1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2004/12/06 12:54:40 $
7 Version: $Revision: 1.148 $
9 Copyright (c) CREATIS (Centre de Recherche et d'Applications en Traitement de
10 l'Image). All rights reserved. See Doc/License.txt or
11 http://www.creatis.insa-lyon.fr/Public/Gdcm/License.html for details.
13 This software is distributed WITHOUT ANY WARRANTY; without even
14 the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
15 PURPOSE. See the above copyright notices for more information.
17 =========================================================================*/
19 #include "gdcmDocument.h"
20 #include "gdcmValEntry.h"
21 #include "gdcmBinEntry.h"
22 #include "gdcmSeqEntry.h"
23 #include "gdcmGlobal.h"
25 #include "gdcmDebug.h"
27 #include "gdcmException.h"
28 #include "gdcmDictSet.h"
29 #include "gdcmRLEFramesInfo.h"
30 #include "gdcmJPEGFragmentsInfo.h"
31 #include "gdcmDocEntrySet.h"
32 #include "gdcmSQItem.h"
38 #if defined(_MSC_VER) || defined(__BORLANDC__)
41 #include <netinet/in.h>
46 //-----------------------------------------------------------------------------
47 static const char *TransferSyntaxStrings[] = {
48 // Implicit VR Little Endian
50 // Implicit VR Little Endian DLX G.E?
52 // Explicit VR Little Endian
53 "1.2.840.10008.1.2.1",
54 // Deflated Explicit VR Little Endian
55 "1.2.840.10008.1.2.1.99",
56 // Explicit VR Big Endian
57 "1.2.840.10008.1.2.2",
58 // JPEG Baseline (Process 1)
59 "1.2.840.10008.1.2.4.50",
60 // JPEG Extended (Process 2 & 4)
61 "1.2.840.10008.1.2.4.51",
62 // JPEG Extended (Process 3 & 5)
63 "1.2.840.10008.1.2.4.52",
64 // JPEG Spectral Selection, Non-Hierarchical (Process 6 & 8)
65 "1.2.840.10008.1.2.4.53",
66 // JPEG Full Progression, Non-Hierarchical (Process 10 & 12)
67 "1.2.840.10008.1.2.4.55",
68 // JPEG Lossless, Non-Hierarchical (Process 14)
69 "1.2.840.10008.1.2.4.57",
70 // JPEG Lossless, Hierarchical, First-Order Prediction (Process 14, [Selection Value 1])
71 "1.2.840.10008.1.2.4.70",
73 "1.2.840.10008.1.2.4.90",
75 "1.2.840.10008.1.2.4.91",
77 "1.2.840.10008.1.2.5",
79 "Unknown Transfer Syntax"
82 //-----------------------------------------------------------------------------
83 // Refer to Document::CheckSwap()
84 //const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
86 // Refer to Document::SetMaxSizeLoadEntry()
87 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
88 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
90 //-----------------------------------------------------------------------------
91 // Constructor / Destructor
95 * @param filename file to be opened for parsing
97 Document::Document( std::string const & filename ) : ElementSet(-1)
99 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
109 dbg.Verbose(0, "Document::Document: starting parsing of file: ",
111 Fp->seekg( 0, std::ios::beg);
113 Fp->seekg(0, std::ios::end);
114 long lgt = Fp->tellg();
116 Fp->seekg( 0, std::ios::beg);
118 long beg = Fp->tellg();
121 ParseDES( this, beg, lgt, false); // le Load sera fait a la volee
123 Fp->seekg( 0, std::ios::beg);
125 // Load 'non string' values
127 std::string PhotometricInterpretation = GetEntryByNumber(0x0028,0x0004);
128 if( PhotometricInterpretation == "PALETTE COLOR " )
130 LoadEntryBinArea(0x0028,0x1200); // gray LUT
131 /// FIXME FIXME FIXME
132 /// The tags refered by the three following lines used to be CORRECTLY
133 /// defined as having an US Value Representation in the public
134 /// dictionnary. BUT the semantics implied by the three following
135 /// lines state that the corresponding tag contents are in fact
136 /// the ones of a BinEntry.
137 /// In order to fix things "Quick and Dirty" the dictionnary was
138 /// altered on PURPOUS but now contains a WRONG value.
139 /// In order to fix things and restore the dictionary to its
140 /// correct value, one needs to decided of the semantics by deciding
141 /// wether the following tags are either:
142 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
143 /// also used as BinEntry, which requires the proper conversion,
144 /// - OW, and hence loaded as BinEntry, but afterwards also used
145 /// as ValEntry, which requires the proper conversion.
146 LoadEntryBinArea(0x0028,0x1201); // R LUT
147 LoadEntryBinArea(0x0028,0x1202); // G LUT
148 LoadEntryBinArea(0x0028,0x1203); // B LUT
150 // Segmented Red Palette Color LUT Data
151 LoadEntryBinArea(0x0028,0x1221);
152 // Segmented Green Palette Color LUT Data
153 LoadEntryBinArea(0x0028,0x1222);
154 // Segmented Blue Palette Color LUT Data
155 LoadEntryBinArea(0x0028,0x1223);
157 //FIXME later : how to use it?
158 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
162 // --------------------------------------------------------------
163 // Specific code to allow gdcm to read ACR-LibIDO formated images
164 // Note: ACR-LibIDO is an extension of the ACR standard that was
165 // used at CREATIS. For the time being (say a couple years)
166 // we keep this kludge to allow a smooth move to gdcm for
167 // CREATIS developpers (sorry folks).
169 // if recognition code tells us we deal with a LibIDO image
170 // we switch lineNumber and columnNumber
173 RecCode = GetEntryByNumber(0x0008, 0x0010); // recognition code
174 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
175 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
176 // with "little-endian strings"
178 Filetype = ACR_LIBIDO;
179 std::string rows = GetEntryByNumber(0x0028, 0x0010);
180 std::string columns = GetEntryByNumber(0x0028, 0x0011);
181 SetEntryByNumber(columns, 0x0028, 0x0010);
182 SetEntryByNumber(rows , 0x0028, 0x0011);
184 // ----------------- End of ACR-LibIDO kludge ------------------
186 PrintLevel = 1; // 'Medium' print level by default
190 * \brief This default constructor doesn't parse the file. You should
191 * then invoke \ref Document::SetFileName and then the parsing.
193 Document::Document() : ElementSet(-1)
197 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
200 Filetype = ExplicitVR;
201 PrintLevel = 1; // 'Medium' print level by default
205 * \brief Canonical destructor.
207 Document::~Document ()
216 //-----------------------------------------------------------------------------
220 * \brief Prints The Dict Entries of THE public Dicom Dictionary
223 void Document::PrintPubDict(std::ostream & os)
225 RefPubDict->Print(os);
229 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
232 void Document::PrintShaDict(std::ostream & os)
234 RefShaDict->Print(os);
237 //-----------------------------------------------------------------------------
240 * \brief Get the public dictionary used
242 Dict* Document::GetPubDict()
248 * \brief Get the shadow dictionary used
250 Dict* Document::GetShaDict()
256 * \brief Set the shadow dictionary used
257 * \param dict dictionary to use in shadow
259 bool Document::SetShaDict(Dict *dict)
266 * \brief Set the shadow dictionary used
267 * \param dictName name of the dictionary to use in shadow
269 bool Document::SetShaDict(DictKey const & dictName)
271 RefShaDict = Global::GetDicts()->GetDict(dictName);
276 * \brief This predicate, based on hopefully reasonable heuristics,
277 * decides whether or not the current Document was properly parsed
278 * and contains the mandatory information for being considered as
279 * a well formed and usable Dicom/Acr File.
280 * @return true when Document is the one of a reasonable Dicom/Acr file,
283 bool Document::IsReadable()
285 if( Filetype == Unknown)
287 dbg.Verbose(0, "Document::IsReadable: wrong filetype");
293 dbg.Verbose(0, "Document::IsReadable: no tags in internal"
302 * \brief Accessor to the Transfer Syntax (when present) of the
303 * current document (it internally handles reading the
304 * value from disk when only parsing occured).
305 * @return The encountered Transfer Syntax of the current document.
307 TransferSyntaxType Document::GetTransferSyntax()
309 DocEntry *entry = GetDocEntryByNumber(0x0002, 0x0010);
315 // The entry might be present but not loaded (parsing and loading
316 // happen at different stages): try loading and proceed with check...
317 LoadDocEntrySafe(entry);
318 if (ValEntry* valEntry = dynamic_cast< ValEntry* >(entry) )
320 std::string transfer = valEntry->GetValue();
321 // The actual transfer (as read from disk) might be padded. We
322 // first need to remove the potential padding. We can make the
323 // weak assumption that padding was not executed with digits...
324 if ( transfer.length() == 0 )
326 // for brain damaged headers
329 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
331 transfer.erase(transfer.length()-1, 1);
333 for (int i = 0; TransferSyntaxStrings[i] != NULL; i++)
335 if ( TransferSyntaxStrings[i] == transfer )
337 return TransferSyntaxType(i);
344 bool Document::IsJPEGLossless()
346 TransferSyntaxType r = GetTransferSyntax();
347 return r == JPEGFullProgressionProcess10_12
348 || r == JPEGLosslessProcess14
349 || r == JPEGLosslessProcess14_1;
353 * \brief Determines if the Transfer Syntax was already encountered
354 * and if it corresponds to a JPEG2000 one
355 * @return True when JPEG2000 (Lossly or LossLess) found. False in all
358 bool Document::IsJPEG2000()
360 TransferSyntaxType r = GetTransferSyntax();
361 return r == JPEG2000Lossless || r == JPEG2000;
365 * \brief Determines if the Transfer Syntax corresponds to any form
366 * of Jpeg encoded Pixel data.
367 * @return True when any form of JPEG found. False otherwise.
369 bool Document::IsJPEG()
371 TransferSyntaxType r = GetTransferSyntax();
372 return r == JPEGBaselineProcess1
373 || r == JPEGExtendedProcess2_4
374 || r == JPEGExtendedProcess3_5
375 || r == JPEGSpectralSelectionProcess6_8
381 * \brief Determines if the Transfer Syntax corresponds to encapsulated
382 * of encoded Pixel Data (as opposed to native).
383 * @return True when encapsulated. False when native.
385 bool Document::IsEncapsulate()
387 TransferSyntaxType r = GetTransferSyntax();
388 return IsJPEG() || r == RLELossless;
392 * \brief Predicate for dicom version 3 file.
393 * @return True when the file is a dicom version 3.
395 bool Document::IsDicomV3()
397 // Checking if Transfert Syntax exists is enough
398 // Anyway, it's to late check if the 'Preamble' was found ...
399 // And ... would it be a rich idea to check ?
400 // (some 'no Preamble' DICOM images exist !)
401 return GetDocEntryByNumber(0x0002, 0x0010) != NULL;
405 * \brief returns the File Type
406 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
407 * @return the FileType code
409 FileType Document::GetFileType()
415 * \brief Tries to open the file \ref Document::Filename and
416 * checks the preamble when existing.
417 * @return The FILE pointer on success.
419 std::ifstream* Document::OpenFile()
421 if (Filename.length() == 0)
429 "Document::OpenFile is already opened when opening: ",
433 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
437 "Document::OpenFile cannot open file: ",
445 Fp->read((char*)&zero, (size_t)2);
452 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
454 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
455 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
456 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
457 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
463 Fp->seekg(126L, std::ios::cur);
465 Fp->read(dicm, (size_t)4);
471 if( memcmp(dicm, "DICM", 4) == 0 )
478 "Document::OpenFile not DICOM/ACR (missing preamble)",
485 * \brief closes the file
486 * @return TRUE if the close was successfull
488 bool Document::CloseFile()
497 return true; //FIXME how do we detect a non-close ifstream ?
501 * \brief Writes in a file all the Header Entries (Dicom Elements)
502 * @param fp file pointer on an already open file
503 * @param filetype Type of the File to be written
504 * (ACR-NEMA, ExplicitVR, ImplicitVR)
505 * \return Always true.
507 void Document::WriteContent(std::ofstream* fp, FileType filetype)
509 /// \todo move the following lines (and a lot of others, to be written)
510 /// to a future function CheckAndCorrectHeader
511 /// (necessary if user wants to write a DICOM V3 file
512 /// starting from an ACR-NEMA (V2) Header
514 if ( filetype == ImplicitVR || filetype == ExplicitVR )
516 // writing Dicom File Preamble
517 char filePreamble[128];
518 memset(filePreamble, 0, 128);
519 fp->write(filePreamble, 128);
520 fp->write("DICM", 4);
524 * \todo rewrite later, if really usefull
525 * - 'Group Length' element is optional in DICOM
526 * - but un-updated odd groups lengthes can causes pb
529 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
530 * UpdateGroupLength(false,filetype);
531 * if ( filetype == ACR)
532 * UpdateGroupLength(true,ACR);
535 ElementSet::WriteContent(fp, filetype); // This one is recursive
539 * \brief Modifies the value of a given Header Entry (Dicom Element)
540 * when it exists. Create it with the given value when unexistant.
541 * @param value (string) Value to be set
542 * @param group Group number of the Entry
543 * @param elem Element number of the Entry
544 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
545 * \return pointer to the modified/created Header Entry (NULL when creation
548 ValEntry* Document::ReplaceOrCreateByNumber(
549 std::string const & value,
554 ValEntry* valEntry = 0;
555 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
559 // check if (group,element) DictEntry exists
560 // if it doesn't, create an entry in DictSet::VirtualEntry
563 // Find out if the tag we received is in the dictionaries:
564 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
565 DictEntry* dictEntry = pubDict->GetDictEntryByNumber(group, elem);
568 currentEntry = NewDocEntryByNumber(group, elem, vr);
572 currentEntry = NewDocEntryByNumber(group, elem);
577 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
578 " NewDocEntryByNumber failed.");
582 valEntry = new ValEntry(currentEntry);
585 if ( !AddEntry(valEntry))
588 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
589 " failed allthough this is a creation.");
595 valEntry = dynamic_cast< ValEntry* >(currentEntry);
596 if ( !valEntry ) // Euuuuh? It wasn't a ValEntry
597 // then we change it to a ValEntry ?
598 // Shouldn't it be considered as an error ?
600 // We need to promote the DocEntry to a ValEntry:
601 valEntry = new ValEntry(currentEntry);
602 if (!RemoveEntry(currentEntry))
605 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
606 " of previous DocEntry failed.");
609 if ( !AddEntry(valEntry))
612 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
613 " promoted ValEntry failed.");
619 SetEntryByNumber(value, group, elem);
625 * \brief Modifies the value of a given Header Entry (Dicom Element)
626 * when it exists. Create it with the given value when unexistant.
627 * A copy of the binArea is made to be kept in the Document.
628 * @param binArea (binary) value to be set
629 * @param Group Group number of the Entry
630 * @param Elem Element number of the Entry
631 * \return pointer to the modified/created Header Entry (NULL when creation
634 BinEntry* Document::ReplaceOrCreateByNumber(
641 BinEntry* binEntry = 0;
642 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
646 // check if (group,element) DictEntry exists
647 // if it doesn't, create an entry in DictSet::VirtualEntry
650 // Find out if the tag we received is in the dictionaries:
651 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
652 DictEntry *dictEntry = pubDict->GetDictEntryByNumber(group, elem);
656 currentEntry = NewDocEntryByNumber(group, elem, vr);
660 currentEntry = NewDocEntryByNumber(group, elem);
664 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
665 " NewDocEntryByNumber failed.");
668 binEntry = new BinEntry(currentEntry);
669 if ( !AddEntry(binEntry))
671 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
672 " failed allthough this is a creation.");
678 binEntry = dynamic_cast< BinEntry* >(currentEntry);
679 if ( !binEntry ) // Euuuuh? It wasn't a BinEntry
680 // then we change it to a BinEntry ?
681 // Shouldn't it be considered as an error ?
683 // We need to promote the DocEntry to a BinEntry:
684 binEntry = new BinEntry(currentEntry);
685 if (!RemoveEntry(currentEntry))
687 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
688 " of previous DocEntry failed.");
691 if ( !AddEntry(binEntry))
693 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
694 " promoted BinEntry failed.");
701 if (lgth>0 && binArea)
703 tmpArea = new uint8_t[lgth];
704 memcpy(tmpArea,binArea,lgth);
710 if (!SetEntryByNumber(tmpArea, lgth, group, elem))
723 * \brief Modifies the value of a given Header Entry (Dicom Element)
724 * when it exists. Create it when unexistant.
725 * @param Group Group number of the Entry
726 * @param Elem Element number of the Entry
727 * \return pointer to the modified/created SeqEntry (NULL when creation
730 SeqEntry* Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem)
733 DocEntry* a = GetDocEntryByNumber( group, elem);
736 a = NewSeqEntryByNumber(group, elem);
742 b = new SeqEntry(a, 1); // FIXME : 1 (Depth)
749 * \brief Set a new value if the invoked element exists
750 * Seems to be useless !!!
751 * @param value new element value
752 * @param group group number of the Entry
753 * @param elem element number of the Entry
756 bool Document::ReplaceIfExistByNumber(std::string const & value,
757 uint16_t group, uint16_t elem )
759 SetEntryByNumber(value, group, elem);
764 std::string Document::GetTransferSyntaxValue(TransferSyntaxType type)
766 return TransferSyntaxStrings[type];
769 //-----------------------------------------------------------------------------
773 * \brief Checks if a given Dicom Element exists within the H table
774 * @param group Group number of the searched Dicom Element
775 * @param element Element number of the searched Dicom Element
776 * @return true is found
778 bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
780 const std::string &key = DictEntry::TranslateToKey(group, element );
781 return TagHT.count(key) != 0;
785 * \brief Searches within Header Entries (Dicom Elements) parsed with
786 * the public and private dictionaries
787 * for the element value of a given tag.
788 * \warning Don't use any longer : use GetPubEntryByName
789 * @param tagName name of the searched element.
790 * @return Corresponding element value when it exists,
791 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
793 std::string Document::GetEntryByName(TagName const & tagName)
795 DictEntry* dictEntry = RefPubDict->GetDictEntryByName(tagName);
801 return GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
805 * \brief Searches within Header Entries (Dicom Elements) parsed with
806 * the public and private dictionaries
807 * for the element value representation of a given tag.
809 * Obtaining the VR (Value Representation) might be needed by caller
810 * to convert the string typed content to caller's native type
811 * (think of C++ vs Python). The VR is actually of a higher level
812 * of semantics than just the native C++ type.
813 * @param tagName name of the searched element.
814 * @return Corresponding element value representation when it exists,
815 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
817 std::string Document::GetEntryVRByName(TagName const& tagName)
819 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
820 if( dictEntry == NULL)
825 DocEntry* elem = GetDocEntryByNumber(dictEntry->GetGroup(),
826 dictEntry->GetElement());
827 return elem->GetVR();
831 * \brief Searches within Header Entries (Dicom Elements) parsed with
832 * the public and private dictionaries
833 * for the element value representation of a given tag.
834 * @param group Group number of the searched tag.
835 * @param element Element number of the searched tag.
836 * @return Corresponding element value representation when it exists,
837 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
839 std::string Document::GetEntryByNumber(uint16_t group, uint16_t element)
841 TagKey key = DictEntry::TranslateToKey(group, element);
842 /// \todo use map methods, instead of multimap JPR
843 if ( !TagHT.count(key))
848 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
852 * \brief Searches within Header Entries (Dicom Elements) parsed with
853 * the public and private dictionaries
854 * for the element value representation of a given tag..
856 * Obtaining the VR (Value Representation) might be needed by caller
857 * to convert the string typed content to caller's native type
858 * (think of C++ vs Python). The VR is actually of a higher level
859 * of semantics than just the native C++ type.
860 * @param group Group number of the searched tag.
861 * @param element Element number of the searched tag.
862 * @return Corresponding element value representation when it exists,
863 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
865 std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element)
867 DocEntry* elem = GetDocEntryByNumber(group, element);
872 return elem->GetVR();
876 * \brief Searches within Header Entries (Dicom Elements) parsed with
877 * the public and private dictionaries
878 * for the value length of a given tag..
879 * @param group Group number of the searched tag.
880 * @param element Element number of the searched tag.
881 * @return Corresponding element length; -2 if not found
883 int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element)
885 DocEntry* elem = GetDocEntryByNumber(group, element);
888 return -2; //magic number
890 return elem->GetLength();
893 * \brief Sets the value (string) of the Header Entry (Dicom Element)
894 * @param content string value of the Dicom Element
895 * @param tagName name of the searched Dicom Element.
896 * @return true when found
898 bool Document::SetEntryByName( std::string const & content,
899 TagName const & tagName)
901 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
907 return SetEntryByNumber(content,dictEntry->GetGroup(),
908 dictEntry->GetElement());
912 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
913 * through it's (group, element) and modifies it's content with
915 * @param content new value (string) to substitute with
916 * @param group group number of the Dicom Element to modify
917 * @param element element number of the Dicom Element to modify
919 bool Document::SetEntryByNumber(std::string const& content,
920 uint16_t group, uint16_t element)
925 ValEntry* valEntry = GetValEntryByNumber(group, element);
928 dbg.Verbose(0, "Document::SetEntryByNumber: no corresponding",
929 " ValEntry (try promotion first).");
932 // Non even content must be padded with a space (020H)...
933 std::string finalContent = Util::DicomString( content.c_str() );
934 assert( !(finalContent.size() % 2) );
935 valEntry->SetValue(finalContent);
937 // Integers have a special treatement for their length:
938 l = finalContent.length();
939 if ( l != 0) // To avoid to be cheated by 'zero length' integers
941 VRKey vr = valEntry->GetVR();
942 if( vr == "US" || vr == "SS" )
944 // for multivaluated items
945 c = Util::CountSubstring(content, "\\") + 1;
948 else if( vr == "UL" || vr == "SL" )
950 // for multivaluated items
951 c = Util::CountSubstring(content, "\\") + 1;
955 valEntry->SetLength(l);
960 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
961 * through it's (group, element) and modifies it's content with
963 * @param content new value (void* -> uint8_t*) to substitute with
964 * @param lgth new value length
965 * @param group group number of the Dicom Element to modify
966 * @param element element number of the Dicom Element to modify
968 bool Document::SetEntryByNumber(uint8_t*content, int lgth,
969 uint16_t group, uint16_t element)
971 (void)lgth; //not used
972 TagKey key = DictEntry::TranslateToKey(group, element);
973 if ( !TagHT.count(key))
978 /* Hope Binary field length is *never* wrong
979 if(lgth%2) // Non even length are padded with a space (020H).
982 //content = content + '\0'; // fing a trick to enlarge a binary field?
985 BinEntry* entry = (BinEntry *)TagHT[key];
986 entry->SetBinArea(content);
987 entry->SetLength(lgth);
988 entry->SetValue(GDCM_BINLOADED);
994 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
995 * in the PubDocEntrySet of this instance
996 * through it's (group, element) and modifies it's length with
998 * \warning Use with extreme caution.
999 * @param l new length to substitute with
1000 * @param group group number of the Entry to modify
1001 * @param element element number of the Entry to modify
1002 * @return true on success, false otherwise.
1004 /*bool Document::SetEntryLengthByNumber(uint32_t l,
1005 uint16_t group, uint16_t element)
1007 /// \todo use map methods, instead of multimap JPR
1008 TagKey key = DictEntry::TranslateToKey(group, element);
1009 if ( !TagHT.count(key) )
1015 l++; // length must be even
1017 ( ((TagHT.equal_range(key)).first)->second )->SetLength(l);
1023 * \brief Gets (from Header) the offset of a 'non string' element value
1024 * (LoadElementValues has already be executed)
1025 * @param group group number of the Entry
1026 * @param elem element number of the Entry
1027 * @return File Offset of the Element Value
1029 /*size_t Document::GetEntryOffsetByNumber(uint16_t group, uint16_t elem)
1031 DocEntry* entry = GetDocEntryByNumber(group, elem);
1034 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry present.");
1037 return entry->GetOffset();
1041 * \brief Gets (from Header) a 'non string' element value
1042 * (LoadElementValues has already be executed)
1043 * @param group group number of the Entry
1044 * @param elem element number of the Entry
1045 * @return Pointer to the 'non string' area
1047 void* Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem)
1049 DocEntry* entry = GetDocEntryByNumber(group, elem);
1052 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry");
1055 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(entry) )
1057 return binEntry->GetBinArea();
1064 * \brief Loads (from disk) the element content
1065 * when a string is not suitable
1066 * @param group group number of the Entry
1067 * @param elem element number of the Entry
1069 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
1071 // Search the corresponding DocEntry
1072 DocEntry *docElement = GetDocEntryByNumber(group, elem);
1076 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
1080 LoadEntryBinArea(binElement);
1084 * \brief Loads (from disk) the element content
1085 * when a string is not suitable
1086 * @param element Entry whose binArea is going to be loaded
1088 void Document::LoadEntryBinArea(BinEntry* element)
1090 if(element->GetBinArea())
1093 bool openFile = !Fp;
1097 size_t o =(size_t)element->GetOffset();
1098 Fp->seekg(o, std::ios::beg);
1100 size_t l = element->GetLength();
1101 uint8_t* a = new uint8_t[l];
1104 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1108 /// \todo check the result
1109 Fp->read((char*)a, l);
1110 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
1116 element->SetBinArea(a);
1123 * \brief Sets a 'non string' value to a given Dicom Element
1124 * @param area area containing the 'non string' value
1125 * @param group Group number of the searched Dicom Element
1126 * @param element Element number of the searched Dicom Element
1129 /*bool Document::SetEntryBinAreaByNumber(uint8_t* area,
1130 uint16_t group, uint16_t element)
1132 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1133 if ( !currentEntry )
1138 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1140 binEntry->SetBinArea( area );
1148 * \brief Update the entries with the shadow dictionary.
1149 * Only non even entries are analyzed
1151 void Document::UpdateShaEntries()
1156 /// \todo TODO : still any use to explore recursively the whole structure?
1158 for(ListTag::iterator it=listEntries.begin();
1159 it!=listEntries.end();
1162 // Odd group => from public dictionary
1163 if((*it)->GetGroup()%2==0)
1166 // Peer group => search the corresponding dict entry
1168 entry=RefShaDict->GetDictEntryByNumber((*it)->GetGroup(),(*it)->GetElement());
1172 if((*it)->IsImplicitVR())
1177 (*it)->SetValue(GetDocEntryUnvalue(*it)); // to go on compiling
1179 // Set the new entry and the new value
1180 (*it)->SetDictEntry(entry);
1181 CheckDocEntryVR(*it,vr);
1183 (*it)->SetValue(GetDocEntryValue(*it)); // to go on compiling
1188 // Remove precedent value transformation
1189 (*it)->SetDictEntry(NewVirtualDictEntry((*it)->GetGroup(),(*it)->GetElement(),vr));
1196 * \brief Searches within the Header Entries for a Dicom Element of
1198 * @param tagName name of the searched Dicom Element.
1199 * @return Corresponding Dicom Element when it exists, and NULL
1202 DocEntry* Document::GetDocEntryByName(TagName const & tagName)
1204 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
1210 return GetDocEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
1214 * \brief retrieves a Dicom Element (the first one) using (group, element)
1215 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1216 * if you think it's NOT UNIQUE, check the count number
1217 * and use iterators to retrieve ALL the Dicoms Elements within
1218 * a given couple (group, element)
1219 * @param group Group number of the searched Dicom Element
1220 * @param element Element number of the searched Dicom Element
1223 DocEntry* Document::GetDocEntryByNumber(uint16_t group, uint16_t element)
1225 TagKey key = DictEntry::TranslateToKey(group, element);
1226 if ( !TagHT.count(key))
1230 return TagHT.find(key)->second;
1234 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1235 * returns a result when the corresponding entry is of type
1237 * @return When present, the corresponding ValEntry.
1239 ValEntry* Document::GetValEntryByNumber(uint16_t group, uint16_t element)
1241 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1242 if ( !currentEntry )
1246 if ( ValEntry* valEntry = dynamic_cast<ValEntry*>(currentEntry) )
1250 dbg.Verbose(0, "Document::GetValEntryByNumber: unfound ValEntry.");
1256 * \brief Loads the element while preserving the current
1257 * underlying file position indicator as opposed to
1258 * to LoadDocEntry that modifies it.
1259 * @param entry Header Entry whose value shall be loaded.
1262 void Document::LoadDocEntrySafe(DocEntry * entry)
1266 long PositionOnEntry = Fp->tellg();
1267 LoadDocEntry(entry);
1268 Fp->seekg(PositionOnEntry, std::ios::beg);
1273 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1275 * @return The properly swaped 32 bits integer.
1277 uint32_t Document::SwapLong(uint32_t a)
1284 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1285 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1289 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1293 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1296 //std::cout << "swapCode= " << SwapCode << std::endl;
1297 dbg.Error(" Document::SwapLong : unset swap code");
1304 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1306 * @return The properly unswaped 32 bits integer.
1308 uint32_t Document::UnswapLong(uint32_t a)
1314 * \brief Swaps the bytes so they agree with the processor order
1315 * @return The properly swaped 16 bits integer.
1317 uint16_t Document::SwapShort(uint16_t a)
1319 if ( SwapCode == 4321 || SwapCode == 2143 )
1321 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1327 * \brief Unswaps the bytes so they agree with the processor order
1328 * @return The properly unswaped 16 bits integer.
1330 uint16_t Document::UnswapShort(uint16_t a)
1332 return SwapShort(a);
1335 //-----------------------------------------------------------------------------
1339 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1340 * @return length of the parsed set.
1342 void Document::ParseDES(DocEntrySet *set, long offset,
1343 long l_max, bool delim_mode)
1345 DocEntry *newDocEntry = 0;
1349 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1353 newDocEntry = ReadNextDocEntry( );
1359 VRKey vr = newDocEntry->GetVR();
1363 if ( Global::GetVR()->IsVROfGdcmStringRepresentable(vr) )
1365 /////////////////////// ValEntry
1366 ValEntry* newValEntry =
1367 new ValEntry( newDocEntry->GetDictEntry() ); //LEAK
1368 newValEntry->Copy( newDocEntry );
1370 // When "set" is a Document, then we are at the top of the
1371 // hierarchy and the Key is simply of the form ( group, elem )...
1372 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1375 newValEntry->SetKey( newValEntry->GetKey() );
1377 // ...but when "set" is a SQItem, we are inserting this new
1378 // valEntry in a sequence item. Hence the key has the
1379 // generalized form (refer to \ref BaseTagKey):
1380 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1382 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1383 + newValEntry->GetKey() );
1386 LoadDocEntry( newValEntry );
1387 bool delimitor=newValEntry->IsItemDelimitor();
1388 if( !set->AddEntry( newValEntry ) )
1390 // If here expect big troubles
1391 delete newValEntry; //otherwise mem leak
1399 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1407 if ( ! Global::GetVR()->IsVROfGdcmBinaryRepresentable(vr) )
1409 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1410 dbg.Verbose(0, "Document::ParseDES: neither Valentry, "
1411 "nor BinEntry. Probably unknown VR.");
1414 //////////////////// BinEntry or UNKOWN VR:
1415 BinEntry* newBinEntry = new BinEntry( newDocEntry ); //LEAK
1417 // When "this" is a Document the Key is simply of the
1418 // form ( group, elem )...
1419 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1422 newBinEntry->SetKey( newBinEntry->GetKey() );
1424 // but when "this" is a SQItem, we are inserting this new
1425 // valEntry in a sequence item, and the kay has the
1426 // generalized form (refer to \ref BaseTagKey):
1427 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1429 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1430 + newBinEntry->GetKey() );
1433 LoadDocEntry( newBinEntry );
1434 if( !set->AddEntry( newBinEntry ) )
1436 //Expect big troubles if here
1441 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1442 && ( newDocEntry->GetElement() == 0x0010 ) )
1444 TransferSyntaxType ts = GetTransferSyntax();
1445 if ( ts == RLELossless )
1447 long PositionOnEntry = Fp->tellg();
1448 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1450 Fp->seekg( PositionOnEntry, std::ios::beg );
1452 else if ( IsJPEG() )
1454 long PositionOnEntry = Fp->tellg();
1455 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1456 ComputeJPEGFragmentInfo();
1457 Fp->seekg( PositionOnEntry, std::ios::beg );
1461 // Just to make sure we are at the beginning of next entry.
1462 SkipToNextDocEntry(newDocEntry);
1463 //delete newDocEntry;
1468 unsigned long l = newDocEntry->GetReadLength();
1469 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1471 if ( l == 0xffffffff )
1480 // no other way to create it ...
1481 SeqEntry* newSeqEntry =
1482 new SeqEntry( newDocEntry->GetDictEntry() );
1483 newSeqEntry->Copy( newDocEntry );
1484 newSeqEntry->SetDelimitorMode( delim_mode );
1486 // At the top of the hierarchy, stands a Document. When "set"
1487 // is a Document, then we are building the first depth level.
1488 // Hence the SeqEntry we are building simply has a depth
1490 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1493 newSeqEntry->SetDepthLevel( 1 );
1494 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1496 // But when "set" is allready a SQItem, we are building a nested
1497 // sequence, and hence the depth level of the new SeqEntry
1498 // we are building, is one level deeper:
1499 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1501 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1502 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1503 + newSeqEntry->GetKey() );
1507 { // Don't try to parse zero-length sequences
1508 ParseSQ( newSeqEntry,
1509 newDocEntry->GetOffset(),
1512 set->AddEntry( newSeqEntry );
1513 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1524 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1525 * @return parsed length for this level
1527 void Document::ParseSQ( SeqEntry* seqEntry,
1528 long offset, long l_max, bool delim_mode)
1530 int SQItemNumber = 0;
1535 DocEntry* newDocEntry = ReadNextDocEntry();
1538 // FIXME Should warn user
1543 if ( newDocEntry->IsSequenceDelimitor() )
1545 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1549 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1555 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1556 std::ostringstream newBase;
1557 newBase << seqEntry->GetKey()
1561 itemSQ->SetBaseTagKey( newBase.str() );
1562 unsigned int l = newDocEntry->GetReadLength();
1564 if ( l == 0xffffffff )
1573 ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1576 seqEntry->AddEntry( itemSQ, SQItemNumber );
1578 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1586 * \brief Loads the element content if its length doesn't exceed
1587 * the value specified with Document::SetMaxSizeLoadEntry()
1588 * @param entry Header Entry (Dicom Element) to be dealt with
1590 void Document::LoadDocEntry(DocEntry* entry)
1592 uint16_t group = entry->GetGroup();
1593 std::string vr = entry->GetVR();
1594 uint32_t length = entry->GetLength();
1596 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1598 // A SeQuence "contains" a set of Elements.
1599 // (fffe e000) tells us an Element is beginning
1600 // (fffe e00d) tells us an Element just ended
1601 // (fffe e0dd) tells us the current SeQuence just ended
1602 if( group == 0xfffe )
1604 // NO more value field for SQ !
1608 // When the length is zero things are easy:
1611 ((ValEntry *)entry)->SetValue("");
1615 // The elements whose length is bigger than the specified upper bound
1616 // are not loaded. Instead we leave a short notice of the offset of
1617 // the element content and it's length.
1619 std::ostringstream s;
1620 if (length > MaxSizeLoadEntry)
1622 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1624 //s << "gdcm::NotLoaded (BinEntry)";
1625 s << GDCM_NOTLOADED;
1626 s << " Address:" << (long)entry->GetOffset();
1627 s << " Length:" << entry->GetLength();
1628 s << " x(" << std::hex << entry->GetLength() << ")";
1629 binEntryPtr->SetValue(s.str());
1631 // Be carefull : a BinEntry IS_A ValEntry ...
1632 else if (ValEntry* valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1634 // s << "gdcm::NotLoaded. (ValEntry)";
1635 s << GDCM_NOTLOADED;
1636 s << " Address:" << (long)entry->GetOffset();
1637 s << " Length:" << entry->GetLength();
1638 s << " x(" << std::hex << entry->GetLength() << ")";
1639 valEntryPtr->SetValue(s.str());
1644 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1645 << "nor a ValEntry ?! Should never print that !" << std::endl;
1648 // to be sure we are at the end of the value ...
1649 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1654 // When we find a BinEntry not very much can be done :
1655 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1657 s << GDCM_BINLOADED;
1658 binEntryPtr->SetValue(s.str());
1659 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1663 /// \todo Any compacter code suggested (?)
1664 if ( IsDocEntryAnInteger(entry) )
1668 // When short integer(s) are expected, read and convert the following
1669 // n *two characters properly i.e. consider them as short integers as
1670 // opposed to strings.
1671 // Elements with Value Multiplicity > 1
1672 // contain a set of integers (not a single one)
1673 if (vr == "US" || vr == "SS")
1676 NewInt = ReadInt16();
1680 for (int i=1; i < nbInt; i++)
1683 NewInt = ReadInt16();
1688 // See above comment on multiple integers (mutatis mutandis).
1689 else if (vr == "UL" || vr == "SL")
1692 NewInt = ReadInt32();
1696 for (int i=1; i < nbInt; i++)
1699 NewInt = ReadInt32();
1704 #ifdef GDCM_NO_ANSI_STRING_STREAM
1705 s << std::ends; // to avoid oddities on Solaris
1706 #endif //GDCM_NO_ANSI_STRING_STREAM
1708 ((ValEntry *)entry)->SetValue(s.str());
1712 // FIXME: We need an additional byte for storing \0 that is not on disk
1713 char *str = new char[length+1];
1714 Fp->read(str, (size_t)length);
1715 str[length] = '\0'; //this is only useful when length is odd
1716 // Special DicomString call to properly handle \0 and even length
1717 std::string newValue;
1720 newValue = Util::DicomString(str, length+1);
1721 //dbg.Verbose(0, "Warning: bad length: ", length );
1722 dbg.Verbose(0, "For string :", newValue.c_str());
1723 // Since we change the length of string update it length
1724 entry->SetReadLength(length+1);
1728 newValue = Util::DicomString(str, length);
1732 if ( ValEntry* valEntry = dynamic_cast<ValEntry* >(entry) )
1734 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1736 dbg.Verbose(1, "Document::LoadDocEntry",
1737 "unread element value");
1738 valEntry->SetValue(GDCM_UNREAD);
1744 // Because of correspondance with the VR dic
1745 valEntry->SetValue(newValue);
1749 valEntry->SetValue(newValue);
1754 dbg.Error(true, "Document::LoadDocEntry"
1755 "Should have a ValEntry, here !");
1761 * \brief Find the value Length of the passed Header Entry
1762 * @param entry Header Entry whose length of the value shall be loaded.
1764 void Document::FindDocEntryLength( DocEntry *entry )
1765 throw ( FormatError )
1767 uint16_t element = entry->GetElement();
1768 std::string vr = entry->GetVR();
1771 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1773 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1775 // The following reserved two bytes (see PS 3.5-2003, section
1776 // "7.1.2 Data element structure with explicit vr", p 27) must be
1777 // skipped before proceeding on reading the length on 4 bytes.
1778 Fp->seekg( 2L, std::ios::cur);
1779 uint32_t length32 = ReadInt32();
1781 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1786 /// \todo rename that to FindDocEntryLengthOBOrOW since
1787 /// the above test is on both OB and OW...
1788 lengthOB = FindDocEntryLengthOB();
1790 catch ( FormatUnexpected )
1792 // Computing the length failed (this happens with broken
1793 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1794 // chance to get the pixels by deciding the element goes
1795 // until the end of the file. Hence we artificially fix the
1796 // the length and proceed.
1797 long currentPosition = Fp->tellg();
1798 Fp->seekg(0L,std::ios::end);
1799 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1800 Fp->seekg(currentPosition, std::ios::beg);
1801 entry->SetLength(lengthUntilEOF);
1804 entry->SetLength(lengthOB);
1807 FixDocEntryFoundLength(entry, length32);
1811 // Length is encoded on 2 bytes.
1812 length16 = ReadInt16();
1814 // We can tell the current file is encoded in big endian (like
1815 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1816 // and it's value is the one of the encoding of a big endian file.
1817 // In order to deal with such big endian encoded files, we have
1818 // (at least) two strategies:
1819 // * when we load the "Transfer Syntax" tag with value of big endian
1820 // encoding, we raise the proper flags. Then we wait for the end
1821 // of the META group (0x0002) among which is "Transfer Syntax",
1822 // before switching the swap code to big endian. We have to postpone
1823 // the switching of the swap code since the META group is fully encoded
1824 // in little endian, and big endian coding only starts at the next
1825 // group. The corresponding code can be hard to analyse and adds
1826 // many additional unnecessary tests for regular tags.
1827 // * the second strategy consists in waiting for trouble, that shall
1828 // appear when we find the first group with big endian encoding. This
1829 // is easy to detect since the length of a "Group Length" tag (the
1830 // ones with zero as element number) has to be of 4 (0x0004). When we
1831 // encounter 1024 (0x0400) chances are the encoding changed and we
1832 // found a group with big endian encoding.
1833 // We shall use this second strategy. In order to make sure that we
1834 // can interpret the presence of an apparently big endian encoded
1835 // length of a "Group Length" without committing a big mistake, we
1836 // add an additional check: we look in the already parsed elements
1837 // for the presence of a "Transfer Syntax" whose value has to be "big
1838 // endian encoding". When this is the case, chances are we have got our
1839 // hands on a big endian encoded file: we switch the swap code to
1840 // big endian and proceed...
1841 if ( element == 0x0000 && length16 == 0x0400 )
1843 TransferSyntaxType ts = GetTransferSyntax();
1844 if ( ts != ExplicitVRBigEndian )
1846 throw FormatError( "Document::FindDocEntryLength()",
1847 " not explicit VR." );
1851 SwitchSwapToBigEndian();
1852 // Restore the unproperly loaded values i.e. the group, the element
1853 // and the dictionary entry depending on them.
1854 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1855 uint16_t correctElem = SwapShort( entry->GetElement() );
1856 DictEntry* newTag = GetDictEntryByNumber( correctGroup,
1860 // This correct tag is not in the dictionary. Create a new one.
1861 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1863 // FIXME this can create a memory leaks on the old entry that be
1864 // left unreferenced.
1865 entry->SetDictEntry( newTag );
1868 // Heuristic: well, some files are really ill-formed.
1869 if ( length16 == 0xffff)
1871 // 0xffff means that we deal with 'Unknown Length' Sequence
1874 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1879 // Either implicit VR or a non DICOM conformal (see note below) explicit
1880 // VR that ommited the VR of (at least) this element. Farts happen.
1881 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1882 // on Data elements "Implicit and Explicit VR Data Elements shall
1883 // not coexist in a Data Set and Data Sets nested within it".]
1884 // Length is on 4 bytes.
1886 FixDocEntryFoundLength( entry, ReadInt32() );
1892 * \brief Find the Value Representation of the current Dicom Element.
1895 void Document::FindDocEntryVR( DocEntry *entry )
1897 if ( Filetype != ExplicitVR )
1904 long positionOnEntry = Fp->tellg();
1905 // Warning: we believe this is explicit VR (Value Representation) because
1906 // we used a heuristic that found "UL" in the first tag. Alas this
1907 // doesn't guarantee that all the tags will be in explicit VR. In some
1908 // cases (see e-film filtered files) one finds implicit VR tags mixed
1909 // within an explicit VR file. Hence we make sure the present tag
1910 // is in explicit VR and try to fix things if it happens not to be
1912 Fp->read (vr, (size_t)2);
1915 if( !CheckDocEntryVR(entry, vr) )
1917 Fp->seekg(positionOnEntry, std::ios::beg);
1918 // When this element is known in the dictionary we shall use, e.g. for
1919 // the semantics (see the usage of IsAnInteger), the VR proposed by the
1920 // dictionary entry. Still we have to flag the element as implicit since
1921 // we know now our assumption on expliciteness is not furfilled.
1923 if ( entry->IsVRUnknown() )
1925 entry->SetVR("Implicit");
1927 entry->SetImplicitVR();
1932 * \brief Check the correspondance between the VR of the header entry
1933 * and the taken VR. If they are different, the header entry is
1934 * updated with the new VR.
1935 * @param entry Header Entry to check
1936 * @param vr Dicom Value Representation
1937 * @return false if the VR is incorrect of if the VR isn't referenced
1938 * otherwise, it returns true
1940 bool Document::CheckDocEntryVR(DocEntry *entry, VRKey vr)
1943 bool realExplicit = true;
1945 // Assume we are reading a falsely explicit VR file i.e. we reached
1946 // a tag where we expect reading a VR but are in fact we read the
1947 // first to bytes of the length. Then we will interogate (through find)
1948 // the dicom_vr dictionary with oddities like "\004\0" which crashes
1949 // both GCC and VC++ implementations of the STL map. Hence when the
1950 // expected VR read happens to be non-ascii characters we consider
1951 // we hit falsely explicit VR tag.
1953 if ( !isalpha((unsigned char)vr[0]) && !isalpha((unsigned char)vr[1]) )
1955 realExplicit = false;
1958 // CLEANME searching the dicom_vr at each occurence is expensive.
1959 // PostPone this test in an optional integrity check at the end
1960 // of parsing or only in debug mode.
1961 if ( realExplicit && !Global::GetVR()->Count(vr) )
1963 realExplicit = false;
1966 if ( !realExplicit )
1968 // We thought this was explicit VR, but we end up with an
1969 // implicit VR tag. Let's backtrack.
1970 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
1971 entry->GetGroup(), entry->GetElement());
1972 dbg.Verbose(1, "Document::FindVR: ", msg.c_str());
1974 if( entry->GetGroup() % 2 && entry->GetElement() == 0x0000)
1976 // Group length is UL !
1977 DictEntry* newEntry = NewVirtualDictEntry(
1978 entry->GetGroup(), entry->GetElement(),
1979 "UL", "FIXME", "Group Length");
1980 entry->SetDictEntry( newEntry );
1985 if ( entry->IsVRUnknown() )
1987 // When not a dictionary entry, we can safely overwrite the VR.
1988 if( entry->GetElement() == 0x0000 )
1990 // Group length is UL !
1998 else if ( entry->GetVR() != vr )
2000 // The VR present in the file and the dictionary disagree. We assume
2001 // the file writer knew best and use the VR of the file. Since it would
2002 // be unwise to overwrite the VR of a dictionary (since it would
2003 // compromise it's next user), we need to clone the actual DictEntry
2004 // and change the VR for the read one.
2005 DictEntry* newEntry = NewVirtualDictEntry(
2006 entry->GetGroup(), entry->GetElement(),
2007 vr, "FIXME", entry->GetName());
2008 entry->SetDictEntry(newEntry);
2015 * \brief Get the transformed value of the header entry. The VR value
2016 * is used to define the transformation to operate on the value
2017 * \warning NOT end user intended method !
2018 * @param entry entry to tranform
2019 * @return Transformed entry value
2021 std::string Document::GetDocEntryValue(DocEntry *entry)
2023 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2025 std::string val = ((ValEntry *)entry)->GetValue();
2026 std::string vr = entry->GetVR();
2027 uint32_t length = entry->GetLength();
2028 std::ostringstream s;
2031 // When short integer(s) are expected, read and convert the following
2032 // n * 2 bytes properly i.e. as a multivaluated strings
2033 // (each single value is separated fromthe next one by '\'
2034 // as usual for standard multivaluated filels
2035 // Elements with Value Multiplicity > 1
2036 // contain a set of short integers (not a single one)
2038 if( vr == "US" || vr == "SS" )
2043 for (int i=0; i < nbInt; i++)
2049 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
2050 newInt16 = SwapShort( newInt16 );
2055 // When integer(s) are expected, read and convert the following
2056 // n * 4 bytes properly i.e. as a multivaluated strings
2057 // (each single value is separated fromthe next one by '\'
2058 // as usual for standard multivaluated filels
2059 // Elements with Value Multiplicity > 1
2060 // contain a set of integers (not a single one)
2061 else if( vr == "UL" || vr == "SL" )
2066 for (int i=0; i < nbInt; i++)
2072 newInt32 = ( val[4*i+0] & 0xFF )
2073 + (( val[4*i+1] & 0xFF ) << 8 )
2074 + (( val[4*i+2] & 0xFF ) << 16 )
2075 + (( val[4*i+3] & 0xFF ) << 24 );
2076 newInt32 = SwapLong( newInt32 );
2080 #ifdef GDCM_NO_ANSI_STRING_STREAM
2081 s << std::ends; // to avoid oddities on Solaris
2082 #endif //GDCM_NO_ANSI_STRING_STREAM
2086 return ((ValEntry *)entry)->GetValue();
2090 * \brief Get the reverse transformed value of the header entry. The VR
2091 * value is used to define the reverse transformation to operate on
2093 * \warning NOT end user intended method !
2094 * @param entry Entry to reverse transform
2095 * @return Reverse transformed entry value
2097 std::string Document::GetDocEntryUnvalue(DocEntry* entry)
2099 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2101 std::string vr = entry->GetVR();
2102 std::vector<std::string> tokens;
2103 std::ostringstream s;
2105 if ( vr == "US" || vr == "SS" )
2109 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
2110 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2111 for (unsigned int i=0; i<tokens.size(); i++)
2113 newInt16 = atoi(tokens[i].c_str());
2114 s << ( newInt16 & 0xFF )
2115 << (( newInt16 >> 8 ) & 0xFF );
2119 if ( vr == "UL" || vr == "SL")
2123 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
2124 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2125 for (unsigned int i=0; i<tokens.size();i++)
2127 newInt32 = atoi(tokens[i].c_str());
2128 s << (char)( newInt32 & 0xFF )
2129 << (char)(( newInt32 >> 8 ) & 0xFF )
2130 << (char)(( newInt32 >> 16 ) & 0xFF )
2131 << (char)(( newInt32 >> 24 ) & 0xFF );
2136 #ifdef GDCM_NO_ANSI_STRING_STREAM
2137 s << std::ends; // to avoid oddities on Solaris
2138 #endif //GDCM_NO_ANSI_STRING_STREAM
2142 return ((ValEntry *)entry)->GetValue();
2146 * \brief Skip a given Header Entry
2147 * \warning NOT end user intended method !
2148 * @param entry entry to skip
2150 void Document::SkipDocEntry(DocEntry *entry)
2152 SkipBytes(entry->GetLength());
2156 * \brief Skips to the begining of the next Header Entry
2157 * \warning NOT end user intended method !
2158 * @param entry entry to skip
2160 void Document::SkipToNextDocEntry(DocEntry *entry)
2162 Fp->seekg((long)(entry->GetOffset()), std::ios::beg);
2163 Fp->seekg( (long)(entry->GetReadLength()), std::ios::cur);
2167 * \brief When the length of an element value is obviously wrong (because
2168 * the parser went Jabberwocky) one can hope improving things by
2169 * applying some heuristics.
2170 * @param entry entry to check
2171 * @param foundLength fist assumption about length
2173 void Document::FixDocEntryFoundLength(DocEntry *entry,
2174 uint32_t foundLength)
2176 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
2177 if ( foundLength == 0xffffffff)
2182 uint16_t gr = entry->GetGroup();
2183 uint16_t el = entry->GetElement();
2185 if ( foundLength % 2)
2187 std::ostringstream s;
2188 s << "Warning : Tag with uneven length "
2190 << " in x(" << std::hex << gr << "," << el <<")" << std::dec;
2191 dbg.Verbose(0, s.str().c_str());
2194 //////// Fix for some naughty General Electric images.
2195 // Allthough not recent many such GE corrupted images are still present
2196 // on Creatis hard disks. Hence this fix shall remain when such images
2197 // are no longer in user (we are talking a few years, here)...
2198 // Note: XMedCom probably uses such a trick since it is able to read
2199 // those pesky GE images ...
2200 if ( foundLength == 13)
2202 // Only happens for this length !
2203 if ( entry->GetGroup() != 0x0008
2204 || ( entry->GetElement() != 0x0070
2205 && entry->GetElement() != 0x0080 ) )
2208 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
2212 //////// Fix for some brain-dead 'Leonardo' Siemens images.
2213 // Occurence of such images is quite low (unless one leaves close to a
2214 // 'Leonardo' source. Hence, one might consider commenting out the
2215 // following fix on efficiency reasons.
2216 else if ( entry->GetGroup() == 0x0009
2217 && ( entry->GetElement() == 0x1113
2218 || entry->GetElement() == 0x1114 ) )
2221 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
2224 else if ( entry->GetVR() == "SQ" )
2226 foundLength = 0; // ReadLength is unchanged
2229 //////// We encountered a 'delimiter' element i.e. a tag of the form
2230 // "fffe|xxxx" which is just a marker. Delimiters length should not be
2231 // taken into account.
2232 else if( entry->GetGroup() == 0xfffe )
2234 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
2235 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
2236 // causes extra troubles...
2237 if( entry->GetElement() != 0x0000 )
2243 entry->SetUsableLength(foundLength);
2247 * \brief Apply some heuristics to predict whether the considered
2248 * element value contains/represents an integer or not.
2249 * @param entry The element value on which to apply the predicate.
2250 * @return The result of the heuristical predicate.
2252 bool Document::IsDocEntryAnInteger(DocEntry *entry)
2254 uint16_t element = entry->GetElement();
2255 uint16_t group = entry->GetGroup();
2256 const std::string & vr = entry->GetVR();
2257 uint32_t length = entry->GetLength();
2259 // When we have some semantics on the element we just read, and if we
2260 // a priori know we are dealing with an integer, then we shall be
2261 // able to swap it's element value properly.
2262 if ( element == 0 ) // This is the group length of the group
2270 // Allthough this should never happen, still some images have a
2271 // corrupted group length [e.g. have a glance at offset x(8336) of
2272 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2273 // Since for dicom compliant and well behaved headers, the present
2274 // test is useless (and might even look a bit paranoid), when we
2275 // encounter such an ill-formed image, we simply display a warning
2276 // message and proceed on parsing (while crossing fingers).
2277 std::ostringstream s;
2278 long filePosition = Fp->tellg();
2279 s << "Erroneous Group Length element length on : (" \
2280 << std::hex << group << " , " << element
2281 << ") -before- position x(" << filePosition << ")"
2282 << "lgt : " << length;
2283 dbg.Verbose(0, "Document::IsDocEntryAnInteger", s.str().c_str() );
2287 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2296 * \brief Find the Length till the next sequence delimiter
2297 * \warning NOT end user intended method !
2301 uint32_t Document::FindDocEntryLengthOB()
2302 throw( FormatUnexpected )
2304 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2305 long positionOnEntry = Fp->tellg();
2306 bool foundSequenceDelimiter = false;
2307 uint32_t totalLength = 0;
2309 while ( !foundSequenceDelimiter )
2315 group = ReadInt16();
2318 catch ( FormatError )
2320 throw FormatError("Document::FindDocEntryLengthOB()",
2321 " group or element not present.");
2324 // We have to decount the group and element we just read
2327 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2329 dbg.Verbose(1, "Document::FindDocEntryLengthOB: neither an Item "
2330 "tag nor a Sequence delimiter tag.");
2331 Fp->seekg(positionOnEntry, std::ios::beg);
2332 throw FormatUnexpected("Document::FindDocEntryLengthOB()",
2333 "Neither an Item tag nor a Sequence "
2337 if ( elem == 0xe0dd )
2339 foundSequenceDelimiter = true;
2342 uint32_t itemLength = ReadInt32();
2343 // We add 4 bytes since we just read the ItemLength with ReadInt32
2344 totalLength += itemLength + 4;
2345 SkipBytes(itemLength);
2347 if ( foundSequenceDelimiter )
2352 Fp->seekg( positionOnEntry, std::ios::beg);
2357 * \brief Reads a supposed to be 16 Bits integer
2358 * (swaps it depending on processor endianity)
2359 * @return read value
2361 uint16_t Document::ReadInt16()
2362 throw( FormatError )
2365 Fp->read ((char*)&g, (size_t)2);
2368 throw FormatError( "Document::ReadInt16()", " file error." );
2372 throw FormatError( "Document::ReadInt16()", "EOF." );
2379 * \brief Reads a supposed to be 32 Bits integer
2380 * (swaps it depending on processor endianity)
2381 * @return read value
2383 uint32_t Document::ReadInt32()
2384 throw( FormatError )
2387 Fp->read ((char*)&g, (size_t)4);
2390 throw FormatError( "Document::ReadInt32()", " file error." );
2394 throw FormatError( "Document::ReadInt32()", "EOF." );
2401 * \brief skips bytes inside the source file
2402 * \warning NOT end user intended method !
2405 void Document::SkipBytes(uint32_t nBytes)
2407 //FIXME don't dump the returned value
2408 Fp->seekg((long)nBytes, std::ios::cur);
2412 * \brief Loads all the needed Dictionaries
2413 * \warning NOT end user intended method !
2415 void Document::Initialise()
2417 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2419 RLEInfo = new RLEFramesInfo;
2420 JPEGInfo = new JPEGFragmentsInfo;
2425 * \brief Discover what the swap code is (among little endian, big endian,
2426 * bad little endian, bad big endian).
2428 * @return false when we are absolutely sure
2429 * it's neither ACR-NEMA nor DICOM
2430 * true when we hope ours assuptions are OK
2432 bool Document::CheckSwap()
2434 // The only guaranted way of finding the swap code is to find a
2435 // group tag since we know it's length has to be of four bytes i.e.
2436 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2437 // occurs when we can't find such group...
2439 uint32_t x = 4; // x : for ntohs
2440 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2446 // First, compare HostByteOrder and NetworkByteOrder in order to
2447 // determine if we shall need to swap bytes (i.e. the Endian type).
2448 if ( x == ntohs(x) )
2457 // The easiest case is the one of a DICOM header, since it possesses a
2458 // file preamble where it suffice to look for the string "DICM".
2461 char *entCur = deb + 128;
2462 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2464 dbg.Verbose(1, "Document::CheckSwap:", "looks like DICOM Version3");
2466 // Next, determine the value representation (VR). Let's skip to the
2467 // first element (0002, 0000) and check there if we find "UL"
2468 // - or "OB" if the 1st one is (0002,0001) -,
2469 // in which case we (almost) know it is explicit VR.
2470 // WARNING: if it happens to be implicit VR then what we will read
2471 // is the length of the group. If this ascii representation of this
2472 // length happens to be "UL" then we shall believe it is explicit VR.
2473 // FIXME: in order to fix the above warning, we could read the next
2474 // element value (or a couple of elements values) in order to make
2475 // sure we are not commiting a big mistake.
2476 // We need to skip :
2477 // * the 128 bytes of File Preamble (often padded with zeroes),
2478 // * the 4 bytes of "DICM" string,
2479 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2480 // i.e. a total of 136 bytes.
2484 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2485 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2486 // *Implicit* VR. -and it is !-
2488 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2489 memcmp(entCur, "OB", (size_t)2) == 0 ||
2490 memcmp(entCur, "UI", (size_t)2) == 0 ||
2491 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2492 // when Write DCM *adds*
2494 // Use Document::dicom_vr to test all the possibilities
2495 // instead of just checking for UL, OB and UI !? group 0000
2497 Filetype = ExplicitVR;
2498 dbg.Verbose(1, "Document::CheckSwap:",
2499 "explicit Value Representation");
2503 Filetype = ImplicitVR;
2504 dbg.Verbose(1, "Document::CheckSwap:",
2505 "not an explicit Value Representation");
2511 dbg.Verbose(1, "Document::CheckSwap:",
2512 "HostByteOrder != NetworkByteOrder");
2517 dbg.Verbose(1, "Document::CheckSwap:",
2518 "HostByteOrder = NetworkByteOrder");
2521 // Position the file position indicator at first tag (i.e.
2522 // after the file preamble and the "DICM" string).
2523 Fp->seekg(0, std::ios::beg);
2524 Fp->seekg ( 132L, std::ios::beg);
2528 // Alas, this is not a DicomV3 file and whatever happens there is no file
2529 // preamble. We can reset the file position indicator to where the data
2530 // is (i.e. the beginning of the file).
2531 dbg.Verbose(1, "Document::CheckSwap:", "not a DICOM Version3 file");
2532 Fp->seekg(0, std::ios::beg);
2534 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2535 // By clean we mean that the length of the first tag is written down.
2536 // If this is the case and since the length of the first group HAS to be
2537 // four (bytes), then determining the proper swap code is straightforward.
2540 // We assume the array of char we are considering contains the binary
2541 // representation of a 32 bits integer. Hence the following dirty
2543 s32 = *((uint32_t *)(entCur));
2564 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2565 // It is time for despaired wild guesses.
2566 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2567 // i.e. the 'group length' element is not present :
2569 // check the supposed-to-be 'group number'
2570 // in ( 0x0001 .. 0x0008 )
2571 // to determine ' SwapCode' value .
2572 // Only 0 or 4321 will be possible
2573 // (no oportunity to check for the formerly well known
2574 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2575 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2576 // the file IS NOT ACR-NEMA nor DICOM V3
2577 // Find a trick to tell it the caller...
2579 s16 = *((uint16_t *)(deb));
2606 dbg.Verbose(0, "Document::CheckSwap:",
2607 "ACR/NEMA unfound swap info (Really hopeless !)");
2611 // Then the only info we have is the net2host one.
2623 * \brief Restore the unproperly loaded values i.e. the group, the element
2624 * and the dictionary entry depending on them.
2626 void Document::SwitchSwapToBigEndian()
2628 dbg.Verbose(1, "Document::SwitchSwapToBigEndian",
2629 "Switching to BigEndian mode.");
2630 if ( SwapCode == 0 )
2634 else if ( SwapCode == 4321 )
2638 else if ( SwapCode == 3412 )
2642 else if ( SwapCode == 2143 )
2649 * \brief during parsing, Header Elements too long are not loaded in memory
2652 void Document::SetMaxSizeLoadEntry(long newSize)
2658 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2660 MaxSizeLoadEntry = 0xffffffff;
2663 MaxSizeLoadEntry = newSize;
2668 * \brief Header Elements too long will not be printed
2669 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2672 void Document::SetMaxSizePrintEntry(long newSize)
2674 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2679 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2681 MaxSizePrintEntry = 0xffffffff;
2684 MaxSizePrintEntry = newSize;
2690 * \brief Handle broken private tag from Philips NTSCAN
2691 * where the endianess is being switch to BigEndian for no
2695 void Document::HandleBrokenEndian(uint16_t group, uint16_t elem)
2697 // Endian reversion. Some files contain groups of tags with reversed endianess.
2698 static int reversedEndian = 0;
2699 // try to fix endian switching in the middle of headers
2700 if ((group == 0xfeff) && (elem == 0x00e0))
2702 // start endian swap mark for group found
2704 SwitchSwapToBigEndian();
2709 else if ((group == 0xfffe) && (elem == 0xe00d) && reversedEndian)
2711 // end of reversed endian group
2713 SwitchSwapToBigEndian();
2719 * \brief Read the next tag but WITHOUT loading it's value
2720 * (read the 'Group Number', the 'Element Number',
2721 * gets the Dict Entry
2722 * gets the VR, gets the length, gets the offset value)
2723 * @return On succes the newly created DocEntry, NULL on failure.
2725 DocEntry* Document::ReadNextDocEntry()
2732 group = ReadInt16();
2735 catch ( FormatError e )
2737 // We reached the EOF (or an error occured) therefore
2738 // header parsing has to be considered as finished.
2743 HandleBrokenEndian(group, elem);
2744 DocEntry *newEntry = NewDocEntryByNumber(group, elem);
2745 FindDocEntryVR(newEntry);
2749 FindDocEntryLength(newEntry);
2751 catch ( FormatError e )
2759 newEntry->SetOffset(Fp->tellg());
2766 * \brief Generate a free TagKey i.e. a TagKey that is not present
2767 * in the TagHt dictionary.
2768 * @param group The generated tag must belong to this group.
2769 * @return The element of tag with given group which is fee.
2771 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2773 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2775 TagKey key = DictEntry::TranslateToKey(group, elem);
2776 if (TagHT.count(key) == 0)
2785 * \brief Assuming the internal file pointer \ref Document::Fp
2786 * is placed at the beginning of a tag check whether this
2787 * tag is (TestGroup, TestElement).
2788 * \warning On success the internal file pointer \ref Document::Fp
2789 * is modified to point after the tag.
2790 * On failure (i.e. when the tag wasn't the expected tag
2791 * (TestGroup, TestElement) the internal file pointer
2792 * \ref Document::Fp is restored to it's original position.
2793 * @param testGroup The expected group of the tag.
2794 * @param testElement The expected Element of the tag.
2795 * @return True on success, false otherwise.
2797 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2799 long positionOnEntry = Fp->tellg();
2800 long currentPosition = Fp->tellg(); // On debugging purposes
2802 //// Read the Item Tag group and element, and make
2803 // sure they are what we expected:
2804 uint16_t itemTagGroup;
2805 uint16_t itemTagElement;
2808 itemTagGroup = ReadInt16();
2809 itemTagElement = ReadInt16();
2811 catch ( FormatError e )
2813 //std::cerr << e << std::endl;
2816 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2818 std::ostringstream s;
2819 s << " We should have found tag (";
2820 s << std::hex << testGroup << "," << testElement << ")" << std::endl;
2821 s << " but instead we encountered tag (";
2822 s << std::hex << itemTagGroup << "," << itemTagElement << ")"
2824 s << " at address: " << (unsigned)currentPosition << std::endl;
2825 dbg.Verbose(0, "Document::ReadItemTagLength: wrong Item Tag found:");
2826 dbg.Verbose(0, s.str().c_str());
2827 Fp->seekg(positionOnEntry, std::ios::beg);
2835 * \brief Assuming the internal file pointer \ref Document::Fp
2836 * is placed at the beginning of a tag (TestGroup, TestElement),
2837 * read the length associated to the Tag.
2838 * \warning On success the internal file pointer \ref Document::Fp
2839 * is modified to point after the tag and it's length.
2840 * On failure (i.e. when the tag wasn't the expected tag
2841 * (TestGroup, TestElement) the internal file pointer
2842 * \ref Document::Fp is restored to it's original position.
2843 * @param testGroup The expected group of the tag.
2844 * @param testElement The expected Element of the tag.
2845 * @return On success returns the length associated to the tag. On failure
2848 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2850 long positionOnEntry = Fp->tellg();
2851 (void)positionOnEntry;
2853 if ( !ReadTag(testGroup, testElement) )
2858 //// Then read the associated Item Length
2859 long currentPosition = Fp->tellg();
2860 uint32_t itemLength = ReadInt32();
2862 std::ostringstream s;
2863 s << "Basic Item Length is: "
2864 << itemLength << std::endl;
2865 s << " at address: " << (unsigned)currentPosition << std::endl;
2866 dbg.Verbose(0, "Document::ReadItemTagLength: ", s.str().c_str());
2872 * \brief When parsing the Pixel Data of an encapsulated file, read
2873 * the basic offset table (when present, and BTW dump it).
2875 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2877 //// Read the Basic Offset Table Item Tag length...
2878 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2880 // When present, read the basic offset table itself.
2881 // Notes: - since the presence of this basic offset table is optional
2882 // we can't rely on it for the implementation, and we will simply
2883 // trash it's content (when present).
2884 // - still, when present, we could add some further checks on the
2885 // lengths, but we won't bother with such fuses for the time being.
2886 if ( itemLength != 0 )
2888 char* basicOffsetTableItemValue = new char[itemLength + 1];
2889 Fp->read(basicOffsetTableItemValue, itemLength);
2892 for (unsigned int i=0; i < itemLength; i += 4 )
2894 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2896 std::ostringstream s;
2897 s << " Read one length: ";
2898 s << std::hex << individualLength << std::endl;
2900 "Document::ReadAndSkipEncapsulatedBasicOffsetTable: ",
2905 delete[] basicOffsetTableItemValue;
2910 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2911 * Compute the RLE extra information and store it in \ref RLEInfo
2912 * for later pixel retrieval usage.
2914 void Document::ComputeRLEInfo()
2916 TransferSyntaxType ts = GetTransferSyntax();
2917 if ( ts != RLELossless )
2922 // Encoded pixel data: for the time being we are only concerned with
2923 // Jpeg or RLE Pixel data encodings.
2924 // As stated in PS 3.5-2003, section 8.2 p44:
2925 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2926 // value representation OB is used".
2927 // Hence we expect an OB value representation. Concerning OB VR,
2928 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2929 // "For the Value Representations OB and OW, the encoding shall meet the
2930 // following specifications depending on the Data element tag:"
2932 // - the first item in the sequence of items before the encoded pixel
2933 // data stream shall be basic offset table item. The basic offset table
2934 // item value, however, is not required to be present"
2936 ReadAndSkipEncapsulatedBasicOffsetTable();
2938 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2939 // Loop on the individual frame[s] and store the information
2940 // on the RLE fragments in a RLEFramesInfo.
2941 // Note: - when only a single frame is present, this is a
2943 // - when more than one frame are present, then we are in
2944 // the case of a multi-frame image.
2946 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2948 // Parse the RLE Header and store the corresponding RLE Segment
2949 // Offset Table information on fragments of this current Frame.
2950 // Note that the fragment pixels themselves are not loaded
2951 // (but just skipped).
2952 long frameOffset = Fp->tellg();
2954 uint32_t nbRleSegments = ReadInt32();
2955 if ( nbRleSegments > 16 )
2957 // There should be at most 15 segments (refer to RLEFrame class)
2958 dbg.Verbose(0, "Document::ComputeRLEInfo: too many segments.");
2961 uint32_t rleSegmentOffsetTable[16];
2962 for( int k = 1; k <= 15; k++ )
2964 rleSegmentOffsetTable[k] = ReadInt32();
2967 // Deduce from both the RLE Header and the frameLength the
2968 // fragment length, and again store this info in a
2970 long rleSegmentLength[15];
2971 // skipping (not reading) RLE Segments
2972 if ( nbRleSegments > 1)
2974 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2976 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2977 - rleSegmentOffsetTable[k];
2978 SkipBytes(rleSegmentLength[k]);
2982 rleSegmentLength[nbRleSegments] = frameLength
2983 - rleSegmentOffsetTable[nbRleSegments];
2984 SkipBytes(rleSegmentLength[nbRleSegments]);
2986 // Store the collected info
2987 RLEFrame* newFrameInfo = new RLEFrame;
2988 newFrameInfo->NumberFragments = nbRleSegments;
2989 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2991 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2992 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2994 RLEInfo->Frames.push_back( newFrameInfo );
2997 // Make sure that at the end of the item we encounter a 'Sequence
2999 if ( !ReadTag(0xfffe, 0xe0dd) )
3001 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
3002 dbg.Verbose(0, " item at end of RLE item sequence");
3007 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
3008 * Compute the jpeg extra information (fragment[s] offset[s] and
3009 * length) and store it[them] in \ref JPEGInfo for later pixel
3012 void Document::ComputeJPEGFragmentInfo()
3014 // If you need to, look for comments of ComputeRLEInfo().
3020 ReadAndSkipEncapsulatedBasicOffsetTable();
3022 // Loop on the fragments[s] and store the parsed information in a
3024 long fragmentLength;
3025 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
3027 long fragmentOffset = Fp->tellg();
3029 // Store the collected info
3030 JPEGFragment* newFragment = new JPEGFragment;
3031 newFragment->Offset = fragmentOffset;
3032 newFragment->Length = fragmentLength;
3033 JPEGInfo->Fragments.push_back( newFragment );
3035 SkipBytes( fragmentLength );
3038 // Make sure that at the end of the item we encounter a 'Sequence
3040 if ( !ReadTag(0xfffe, 0xe0dd) )
3042 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
3043 dbg.Verbose(0, " item at end of JPEG item sequence");
3048 * \brief Walk recursively the given \ref DocEntrySet, and feed
3049 * the given hash table (\ref TagDocEntryHT) with all the
3050 * \ref DocEntry (Dicom entries) encountered.
3051 * This method does the job for \ref BuildFlatHashTable.
3052 * @param builtHT Where to collect all the \ref DocEntry encountered
3053 * when recursively walking the given set.
3054 * @param set The structure to be traversed (recursively).
3056 void Document::BuildFlatHashTableRecurse( TagDocEntryHT& builtHT,
3059 if (ElementSet* elementSet = dynamic_cast< ElementSet* > ( set ) )
3061 TagDocEntryHT const & currentHT = elementSet->GetTagHT();
3062 for( TagDocEntryHT::const_iterator i = currentHT.begin();
3063 i != currentHT.end();
3066 DocEntry* entry = i->second;
3067 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3069 const ListSQItem& items = seqEntry->GetSQItems();
3070 for( ListSQItem::const_iterator item = items.begin();
3071 item != items.end();
3074 BuildFlatHashTableRecurse( builtHT, *item );
3078 builtHT[entry->GetKey()] = entry;
3083 if (SQItem* SQItemSet = dynamic_cast< SQItem* > ( set ) )
3085 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
3086 for (ListDocEntry::const_iterator i = currentList.begin();
3087 i != currentList.end();
3090 DocEntry* entry = *i;
3091 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3093 const ListSQItem& items = seqEntry->GetSQItems();
3094 for( ListSQItem::const_iterator item = items.begin();
3095 item != items.end();
3098 BuildFlatHashTableRecurse( builtHT, *item );
3102 builtHT[entry->GetKey()] = entry;
3109 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
3112 * The structure used by a Document (through \ref ElementSet),
3113 * in order to hold the parsed entries of a Dicom header, is a recursive
3114 * one. This is due to the fact that the sequences (when present)
3115 * can be nested. Additionaly, the sequence items (represented in
3116 * gdcm as \ref SQItem) add an extra complexity to the data
3117 * structure. Hence, a gdcm user whishing to visit all the entries of
3118 * a Dicom header will need to dig in the gdcm internals (which
3119 * implies exposing all the internal data structures to the API).
3120 * In order to avoid this burden to the user, \ref BuildFlatHashTable
3121 * recursively builds a temporary hash table, which holds all the
3122 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
3124 * \warning Of course there is NO integrity constrain between the
3125 * returned \ref TagDocEntryHT and the \ref ElementSet used
3126 * to build it. Hence if the underlying \ref ElementSet is
3127 * altered, then it is the caller responsability to invoke
3128 * \ref BuildFlatHashTable again...
3129 * @return The flat std::map<> we juste build.
3131 TagDocEntryHT* Document::BuildFlatHashTable()
3133 TagDocEntryHT* FlatHT = new TagDocEntryHT;
3134 BuildFlatHashTableRecurse( *FlatHT, this );
3141 * \brief Compares two documents, according to \ref DicomDir rules
3142 * \warning Does NOT work with ACR-NEMA files
3143 * \todo Find a trick to solve the pb (use RET fields ?)
3145 * @return true if 'smaller'
3147 bool Document::operator<(Document &document)
3150 std::string s1 = GetEntryByNumber(0x0010,0x0010);
3151 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
3163 s1 = GetEntryByNumber(0x0010,0x0020);
3164 s2 = document.GetEntryByNumber(0x0010,0x0020);
3175 // Study Instance UID
3176 s1 = GetEntryByNumber(0x0020,0x000d);
3177 s2 = document.GetEntryByNumber(0x0020,0x000d);
3188 // Serie Instance UID
3189 s1 = GetEntryByNumber(0x0020,0x000e);
3190 s2 = document.GetEntryByNumber(0x0020,0x000e);
3205 } // end namespace gdcm
3207 //-----------------------------------------------------------------------------