ostream & os)
226 RefShaDict->Print(os);
229 //-----------------------------------------------------------------------------
232 * \brief Get the public dictionary used
234 Dict* Document::GetPubDict()
240 * \brief Get the shadow dictionary used
242 Dict* Document::GetShaDict()
248 * \brief Set the shadow dictionary used
249 * \param dict dictionary to use in shadow
251 bool Document::SetShaDict(Dict *dict)
258 * \brief Set the shadow dictionary used
259 * \param dictName name of the dictionary to use in shadow
261 bool Document::SetShaDict(DictKey const & dictName)
263 RefShaDict = Global::GetDicts()->GetDict(dictName);
268 * \brief This predicate, based on hopefully reasonable heuristics,
269 * decides whether or not the current Document was properly parsed
270 * and contains the mandatory information for being considered as
271 * a well formed and usable Dicom/Acr File.
272 * @return true when Document is the one of a reasonable Dicom/Acr file,
275 bool Document::IsReadable()
277 if( Filetype == Unknown)
279 dbg.Verbose(0, "Document::IsReadable: wrong filetype");
285 dbg.Verbose(0, "Document::IsReadable: no tags in internal"
294 * \brief Accessor to the Transfer Syntax (when present) of the
295 * current document (it internally handles reading the
296 * value from disk when only parsing occured).
297 * @return The encountered Transfer Syntax of the current document.
299 TransferSyntaxType Document::GetTransferSyntax()
301 DocEntry *entry = GetDocEntryByNumber(0x0002, 0x0010);
307 // The entry might be present but not loaded (parsing and loading
308 // happen at different stages): try loading and proceed with check...
309 LoadDocEntrySafe(entry);
310 if (ValEntry* valEntry = dynamic_cast< ValEntry* >(entry) )
312 std::string transfer = valEntry->GetValue();
313 // The actual transfer (as read from disk) might be padded. We
314 // first need to remove the potential padding. We can make the
315 // weak assumption that padding was not executed with digits...
316 if ( transfer.length() == 0 )
318 // for brain damaged headers
321 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
323 transfer.erase(transfer.length()-1, 1);
325 for (int i = 0; TransferSyntaxStrings[i] != NULL; i++)
327 if ( TransferSyntaxStrings[i] == transfer )
329 return TransferSyntaxType(i);
336 bool Document::IsJPEGLossless()
338 TransferSyntaxType r = GetTransferSyntax();
339 return r == JPEGFullProgressionProcess10_12
340 || r == JPEGLosslessProcess14
341 || r == JPEGLosslessProcess14_1;
345 * \brief Determines if the Transfer Syntax was already encountered
346 * and if it corresponds to a JPEG2000 one
347 * @return True when JPEG2000 (Lossly or LossLess) found. False in all
350 bool Document::IsJPEG2000()
352 TransferSyntaxType r = GetTransferSyntax();
353 return r == JPEG2000Lossless || r == JPEG2000;
357 * \brief Determines if the Transfer Syntax corresponds to any form
358 * of Jpeg encoded Pixel data.
359 * @return True when any form of JPEG found. False otherwise.
361 bool Document::IsJPEG()
363 TransferSyntaxType r = GetTransferSyntax();
364 return r == JPEGBaselineProcess1
365 || r == JPEGExtendedProcess2_4
366 || r == JPEGExtendedProcess3_5
367 || r == JPEGSpectralSelectionProcess6_8
373 * \brief Determines if the Transfer Syntax corresponds to encapsulated
374 * of encoded Pixel Data (as opposed to native).
375 * @return True when encapsulated. False when native.
377 bool Document::IsEncapsulate()
379 TransferSyntaxType r = GetTransferSyntax();
380 return IsJPEG() || r == RLELossless;
384 * \brief Predicate for dicom version 3 file.
385 * @return True when the file is a dicom version 3.
387 bool Document::IsDicomV3()
389 // Checking if Transfert Syntax exists is enough
390 // Anyway, it's to late check if the 'Preamble' was found ...
391 // And ... would it be a rich idea to check ?
392 // (some 'no Preamble' DICOM images exist !)
393 return GetDocEntryByNumber(0x0002, 0x0010) != NULL;
397 * \brief returns the File Type
398 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
399 * @return the FileType code
401 FileType Document::GetFileType()
407 * \brief Tries to open the file \ref Document::Filename and
408 * checks the preamble when existing.
409 * @return The FILE pointer on success.
411 std::ifstream* Document::OpenFile()
413 if (Filename.length() == 0) return 0;
417 "Document::OpenFile is already opened when opening: ",
421 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
426 "Document::OpenFile cannot open file: ",
432 Fp->read((char*)&zero, (size_t)2 );
434 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
436 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
437 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
438 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
439 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
445 Fp->seekg(126L, std::ios_base::cur);
447 Fp->read(dicm, (size_t)4);
448 if( memcmp(dicm, "DICM", 4) == 0 )
455 "Document::OpenFile not DICOM/ACR (missing preamble)",
462 * \brief closes the file
463 * @return TRUE if the close was successfull
465 bool Document::CloseFile()
474 return true; //FIXME how do we detect a non-close ifstream ?
478 * \brief Writes in a file all the Header Entries (Dicom Elements)
479 * @param fp file pointer on an already open file
480 * @param filetype Type of the File to be written
481 * (ACR-NEMA, ExplicitVR, ImplicitVR)
482 * \return Always true.
484 void Document::Write(std::ofstream* fp, FileType filetype)
486 /// \todo move the following lines (and a lot of others, to be written)
487 /// to a future function CheckAndCorrectHeader
488 /// (necessary if user wants to write a DICOM V3 file
489 /// starting from an ACR-NEMA (V2) Header
491 if (filetype == ImplicitVR)
494 Util::DicomString( TransferSyntaxStrings[ImplicitVRLittleEndian] );
495 ReplaceOrCreateByNumber(ts, 0x0002, 0x0010);
497 /// \todo Refer to standards on page 21, chapter 6.2
498 /// "Value representation": values with a VR of UI shall be
499 /// padded with a single trailing null
500 /// in the following case we have to padd manually with a 0
502 SetEntryLengthByNumber(18, 0x0002, 0x0010);
505 if (filetype == ExplicitVR)
508 Util::DicomString( TransferSyntaxStrings[ExplicitVRLittleEndian] );
509 ReplaceOrCreateByNumber(ts, 0x0002, 0x0010); //LEAK
511 /// \todo Refer to standards on page 21, chapter 6.2
512 /// "Value representation": values with a VR of UI shall be
513 /// padded with a single trailing null
514 /// Dans le cas suivant on doit pader manuellement avec un 0
516 SetEntryLengthByNumber(20, 0x0002, 0x0010);
520 * \todo rewrite later, if really usefull
521 * - 'Group Length' element is optional in DICOM
522 * - but un-updated odd groups lengthes can causes pb
525 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
526 * UpdateGroupLength(false,filetype);
527 * if ( filetype == ACR)
528 * UpdateGroupLength(true,ACR);
531 ElementSet::Write(fp, filetype); // This one is recursive
536 * \brief Modifies the value of a given Header Entry (Dicom Element)
537 * when it exists. Create it with the given value when unexistant.
538 * @param value (string) Value to be set
539 * @param group Group number of the Entry
540 * @param elem Element number of the Entry
541 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
542 * \return pointer to the modified/created Header Entry (NULL when creation
545 ValEntry* Document::ReplaceOrCreateByNumber(
546 std::string const & value,
551 ValEntry* valEntry = 0;
552 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
556 // check if (group,element) DictEntry exists
557 // if it doesn't, create an entry in DictSet::VirtualEntry
560 // Find out if the tag we received is in the dictionaries:
561 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
562 DictEntry* dictEntry = pubDict->GetDictEntryByNumber(group, elem);
565 currentEntry = NewDocEntryByNumber(group, elem, vr);
569 currentEntry = NewDocEntryByNumber(group, elem);
574 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
575 " NewDocEntryByNumber failed.");
579 valEntry = new ValEntry(currentEntry);
582 if ( !AddEntry(valEntry))
585 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
586 " failed allthough this is a creation.");
592 valEntry = dynamic_cast< ValEntry* >(currentEntry);
593 if ( !valEntry ) // Euuuuh? It wasn't a ValEntry
594 // then we change it to a ValEntry ?
595 // Shouldn't it be considered as an error ?
597 // We need to promote the DocEntry to a ValEntry:
598 valEntry = new ValEntry(currentEntry);
599 if (!RemoveEntry(currentEntry))
602 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
603 " of previous DocEntry failed.");
606 if ( !AddEntry(valEntry))
609 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
610 " promoted ValEntry failed.");
616 SetEntryByNumber(value, group, elem);
622 * \brief Modifies the value of a given Header Entry (Dicom Element)
623 * when it exists. Create it with the given value when unexistant.
624 * A copy of the binArea is made to be kept in the Document.
625 * @param binArea (binary) value to be set
626 * @param Group Group number of the Entry
627 * @param Elem Element number of the Entry
628 * \return pointer to the modified/created Header Entry (NULL when creation
631 BinEntry* Document::ReplaceOrCreateByNumber(
638 BinEntry* binEntry = 0;
639 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
643 // check if (group,element) DictEntry exists
644 // if it doesn't, create an entry in DictSet::VirtualEntry
647 // Find out if the tag we received is in the dictionaries:
648 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
649 DictEntry *dictEntry = pubDict->GetDictEntryByNumber(group, elem);
653 currentEntry = NewDocEntryByNumber(group, elem, vr);
657 currentEntry = NewDocEntryByNumber(group, elem);
661 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
662 " NewDocEntryByNumber failed.");
665 binEntry = new BinEntry(currentEntry);
666 if ( !AddEntry(binEntry))
668 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
669 " failed allthough this is a creation.");
674 binEntry = dynamic_cast< BinEntry* >(currentEntry);
675 if ( !binEntry ) // Euuuuh? It wasn't a BinEntry
676 // then we change it to a BinEntry ?
677 // Shouldn't it be considered as an error ?
679 // We need to promote the DocEntry to a BinEntry:
680 binEntry = new BinEntry(currentEntry);
681 if (!RemoveEntry(currentEntry))
683 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
684 " of previous DocEntry failed.");
687 if ( !AddEntry(binEntry))
689 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
690 " promoted BinEntry failed.");
697 if (lgth>0 && binArea)
699 tmpArea = new uint8_t[lgth];
700 memcpy(tmpArea,binArea,lgth);
706 if (!SetEntryByNumber(tmpArea, lgth, group, elem))
719 * \brief Modifies the value of a given Header Entry (Dicom Element)
720 * when it exists. Create it when unexistant.
721 * @param Group Group number of the Entry
722 * @param Elem Element number of the Entry
723 * \return pointer to the modified/created SeqEntry (NULL when creation
726 SeqEntry* Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem)
729 DocEntry* a = GetDocEntryByNumber( group, elem);
732 a = NewSeqEntryByNumber(group, elem);
738 b = new SeqEntry(a, 1); // FIXME : 1 (Depth)
745 * \brief Set a new value if the invoked element exists
746 * Seems to be useless !!!
747 * @param value new element value
748 * @param group group number of the Entry
749 * @param elem element number of the Entry
752 bool Document::ReplaceIfExistByNumber(std::string const & value,
753 uint16_t group, uint16_t elem )
755 SetEntryByNumber(value, group, elem);
760 //-----------------------------------------------------------------------------
764 * \brief Checks if a given Dicom Element exists within the H table
765 * @param group Group number of the searched Dicom Element
766 * @param element Element number of the searched Dicom Element
767 * @return true is found
769 bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
771 const std::string &key = DictEntry::TranslateToKey(group, element );
772 return TagHT.count(key) != 0;
776 * \brief Searches within Header Entries (Dicom Elements) parsed with
777 * the public and private dictionaries
778 * for the element value of a given tag.
779 * \warning Don't use any longer : use GetPubEntryByName
780 * @param tagName name of the searched element.
781 * @return Corresponding element value when it exists,
782 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
784 std::string Document::GetEntryByName(TagName const & tagName)
786 DictEntry* dictEntry = RefPubDict->GetDictEntryByName(tagName);
792 return GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
796 * \brief Searches within Header Entries (Dicom Elements) parsed with
797 * the public and private dictionaries
798 * for the element value representation of a given tag.
800 * Obtaining the VR (Value Representation) might be needed by caller
801 * to convert the string typed content to caller's native type
802 * (think of C++ vs Python). The VR is actually of a higher level
803 * of semantics than just the native C++ type.
804 * @param tagName name of the searched element.
805 * @return Corresponding element value representation when it exists,
806 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
808 std::string Document::GetEntryVRByName(TagName const& tagName)
810 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
811 if( dictEntry == NULL)
816 DocEntry* elem = GetDocEntryByNumber(dictEntry->GetGroup(),
817 dictEntry->GetElement());
818 return elem->GetVR();
822 * \brief Searches within Header Entries (Dicom Elements) parsed with
823 * the public and private dictionaries
824 * for the element value representation of a given tag.
825 * @param group Group number of the searched tag.
826 * @param element Element number of the searched tag.
827 * @return Corresponding element value representation when it exists,
828 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
830 std::string Document::GetEntryByNumber(uint16_t group, uint16_t element)
832 TagKey key = DictEntry::TranslateToKey(group, element);
833 /// \todo use map methods, instead of multimap JPR
834 if ( !TagHT.count(key))
839 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
843 * \brief Searches within Header Entries (Dicom Elements) parsed with
844 * the public and private dictionaries
845 * for the element value representation of a given tag..
847 * Obtaining the VR (Value Representation) might be needed by caller
848 * to convert the string typed content to caller's native type
849 * (think of C++ vs Python). The VR is actually of a higher level
850 * of semantics than just the native C++ type.
851 * @param group Group number of the searched tag.
852 * @param element Element number of the searched tag.
853 * @return Corresponding element value representation when it exists,
854 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
856 std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element)
858 DocEntry* elem = GetDocEntryByNumber(group, element);
863 return elem->GetVR();
867 * \brief Searches within Header Entries (Dicom Elements) parsed with
868 * the public and private dictionaries
869 * for the value length of a given tag..
870 * @param group Group number of the searched tag.
871 * @param element Element number of the searched tag.
872 * @return Corresponding element length; -2 if not found
874 int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element)
876 DocEntry* elem = GetDocEntryByNumber(group, element);
879 return -2; //magic number
881 return elem->GetLength();
884 * \brief Sets the value (string) of the Header Entry (Dicom Element)
885 * @param content string value of the Dicom Element
886 * @param tagName name of the searched Dicom Element.
887 * @return true when found
889 bool Document::SetEntryByName( std::string const & content,
890 TagName const & tagName)
892 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
898 return SetEntryByNumber(content,dictEntry->GetGroup(),
899 dictEntry->GetElement());
903 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
904 * through it's (group, element) and modifies it's content with
906 * @param content new value (string) to substitute with
907 * @param group group number of the Dicom Element to modify
908 * @param element element number of the Dicom Element to modify
910 bool Document::SetEntryByNumber(std::string const& content,
911 uint16_t group, uint16_t element)
916 ValEntry* valEntry = GetValEntryByNumber(group, element);
919 dbg.Verbose(0, "Document::SetEntryByNumber: no corresponding",
920 " ValEntry (try promotion first).");
923 // Non even content must be padded with a space (020H)...
924 std::string finalContent = Util::DicomString( content.c_str() );
925 assert( !(finalContent.size() % 2) );
926 valEntry->SetValue(finalContent);
928 // Integers have a special treatement for their length:
930 l = finalContent.length();
931 if ( l != 0) // To avoid to be cheated by 'zero length' integers
933 VRKey vr = valEntry->GetVR();
934 if( vr == "US" || vr == "SS" )
936 // for multivaluated items
937 c = Util::CountSubstring(content, "\\") + 1;
940 else if( vr == "UL" || vr == "SL" )
942 // for multivaluated items
943 c = Util::CountSubstring(content, "\\") + 1;
947 valEntry->SetLength(l);
952 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
953 * through it's (group, element) and modifies it's content with
955 * @param content new value (void* -> uint8_t*) to substitute with
956 * @param lgth new value length
957 * @param group group number of the Dicom Element to modify
958 * @param element element number of the Dicom Element to modify
960 bool Document::SetEntryByNumber(uint8_t*content, int lgth,
961 uint16_t group, uint16_t element)
963 (void)lgth; //not used
964 TagKey key = DictEntry::TranslateToKey(group, element);
965 if ( !TagHT.count(key))
970 /* Hope Binary field length is *never* wrong
971 if(lgth%2) // Non even length are padded with a space (020H).
974 //content = content + '\0'; // fing a trick to enlarge a binary field?
977 BinEntry* entry = (BinEntry *)TagHT[key];
978 entry->SetBinArea(content);
979 entry->SetLength(lgth);
980 entry->SetValue(GDCM_BINLOADED);
986 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
987 * in the PubDocEntrySet of this instance
988 * through it's (group, element) and modifies it's length with
990 * \warning Use with extreme caution.
991 * @param l new length to substitute with
992 * @param group group number of the Entry to modify
993 * @param element element number of the Entry to modify
994 * @return true on success, false otherwise.
996 bool Document::SetEntryLengthByNumber(uint32_t l,
997 uint16_t group, uint16_t element)
999 /// \todo use map methods, instead of multimap JPR
1000 TagKey key = DictEntry::TranslateToKey(group, element);
1001 if ( !TagHT.count(key) )
1007 l++; // length must be even
1009 ( ((TagHT.equal_range(key)).first)->second )->SetLength(l);
1015 * \brief Gets (from Header) the offset of a 'non string' element value
1016 * (LoadElementValues has already be executed)
1017 * @param group group number of the Entry
1018 * @param elem element number of the Entry
1019 * @return File Offset of the Element Value
1021 size_t Document::GetEntryOffsetByNumber(uint16_t group, uint16_t elem)
1023 DocEntry* entry = GetDocEntryByNumber(group, elem);
1026 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry present.");
1029 return entry->GetOffset();
1033 * \brief Gets (from Header) a 'non string' element value
1034 * (LoadElementValues has already be executed)
1035 * @param group group number of the Entry
1036 * @param elem element number of the Entry
1037 * @return Pointer to the 'non string' area
1039 void* Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem)
1041 DocEntry* entry = GetDocEntryByNumber(group, elem);
1044 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry");
1047 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(entry) )
1049 return binEntry->GetBinArea();
1056 * \brief Loads (from disk) the element content
1057 * when a string is not suitable
1058 * @param group group number of the Entry
1059 * @param elem element number of the Entry
1061 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
1063 // Search the corresponding DocEntry
1064 DocEntry *docElement = GetDocEntryByNumber(group, elem);
1068 size_t o =(size_t)docElement->GetOffset();
1069 Fp->seekg( o, std::ios_base::beg);
1070 size_t l = docElement->GetLength();
1071 uint8_t* a = new uint8_t[l];
1074 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1079 Fp->read((char*)a, l);
1080 if( Fp->fail() || Fp->eof() )//Fp->gcount() == 1
1086 // Set the value to the DocEntry
1087 if( !SetEntryBinAreaByNumber( a, group, elem ) )
1090 dbg.Verbose(0, "Document::LoadEntryBinArea setting failed.");
1094 * \brief Loads (from disk) the element content
1095 * when a string is not suitable
1096 * @param element Entry whose binArea is going to be loaded
1098 void Document::LoadEntryBinArea(BinEntry* element)
1100 size_t o =(size_t)element->GetOffset();
1101 Fp->seekg(o, std::ios_base::beg);
1102 size_t l = element->GetLength();
1103 uint8_t* a = new uint8_t[l];
1106 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1110 /// \todo check the result
1111 Fp->read((char*)a, l);
1112 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
1118 element->SetBinArea(a);
1122 * \brief Sets a 'non string' value to a given Dicom Element
1123 * @param area area containing the 'non string' value
1124 * @param group Group number of the searched Dicom Element
1125 * @param element Element number of the searched Dicom Element
1128 bool Document::SetEntryBinAreaByNumber(uint8_t* area,
1129 uint16_t group, uint16_t element)
1131 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1132 if ( !currentEntry )
1137 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1139 binEntry->SetBinArea( area );
1147 * \brief Update the entries with the shadow dictionary.
1148 * Only non even entries are analyzed
1150 void Document::UpdateShaEntries()
1155 /// \todo TODO : still any use to explore recursively the whole structure?
1157 for(ListTag::iterator it=listEntries.begin();
1158 it!=listEntries.end();
1161 // Odd group => from public dictionary
1162 if((*it)->GetGroup()%2==0)
1165 // Peer group => search the corresponding dict entry
1167 entry=RefShaDict->GetDictEntryByNumber((*it)->GetGroup(),(*it)->GetElement());
1171 if((*it)->IsImplicitVR())
1176 (*it)->SetValue(GetDocEntryUnvalue(*it)); // to go on compiling
1178 // Set the new entry and the new value
1179 (*it)->SetDictEntry(entry);
1180 CheckDocEntryVR(*it,vr);
1182 (*it)->SetValue(GetDocEntryValue(*it)); // to go on compiling
1187 // Remove precedent value transformation
1188 (*it)->SetDictEntry(NewVirtualDictEntry((*it)->GetGroup(),(*it)->GetElement(),vr));
1195 * \brief Searches within the Header Entries for a Dicom Element of
1197 * @param tagName name of the searched Dicom Element.
1198 * @return Corresponding Dicom Element when it exists, and NULL
1201 DocEntry* Document::GetDocEntryByName(TagName const & tagName)
1203 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
1209 return GetDocEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
1213 * \brief retrieves a Dicom Element (the first one) using (group, element)
1214 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1215 * if you think it's NOT UNIQUE, check the count number
1216 * and use iterators to retrieve ALL the Dicoms Elements within
1217 * a given couple (group, element)
1218 * @param group Group number of the searched Dicom Element
1219 * @param element Element number of the searched Dicom Element
1222 DocEntry* Document::GetDocEntryByNumber(uint16_t group, uint16_t element)
1224 TagKey key = DictEntry::TranslateToKey(group, element);
1225 if ( !TagHT.count(key))
1229 return TagHT.find(key)->second;
1233 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1234 * returns a result when the corresponding entry is of type
1236 * @return When present, the corresponding ValEntry.
1238 ValEntry* Document::GetValEntryByNumber(uint16_t group, uint16_t element)
1240 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1241 if ( !currentEntry )
1245 if ( ValEntry* valEntry = dynamic_cast<ValEntry*>(currentEntry) )
1249 dbg.Verbose(0, "Document::GetValEntryByNumber: unfound ValEntry.");
1255 * \brief Loads the element while preserving the current
1256 * underlying file position indicator as opposed to
1257 * to LoadDocEntry that modifies it.
1258 * @param entry Header Entry whose value shall be loaded.
1261 void Document::LoadDocEntrySafe(DocEntry * entry)
1263 long PositionOnEntry = Fp->tellg();
1264 LoadDocEntry(entry);
1265 Fp->seekg(PositionOnEntry, std::ios_base::beg);
1269 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1271 * @return The properly swaped 32 bits integer.
1273 uint32_t Document::SwapLong(uint32_t a)
1280 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1281 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1285 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1289 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1292 //std::cout << "swapCode= " << SwapCode << std::endl;
1293 dbg.Error(" Document::SwapLong : unset swap code");
1300 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1302 * @return The properly unswaped 32 bits integer.
1304 uint32_t Document::UnswapLong(uint32_t a)
1310 * \brief Swaps the bytes so they agree with the processor order
1311 * @return The properly swaped 16 bits integer.
1313 uint16_t Document::SwapShort(uint16_t a)
1315 if ( SwapCode == 4321 || SwapCode == 2143 )
1317 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1323 * \brief Unswaps the bytes so they agree with the processor order
1324 * @return The properly unswaped 16 bits integer.
1326 uint16_t Document::UnswapShort(uint16_t a)
1328 return SwapShort(a);
1331 //-----------------------------------------------------------------------------
1335 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1336 * @return length of the parsed set.
1338 void Document::ParseDES(DocEntrySet *set, long offset,
1339 long l_max, bool delim_mode)
1341 DocEntry *newDocEntry = 0;
1345 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1349 newDocEntry = ReadNextDocEntry( );
1355 VRKey vr = newDocEntry->GetVR();
1359 if ( Global::GetVR()->IsVROfGdcmStringRepresentable(vr) )
1361 /////////////////////// ValEntry
1362 ValEntry* newValEntry =
1363 new ValEntry( newDocEntry->GetDictEntry() ); //LEAK
1364 newValEntry->Copy( newDocEntry );
1366 // When "set" is a Document, then we are at the top of the
1367 // hierarchy and the Key is simply of the form ( group, elem )...
1368 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1371 newValEntry->SetKey( newValEntry->GetKey() );
1373 // ...but when "set" is a SQItem, we are inserting this new
1374 // valEntry in a sequence item. Hence the key has the
1375 // generalized form (refer to \ref BaseTagKey):
1376 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1378 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1379 + newValEntry->GetKey() );
1382 LoadDocEntry( newValEntry );
1383 bool delimitor=newValEntry->IsItemDelimitor();
1384 if( !set->AddEntry( newValEntry ) )
1386 // If here expect big troubles
1387 delete newValEntry; //otherwise mem leak
1394 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1401 if ( ! Global::GetVR()->IsVROfGdcmBinaryRepresentable(vr) )
1403 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1404 dbg.Verbose(0, "Document::ParseDES: neither Valentry, "
1405 "nor BinEntry. Probably unknown VR.");
1408 //////////////////// BinEntry or UNKOWN VR:
1409 /* BinEntry* newBinEntry =
1410 new BinEntry( newDocEntry->GetDictEntry() ); //LEAK
1411 newBinEntry->Copy( newDocEntry );*/
1412 BinEntry* newBinEntry = new BinEntry( newDocEntry ); //LEAK
1414 // When "this" is a Document the Key is simply of the
1415 // form ( group, elem )...
1416 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1419 newBinEntry->SetKey( newBinEntry->GetKey() );
1421 // but when "this" is a SQItem, we are inserting this new
1422 // valEntry in a sequence item, and the kay has the
1423 // generalized form (refer to \ref BaseTagKey):
1424 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1426 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1427 + newBinEntry->GetKey() );
1430 LoadDocEntry( newBinEntry );
1431 if( !set->AddEntry( newBinEntry ) )
1433 //Expect big troubles if here
1438 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1439 && ( newDocEntry->GetElement() == 0x0010 ) )
1441 TransferSyntaxType ts = GetTransferSyntax();
1442 if ( ts == RLELossless )
1444 long PositionOnEntry = Fp->tellg();
1445 Fp->seekg( newDocEntry->GetOffset(), std::ios_base::beg );
1447 Fp->seekg( PositionOnEntry, std::ios_base::beg );
1449 else if ( IsJPEG() )
1451 long PositionOnEntry = Fp->tellg();
1452 Fp->seekg( newDocEntry->GetOffset(), std::ios_base::beg );
1453 ComputeJPEGFragmentInfo();
1454 Fp->seekg( PositionOnEntry, std::ios_base::beg );
1458 // Just to make sure we are at the beginning of next entry.
1459 SkipToNextDocEntry(newDocEntry);
1464 unsigned long l = newDocEntry->GetReadLength();
1465 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1467 if ( l == 0xffffffff )
1476 // no other way to create it ...
1477 SeqEntry* newSeqEntry =
1478 new SeqEntry( newDocEntry->GetDictEntry() );
1479 newSeqEntry->Copy( newDocEntry );
1480 newSeqEntry->SetDelimitorMode( delim_mode );
1482 // At the top of the hierarchy, stands a Document. When "set"
1483 // is a Document, then we are building the first depth level.
1484 // Hence the SeqEntry we are building simply has a depth
1486 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1489 newSeqEntry->SetDepthLevel( 1 );
1490 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1492 // But when "set" is allready a SQItem, we are building a nested
1493 // sequence, and hence the depth level of the new SeqEntry
1494 // we are building, is one level deeper:
1495 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1497 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1498 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1499 + newSeqEntry->GetKey() );
1503 { // Don't try to parse zero-length sequences
1504 ParseSQ( newSeqEntry,
1505 newDocEntry->GetOffset(),
1508 set->AddEntry( newSeqEntry );
1509 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1519 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1520 * @return parsed length for this level
1522 void Document::ParseSQ( SeqEntry* seqEntry,
1523 long offset, long l_max, bool delim_mode)
1525 int SQItemNumber = 0;
1530 DocEntry* newDocEntry = ReadNextDocEntry();
1533 // FIXME Should warn user
1538 if ( newDocEntry->IsSequenceDelimitor() )
1540 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1544 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1549 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1550 std::ostringstream newBase;
1551 newBase << seqEntry->GetKey()
1555 itemSQ->SetBaseTagKey( newBase.str() );
1556 unsigned int l = newDocEntry->GetReadLength();
1558 if ( l == 0xffffffff )
1567 ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1569 seqEntry->AddEntry( itemSQ, SQItemNumber );
1571 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1579 * \brief Loads the element content if its length doesn't exceed
1580 * the value specified with Document::SetMaxSizeLoadEntry()
1581 * @param entry Header Entry (Dicom Element) to be dealt with
1583 void Document::LoadDocEntry(DocEntry* entry)
1585 uint16_t group = entry->GetGroup();
1586 std::string vr = entry->GetVR();
1587 uint32_t length = entry->GetLength();
1589 Fp->seekg((long)entry->GetOffset(), std::ios_base::beg);
1591 // A SeQuence "contains" a set of Elements.
1592 // (fffe e000) tells us an Element is beginning
1593 // (fffe e00d) tells us an Element just ended
1594 // (fffe e0dd) tells us the current SeQuence just ended
1595 if( group == 0xfffe )
1597 // NO more value field for SQ !
1601 // When the length is zero things are easy:
1604 ((ValEntry *)entry)->SetValue("");
1608 // The elements whose length is bigger than the specified upper bound
1609 // are not loaded. Instead we leave a short notice of the offset of
1610 // the element content and it's length.
1612 std::ostringstream s;
1613 if (length > MaxSizeLoadEntry)
1615 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1617 //s << "gdcm::NotLoaded (BinEntry)";
1618 s << GDCM_NOTLOADED;
1619 s << " Address:" << (long)entry->GetOffset();
1620 s << " Length:" << entry->GetLength();
1621 s << " x(" << std::hex << entry->GetLength() << ")";
1622 binEntryPtr->SetValue(s.str());
1624 // Be carefull : a BinEntry IS_A ValEntry ...
1625 else if (ValEntry* valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1627 // s << "gdcm::NotLoaded. (ValEntry)";
1628 s << GDCM_NOTLOADED;
1629 s << " Address:" << (long)entry->GetOffset();
1630 s << " Length:" << entry->GetLength();
1631 s << " x(" << std::hex << entry->GetLength() << ")";
1632 valEntryPtr->SetValue(s.str());
1637 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1638 << "nor a ValEntry ?! Should never print that !" << std::endl;
1641 // to be sure we are at the end of the value ...
1642 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1643 std::ios_base::beg);
1647 // When we find a BinEntry not very much can be done :
1648 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1650 s << GDCM_BINLOADED;
1651 binEntryPtr->SetValue(s.str());
1652 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1656 /// \todo Any compacter code suggested (?)
1657 if ( IsDocEntryAnInteger(entry) )
1661 // When short integer(s) are expected, read and convert the following
1662 // n *two characters properly i.e. consider them as short integers as
1663 // opposed to strings.
1664 // Elements with Value Multiplicity > 1
1665 // contain a set of integers (not a single one)
1666 if (vr == "US" || vr == "SS")
1669 NewInt = ReadInt16();
1673 for (int i=1; i < nbInt; i++)
1676 NewInt = ReadInt16();
1681 // See above comment on multiple integers (mutatis mutandis).
1682 else if (vr == "UL" || vr == "SL")
1685 NewInt = ReadInt32();
1689 for (int i=1; i < nbInt; i++)
1692 NewInt = ReadInt32();
1697 #ifdef GDCM_NO_ANSI_STRING_STREAM
1698 s << std::ends; // to avoid oddities on Solaris
1699 #endif //GDCM_NO_ANSI_STRING_STREAM
1701 ((ValEntry *)entry)->SetValue(s.str());
1705 // FIXME: We need an additional byte for storing \0 that is not on disk
1706 char *str = new char[length+1];
1707 Fp->read(str, (size_t)length);
1708 str[length] = '\0'; //this is only useful when length is odd
1709 // Special DicomString call to properly handle \0 and even length
1710 std::string newValue;
1713 newValue = Util::DicomString(str, length+1);
1714 //dbg.Verbose(0, "Warning: bad length: ", length );
1715 dbg.Verbose(0, "For string :", newValue.c_str());
1716 // Since we change the length of string update it length
1717 entry->SetReadLength(length+1);
1721 newValue = Util::DicomString(str, length);
1725 if ( ValEntry* valEntry = dynamic_cast<ValEntry* >(entry) )
1727 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1729 dbg.Verbose(1, "Document::LoadDocEntry",
1730 "unread element value");
1731 valEntry->SetValue(GDCM_UNREAD);
1737 // Because of correspondance with the VR dic
1738 valEntry->SetValue(newValue);
1742 valEntry->SetValue(newValue);
1747 dbg.Error(true, "Document::LoadDocEntry"
1748 "Should have a ValEntry, here !");
1754 * \brief Find the value Length of the passed Header Entry
1755 * @param entry Header Entry whose length of the value shall be loaded.
1757 void Document::FindDocEntryLength( DocEntry *entry )
1758 throw ( FormatError )
1760 uint16_t element = entry->GetElement();
1761 std::string vr = entry->GetVR();
1764 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1766 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1768 // The following reserved two bytes (see PS 3.5-2003, section
1769 // "7.1.2 Data element structure with explicit vr", p 27) must be
1770 // skipped before proceeding on reading the length on 4 bytes.
1771 Fp->seekg( 2L, std::ios_base::cur);
1772 uint32_t length32 = ReadInt32();
1774 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1779 /// \todo rename that to FindDocEntryLengthOBOrOW since
1780 /// the above test is on both OB and OW...
1781 lengthOB = FindDocEntryLengthOB();
1783 catch ( FormatUnexpected )
1785 // Computing the length failed (this happens with broken
1786 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1787 // chance to get the pixels by deciding the element goes
1788 // until the end of the file. Hence we artificially fix the
1789 // the length and proceed.
1790 long currentPosition = Fp->tellg();
1791 Fp->seekg(0L,std::ios_base::end);
1792 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1793 Fp->seekg(currentPosition, std::ios_base::beg);
1794 entry->SetLength(lengthUntilEOF);
1797 entry->SetLength(lengthOB);
1800 FixDocEntryFoundLength(entry, length32);
1804 // Length is encoded on 2 bytes.
1805 length16 = ReadInt16();
1807 // We can tell the current file is encoded in big endian (like
1808 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1809 // and it's value is the one of the encoding of a big endian file.
1810 // In order to deal with such big endian encoded files, we have
1811 // (at least) two strategies:
1812 // * when we load the "Transfer Syntax" tag with value of big endian
1813 // encoding, we raise the proper flags. Then we wait for the end
1814 // of the META group (0x0002) among which is "Transfer Syntax",
1815 // before switching the swap code to big endian. We have to postpone
1816 // the switching of the swap code since the META group is fully encoded
1817 // in little endian, and big endian coding only starts at the next
1818 // group. The corresponding code can be hard to analyse and adds
1819 // many additional unnecessary tests for regular tags.
1820 // * the second strategy consists in waiting for trouble, that shall
1821 // appear when we find the first group with big endian encoding. This
1822 // is easy to detect since the length of a "Group Length" tag (the
1823 // ones with zero as element number) has to be of 4 (0x0004). When we
1824 // encounter 1024 (0x0400) chances are the encoding changed and we
1825 // found a group with big endian encoding.
1826 // We shall use this second strategy. In order to make sure that we
1827 // can interpret the presence of an apparently big endian encoded
1828 // length of a "Group Length" without committing a big mistake, we
1829 // add an additional check: we look in the already parsed elements
1830 // for the presence of a "Transfer Syntax" whose value has to be "big
1831 // endian encoding". When this is the case, chances are we have got our
1832 // hands on a big endian encoded file: we switch the swap code to
1833 // big endian and proceed...
1834 if ( element == 0x0000 && length16 == 0x0400 )
1836 TransferSyntaxType ts = GetTransferSyntax();
1837 if ( ts != ExplicitVRBigEndian )
1839 throw FormatError( "Document::FindDocEntryLength()",
1840 " not explicit VR." );
1844 SwitchSwapToBigEndian();
1845 // Restore the unproperly loaded values i.e. the group, the element
1846 // and the dictionary entry depending on them.
1847 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1848 uint16_t correctElem = SwapShort( entry->GetElement() );
1849 DictEntry* newTag = GetDictEntryByNumber( correctGroup,
1853 // This correct tag is not in the dictionary. Create a new one.
1854 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1856 // FIXME this can create a memory leaks on the old entry that be
1857 // left unreferenced.
1858 entry->SetDictEntry( newTag );
1861 // Heuristic: well, some files are really ill-formed.
1862 if ( length16 == 0xffff)
1864 // 0xffff means that we deal with 'Unknown Length' Sequence
1867 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1872 // Either implicit VR or a non DICOM conformal (see note below) explicit
1873 // VR that ommited the VR of (at least) this element. Farts happen.
1874 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1875 // on Data elements "Implicit and Explicit VR Data Elements shall
1876 // not coexist in a Data Set and Data Sets nested within it".]
1877 // Length is on 4 bytes.
1879 FixDocEntryFoundLength( entry, ReadInt32() );
1885 * \brief Find the Value Representation of the current Dicom Element.
1888 void Document::FindDocEntryVR( DocEntry *entry )
1890 if ( Filetype != ExplicitVR )
1897 long positionOnEntry = Fp->tellg();
1898 // Warning: we believe this is explicit VR (Value Representation) because
1899 // we used a heuristic that found "UL" in the first tag. Alas this
1900 // doesn't guarantee that all the tags will be in explicit VR. In some
1901 // cases (see e-film filtered files) one finds implicit VR tags mixed
1902 // within an explicit VR file. Hence we make sure the present tag
1903 // is in explicit VR and try to fix things if it happens not to be
1906 Fp->read (vr, (size_t)2);
1909 if( !CheckDocEntryVR(entry, vr) )
1911 Fp->seekg(positionOnEntry, std::ios_base::beg);
1912 // When this element is known in the dictionary we shall use, e.g. for
1913 // the semantics (see the usage of IsAnInteger), the VR proposed by the
1914 // dictionary entry. Still we have to flag the element as implicit since
1915 // we know now our assumption on expliciteness is not furfilled.
1917 if ( entry->IsVRUnknown() )
1919 entry->SetVR("Implicit");
1921 entry->SetImplicitVR();
1926 * \brief Check the correspondance between the VR of the header entry
1927 * and the taken VR. If they are different, the header entry is
1928 * updated with the new VR.
1929 * @param entry Header Entry to check
1930 * @param vr Dicom Value Representation
1931 * @return false if the VR is incorrect of if the VR isn't referenced
1932 * otherwise, it returns true
1934 bool Document::CheckDocEntryVR(DocEntry *entry, VRKey vr)
1937 bool realExplicit = true;
1939 // Assume we are reading a falsely explicit VR file i.e. we reached
1940 // a tag where we expect reading a VR but are in fact we read the
1941 // first to bytes of the length. Then we will interogate (through find)
1942 // the dicom_vr dictionary with oddities like "\004\0" which crashes
1943 // both GCC and VC++ implementations of the STL map. Hence when the
1944 // expected VR read happens to be non-ascii characters we consider
1945 // we hit falsely explicit VR tag.
1947 if ( !isalpha((unsigned char)vr[0]) && !isalpha((unsigned char)vr[1]) )
1949 realExplicit = false;
1952 // CLEANME searching the dicom_vr at each occurence is expensive.
1953 // PostPone this test in an optional integrity check at the end
1954 // of parsing or only in debug mode.
1955 if ( realExplicit && !Global::GetVR()->Count(vr) )
1957 realExplicit = false;
1960 if ( !realExplicit )
1962 // We thought this was explicit VR, but we end up with an
1963 // implicit VR tag. Let's backtrack.
1964 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
1965 entry->GetGroup(), entry->GetElement());
1966 dbg.Verbose(1, "Document::FindVR: ", msg.c_str());
1968 if( entry->GetGroup() % 2 && entry->GetElement() == 0x0000)
1970 // Group length is UL !
1971 DictEntry* newEntry = NewVirtualDictEntry(
1972 entry->GetGroup(), entry->GetElement(),
1973 "UL", "FIXME", "Group Length");
1974 entry->SetDictEntry( newEntry );
1979 if ( entry->IsVRUnknown() )
1981 // When not a dictionary entry, we can safely overwrite the VR.
1982 if( entry->GetElement() == 0x0000 )
1984 // Group length is UL !
1992 else if ( entry->GetVR() != vr )
1994 // The VR present in the file and the dictionary disagree. We assume
1995 // the file writer knew best and use the VR of the file. Since it would
1996 // be unwise to overwrite the VR of a dictionary (since it would
1997 // compromise it's next user), we need to clone the actual DictEntry
1998 // and change the VR for the read one.
1999 DictEntry* newEntry = NewVirtualDictEntry(
2000 entry->GetGroup(), entry->GetElement(),
2001 vr, "FIXME", entry->GetName());
2002 entry->SetDictEntry(newEntry);
2009 * \brief Get the transformed value of the header entry. The VR value
2010 * is used to define the transformation to operate on the value
2011 * \warning NOT end user intended method !
2012 * @param entry entry to tranform
2013 * @return Transformed entry value
2015 std::string Document::GetDocEntryValue(DocEntry *entry)
2017 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2019 std::string val = ((ValEntry *)entry)->GetValue();
2020 std::string vr = entry->GetVR();
2021 uint32_t length = entry->GetLength();
2022 std::ostringstream s;
2025 // When short integer(s) are expected, read and convert the following
2026 // n * 2 bytes properly i.e. as a multivaluated strings
2027 // (each single value is separated fromthe next one by '\'
2028 // as usual for standard multivaluated filels
2029 // Elements with Value Multiplicity > 1
2030 // contain a set of short integers (not a single one)
2032 if( vr == "US" || vr == "SS" )
2037 for (int i=0; i < nbInt; i++)
2043 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
2044 newInt16 = SwapShort( newInt16 );
2049 // When integer(s) are expected, read and convert the following
2050 // n * 4 bytes properly i.e. as a multivaluated strings
2051 // (each single value is separated fromthe next one by '\'
2052 // as usual for standard multivaluated filels
2053 // Elements with Value Multiplicity > 1
2054 // contain a set of integers (not a single one)
2055 else if( vr == "UL" || vr == "SL" )
2060 for (int i=0; i < nbInt; i++)
2066 newInt32 = ( val[4*i+0] & 0xFF )
2067 + (( val[4*i+1] & 0xFF ) << 8 )
2068 + (( val[4*i+2] & 0xFF ) << 16 )
2069 + (( val[4*i+3] & 0xFF ) << 24 );
2070 newInt32 = SwapLong( newInt32 );
2074 #ifdef GDCM_NO_ANSI_STRING_STREAM
2075 s << std::ends; // to avoid oddities on Solaris
2076 #endif //GDCM_NO_ANSI_STRING_STREAM
2080 return ((ValEntry *)entry)->GetValue();
2084 * \brief Get the reverse transformed value of the header entry. The VR
2085 * value is used to define the reverse transformation to operate on
2087 * \warning NOT end user intended method !
2088 * @param entry Entry to reverse transform
2089 * @return Reverse transformed entry value
2091 std::string Document::GetDocEntryUnvalue(DocEntry* entry)
2093 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2095 std::string vr = entry->GetVR();
2096 std::vector<std::string> tokens;
2097 std::ostringstream s;
2099 if ( vr == "US" || vr == "SS" )
2103 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
2104 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2105 for (unsigned int i=0; i<tokens.size(); i++)
2107 newInt16 = atoi(tokens[i].c_str());
2108 s << ( newInt16 & 0xFF )
2109 << (( newInt16 >> 8 ) & 0xFF );
2113 if ( vr == "UL" || vr == "SL")
2117 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
2118 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2119 for (unsigned int i=0; i<tokens.size();i++)
2121 newInt32 = atoi(tokens[i].c_str());
2122 s << (char)( newInt32 & 0xFF )
2123 << (char)(( newInt32 >> 8 ) & 0xFF )
2124 << (char)(( newInt32 >> 16 ) & 0xFF )
2125 << (char)(( newInt32 >> 24 ) & 0xFF );
2130 #ifdef GDCM_NO_ANSI_STRING_STREAM
2131 s << std::ends; // to avoid oddities on Solaris
2132 #endif //GDCM_NO_ANSI_STRING_STREAM
2136 return ((ValEntry *)entry)->GetValue();
2140 * \brief Skip a given Header Entry
2141 * \warning NOT end user intended method !
2142 * @param entry entry to skip
2144 void Document::SkipDocEntry(DocEntry *entry)
2146 SkipBytes(entry->GetLength());
2150 * \brief Skips to the begining of the next Header Entry
2151 * \warning NOT end user intended method !
2152 * @param entry entry to skip
2154 void Document::SkipToNextDocEntry(DocEntry *entry)
2156 Fp->seekg((long)(entry->GetOffset()), std::ios_base::beg);
2157 Fp->seekg( (long)(entry->GetReadLength()), std::ios_base::cur);
2161 * \brief When the length of an element value is obviously wrong (because
2162 * the parser went Jabberwocky) one can hope improving things by
2163 * applying some heuristics.
2164 * @param entry entry to check
2165 * @param foundLength fist assumption about length
2167 void Document::FixDocEntryFoundLength(DocEntry *entry,
2168 uint32_t foundLength)
2170 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
2171 if ( foundLength == 0xffffffff)
2176 uint16_t gr = entry->GetGroup();
2177 uint16_t el = entry->GetElement();
2179 if ( foundLength % 2)
2181 std::ostringstream s;
2182 s << "Warning : Tag with uneven length "
2184 << " in x(" << std::hex << gr << "," << el <<")" << std::dec;
2185 dbg.Verbose(0, s.str().c_str());
2188 //////// Fix for some naughty General Electric images.
2189 // Allthough not recent many such GE corrupted images are still present
2190 // on Creatis hard disks. Hence this fix shall remain when such images
2191 // are no longer in user (we are talking a few years, here)...
2192 // Note: XMedCom probably uses such a trick since it is able to read
2193 // those pesky GE images ...
2194 if ( foundLength == 13)
2196 // Only happens for this length !
2197 if ( entry->GetGroup() != 0x0008
2198 || ( entry->GetElement() != 0x0070
2199 && entry->GetElement() != 0x0080 ) )
2202 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
2206 //////// Fix for some brain-dead 'Leonardo' Siemens images.
2207 // Occurence of such images is quite low (unless one leaves close to a
2208 // 'Leonardo' source. Hence, one might consider commenting out the
2209 // following fix on efficiency reasons.
2210 else if ( entry->GetGroup() == 0x0009
2211 && ( entry->GetElement() == 0x1113
2212 || entry->GetElement() == 0x1114 ) )
2215 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
2218 else if ( entry->GetVR() == "SQ" )
2220 foundLength = 0; // ReadLength is unchanged
2223 //////// We encountered a 'delimiter' element i.e. a tag of the form
2224 // "fffe|xxxx" which is just a marker. Delimiters length should not be
2225 // taken into account.
2226 else if( entry->GetGroup() == 0xfffe )
2228 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
2229 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
2230 // causes extra troubles...
2231 if( entry->GetElement() != 0x0000 )
2237 entry->SetUsableLength(foundLength);
2241 * \brief Apply some heuristics to predict whether the considered
2242 * element value contains/represents an integer or not.
2243 * @param entry The element value on which to apply the predicate.
2244 * @return The result of the heuristical predicate.
2246 bool Document::IsDocEntryAnInteger(DocEntry *entry)
2248 uint16_t element = entry->GetElement();
2249 uint16_t group = entry->GetGroup();
2250 const std::string & vr = entry->GetVR();
2251 uint32_t length = entry->GetLength();
2253 // When we have some semantics on the element we just read, and if we
2254 // a priori know we are dealing with an integer, then we shall be
2255 // able to swap it's element value properly.
2256 if ( element == 0 ) // This is the group length of the group
2264 // Allthough this should never happen, still some images have a
2265 // corrupted group length [e.g. have a glance at offset x(8336) of
2266 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2267 // Since for dicom compliant and well behaved headers, the present
2268 // test is useless (and might even look a bit paranoid), when we
2269 // encounter such an ill-formed image, we simply display a warning
2270 // message and proceed on parsing (while crossing fingers).
2271 std::ostringstream s;
2272 long filePosition = Fp->tellg();
2273 s << "Erroneous Group Length element length on : (" \
2274 << std::hex << group << " , " << element
2275 << ") -before- position x(" << filePosition << ")"
2276 << "lgt : " << length;
2277 dbg.Verbose(0, "Document::IsDocEntryAnInteger", s.str().c_str() );
2281 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2290 * \brief Find the Length till the next sequence delimiter
2291 * \warning NOT end user intended method !
2295 uint32_t Document::FindDocEntryLengthOB()
2296 throw( FormatUnexpected )
2298 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2299 long positionOnEntry = Fp->tellg();
2300 bool foundSequenceDelimiter = false;
2301 uint32_t totalLength = 0;
2303 while ( !foundSequenceDelimiter )
2309 group = ReadInt16();
2312 catch ( FormatError )
2314 throw FormatError("Document::FindDocEntryLengthOB()",
2315 " group or element not present.");
2318 // We have to decount the group and element we just read
2321 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2323 dbg.Verbose(1, "Document::FindDocEntryLengthOB: neither an Item "
2324 "tag nor a Sequence delimiter tag.");
2325 Fp->seekg(positionOnEntry, std::ios_base::beg);
2326 throw FormatUnexpected("Document::FindDocEntryLengthOB()",
2327 "Neither an Item tag nor a Sequence "
2331 if ( elem == 0xe0dd )
2333 foundSequenceDelimiter = true;
2336 uint32_t itemLength = ReadInt32();
2337 // We add 4 bytes since we just read the ItemLength with ReadInt32
2338 totalLength += itemLength + 4;
2339 SkipBytes(itemLength);
2341 if ( foundSequenceDelimiter )
2346 Fp->seekg( positionOnEntry, std::ios_base::beg);
2351 * \brief Reads a supposed to be 16 Bits integer
2352 * (swaps it depending on processor endianity)
2353 * @return read value
2355 uint16_t Document::ReadInt16()
2356 throw( FormatError )
2359 Fp->read ((char*)&g, (size_t)2);
2362 throw FormatError( "Document::ReadInt16()", " file error." );
2366 throw FormatError( "Document::ReadInt16()", "EOF." );
2373 * \brief Reads a supposed to be 32 Bits integer
2374 * (swaps it depending on processor endianity)
2375 * @return read value
2377 uint32_t Document::ReadInt32()
2378 throw( FormatError )
2381 Fp->read ((char*)&g, (size_t)4);
2384 throw FormatError( "Document::ReadInt32()", " file error." );
2388 throw FormatError( "Document::ReadInt32()", "EOF." );
2395 * \brief skips bytes inside the source file
2396 * \warning NOT end user intended method !
2399 void Document::SkipBytes(uint32_t nBytes)
2401 //FIXME don't dump the returned value
2402 Fp->seekg((long)nBytes, std::ios_base::cur);
2406 * \brief Loads all the needed Dictionaries
2407 * \warning NOT end user intended method !
2409 void Document::Initialise()
2411 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2413 RLEInfo = new RLEFramesInfo;
2414 JPEGInfo = new JPEGFragmentsInfo;
2418 * \brief Discover what the swap code is (among little endian, big endian,
2419 * bad little endian, bad big endian).
2421 * @return false when we are absolutely sure
2422 * it's neither ACR-NEMA nor DICOM
2423 * true when we hope ours assuptions are OK
2425 bool Document::CheckSwap()
2427 // The only guaranted way of finding the swap code is to find a
2428 // group tag since we know it's length has to be of four bytes i.e.
2429 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2430 // occurs when we can't find such group...
2432 uint32_t x = 4; // x : for ntohs
2433 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2437 char deb[256]; //HEADER_LENGTH_TO_READ];
2439 // First, compare HostByteOrder and NetworkByteOrder in order to
2440 // determine if we shall need to swap bytes (i.e. the Endian type).
2441 if ( x == ntohs(x) )
2450 // The easiest case is the one of a DICOM header, since it possesses a
2451 // file preamble where it suffice to look for the string "DICM".
2452 Fp->read(deb, HEADER_LENGTH_TO_READ);
2454 char *entCur = deb + 128;
2455 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2457 dbg.Verbose(1, "Document::CheckSwap:", "looks like DICOM Version3");
2459 // Next, determine the value representation (VR). Let's skip to the
2460 // first element (0002, 0000) and check there if we find "UL"
2461 // - or "OB" if the 1st one is (0002,0001) -,
2462 // in which case we (almost) know it is explicit VR.
2463 // WARNING: if it happens to be implicit VR then what we will read
2464 // is the length of the group. If this ascii representation of this
2465 // length happens to be "UL" then we shall believe it is explicit VR.
2466 // FIXME: in order to fix the above warning, we could read the next
2467 // element value (or a couple of elements values) in order to make
2468 // sure we are not commiting a big mistake.
2469 // We need to skip :
2470 // * the 128 bytes of File Preamble (often padded with zeroes),
2471 // * the 4 bytes of "DICM" string,
2472 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2473 // i.e. a total of 136 bytes.
2477 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2478 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2479 // *Implicit* VR. -and it is !-
2481 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2482 memcmp(entCur, "OB", (size_t)2) == 0 ||
2483 memcmp(entCur, "UI", (size_t)2) == 0 ||
2484 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2485 // when Write DCM *adds*
2487 // Use Document::dicom_vr to test all the possibilities
2488 // instead of just checking for UL, OB and UI !? group 0000
2490 Filetype = ExplicitVR;
2491 dbg.Verbose(1, "Document::CheckSwap:",
2492 "explicit Value Representation");
2496 Filetype = ImplicitVR;
2497 dbg.Verbose(1, "Document::CheckSwap:",
2498 "not an explicit Value Representation");
2504 dbg.Verbose(1, "Document::CheckSwap:",
2505 "HostByteOrder != NetworkByteOrder");
2510 dbg.Verbose(1, "Document::CheckSwap:",
2511 "HostByteOrder = NetworkByteOrder");
2514 // Position the file position indicator at first tag (i.e.
2515 // after the file preamble and the "DICM" string).
2516 Fp->seekg(0, std::ios_base::beg);
2517 Fp->seekg ( 132L, std::ios_base::beg);
2521 // Alas, this is not a DicomV3 file and whatever happens there is no file
2522 // preamble. We can reset the file position indicator to where the data
2523 // is (i.e. the beginning of the file).
2524 dbg.Verbose(1, "Document::CheckSwap:", "not a DICOM Version3 file");
2525 Fp->seekg(0, std::ios_base::beg);
2527 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2528 // By clean we mean that the length of the first tag is written down.
2529 // If this is the case and since the length of the first group HAS to be
2530 // four (bytes), then determining the proper swap code is straightforward.
2533 // We assume the array of char we are considering contains the binary
2534 // representation of a 32 bits integer. Hence the following dirty
2536 s32 = *((uint32_t *)(entCur));
2557 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2558 // It is time for despaired wild guesses.
2559 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2560 // i.e. the 'group length' element is not present :
2562 // check the supposed-to-be 'group number'
2563 // in ( 0x0001 .. 0x0008 )
2564 // to determine ' SwapCode' value .
2565 // Only 0 or 4321 will be possible
2566 // (no oportunity to check for the formerly well known
2567 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2568 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2569 // the file IS NOT ACR-NEMA nor DICOM V3
2570 // Find a trick to tell it the caller...
2572 s16 = *((uint16_t *)(deb));
2599 dbg.Verbose(0, "Document::CheckSwap:",
2600 "ACR/NEMA unfound swap info (Really hopeless !)");
2604 // Then the only info we have is the net2host one.
2616 * \brief Restore the unproperly loaded values i.e. the group, the element
2617 * and the dictionary entry depending on them.
2619 void Document::SwitchSwapToBigEndian()
2621 dbg.Verbose(1, "Document::SwitchSwapToBigEndian",
2622 "Switching to BigEndian mode.");
2623 if ( SwapCode == 0 )
2627 else if ( SwapCode == 4321 )
2631 else if ( SwapCode == 3412 )
2635 else if ( SwapCode == 2143 )
2642 * \brief during parsing, Header Elements too long are not loaded in memory
2645 void Document::SetMaxSizeLoadEntry(long newSize)
2651 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2653 MaxSizeLoadEntry = 0xffffffff;
2656 MaxSizeLoadEntry = newSize;
2661 * \brief Header Elements too long will not be printed
2662 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2665 void Document::SetMaxSizePrintEntry(long newSize)
2667 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2672 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2674 MaxSizePrintEntry = 0xffffffff;
2677 MaxSizePrintEntry = newSize;
2683 * \brief Handle broken private tag from Philips NTSCAN
2684 * where the endianess is being switch to BigEndian for no
2688 void Document::HandleBrokenEndian(uint16_t group, uint16_t elem)
2690 // Endian reversion. Some files contain groups of tags with reversed endianess.
2691 static int reversedEndian = 0;
2692 // try to fix endian switching in the middle of headers
2693 if ((group == 0xfeff) && (elem == 0x00e0))
2695 // start endian swap mark for group found
2697 SwitchSwapToBigEndian();
2702 else if ((group == 0xfffe) && (elem == 0xe00d) && reversedEndian)
2704 // end of reversed endian group
2706 SwitchSwapToBigEndian();
2712 * \brief Read the next tag but WITHOUT loading it's value
2713 * (read the 'Group Number', the 'Element Number',
2714 * gets the Dict Entry
2715 * gets the VR, gets the length, gets the offset value)
2716 * @return On succes the newly created DocEntry, NULL on failure.
2718 DocEntry* Document::ReadNextDocEntry()
2725 group = ReadInt16();
2728 catch ( FormatError e )
2730 // We reached the EOF (or an error occured) therefore
2731 // header parsing has to be considered as finished.
2736 HandleBrokenEndian(group, elem);
2737 DocEntry *newEntry = NewDocEntryByNumber(group, elem);
2738 FindDocEntryVR(newEntry);
2742 FindDocEntryLength(newEntry);
2744 catch ( FormatError e )
2752 newEntry->SetOffset(Fp->tellg());
2759 * \brief Generate a free TagKey i.e. a TagKey that is not present
2760 * in the TagHt dictionary.
2761 * @param group The generated tag must belong to this group.
2762 * @return The element of tag with given group which is fee.
2764 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2766 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2768 TagKey key = DictEntry::TranslateToKey(group, elem);
2769 if (TagHT.count(key) == 0)
2778 * \brief Assuming the internal file pointer \ref Document::Fp
2779 * is placed at the beginning of a tag check whether this
2780 * tag is (TestGroup, TestElement).
2781 * \warning On success the internal file pointer \ref Document::Fp
2782 * is modified to point after the tag.
2783 * On failure (i.e. when the tag wasn't the expected tag
2784 * (TestGroup, TestElement) the internal file pointer
2785 * \ref Document::Fp is restored to it's original position.
2786 * @param testGroup The expected group of the tag.
2787 * @param testElement The expected Element of the tag.
2788 * @return True on success, false otherwise.
2790 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2792 long positionOnEntry = Fp->tellg();
2793 long currentPosition = Fp->tellg(); // On debugging purposes
2795 //// Read the Item Tag group and element, and make
2796 // sure they are what we expected:
2797 uint16_t itemTagGroup;
2798 uint16_t itemTagElement;
2801 itemTagGroup = ReadInt16();
2802 itemTagElement = ReadInt16();
2804 catch ( FormatError e )
2806 //std::cerr << e << std::endl;
2809 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2811 std::ostringstream s;
2812 s << " We should have found tag (";
2813 s << std::hex << testGroup << "," << testElement << ")" << std::endl;
2814 s << " but instead we encountered tag (";
2815 s << std::hex << itemTagGroup << "," << itemTagElement << ")"
2817 s << " at address: " << (unsigned)currentPosition << std::endl;
2818 dbg.Verbose(0, "Document::ReadItemTagLength: wrong Item Tag found:");
2819 dbg.Verbose(0, s.str().c_str());
2820 Fp->seekg(positionOnEntry, std::ios_base::beg);
2828 * \brief Assuming the internal file pointer \ref Document::Fp
2829 * is placed at the beginning of a tag (TestGroup, TestElement),
2830 * read the length associated to the Tag.
2831 * \warning On success the internal file pointer \ref Document::Fp
2832 * is modified to point after the tag and it's length.
2833 * On failure (i.e. when the tag wasn't the expected tag
2834 * (TestGroup, TestElement) the internal file pointer
2835 * \ref Document::Fp is restored to it's original position.
2836 * @param testGroup The expected group of the tag.
2837 * @param testElement The expected Element of the tag.
2838 * @return On success returns the length associated to the tag. On failure
2841 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2843 long positionOnEntry = Fp->tellg();
2844 (void)positionOnEntry;
2846 if ( !ReadTag(testGroup, testElement) )
2851 //// Then read the associated Item Length
2852 long currentPosition = Fp->tellg();
2853 uint32_t itemLength = ReadInt32();
2855 std::ostringstream s;
2856 s << "Basic Item Length is: "
2857 << itemLength << std::endl;
2858 s << " at address: " << (unsigned)currentPosition << std::endl;
2859 dbg.Verbose(0, "Document::ReadItemTagLength: ", s.str().c_str());
2865 * \brief When parsing the Pixel Data of an encapsulated file, read
2866 * the basic offset table (when present, and BTW dump it).
2868 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2870 //// Read the Basic Offset Table Item Tag length...
2871 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2873 // When present, read the basic offset table itself.
2874 // Notes: - since the presence of this basic offset table is optional
2875 // we can't rely on it for the implementation, and we will simply
2876 // trash it's content (when present).
2877 // - still, when present, we could add some further checks on the
2878 // lengths, but we won't bother with such fuses for the time being.
2879 if ( itemLength != 0 )
2881 char* basicOffsetTableItemValue = new char[itemLength + 1];
2882 Fp->read(basicOffsetTableItemValue, itemLength);
2885 for (unsigned int i=0; i < itemLength; i += 4 )
2887 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2889 std::ostringstream s;
2890 s << " Read one length: ";
2891 s << std::hex << individualLength << std::endl;
2893 "Document::ReadAndSkipEncapsulatedBasicOffsetTable: ",
2898 delete[] basicOffsetTableItemValue;
2903 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2904 * Compute the RLE extra information and store it in \ref RLEInfo
2905 * for later pixel retrieval usage.
2907 void Document::ComputeRLEInfo()
2909 TransferSyntaxType ts = GetTransferSyntax();
2910 if ( ts != RLELossless )
2915 // Encoded pixel data: for the time being we are only concerned with
2916 // Jpeg or RLE Pixel data encodings.
2917 // As stated in PS 3.5-2003, section 8.2 p44:
2918 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2919 // value representation OB is used".
2920 // Hence we expect an OB value representation. Concerning OB VR,
2921 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2922 // "For the Value Representations OB and OW, the encoding shall meet the
2923 // following specifications depending on the Data element tag:"
2925 // - the first item in the sequence of items before the encoded pixel
2926 // data stream shall be basic offset table item. The basic offset table
2927 // item value, however, is not required to be present"
2929 ReadAndSkipEncapsulatedBasicOffsetTable();
2931 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2932 // Loop on the individual frame[s] and store the information
2933 // on the RLE fragments in a RLEFramesInfo.
2934 // Note: - when only a single frame is present, this is a
2936 // - when more than one frame are present, then we are in
2937 // the case of a multi-frame image.
2939 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2941 // Parse the RLE Header and store the corresponding RLE Segment
2942 // Offset Table information on fragments of this current Frame.
2943 // Note that the fragment pixels themselves are not loaded
2944 // (but just skipped).
2945 long frameOffset = Fp->tellg();
2947 uint32_t nbRleSegments = ReadInt32();
2948 if ( nbRleSegments > 16 )
2950 // There should be at most 15 segments (refer to RLEFrame class)
2951 dbg.Verbose(0, "Document::ComputeRLEInfo: too many segments.");
2954 uint32_t rleSegmentOffsetTable[16];
2955 for( int k = 1; k <= 15; k++ )
2957 rleSegmentOffsetTable[k] = ReadInt32();
2960 // Deduce from both the RLE Header and the frameLength the
2961 // fragment length, and again store this info in a
2963 long rleSegmentLength[15];
2964 // skipping (not reading) RLE Segments
2965 if ( nbRleSegments > 1)
2967 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2969 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2970 - rleSegmentOffsetTable[k];
2971 SkipBytes(rleSegmentLength[k]);
2975 rleSegmentLength[nbRleSegments] = frameLength
2976 - rleSegmentOffsetTable[nbRleSegments];
2977 SkipBytes(rleSegmentLength[nbRleSegments]);
2979 // Store the collected info
2980 RLEFrame* newFrameInfo = new RLEFrame;
2981 newFrameInfo->NumberFragments = nbRleSegments;
2982 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2984 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2985 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2987 RLEInfo->Frames.push_back( newFrameInfo );
2990 // Make sure that at the end of the item we encounter a 'Sequence
2992 if ( !ReadTag(0xfffe, 0xe0dd) )
2994 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2995 dbg.Verbose(0, " item at end of RLE item sequence");
3000 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
3001 * Compute the jpeg extra information (fragment[s] offset[s] and
3002 * length) and store it[them] in \ref JPEGInfo for later pixel
3005 void Document::ComputeJPEGFragmentInfo()
3007 // If you need to, look for comments of ComputeRLEInfo().
3013 ReadAndSkipEncapsulatedBasicOffsetTable();
3015 // Loop on the fragments[s] and store the parsed information in a
3017 long fragmentLength;
3018 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
3020 long fragmentOffset = Fp->tellg();
3022 // Store the collected info
3023 JPEGFragment* newFragment = new JPEGFragment;
3024 newFragment->Offset = fragmentOffset;
3025 newFragment->Length = fragmentLength;
3026 JPEGInfo->Fragments.push_back( newFragment );
3028 SkipBytes( fragmentLength );
3031 // Make sure that at the end of the item we encounter a 'Sequence
3033 if ( !ReadTag(0xfffe, 0xe0dd) )
3035 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
3036 dbg.Verbose(0, " item at end of JPEG item sequence");
3041 * \brief Walk recursively the given \ref DocEntrySet, and feed
3042 * the given hash table (\ref TagDocEntryHT) with all the
3043 * \ref DocEntry (Dicom entries) encountered.
3044 * This method does the job for \ref BuildFlatHashTable.
3045 * @param builtHT Where to collect all the \ref DocEntry encountered
3046 * when recursively walking the given set.
3047 * @param set The structure to be traversed (recursively).
3049 void Document::BuildFlatHashTableRecurse( TagDocEntryHT& builtHT,
3052 if (ElementSet* elementSet = dynamic_cast< ElementSet* > ( set ) )
3054 TagDocEntryHT const & currentHT = elementSet->GetTagHT();
3055 for( TagDocEntryHT::const_iterator i = currentHT.begin();
3056 i != currentHT.end();
3059 DocEntry* entry = i->second;
3060 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3062 const ListSQItem& items = seqEntry->GetSQItems();
3063 for( ListSQItem::const_iterator item = items.begin();
3064 item != items.end();
3067 BuildFlatHashTableRecurse( builtHT, *item );
3071 builtHT[entry->GetKey()] = entry;
3076 if (SQItem* SQItemSet = dynamic_cast< SQItem* > ( set ) )
3078 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
3079 for (ListDocEntry::const_iterator i = currentList.begin();
3080 i != currentList.end();
3083 DocEntry* entry = *i;
3084 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3086 const ListSQItem& items = seqEntry->GetSQItems();
3087 for( ListSQItem::const_iterator item = items.begin();
3088 item != items.end();
3091 BuildFlatHashTableRecurse( builtHT, *item );
3095 builtHT[entry->GetKey()] = entry;
3102 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
3105 * The structure used by a Document (through \ref ElementSet),
3106 * in order to hold the parsed entries of a Dicom header, is a recursive
3107 * one. This is due to the fact that the sequences (when present)
3108 * can be nested. Additionaly, the sequence items (represented in
3109 * gdcm as \ref SQItem) add an extra complexity to the data
3110 * structure. Hence, a gdcm user whishing to visit all the entries of
3111 * a Dicom header will need to dig in the gdcm internals (which
3112 * implies exposing all the internal data structures to the API).
3113 * In order to avoid this burden to the user, \ref BuildFlatHashTable
3114 * recursively builds a temporary hash table, which holds all the
3115 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
3117 * \warning Of course there is NO integrity constrain between the
3118 * returned \ref TagDocEntryHT and the \ref ElementSet used
3119 * to build it. Hence if the underlying \ref ElementSet is
3120 * altered, then it is the caller responsability to invoke
3121 * \ref BuildFlatHashTable again...
3122 * @return The flat std::map<> we juste build.
3124 TagDocEntryHT* Document::BuildFlatHashTable()
3126 TagDocEntryHT* FlatHT = new TagDocEntryHT;
3127 BuildFlatHashTableRecurse( *FlatHT, this );
3134 * \brief Compares two documents, according to \ref DicomDir rules
3135 * \warning Does NOT work with ACR-NEMA files
3136 * \todo Find a trick to solve the pb (use RET fields ?)
3138 * @return true if 'smaller'
3140 bool Document::operator<(Document &document)
3143 std::string s1 = GetEntryByNumber(0x0010,0x0010);
3144 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
3156 s1 = GetEntryByNumber(0x0010,0x0020);
3157 s2 = document.GetEntryByNumber(0x0010,0x0020);
3168 // Study Instance UID
3169 s1 = GetEntryByNumber(0x0020,0x000d);
3170 s2 = document.GetEntryByNumber(0x0020,0x000d);
3181 // Serie Instance UID
3182 s1 = GetEntryByNumber(0x0020,0x000e);
3183 s2 = document.GetEntryByNumber(0x0020,0x000e);
3198 } // end namespace gdcm
3200 //-----------------------------------------------------------------------------