2 //-----------------------------------------------------------------------------
3 #include "gdcmParser.h"
11 #include <netinet/in.h>
14 #ifdef GDCM_NO_ANSI_STRING_STREAM
16 # define ostringstream ostrstream
22 #define UI1_2_840_10008_1_2 "1.2.840.10008.1.2"
23 #define UI1_2_840_10008_1_2_1 "1.2.840.10008.1.2.1"
24 #define UI1_2_840_10008_1_2_2 "1.2.840.10008.1.2.2"
25 #define UI1_2_840_10008_1_2_1_99 "1.2.840.10008.1.2.1.99"
29 // ---> Warning : This fourth field is NOT part
30 // of the 'official' Dicom Dictionnary
31 // and should NOT be used.
32 // (Not defined for all the groups
33 // may be removed in a future release)
36 // META Meta Information
48 // NMI Nuclear Medicine
50 // BFS Basic Film Session
52 // BIB Basic Image Box
67 //-----------------------------------------------------------------------------
68 // Refer to gdcmParser::CheckSwap()
69 const unsigned int gdcmParser::HEADER_LENGTH_TO_READ = 256;
71 // Refer to gdcmParser::SetMaxSizeLoadEntry()
72 const unsigned int gdcmParser::MAX_SIZE_LOAD_ELEMENT_VALUE = 4096;
74 // Refer to gdcmParser::SetMaxSizePrintEntry()
75 // TODO : Right now, better see "define, in gdcmHederEntry.cxx
76 const unsigned int gdcmParser::MAX_SIZE_PRINT_ELEMENT_VALUE = 64;
78 //-----------------------------------------------------------------------------
79 // Constructor / Destructor
84 * @param exception_on_error
85 * @param enable_sequences = true to allow the header
86 * to be parsed *inside* the SeQuences,
87 * when they have an actual length
88 * @param ignore_shadow to allow skipping the shadow elements,
89 * to save memory space.
90 * \warning The TRUE value for this param has to be used
91 * with a FALSE value for the 'enable_sequence' param.
92 * ('public elements' may be embedded in 'shadow Sequences')
94 gdcmParser::gdcmParser(const char *InFilename,
95 bool exception_on_error,
96 bool enable_sequences,
98 enableSequences=enable_sequences;
99 ignoreShadow =ignore_shadow;
101 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
102 filename = InFilename;
105 if ( !OpenFile(exception_on_error))
112 wasUpdated = 0; // will be set to 1 if user adds an entry
113 printLevel = 1; // 'Medium' print level by default
117 * \ingroup gdcmParser
119 * @param exception_on_error
121 gdcmParser::gdcmParser(bool exception_on_error) {
124 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
127 wasUpdated = 0; // will be set to 1 if user adds an entry
128 printLevel = 1; // 'Medium' print level by default
132 * \ingroup gdcmParser
133 * \brief Canonical destructor.
135 gdcmParser::~gdcmParser (void) {
140 //-----------------------------------------------------------------------------
143 * \ingroup gdcmParser
144 * \brief Prints the Header Entries (Dicom Elements)
145 * both from the H Table and the chained list
148 void gdcmParser::PrintEntry(std::ostream & os) {
149 std::ostringstream s;
151 for (ListTag::iterator i = listEntries.begin();
152 i != listEntries.end();
155 (*i)->SetPrintLevel(printLevel);
162 * \ingroup gdcmParser
163 * \brief Prints The Dict Entries of THE public Dicom Dictionnry
166 void gdcmParser::PrintPubDict(std::ostream & os) {
167 RefPubDict->Print(os);
171 * \ingroup gdcmParser
172 * \brief Prints The Dict Entries of THE shadow Dicom Dictionnry
175 void gdcmParser::PrintShaDict(std::ostream & os) {
176 RefShaDict->Print(os);
179 //-----------------------------------------------------------------------------
182 * \ingroup gdcmParser
183 * \brief Get the public dictionary used
185 gdcmDict *gdcmParser::GetPubDict(void) {
190 * \ingroup gdcmParser
191 * \brief Get the shadow dictionary used
193 gdcmDict *gdcmParser::GetShaDict(void) {
198 * \ingroup gdcmParser
199 * \brief Set the shadow dictionary used
200 * \param dict dictionary to use in shadow
202 bool gdcmParser::SetShaDict(gdcmDict *dict){
208 * \ingroup gdcmParser
209 * \brief Set the shadow dictionary used
210 * \param dictName name of the dictionary to use in shadow
212 bool gdcmParser::SetShaDict(DictKey dictName){
213 RefShaDict=gdcmGlobal::GetDicts()->GetDict(dictName);
218 * \ingroup gdcmParser
219 * \brief This predicate, based on hopefully reasonable heuristics,
220 * decides whether or not the current gdcmParser was properly parsed
221 * and contains the mandatory information for being considered as
222 * a well formed and usable Dicom/Acr File.
223 * @return true when gdcmParser is the one of a reasonable Dicom/Acr file,
226 bool gdcmParser::IsReadable(void) {
227 if(filetype==Unknown)
229 if(listEntries.size()<=0)
236 * \ingroup gdcmParser
237 * \brief Determines if the Transfer Syntax was already encountered
238 * and if it corresponds to a ImplicitVRLittleEndian one.
240 * @return True when ImplicitVRLittleEndian found. False in all other cases.
242 bool gdcmParser::IsImplicitVRLittleEndianTransferSyntax(void) {
243 gdcmHeaderEntry *Element = GetHeaderEntryByNumber(0x0002, 0x0010);
246 LoadHeaderEntrySafe(Element);
248 std::string Transfer = Element->GetValue();
249 if ( Transfer == UI1_2_840_10008_1_2 )
255 * \ingroup gdcmParser
256 * \brief Determines if the Transfer Syntax was already encountered
257 * and if it corresponds to a ExplicitVRLittleEndian one.
259 * @return True when ExplicitVRLittleEndian found. False in all other cases.
261 bool gdcmParser::IsExplicitVRLittleEndianTransferSyntax(void) {
262 gdcmHeaderEntry* Element = GetHeaderEntryByNumber(0x0002, 0x0010);
265 LoadHeaderEntrySafe(Element);
267 std::string Transfer = Element->GetValue();
268 if ( Transfer == UI1_2_840_10008_1_2_1 )
274 * \ingroup gdcmParser
275 * \brief Determines if the Transfer Syntax was already encountered
276 * and if it corresponds to a DeflatedExplicitVRLittleEndian one.
278 * @return True when DeflatedExplicitVRLittleEndian found. False in all other cases.
280 bool gdcmParser::IsDeflatedExplicitVRLittleEndianTransferSyntax(void) {
281 gdcmHeaderEntry* Element = GetHeaderEntryByNumber(0x0002, 0x0010);
284 LoadHeaderEntrySafe(Element);
286 std::string Transfer = Element->GetValue();
287 if ( Transfer == UI1_2_840_10008_1_2_1_99 )
293 * \ingroup gdcmParser
294 * \brief Determines if the Transfer Syntax was already encountered
295 * and if it corresponds to a Explicit VR Big Endian one.
297 * @return True when big endian found. False in all other cases.
299 bool gdcmParser::IsExplicitVRBigEndianTransferSyntax(void) {
300 gdcmHeaderEntry* Element = GetHeaderEntryByNumber(0x0002, 0x0010);
303 LoadHeaderEntrySafe(Element);
305 std::string Transfer = Element->GetValue();
306 if ( Transfer == UI1_2_840_10008_1_2_2 ) //1.2.2 ??? A verifier !
312 * \ingroup gdcmParser
313 * \brief returns the File Type
314 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
317 FileType gdcmParser::GetFileType(void) {
322 * \ingroup gdcmParser
323 * \brief opens the file
324 * @param exception_on_error
327 FILE *gdcmParser::OpenFile(bool exception_on_error)
330 fp=fopen(filename.c_str(),"rb");
331 if(exception_on_error)
334 throw gdcmFileError("gdcmParser::gdcmParser(const char *, bool)");
340 fread(&zero, (size_t)2, (size_t)1, fp);
342 //ACR -- or DICOM with no Preamble --
343 if( zero == 0x0008 || zero == 0x0800 || zero == 0x0002 || zero == 0x0200)
347 fseek(fp, 126L, SEEK_CUR);
349 fread(dicm, (size_t)4, (size_t)1, fp);
350 if( memcmp(dicm, "DICM", 4) == 0 )
354 dbg.Verbose(0, "gdcmParser::gdcmParser not DICOM/ACR", filename.c_str());
357 dbg.Verbose(0, "gdcmParser::gdcmParser cannot open file", filename.c_str());
363 * \ingroup gdcmParser
364 * \brief closes the file
365 * @return TRUE if the close was successfull
367 bool gdcmParser::CloseFile(void) {
368 int closed = fclose(fp);
376 * \ingroup gdcmParser
378 * @param fp file pointer on an already open file
379 * @param type type of the File to be written
380 * (ACR-NEMA, ExplicitVR, ImplicitVR)
381 * @return always "True" ?!
383 bool gdcmParser::Write(FILE *fp, FileType type) {
385 // TODO The stuff was rewritten using the chained list instead
387 // so we could remove the GroupHT from the gdcmParser
391 // TODO : move the following lines (and a lot of others, to be written)
392 // to a future function CheckAndCorrectHeader
395 // Comment pourrait-on savoir si le DcmHeader vient d'un fichier DicomV3 ou non
396 // (FileType est un champ de gdcmParser ...)
397 // WARNING : Si on veut ecrire du DICOM V3 a partir d'un DcmHeader ACR-NEMA
399 // a moins de se livrer a un tres complique ajout des champs manquants.
400 // faire un CheckAndCorrectHeader (?)
403 if (type == ImplicitVR)
405 std::string implicitVRTransfertSyntax = UI1_2_840_10008_1_2;
406 ReplaceOrCreateByNumber(implicitVRTransfertSyntax,0x0002, 0x0010);
408 //FIXME Refer to standards on page 21, chapter 6.2 "Value representation":
409 // values with a VR of UI shall be padded with a single trailing null
410 // Dans le cas suivant on doit pader manuellement avec un 0
412 SetEntryLengthByNumber(18, 0x0002, 0x0010);
415 if (type == ExplicitVR)
417 std::string explicitVRTransfertSyntax = UI1_2_840_10008_1_2_1;
418 ReplaceOrCreateByNumber(explicitVRTransfertSyntax,0x0002, 0x0010);
420 //FIXME Refer to standards on page 21, chapter 6.2 "Value representation":
421 // values with a VR of UI shall be padded with a single trailing null
422 // Dans le cas suivant on doit pader manuellement avec un 0
424 SetEntryLengthByNumber(20, 0x0002, 0x0010);
427 /* TODO : rewrite later
429 if ( (type == ImplicitVR) || (type == ExplicitVR) )
430 UpdateGroupLength(false,type);
432 UpdateGroupLength(true,ACR);
435 WriteEntries(type, fp);
440 * \ingroup gdcmParser
441 * \brief Modifies the value of a given Header Entry (Dicom Element)
442 * if it exists; Creates it with the given value if it doesn't
443 * @param Value passed as a std::string
448 bool gdcmParser::ReplaceOrCreateByNumber(std::string Value,
451 if (CheckIfEntryExistByNumber(Group, Elem) == 0) {
452 gdcmHeaderEntry *a =NewHeaderEntryByNumber(Group, Elem);
457 SetEntryByNumber(Value, Group, Elem);
462 * \ingroup gdcmParser
463 * \brief Modifies the value of a given Header Entry (Dicom Element)
464 * if it exists; Creates it with the given value if it doesn't
465 * @param Value passed as a char*
471 bool gdcmParser::ReplaceOrCreateByNumber(char* Value, guint16 Group, guint16 Elem ) {
472 gdcmHeaderEntry* nvHeaderEntry=NewHeaderEntryByNumber(Group, Elem);
477 AddHeaderEntry(nvHeaderEntry);
479 std::string v = Value;
480 SetEntryByNumber(v, Group, Elem);
485 * \ingroup gdcmParser
486 * \brief Set a new value if the invoked element exists
487 * Seems to be useless !!!
493 bool gdcmParser::ReplaceIfExistByNumber(char* Value, guint16 Group, guint16 Elem )
495 std::string v = Value;
496 SetEntryByNumber(v, Group, Elem);
500 //-----------------------------------------------------------------------------
503 * \ingroup gdcmParser
504 * \brief Checks if a given Dicom Element exists
505 * \ within the H table
506 * @param group Group number of the searched Dicom Element
507 * @param element Element number of the searched Dicom Element
508 * @return number of occurences
510 int gdcmParser::CheckIfEntryExistByNumber(guint16 group, guint16 element ) {
511 std::string key = gdcmDictEntry::TranslateToKey(group, element );
512 return (tagHT.count(key));
516 * \ingroup gdcmParser
517 * \brief Searches within Header Entries (Dicom Elements) parsed with
518 * the public and private dictionaries
519 * for the element value of a given tag.
520 * \warning Don't use any longer : use GetPubEntryByName
521 * @param tagName name of the searched element.
522 * @return Corresponding element value when it exists,
523 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
525 std::string gdcmParser::GetEntryByName(std::string tagName) {
526 gdcmDictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
527 if( dictEntry == NULL)
530 return(GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement()));
534 * \ingroup gdcmParser
535 * \brief Searches within Header Entries (Dicom Elements) parsed with
536 * the public and private dictionaries
537 * for the element value representation of a given tag.
539 * Obtaining the VR (Value Representation) might be needed by caller
540 * to convert the string typed content to caller's native type
541 * (think of C++ vs Python). The VR is actually of a higher level
542 * of semantics than just the native C++ type.
543 * @param tagName name of the searched element.
544 * @return Corresponding element value representation when it exists,
545 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
547 std::string gdcmParser::GetEntryVRByName(std::string tagName) {
548 gdcmDictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
549 if( dictEntry == NULL)
552 gdcmHeaderEntry* elem = GetHeaderEntryByNumber(dictEntry->GetGroup(),
553 dictEntry->GetElement());
554 return elem->GetVR();
558 * \ingroup gdcmParser
559 * \brief Searches within Header Entries (Dicom Elements) parsed with
560 * the public and private dictionaries
561 * for the element value representation of a given tag.
562 * @param group Group of the searched tag.
563 * @param element Element of the searched tag.
564 * @return Corresponding element value representation when it exists,
565 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
567 std::string gdcmParser::GetEntryByNumber(guint16 group, guint16 element){
568 TagKey key = gdcmDictEntry::TranslateToKey(group, element);
569 if ( ! tagHT.count(key))
571 return tagHT.find(key)->second->GetValue();
575 * \ingroup gdcmParser
576 * \brief Searches within Header Entries (Dicom Elements) parsed with
577 * the public and private dictionaries
578 * for the element value representation of a given tag..
580 * Obtaining the VR (Value Representation) might be needed by caller
581 * to convert the string typed content to caller's native type
582 * (think of C++ vs Python). The VR is actually of a higher level
583 * of semantics than just the native C++ type.
584 * @param group Group of the searched tag.
585 * @param element Element of the searched tag.
586 * @return Corresponding element value representation when it exists,
587 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
589 std::string gdcmParser::GetEntryVRByNumber(guint16 group, guint16 element) {
590 gdcmHeaderEntry* elem = GetHeaderEntryByNumber(group, element);
593 return elem->GetVR();
597 * \ingroup gdcmParser
598 * \brief Sets the value (string) of the Header Entry (Dicom Element)
599 * @param content string value of the Dicom Element
600 * @param tagName name of the searched Dicom Element.
601 * @return true when found
603 bool gdcmParser::SetEntryByName(std::string content,std::string tagName) {
604 gdcmDictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
605 if( dictEntry == NULL)
608 return(SetEntryByNumber(content,dictEntry->GetGroup(),
609 dictEntry->GetElement()));
613 * \ingroup gdcmParser
614 * \brief Accesses an existing gdcmHeaderEntry (i.e. a Dicom Element)
615 * through it's (group, element) and modifies it's content with
617 * \warning Don't use any longer : use SetPubEntryByNumber
618 * @param content new value to substitute with
619 * @param group group of the Dicom Element to modify
620 * @param element element of the Dicom Element to modify
622 bool gdcmParser::SetEntryByNumber(std::string content,
626 TagKey key = gdcmDictEntry::TranslateToKey(group, element);
627 if ( ! tagHT.count(key))
629 int l = content.length();
630 if(l%2) // Odd length are padded with a space (020H).
633 content = content + '\0';
638 TagHeaderEntryHT::iterator p2;
639 // DO NOT remove the following lines : they explain the stuff
640 //p= tagHT.equal_range(key); // get a pair of iterators first-last synonym
641 //p2=p.first; // iterator on the first synonym
642 //a=p2->second; // H Table target column (2-nd col)
645 a = ((tagHT.equal_range(key)).first)->second;
647 a-> SetValue(content);
649 std::string vr = a->GetVR();
652 if( (vr == "US") || (vr == "SS") )
654 else if( (vr == "UL") || (vr == "SL") )
664 * \ingroup gdcmParser
665 * \brief Accesses an existing gdcmHeaderEntry (i.e. a Dicom Element)
666 * in the PubHeaderEntrySet of this instance
667 * through it's (group, element) and modifies it's length with
669 * \warning Use with extreme caution.
670 * @param length new length to substitute with
671 * @param group group of the entry to modify
672 * @param element element of the Entry to modify
673 * @return 1 on success, 0 otherwise.
676 bool gdcmParser::SetEntryLengthByNumber(guint32 length,
680 TagKey key = gdcmDictEntry::TranslateToKey(group, element);
681 if ( ! tagHT.count(key))
683 if (length%2) length++; // length must be even
684 ( ((tagHT.equal_range(key)).first)->second )->SetLength(length);
690 * \ingroup gdcmParser
691 * \brief Gets (from Header) the offset of a 'non string' element value
692 * \ (LoadElementValues has already be executed)
695 * @return File Offset of the Element Value
697 size_t gdcmParser::GetEntryOffsetByNumber(guint16 Group, guint16 Elem)
699 gdcmHeaderEntry* Entry = GetHeaderEntryByNumber(Group, Elem);
702 dbg.Verbose(1, "gdcmParser::GetHeaderEntryByNumber",
703 "failed to Locate gdcmHeaderEntry");
706 return Entry->GetOffset();
710 * \ingroup gdcmParser
711 * \brief Gets (from Header) a 'non string' element value
712 * \ (LoadElementValues has already be executed)
715 * @return Pointer to the 'non string' area
717 void * gdcmParser::GetEntryVoidAreaByNumber(guint16 Group, guint16 Elem)
719 gdcmHeaderEntry* Entry = GetHeaderEntryByNumber(Group, Elem);
722 dbg.Verbose(1, "gdcmParser::GetHeaderEntryByNumber",
723 "failed to Locate gdcmHeaderEntry");
726 return Entry->GetVoidArea();
730 * \ingroup gdcmParser
731 * \brief Loads (from disk) the element content
732 * when a string is not suitable
736 void *gdcmParser::LoadEntryVoidArea(guint16 Group, guint16 Elem)
738 gdcmHeaderEntry * Element= GetHeaderEntryByNumber(Group, Elem);
741 size_t o =(size_t)Element->GetOffset();
742 fseek(fp, o, SEEK_SET);
743 int l=Element->GetLength();
744 void * a = malloc(l);
748 SetEntryVoidAreaByNumber(a, Group, Elem);
749 // TODO check the result
750 size_t l2 = fread(a, 1, l ,fp);
760 * \ingroup gdcmParser
761 * \brief Sets a 'non string' value to a given Dicom Element
763 * @param group Group number of the searched Dicom Element
764 * @param element Element number of the searched Dicom Element
767 bool gdcmParser::SetEntryVoidAreaByNumber(void * area,
771 TagKey key = gdcmDictEntry::TranslateToKey(group, element);
772 if ( ! tagHT.count(key))
774 ( ((tagHT.equal_range(key)).first)->second )->SetVoidArea(area);
779 * \ingroup gdcmParser
780 * \brief Update the entries with the shadow dictionary. Only odd entries are
783 void gdcmParser::UpdateShaEntries(void) {
784 gdcmDictEntry *entry;
787 for(ListTag::iterator it=listEntries.begin();
788 it!=listEntries.end();
791 // Odd group => from public dictionary
792 if((*it)->GetGroup()%2==0)
795 // Peer group => search the corresponding dict entry
797 entry=RefShaDict->GetDictEntryByNumber((*it)->GetGroup(),(*it)->GetElement());
801 if((*it)->IsImplicitVR())
806 (*it)->SetValue(GetHeaderEntryUnvalue(*it));
808 // Set the new entry and the new value
809 (*it)->SetDictEntry(entry);
810 CheckHeaderEntryVR(*it,vr);
812 (*it)->SetValue(GetHeaderEntryValue(*it));
816 // Remove precedent value transformation
817 (*it)->SetDictEntry(NewVirtualDictEntry((*it)->GetGroup(),(*it)->GetElement(),vr));
823 * \ingroup gdcmParser
824 * \brief Searches within the Header Entries for a Dicom Element of
826 * @param tagName name of the searched Dicom Element.
827 * @return Corresponding Dicom Element when it exists, and NULL
830 gdcmHeaderEntry *gdcmParser::GetHeaderEntryByName(std::string tagName) {
831 gdcmDictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
832 if( dictEntry == NULL)
835 return(GetHeaderEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement()));
839 * \ingroup gdcmParser
840 * \brief retrieves a Dicom Element (the first one) using (group, element)
841 * \ warning (group, element) IS NOT an identifier inside the Dicom Header
842 * if you think it's NOT UNIQUE, check the count number
843 * and use iterators to retrieve ALL the Dicoms Elements within
844 * a given couple (group, element)
845 * @param group Group number of the searched Dicom Element
846 * @param element Element number of the searched Dicom Element
849 gdcmHeaderEntry* gdcmParser::GetHeaderEntryByNumber(guint16 group, guint16 element)
851 TagKey key = gdcmDictEntry::TranslateToKey(group, element);
852 if ( ! tagHT.count(key))
854 return tagHT.find(key)->second;
858 * \ingroup gdcmParser
859 * \brief retrieves the Dicom Elements (all of them) using (group, element)
860 * @param group Group number of the searched Dicom Element.
861 * @param element Element number of the searched Dicom Element.
862 * @return a range (i.e.pair<,>) containing all elements whose key is group|element)
865 IterHT gdcmParser::GetHeaderEntrySameNumber(guint16 group, guint16 element){
866 TagKey key = gdcmDictEntry::TranslateToKey(group, element);
867 return (tagHT.equal_range(key));
871 * \ingroup gdcmParser
872 * \brief Loads the element while preserving the current
873 * underlying file position indicator as opposed to
874 * to LoadHeaderEntry that modifies it.
875 * @param entry Header Entry whose value shall be loaded.
878 void gdcmParser::LoadHeaderEntrySafe(gdcmHeaderEntry * entry) {
879 long PositionOnEntry = ftell(fp);
880 LoadHeaderEntry(entry);
881 fseek(fp, PositionOnEntry, SEEK_SET);
885 * \ingroup gdcmParser
886 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
887 * \warning : to be re-written using the chained list instead of the H table.
888 * \warning : DO NOT use (doesn't work any longer because of the multimap)
889 * \todo : to be re-written using the chained list instead of the H table
890 * @param SkipSequence TRUE if we don't want to write Sequences (ACR-NEMA Files)
891 * @param type Type of the File (ExplicitVR,ImplicitVR, ACR, ...)
893 void gdcmParser::UpdateGroupLength(bool SkipSequence, FileType type) {
897 gdcmHeaderEntry *elem;
899 std::string str_trash;
902 GroupHT groupHt; // to hold the length of each group
905 // typedef std::map<GroupKey, int> GroupHT;
907 gdcmHeaderEntry *elemZ;
909 // for each Tag in the DCM Header
911 for (TagHeaderEntryHT::iterator tag2 = tagHT.begin();
916 gr = elem->GetGroup();
917 el = elem->GetElement();
920 sprintf(trash, "%04x", gr);
921 key = trash; // generate 'group tag'
923 // if the caller decided not to take SEQUENCEs into account
924 // e.g : he wants to write an ACR-NEMA File
926 if (SkipSequence && vr == "SQ")
929 // Still unsolved problem :
930 // we cannot find the 'Sequence Delimitation Item'
931 // since it's at the end of the Hash Table
934 // pas SEQUENCE en ACR-NEMA
936 // --> la descente a l'interieur' des SQ
937 // devrait etre faite avec une liste chainee, pas avec une HTable...
939 if ( groupHt.count(key) == 0) // we just read the first elem of a given group
941 if (el == 0x0000) // the first elem is 0x0000
943 groupHt[key] = 0; // initialize group length
947 groupHt[key] = 2 + 2 + 4 + elem->GetLength(); // non 0x0000 first group elem
950 else // any elem but the first
952 if (type == ExplicitVR)
954 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
956 groupHt[key] += 4; // explicit VR AND OB, OW, SQ : 4 more bytes
959 groupHt[key] += 2 + 2 + 4 + elem->GetLength();
963 unsigned short int gr_bid;
965 for (GroupHT::iterator g = groupHt.begin(); // for each group we found
969 // FIXME: g++ -Wall -Wstrict-prototypes reports on following line:
970 // warning: unsigned int format, different type arg
971 sscanf(g->first.c_str(),"%x",&gr_bid);
972 tk = g->first + "|0000"; // generate the element full tag
974 if ( tagHT.count(tk) == 0) // if element 0x0000 not found
976 gdcmDictEntry * tagZ = new gdcmDictEntry(gr_bid, 0x0000, "UL");
977 elemZ = new gdcmHeaderEntry(tagZ);
979 AddHeaderEntry(elemZ); // create it
983 elemZ=GetHeaderEntryByNumber(gr_bid, 0x0000);
985 sprintf(trash ,"%d",g->second);
987 elemZ->SetValue(str_trash);
992 * \ingroup gdcmParser
993 * \brief writes on disc according to the requested format
994 * \ (ACR-NEMA, ExplicitVR, ImplicitVR) the image
995 * \ warning does NOT add the missing elements in the header :
996 * \ it's up to the user doing it !
997 * \ (function CheckHeaderCoherence to be written)
998 * \ warning DON'T try, right now, to write a DICOM image
999 * \ from an ACR Header (meta elements will be missing!)
1000 * @param type type of the File to be written
1001 * (ACR-NEMA, ExplicitVR, ImplicitVR)
1002 * @param _fp already open file pointer
1004 void gdcmParser::WriteEntries(FileType type, FILE * _fp)
1013 std::vector<std::string> tokens;
1015 // TODO : function CheckHeaderCoherence to be written
1017 // uses now listEntries to iterate, not TagHt!
1019 // pb : gdcmParser.Add does NOT update listEntries
1020 // TODO : find a trick (in STL?) to do it, at low cost !
1024 // TODO (?) tester les echecs en ecriture (apres chaque fwrite)
1027 for (ListTag::iterator tag2=listEntries.begin();
1028 tag2 != listEntries.end();
1031 gr = (*tag2)->GetGroup();
1032 el = (*tag2)->GetElement();
1033 lgr = (*tag2)->GetReadLength();
1034 val = (*tag2)->GetValue().c_str();
1035 vr = (*tag2)->GetVR();
1038 if (gr < 0x0008) continue; // ignore pure DICOM V3 groups
1039 if (gr %2) continue; // ignore shadow groups
1040 if (vr == "SQ" ) continue; // ignore Sequences
1041 // TODO : find a trick to *skip* the SeQuences !
1042 // Not only ignore the SQ element
1043 if (gr == 0xfffe ) continue; // ignore delimiters
1046 fwrite ( &gr,(size_t)2 ,(size_t)1 ,_fp); //group
1047 fwrite ( &el,(size_t)2 ,(size_t)1 ,_fp); //element
1049 // === Deal with the length
1050 // --------------------
1052 // if ( (type == ExplicitVR) && (gr <= 0x0002) ) // ?!? < 2
1053 if ( (type == ExplicitVR) || (type == DICOMDIR) )
1056 guint16 z=0, shortLgr;
1057 if (vr == "Unknown") { // Unknown was 'written'
1059 fwrite ( &shortLgr,(size_t)2 ,(size_t)1 ,_fp);
1060 fwrite ( &z, (size_t)2 ,(size_t)1 ,_fp);
1062 if (gr != 0xfffe) { // NO value for 'delimiters'
1063 if (vr == "Unknown") // Unknown was 'written'
1064 fwrite(&z,(size_t)2 ,(size_t)1 ,_fp);
1066 fwrite (vr.c_str(),(size_t)2 ,(size_t)1 ,_fp);
1069 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") || gr == 0xfffe) // JPR
1072 fwrite ( &z, (size_t)2 ,(size_t)1 ,_fp);
1073 fwrite ( &lgr,(size_t)4 ,(size_t)1 ,_fp);
1078 fwrite ( &shortLgr,(size_t)2 ,(size_t)1 ,_fp);
1084 fwrite ( &lgr,(size_t)4 ,(size_t)1 ,_fp);
1087 // === Deal with the value
1088 // -------------------
1089 if (vr == "SQ") continue; // no "value" to write for the SEQuences
1090 if (gr == 0xfffe)continue;
1092 if (vr == "US" || vr == "SS")
1094 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1095 Tokenize ((*tag2)->GetValue(), tokens, "\\");
1096 for (unsigned int i=0; i<tokens.size();i++)
1098 val_uint16 = atoi(tokens[i].c_str());
1100 fwrite ( ptr,(size_t)2 ,(size_t)1 ,_fp);
1105 if (vr == "UL" || vr == "SL")
1107 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1108 Tokenize ((*tag2)->GetValue(), tokens, "\\");
1109 for (unsigned int i=0; i<tokens.size();i++)
1111 val_uint32 = atoi(tokens[i].c_str());
1113 fwrite ( ptr,(size_t)4 ,(size_t)1 ,_fp);
1119 // Pixels are never loaded in the element !
1120 // we stop writting when Pixel are processed
1121 // FIX : we loose trailing elements (RAB, right now)
1123 if ((gr == GrPixel) && (el == NumPixel) ) {
1125 if (compte == countGrPixel) // we passed *all* the GrPixel,NumPixel
1128 fwrite ( val,(size_t)lgr ,(size_t)1 ,_fp); // Elem value
1133 * \ingroup gdcmParser
1134 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1136 * @return The properly swaped 32 bits integer.
1138 guint32 gdcmParser::SwapLong(guint32 a) {
1143 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1144 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1148 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1152 a=( ((a<<8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1155 dbg.Error(" gdcmParser::SwapLong : unset swap code");
1162 * \ingroup gdcmParser
1163 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1165 * @return The properly unswaped 32 bits integer.
1167 guint32 gdcmParser::UnswapLong(guint32 a) {
1168 return (SwapLong(a));
1172 * \ingroup gdcmParser
1173 * \brief Swaps the bytes so they agree with the processor order
1174 * @return The properly swaped 16 bits integer.
1176 guint16 gdcmParser::SwapShort(guint16 a) {
1177 if ( (sw==4321) || (sw==2143) )
1178 a =(((a<<8) & 0x0ff00) | ((a>>8)&0x00ff));
1183 * \ingroup gdcmParser
1184 * \brief Unswaps the bytes so they agree with the processor order
1185 * @return The properly unswaped 16 bits integer.
1187 guint16 gdcmParser::UnswapShort(guint16 a) {
1188 return (SwapShort(a));
1191 //-----------------------------------------------------------------------------
1194 * \ingroup gdcmParser
1195 * \brief Parses the header of the file but WITHOUT loading element values.
1196 * @return false if file is not ACR-NEMA / DICOM
1198 bool gdcmParser::ParseHeader(bool exception_on_error) throw(gdcmFormatError) {
1204 gdcmHeaderEntry *newHeaderEntry = (gdcmHeaderEntry *)0;
1205 while ( (newHeaderEntry = ReadNextHeaderEntry()) ) {
1206 SkipHeaderEntry(newHeaderEntry);
1207 if ( (ignoreShadow==0) || (newHeaderEntry->GetGroup()%2) == 0) {
1208 AddHeaderEntry(newHeaderEntry);
1215 * \ingroup gdcmParser
1216 * \brief Loads the element values of all the Header Entries pointed in the
1217 * public Chained List.
1219 void gdcmParser::LoadHeaderEntries(void) {
1221 for (ListTag::iterator i = GetListEntry().begin();
1222 i != GetListEntry().end();
1225 LoadHeaderEntry(*i);
1230 // Load 'non string' values
1231 std::string PhotometricInterpretation = GetEntryByNumber(0x0028,0x0004);
1232 if( PhotometricInterpretation == "PALETTE COLOR " ) {
1233 LoadEntryVoidArea(0x0028,0x1200); // gray LUT
1234 LoadEntryVoidArea(0x0028,0x1201); // R LUT
1235 LoadEntryVoidArea(0x0028,0x1202); // G LUT
1236 LoadEntryVoidArea(0x0028,0x1203); // B LUT
1238 LoadEntryVoidArea(0x0028,0x1221); // Segmented Red Palette Color LUT Data
1239 LoadEntryVoidArea(0x0028,0x1222); // Segmented Green Palette Color LUT Data
1240 LoadEntryVoidArea(0x0028,0x1223); // Segmented Blue Palette Color LUT Data
1243 // --------------------------------------------------------------
1244 // Special Patch to allow gdcm to read ACR-LibIDO formated images
1246 // if recognition code tells us we deal with a LibIDO image
1247 // we switch lineNumber and columnNumber
1249 std::string RecCode;
1250 RecCode = GetEntryByNumber(0x0008, 0x0010); // recognition code
1251 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
1252 RecCode == "CANRME_AILIBOD1_1." )
1254 filetype = ACR_LIBIDO;
1255 std::string rows = GetEntryByNumber(0x0028, 0x0010);
1256 std::string columns = GetEntryByNumber(0x0028, 0x0011);
1257 SetEntryByNumber(columns, 0x0028, 0x0010);
1258 SetEntryByNumber(rows , 0x0028, 0x0011);
1260 // ----------------- End of Special Patch ----------------
1264 * \ingroup gdcmParser
1265 * \brief Loads the element content if its length doesn't exceed
1266 * the value specified with gdcmParser::SetMaxSizeLoadEntry()
1267 * @param Entry Header Entry (Dicom Element) to be dealt with
1269 void gdcmParser::LoadHeaderEntry(gdcmHeaderEntry *Entry) {
1271 guint16 group = Entry->GetGroup();
1272 std::string vr= Entry->GetVR();
1273 guint32 length = Entry->GetLength();
1274 bool SkipLoad = false;
1276 fseek(fp, (long)Entry->GetOffset(), SEEK_SET);
1278 // the test was commented out to 'go inside' the SeQuences
1279 // we don't any longer skip them !
1281 // if( vr == "SQ" ) // (DO NOT remove this comment)
1284 // A SeQuence "contains" a set of Elements.
1285 // (fffe e000) tells us an Element is beginning
1286 // (fffe e00d) tells us an Element just ended
1287 // (fffe e0dd) tells us the current SeQuence just ended
1288 if( group == 0xfffe )
1292 Entry->SetLength(0);
1293 Entry->SetValue("gdcm::Skipped");
1297 // When the length is zero things are easy:
1298 if ( length == 0 ) {
1299 Entry->SetValue("");
1303 // The elements whose length is bigger than the specified upper bound
1304 // are not loaded. Instead we leave a short notice of the offset of
1305 // the element content and it's length.
1306 if (length > MaxSizeLoadEntry) {
1307 std::ostringstream s;
1308 s << "gdcm::NotLoaded.";
1309 s << " Address:" << (long)Entry->GetOffset();
1310 s << " Length:" << Entry->GetLength();
1311 s << " x(" << std::hex << Entry->GetLength() << ")";
1312 Entry->SetValue(s.str());
1316 // When integer(s) are expected, read and convert the following
1317 // n *(two or four bytes)
1318 // properly i.e. as integers as opposed to strings.
1319 // Elements with Value Multiplicity > 1
1320 // contain a set of integers (not a single one)
1322 // Any compacter code suggested (?)
1323 if ( IsHeaderEntryAnInteger(Entry) ) {
1325 std::ostringstream s;
1327 if (vr == "US" || vr == "SS") {
1329 NewInt = ReadInt16();
1332 for (int i=1; i < nbInt; i++) {
1334 NewInt = ReadInt16();
1339 else if (vr == "UL" || vr == "SL") {
1341 NewInt = ReadInt32();
1344 for (int i=1; i < nbInt; i++) {
1346 NewInt = ReadInt32();
1351 #ifdef GDCM_NO_ANSI_STRING_STREAM
1352 s << std::ends; // to avoid oddities on Solaris
1353 #endif //GDCM_NO_ANSI_STRING_STREAM
1355 Entry->SetValue(s.str());
1359 // We need an additional byte for storing \0 that is not on disk
1360 std::string NewValue(length,0);
1361 item_read = fread(&(NewValue[0]), (size_t)length, (size_t)1, fp);
1362 if ( item_read != 1 ) {
1363 dbg.Verbose(1, "gdcmParser::LoadElementValue","unread element value");
1364 Entry->SetValue("gdcm::UnRead");
1368 if( (vr == "UI") ) // Because of correspondance with the VR dic
1369 Entry->SetValue(NewValue.c_str()); // ??? JPR ???
1371 Entry->SetValue(NewValue);
1375 * \ingroup gdcmParser
1376 * \brief add a new Dicom Element pointer to
1377 * the H Table and to the chained List
1378 * \warning push_bash in listEntries ONLY during ParseHeader
1379 * \TODO something to allow further Elements addition,
1380 * \ when position to be taken care of
1381 * @param newHeaderEntry
1383 void gdcmParser::AddHeaderEntry(gdcmHeaderEntry *newHeaderEntry) {
1384 tagHT.insert( PairHT( newHeaderEntry->GetKey(),newHeaderEntry) );
1385 listEntries.push_back(newHeaderEntry);
1390 * \ingroup gdcmParser
1392 * @param Entry Header Entry whose length of the value shall be loaded.
1396 void gdcmParser::FindHeaderEntryLength (gdcmHeaderEntry *Entry) {
1397 guint16 element = Entry->GetElement();
1398 guint16 group = Entry->GetGroup();
1399 std::string vr = Entry->GetVR();
1401 if( (element == NumPixel) && (group == GrPixel) )
1404 dbg.Verbose(2, "gdcmParser::FindLength: ",
1405 "we reached (GrPixel,NumPixel)");
1408 if ( (filetype == ExplicitVR) && (! Entry->IsImplicitVR()) )
1410 if ( (vr=="OB") || (vr=="OW") || (vr=="SQ") || (vr=="UN") )
1412 // The following reserved two bytes (see PS 3.5-2001, section
1413 // 7.1.2 Data element structure with explicit vr p27) must be
1414 // skipped before proceeding on reading the length on 4 bytes.
1415 fseek(fp, 2L, SEEK_CUR);
1416 guint32 length32 = ReadInt32();
1418 if ( (vr == "OB") && (length32 == 0xffffffff) )
1420 Entry->SetLength(FindHeaderEntryLengthOB());
1423 FixHeaderEntryFoundLength(Entry, length32);
1427 // Length is encoded on 2 bytes.
1428 length16 = ReadInt16();
1430 // We can tell the current file is encoded in big endian (like
1431 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1432 // and it's value is the one of the encoding of a big endian file.
1433 // In order to deal with such big endian encoded files, we have
1434 // (at least) two strategies:
1435 // * when we load the "Transfer Syntax" tag with value of big endian
1436 // encoding, we raise the proper flags. Then we wait for the end
1437 // of the META group (0x0002) among which is "Transfer Syntax",
1438 // before switching the swap code to big endian. We have to postpone
1439 // the switching of the swap code since the META group is fully encoded
1440 // in little endian, and big endian coding only starts at the next
1441 // group. The corresponding code can be hard to analyse and adds
1442 // many additional unnecessary tests for regular tags.
1443 // * the second strategy consists in waiting for trouble, that shall
1444 // appear when we find the first group with big endian encoding. This
1445 // is easy to detect since the length of a "Group Length" tag (the
1446 // ones with zero as element number) has to be of 4 (0x0004). When we
1447 // encounter 1024 (0x0400) chances are the encoding changed and we
1448 // found a group with big endian encoding.
1449 // We shall use this second strategy. In order to make sure that we
1450 // can interpret the presence of an apparently big endian encoded
1451 // length of a "Group Length" without committing a big mistake, we
1452 // add an additional check: we look in the already parsed elements
1453 // for the presence of a "Transfer Syntax" whose value has to be "big
1454 // endian encoding". When this is the case, chances are we have got our
1455 // hands on a big endian encoded file: we switch the swap code to
1456 // big endian and proceed...
1457 if ( (element == 0x0000) && (length16 == 0x0400) )
1459 if ( ! IsExplicitVRBigEndianTransferSyntax() )
1461 dbg.Verbose(0, "gdcmParser::FindLength", "not explicit VR");
1466 SwitchSwapToBigEndian();
1467 // Restore the unproperly loaded values i.e. the group, the element
1468 // and the dictionary entry depending on them.
1469 guint16 CorrectGroup = SwapShort(Entry->GetGroup());
1470 guint16 CorrectElem = SwapShort(Entry->GetElement());
1471 gdcmDictEntry * NewTag = GetDictEntryByNumber(CorrectGroup,
1475 // This correct tag is not in the dictionary. Create a new one.
1476 NewTag = NewVirtualDictEntry(CorrectGroup, CorrectElem);
1478 // FIXME this can create a memory leaks on the old entry that be
1479 // left unreferenced.
1480 Entry->SetDictEntry(NewTag);
1483 // Heuristic: well some files are really ill-formed.
1484 if ( length16 == 0xffff)
1487 //dbg.Verbose(0, "gdcmParser::FindLength",
1488 // "Erroneous element length fixed.");
1489 // Actually, length= 0xffff means that we deal with
1490 // Unknown Sequence Length
1493 FixHeaderEntryFoundLength(Entry, (guint32)length16);
1498 // Either implicit VR or a non DICOM conformal (see not below) explicit
1499 // VR that ommited the VR of (at least) this element. Farts happen.
1500 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1501 // on Data elements "Implicit and Explicit VR Data Elements shall
1502 // not coexist in a Data Set and Data Sets nested within it".]
1503 // Length is on 4 bytes.
1504 FixHeaderEntryFoundLength(Entry, ReadInt32());
1510 * \ingroup gdcmParser
1511 * \brief Find the Value Representation of the current Dicom Element.
1514 void gdcmParser::FindHeaderEntryVR( gdcmHeaderEntry *Entry)
1516 if (filetype != ExplicitVR)
1522 long PositionOnEntry = ftell(fp);
1523 // Warning: we believe this is explicit VR (Value Representation) because
1524 // we used a heuristic that found "UL" in the first tag. Alas this
1525 // doesn't guarantee that all the tags will be in explicit VR. In some
1526 // cases (see e-film filtered files) one finds implicit VR tags mixed
1527 // within an explicit VR file. Hence we make sure the present tag
1528 // is in explicit VR and try to fix things if it happens not to be
1531 lgrLue=fread (&VR, (size_t)2,(size_t)1, fp);
1533 if(!CheckHeaderEntryVR(Entry,VR))
1535 fseek(fp, PositionOnEntry, SEEK_SET);
1536 // When this element is known in the dictionary we shall use, e.g. for
1537 // the semantics (see the usage of IsAnInteger), the VR proposed by the
1538 // dictionary entry. Still we have to flag the element as implicit since
1539 // we know now our assumption on expliciteness is not furfilled.
1541 if ( Entry->IsVRUnknown() )
1542 Entry->SetVR("Implicit");
1543 Entry->SetImplicitVR();
1548 * \ingroup gdcmParser
1549 * \brief Check the correspondance between the VR of the header entry
1550 * and the taken VR. If they are different, the header entry is
1551 * updated with the new VR.
1554 * @return false if the VR is incorrect of if the VR isn't referenced
1555 * otherwise, it returns true
1557 bool gdcmParser::CheckHeaderEntryVR(gdcmHeaderEntry *Entry, VRKey vr)
1559 char msg[100]; // for sprintf
1560 bool RealExplicit = true;
1562 // Assume we are reading a falsely explicit VR file i.e. we reached
1563 // a tag where we expect reading a VR but are in fact we read the
1564 // first to bytes of the length. Then we will interogate (through find)
1565 // the dicom_vr dictionary with oddities like "\004\0" which crashes
1566 // both GCC and VC++ implementations of the STL map. Hence when the
1567 // expected VR read happens to be non-ascii characters we consider
1568 // we hit falsely explicit VR tag.
1570 if ( (!isalpha(vr[0])) && (!isalpha(vr[1])) )
1571 RealExplicit = false;
1573 // CLEANME searching the dicom_vr at each occurence is expensive.
1574 // PostPone this test in an optional integrity check at the end
1575 // of parsing or only in debug mode.
1576 if ( RealExplicit && !gdcmGlobal::GetVR()->Count(vr) )
1577 RealExplicit= false;
1579 if ( !RealExplicit )
1581 // We thought this was explicit VR, but we end up with an
1582 // implicit VR tag. Let's backtrack.
1583 sprintf(msg,"Falsely explicit vr file (%04x,%04x)\n",
1584 Entry->GetGroup(),Entry->GetElement());
1585 dbg.Verbose(1, "gdcmParser::FindVR: ",msg);
1590 if ( Entry->IsVRUnknown() )
1592 // When not a dictionary entry, we can safely overwrite the VR.
1595 else if ( Entry->GetVR() != vr )
1597 // The VR present in the file and the dictionary disagree. We assume
1598 // the file writer knew best and use the VR of the file. Since it would
1599 // be unwise to overwrite the VR of a dictionary (since it would
1600 // compromise it's next user), we need to clone the actual DictEntry
1601 // and change the VR for the read one.
1602 gdcmDictEntry* NewEntry = NewVirtualDictEntry(
1603 Entry->GetGroup(),Entry->GetElement(),
1604 vr,"FIXME",Entry->GetName());
1605 Entry->SetDictEntry(NewEntry);
1611 * \ingroup gdcmParser
1612 * \brief Get the transformed value of the header entry. The VR value
1613 * is used to define the transformation to operate on the value
1614 * \warning NOT end user intended method !
1616 * @return Transformed entry value
1618 std::string gdcmParser::GetHeaderEntryValue(gdcmHeaderEntry *Entry)
1620 if ( (IsHeaderEntryAnInteger(Entry)) && (Entry->IsImplicitVR()) )
1622 std::string val=Entry->GetValue();
1623 std::string vr=Entry->GetVR();
1624 guint32 length = Entry->GetLength();
1625 std::ostringstream s;
1628 if (vr == "US" || vr == "SS")
1633 for (int i=0; i < nbInt; i++)
1637 NewInt16 = (val[2*i+0]&0xFF)+((val[2*i+1]&0xFF)<<8);
1638 NewInt16 = SwapShort(NewInt16);
1643 else if (vr == "UL" || vr == "SL")
1648 for (int i=0; i < nbInt; i++)
1652 NewInt32= (val[4*i+0]&0xFF)+((val[4*i+1]&0xFF)<<8)+
1653 ((val[4*i+2]&0xFF)<<16)+((val[4*i+3]&0xFF)<<24);
1654 NewInt32=SwapLong(NewInt32);
1659 #ifdef GDCM_NO_ANSI_STRING_STREAM
1660 s << std::ends; // to avoid oddities on Solaris
1661 #endif //GDCM_NO_ANSI_STRING_STREAM
1665 return(Entry->GetValue());
1669 * \ingroup gdcmParser
1670 * \brief Get the reverse transformed value of the header entry. The VR
1671 * value is used to define the reverse transformation to operate on
1673 * \warning NOT end user intended method !
1675 * @return Reverse transformed entry value
1677 std::string gdcmParser::GetHeaderEntryUnvalue(gdcmHeaderEntry *Entry)
1679 if ( (IsHeaderEntryAnInteger(Entry)) && (Entry->IsImplicitVR()) )
1681 std::string vr=Entry->GetVR();
1682 std::ostringstream s;
1683 std::vector<std::string> tokens;
1685 if (vr == "US" || vr == "SS")
1689 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1690 Tokenize (Entry->GetValue(), tokens, "\\");
1691 for (unsigned int i=0; i<tokens.size();i++)
1693 NewInt16 = atoi(tokens[i].c_str());
1694 s<<(NewInt16&0xFF)<<((NewInt16>>8)&0xFF);
1698 if (vr == "UL" || vr == "SL")
1702 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1703 Tokenize (Entry->GetValue(), tokens, "\\");
1704 for (unsigned int i=0; i<tokens.size();i++)
1706 NewInt32 = atoi(tokens[i].c_str());
1707 s<<(char)(NewInt32&0xFF)<<(char)((NewInt32>>8)&0xFF)
1708 <<(char)((NewInt32>>16)&0xFF)<<(char)((NewInt32>>24)&0xFF);
1713 #ifdef GDCM_NO_ANSI_STRING_STREAM
1714 s << std::ends; // to avoid oddities on Solaris
1715 #endif //GDCM_NO_ANSI_STRING_STREAM
1719 return(Entry->GetValue());
1723 * \ingroup gdcmParser
1724 * \brief Skip a given Header Entry
1725 * \warning NOT end user intended method !
1728 void gdcmParser::SkipHeaderEntry(gdcmHeaderEntry *entry)
1730 SkipBytes(entry->GetLength());
1734 * \ingroup gdcmParser
1735 * \brief When the length of an element value is obviously wrong (because
1736 * the parser went Jabberwocky) one can hope improving things by
1737 * applying this heuristic.
1739 void gdcmParser::FixHeaderEntryFoundLength(gdcmHeaderEntry *Entry, guint32 FoundLength)
1741 Entry->SetReadLength(FoundLength); // will be updated only if a bug is found
1743 if ( FoundLength == 0xffffffff) {
1747 guint16 gr =Entry->GetGroup();
1748 guint16 el =Entry->GetElement();
1750 if (FoundLength%2) {
1751 std::cout << "Warning : Tag with uneven length " << FoundLength
1752 << "in x(" << hex << gr << "," << el <<")" << std::endl;
1755 // Sorry for the patch!
1756 // XMedCom did the trick to read some nasty GE images ...
1757 if (FoundLength == 13) {
1758 // The following 'if' will be removed when there is no more
1759 // images on Creatis HDs with a 13 length for Manufacturer...
1760 if ( (Entry->GetGroup() != 0x0008) ||
1761 ( (Entry->GetElement() != 0x0070) && (Entry->GetElement() != 0x0080) ) ){
1762 // end of remove area
1764 Entry->SetReadLength(10); // a bug is to be fixed
1768 // to fix some garbage 'Leonardo' Siemens images
1769 // May be commented out to avoid overhead
1770 else if ( (Entry->GetGroup() == 0x0009) &&
1771 ( (Entry->GetElement() == 0x1113) || (Entry->GetElement() == 0x1114) ) ){
1773 Entry->SetReadLength(4); // a bug is to be fixed
1777 // to try to 'go inside' SeQuences (with length), and not to skip them
1778 else if ( Entry->GetVR() == "SQ")
1780 if (enableSequences) // only if the user does want to !
1781 FoundLength =0; // ReadLength is unchanged
1784 // a SeQuence Element is beginning
1785 // Let's forget it's length
1786 // (we want to 'go inside')
1788 // Pb : *normaly* fffe|e000 is just a marker, its length *should be* zero
1789 // in gdcm-MR-PHILIPS-16-Multi-Seq.dcm we find lengthes as big as 28800
1790 // if we set the length to zero IsHeaderEntryAnInteger() breaks...
1791 // if we don't, we lost 28800 characters from the Header :-(
1793 else if(Entry->GetGroup() == 0xfffe)
1795 // sometimes, length seems to be wrong
1796 FoundLength =0; // some more clever checking to be done !
1798 // only gdcm-MR-PHILIPS-16-Multi-Seq.dcm
1799 // causes troubles :-(
1802 Entry->SetUsableLength(FoundLength);
1806 * \ingroup gdcmParser
1807 * \brief Apply some heuristics to predict wether the considered
1808 * element value contains/represents an integer or not.
1809 * @param Entry The element value on which to apply the predicate.
1810 * @return The result of the heuristical predicate.
1812 bool gdcmParser::IsHeaderEntryAnInteger(gdcmHeaderEntry *Entry) {
1813 guint16 element = Entry->GetElement();
1814 guint16 group = Entry->GetGroup();
1815 std::string vr = Entry->GetVR();
1816 guint32 length = Entry->GetLength();
1817 // When we have some semantics on the element we just read, and if we
1818 // a priori know we are dealing with an integer, then we shall be
1819 // able to swap it's element value properly.
1820 if ( element == 0 ) // This is the group length of the group
1826 std::ostringstream s;
1827 int filePosition = ftell(fp);
1828 s << "Erroneous Group Length element length on : (" \
1829 << std::hex << group << " , " << element
1830 << ") -before- position x(" << filePosition << ")"
1831 << "lgt : " << length;
1832 // These 2 lines commented out : a *very dirty* patch
1833 // to go on PrintHeader'ing gdcm-MR-PHILIPS-16-Multi-Seq.dcm.
1834 // have a glance at offset x(8336) ...
1835 // For *regular* headers, the test is useless..
1836 // lets's print a warning message and go on,
1837 // instead of giving up with an error message
1838 std::cout << s.str().c_str() << std::endl;
1839 // dbg.Error("gdcmParser::IsHeaderEntryAnInteger",
1840 // s.str().c_str());
1843 if ( (vr == "UL") || (vr == "US") || (vr == "SL") || (vr == "SS") )
1850 * \ingroup gdcmParser
1855 guint32 gdcmParser::FindHeaderEntryLengthOB(void) {
1856 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1859 long PositionOnEntry = ftell(fp);
1860 bool FoundSequenceDelimiter = false;
1861 guint32 TotalLength = 0;
1864 while ( ! FoundSequenceDelimiter)
1870 TotalLength += 4; // We even have to decount the group and element
1872 if ( g != 0xfffe && g!=0xb00c ) /*for bogus header */
1874 char msg[100]; // for sprintf. Sorry
1875 sprintf(msg,"wrong group (%04x) for an item sequence (%04x,%04x)\n",g, g,n);
1876 dbg.Verbose(1, "gdcmParser::FindLengthOB: ",msg);
1880 if ( n == 0xe0dd || ( g==0xb00c && n==0x0eb6 ) ) /* for bogus header */
1881 FoundSequenceDelimiter = true;
1882 else if ( n != 0xe000 )
1884 char msg[100]; // for sprintf. Sorry
1885 sprintf(msg,"wrong element (%04x) for an item sequence (%04x,%04x)\n",
1887 dbg.Verbose(1, "gdcmParser::FindLengthOB: ",msg);
1891 ItemLength = ReadInt32();
1892 TotalLength += ItemLength + 4; // We add 4 bytes since we just read
1893 // the ItemLength with ReadInt32
1894 SkipBytes(ItemLength);
1896 fseek(fp, PositionOnEntry, SEEK_SET);
1901 * \ingroup gdcmParser
1902 * \brief Reads a supposed to be 16 Bits integer
1903 * \ (swaps it depending on processor endianity)
1905 * @return read value
1907 guint16 gdcmParser::ReadInt16(void) {
1910 item_read = fread (&g, (size_t)2,(size_t)1, fp);
1911 if ( item_read != 1 ) {
1913 dbg.Verbose(0, "gdcmParser::ReadInt16", " File Error");
1923 * \ingroup gdcmParser
1924 * \brief Reads a supposed to be 32 Bits integer
1925 * \ (swaps it depending on processor endianity)
1927 * @return read value
1929 guint32 gdcmParser::ReadInt32(void) {
1932 item_read = fread (&g, (size_t)4,(size_t)1, fp);
1933 if ( item_read != 1 ) {
1935 dbg.Verbose(0, "gdcmParser::ReadInt32", " File Error");
1945 * \ingroup gdcmParser
1950 void gdcmParser::SkipBytes(guint32 NBytes) {
1951 //FIXME don't dump the returned value
1952 (void)fseek(fp, (long)NBytes, SEEK_CUR);
1956 * \ingroup gdcmParser
1959 void gdcmParser::Initialise(void)
1961 RefPubDict = gdcmGlobal::GetDicts()->GetDefaultPubDict();
1962 RefShaDict = (gdcmDict*)0;
1966 * \ingroup gdcmParser
1967 * \brief Discover what the swap code is (among little endian, big endian,
1968 * bad little endian, bad big endian).
1970 * @return false when we are absolutely sure
1971 * it's neither ACR-NEMA nor DICOM
1972 * true when we hope ours assuptions are OK
1974 bool gdcmParser::CheckSwap() {
1976 // The only guaranted way of finding the swap code is to find a
1977 // group tag since we know it's length has to be of four bytes i.e.
1978 // 0x00000004. Finding the swap code in then straigthforward. Trouble
1979 // occurs when we can't find such group...
1981 guint32 x=4; // x : for ntohs
1982 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
1988 char deb[HEADER_LENGTH_TO_READ];
1990 // First, compare HostByteOrder and NetworkByteOrder in order to
1991 // determine if we shall need to swap bytes (i.e. the Endian type).
1997 // The easiest case is the one of a DICOM header, since it possesses a
1998 // file preamble where it suffice to look for the string "DICM".
1999 lgrLue = fread(deb, 1, HEADER_LENGTH_TO_READ, fp);
2002 if(memcmp(entCur, "DICM", (size_t)4) == 0) {
2003 dbg.Verbose(1, "gdcmParser::CheckSwap:", "looks like DICOM Version3");
2005 // Next, determine the value representation (VR). Let's skip to the
2006 // first element (0002, 0000) and check there if we find "UL"
2007 // - or "OB" if the 1st one is (0002,0001) -,
2008 // in which case we (almost) know it is explicit VR.
2009 // WARNING: if it happens to be implicit VR then what we will read
2010 // is the length of the group. If this ascii representation of this
2011 // length happens to be "UL" then we shall believe it is explicit VR.
2012 // FIXME: in order to fix the above warning, we could read the next
2013 // element value (or a couple of elements values) in order to make
2014 // sure we are not commiting a big mistake.
2015 // We need to skip :
2016 // * the 128 bytes of File Preamble (often padded with zeroes),
2017 // * the 4 bytes of "DICM" string,
2018 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2019 // i.e. a total of 136 bytes.
2022 // Use gdcmParser::dicom_vr to test all the possibilities
2023 // instead of just checking for UL, OB and UI !?
2026 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2027 // but elem 0002,0010 (Transfert Syntax) tells us the file is *Implicit* VR.
2030 // The following test is *absolutely useless*, since everything *goes right*
2031 // with a *100 % wrong* assumption !!!
2033 if( (memcmp(entCur, "UL", (size_t)2) == 0) ||
2034 (memcmp(entCur, "OB", (size_t)2) == 0) ||
2035 (memcmp(entCur, "UI", (size_t)2) == 0) )
2037 filetype = ExplicitVR;
2038 dbg.Verbose(1, "gdcmParser::CheckSwap:",
2039 "explicit Value Representation");
2043 filetype = ImplicitVR;
2044 dbg.Verbose(1, "gdcmParser::CheckSwap:",
2045 "not an explicit Value Representation");
2050 dbg.Verbose(1, "gdcmParser::CheckSwap:",
2051 "HostByteOrder != NetworkByteOrder");
2056 dbg.Verbose(1, "gdcmParser::CheckSwap:",
2057 "HostByteOrder = NetworkByteOrder");
2060 // Position the file position indicator at first tag (i.e.
2061 // after the file preamble and the "DICM" string).
2063 fseek (fp, 132L, SEEK_SET);
2067 // Alas, this is not a DicomV3 file and whatever happens there is no file
2068 // preamble. We can reset the file position indicator to where the data
2069 // is (i.e. the beginning of the file).
2070 dbg.Verbose(1, "gdcmParser::CheckSwap:", "not a DICOM Version3 file");
2073 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2074 // By clean we mean that the length of the first tag is written down.
2075 // If this is the case and since the length of the first group HAS to be
2076 // four (bytes), then determining the proper swap code is straightforward.
2079 // We assume the array of char we are considering contains the binary
2080 // representation of a 32 bits integer. Hence the following dirty
2082 s32 = *((guint32 *)(entCur));
2103 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2104 // It is time for despaired wild guesses.
2105 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2106 // i.e. the 'group length' element is not present :
2108 // check the supposed to be 'group number'
2109 // 0x0002 or 0x0004 or 0x0008
2110 // to determine ' sw' value .
2111 // Only 0 or 4321 will be possible
2112 // (no oportunity to check for the formerly well known
2113 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2114 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -4, 8-)
2115 // the file IS NOT ACR-NEMA nor DICOM V3
2116 // Find a trick to tell it the caller...
2118 s16 = *((guint16 *)(deb));
2134 dbg.Verbose(0, "gdcmParser::CheckSwap:",
2135 "ACR/NEMA unfound swap info (Really hopeless !)");
2140 // Then the only info we have is the net2host one.
2150 * \ingroup gdcmParser
2153 void gdcmParser::SwitchSwapToBigEndian(void)
2155 dbg.Verbose(1, "gdcmParser::SwitchSwapToBigEndian",
2156 "Switching to BigEndian mode.");
2177 * \ingroup gdcmParser
2181 void gdcmParser::SetMaxSizeLoadEntry(long NewSize)
2185 if ((guint32)NewSize >= (guint32)0xffffffff)
2187 MaxSizeLoadEntry = 0xffffffff;
2190 MaxSizeLoadEntry = NewSize;
2195 * \ingroup gdcmParser
2197 * \warning TODO : not yet usable
2198 * (see MAX_SIZE_PRINT_ELEMENT_VALUE
2199 * in gdcmHeaderEntry gdcmLoadEntry)
2203 void gdcmParser::SetMaxSizePrintEntry(long NewSize)
2207 if ((guint32)NewSize >= (guint32)0xffffffff)
2209 MaxSizePrintEntry = 0xffffffff;
2212 MaxSizePrintEntry = NewSize;
2216 * \ingroup gdcmParser
2217 * \brief Searches both the public and the shadow dictionary (when they
2218 * exist) for the presence of the DictEntry with given name.
2219 * The public dictionary has precedence on the shadow one.
2220 * @param Name name of the searched DictEntry
2221 * @return Corresponding DictEntry when it exists, NULL otherwise.
2223 gdcmDictEntry *gdcmParser::GetDictEntryByName(std::string Name)
2225 gdcmDictEntry *found = (gdcmDictEntry *)0;
2226 if (!RefPubDict && !RefShaDict)
2228 dbg.Verbose(0, "gdcmParser::GetDictEntry",
2229 "we SHOULD have a default dictionary");
2233 found = RefPubDict->GetDictEntryByName(Name);
2239 found = RefShaDict->GetDictEntryByName(Name);
2247 * \ingroup gdcmParser
2248 * \brief Searches both the public and the shadow dictionary (when they
2249 * exist) for the presence of the DictEntry with given
2250 * group and element. The public dictionary has precedence on the
2252 * @param group group of the searched DictEntry
2253 * @param element element of the searched DictEntry
2254 * @return Corresponding DictEntry when it exists, NULL otherwise.
2256 gdcmDictEntry *gdcmParser::GetDictEntryByNumber(guint16 group,guint16 element)
2258 gdcmDictEntry *found = (gdcmDictEntry *)0;
2259 if (!RefPubDict && !RefShaDict)
2261 dbg.Verbose(0, "gdcmParser::GetDictEntry",
2262 "we SHOULD have a default dictionary");
2266 found = RefPubDict->GetDictEntryByNumber(group, element);
2272 found = RefShaDict->GetDictEntryByNumber(group, element);
2280 * \ingroup gdcmParser
2281 * \brief Read the next tag but WITHOUT loading it's value
2282 * @return On succes the newly created HeaderEntry, NULL on failure.
2284 gdcmHeaderEntry *gdcmParser::ReadNextHeaderEntry(void) {
2286 gdcmHeaderEntry *NewEntry;
2292 // We reached the EOF (or an error occured) therefore
2293 // header parsing has to be considered as finished.
2294 return (gdcmHeaderEntry *)0;
2296 /* Pb : how to propagate the element length (used in SkipHeaderEntry)
2297 // direct call to SkipBytes ?
2299 if (ignoreShadow == 1 && g%2 ==1) //JPR
2300 // if user wants to skip shadow groups
2301 // and current element *is* a shadow element
2302 // we don't create anything
2303 return (gdcmHeaderEntry *)1; // to tell caller it's NOT finished
2305 NewEntry = NewHeaderEntryByNumber(g, n);
2306 FindHeaderEntryVR(NewEntry);
2307 FindHeaderEntryLength(NewEntry);
2313 NewEntry->SetOffset(ftell(fp));
2318 * \ingroup gdcmParser
2319 * \brief Build a new Element Value from all the low level arguments.
2320 * Check for existence of dictionary entry, and build
2321 * a default one when absent.
2322 * @param Name Name of the underlying DictEntry
2324 gdcmHeaderEntry *gdcmParser::NewHeaderEntryByName(std::string Name)
2326 gdcmDictEntry *NewTag = GetDictEntryByName(Name);
2328 NewTag = NewVirtualDictEntry(0xffff, 0xffff, "LO", "Unknown", Name);
2330 gdcmHeaderEntry* NewEntry = new gdcmHeaderEntry(NewTag);
2333 dbg.Verbose(1, "gdcmParser::ObtainHeaderEntryByName",
2334 "failed to allocate gdcmHeaderEntry");
2335 return (gdcmHeaderEntry *)0;
2341 * \ingroup gdcmParser
2342 * \brief Request a new virtual dict entry to the dict set
2343 * @param group group of the underlying DictEntry
2344 * @param elem element of the underlying DictEntry
2345 * @param vr VR of the underlying DictEntry
2346 * @param fourth owner group
2347 * @param name english name
2349 gdcmDictEntry *gdcmParser::NewVirtualDictEntry(guint16 group, guint16 element,
2354 return gdcmGlobal::GetDicts()->NewVirtualDictEntry(group,element,vr,fourth,name);
2358 * \ingroup gdcmParser
2359 * \brief Build a new Element Value from all the low level arguments.
2360 * Check for existence of dictionary entry, and build
2361 * a default one when absent.
2362 * @param Group group of the underlying DictEntry
2363 * @param Elem element of the underlying DictEntry
2365 gdcmHeaderEntry *gdcmParser::NewHeaderEntryByNumber(guint16 Group, guint16 Elem)
2367 // Find out if the tag we encountered is in the dictionaries:
2368 gdcmDictEntry *DictEntry = GetDictEntryByNumber(Group, Elem);
2370 DictEntry = NewVirtualDictEntry(Group, Elem);
2372 gdcmHeaderEntry *NewEntry = new gdcmHeaderEntry(DictEntry);
2375 dbg.Verbose(1, "gdcmParser::NewHeaderEntryByNumber",
2376 "failed to allocate gdcmHeaderEntry");
2382 // Never used; commented out, waiting for removal.
2384 * \ingroup gdcmParser
2385 * \brief Small utility function that creates a new manually crafted
2386 * (as opposed as read from the file) gdcmHeaderEntry with user
2387 * specified name and adds it to the public tag hash table.
2388 * \note A fake TagKey is generated so the PubDict can keep it's coherence.
2389 * @param NewTagName The name to be given to this new tag.
2390 * @param VR The Value Representation to be given to this new tag.
2391 * @return The newly hand crafted Element Value.
2393 //gdcmHeaderEntry *gdcmParser::NewManualHeaderEntryToPubDict(std::string NewTagName,
2396 // gdcmHeaderEntry *NewEntry = NULL;
2397 // guint32 StuffGroup = 0xffff; // Group to be stuffed with additional info
2398 // guint32 FreeElem = 0;
2399 // gdcmDictEntry *DictEntry = NULL;
2401 // FreeElem = GenerateFreeTagKeyInGroup(StuffGroup);
2402 // if (FreeElem == UINT32_MAX)
2404 // dbg.Verbose(1, "gdcmHeader::NewManualHeaderEntryToPubDict",
2405 // "Group 0xffff in Public Dict is full");
2409 // DictEntry = NewVirtualDictEntry(StuffGroup, FreeElem,
2410 // VR, "GDCM", NewTagName);
2411 // NewEntry = new gdcmHeaderEntry(DictEntry);
2412 // AddHeaderEntry(NewEntry);
2417 * \ingroup gdcmParser
2418 * \brief Generate a free TagKey i.e. a TagKey that is not present
2419 * in the TagHt dictionary.
2420 * @param group The generated tag must belong to this group.
2421 * @return The element of tag with given group which is fee.
2423 guint32 gdcmParser::GenerateFreeTagKeyInGroup(guint16 group)
2425 for (guint32 elem = 0; elem < UINT32_MAX; elem++)
2427 TagKey key = gdcmDictEntry::TranslateToKey(group, elem);
2428 if (tagHT.count(key) == 0)
2434 //-----------------------------------------------------------------------------