1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2004/09/22 10:34:53 $
7 Version: $Revision: 1.86 $
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.htm 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"
24 #include "gdcmGlobal.h"
26 #include "gdcmDebug.h"
34 #include <netinet/in.h>
39 // Implicit VR Little Endian
40 #define UI1_2_840_10008_1_2 "1.2.840.10008.1.2"
41 // Explicit VR Little Endian
42 #define UI1_2_840_10008_1_2_1 "1.2.840.10008.1.2.1"
43 // Deflated Explicit VR Little Endian
44 #define UI1_2_840_10008_1_2_1_99 "1.2.840.10008.1.2.1.99"
45 // Explicit VR Big Endian
46 #define UI1_2_840_10008_1_2_2 "1.2.840.10008.1.2.2"
47 // JPEG Baseline (Process 1)
48 #define UI1_2_840_10008_1_2_4_50 "1.2.840.10008.1.2.4.50"
49 // JPEG Extended (Process 2 & 4)
50 #define UI1_2_840_10008_1_2_4_51 "1.2.840.10008.1.2.4.51"
51 // JPEG Extended (Process 3 & 5)
52 #define UI1_2_840_10008_1_2_4_52 "1.2.840.10008.1.2.4.52"
53 // JPEG Spectral Selection, Non-Hierarchical (Process 6 & 8)
54 #define UI1_2_840_10008_1_2_4_53 "1.2.840.10008.1.2.4.53"
55 // JPEG Full Progression, Non-Hierarchical (Process 10 & 12)
56 #define UI1_2_840_10008_1_2_4_55 "1.2.840.10008.1.2.4.55"
57 // JPEG Lossless, Non-Hierarchical (Process 14)
58 #define UI1_2_840_10008_1_2_4_57 "1.2.840.10008.1.2.4.57"
59 // JPEG Lossless, Hierarchical, First-Order Prediction (Process 14,
60 // [Selection Value 1])
61 #define UI1_2_840_10008_1_2_4_70 "1.2.840.10008.1.2.4.70"
63 #define UI1_2_840_10008_1_2_4_90 "1.2.840.10008.1.2.4.90"
65 #define UI1_2_840_10008_1_2_4_91 "1.2.840.10008.1.2.4.91"
67 #define UI1_2_840_10008_1_2_5 "1.2.840.10008.1.2.5"
68 // UI1_1_2_840_10008_1_2_5
69 #define str2num(str, typeNum) *((typeNum *)(str))
71 //-----------------------------------------------------------------------------
72 // Refer to gdcmDocument::CheckSwap()
73 const unsigned int gdcmDocument::HEADER_LENGTH_TO_READ = 256;
75 // Refer to gdcmDocument::SetMaxSizeLoadEntry()
76 const unsigned int gdcmDocument::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
77 const unsigned int gdcmDocument::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
79 //-----------------------------------------------------------------------------
80 // Constructor / Destructor
84 * @param filename file to be opened for parsing
86 gdcmDocument::gdcmDocument( std::string const & filename )
89 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
98 dbg.Verbose(0, "gdcmDocument::gdcmDocument: starting parsing of file: ",
102 fseek(Fp,0L,SEEK_END);
103 long lgt = ftell(Fp);
107 long beg = ftell(Fp);
110 (void)ParseDES( this, beg, lgt, false); // le Load sera fait a la volee
114 // Load 'non string' values
116 std::string PhotometricInterpretation = GetEntryByNumber(0x0028,0x0004);
117 if( PhotometricInterpretation == "PALETTE COLOR " )
119 LoadEntryVoidArea(0x0028,0x1200); // gray LUT
120 /// FIXME FIXME FIXME
121 /// The tags refered by the three following lines used to be CORRECTLY
122 /// defined as having an US Value Representation in the public
123 /// dictionnary. BUT the semantics implied by the three following
124 /// lines state that the corresponding tag contents are in fact
125 /// the ones of a gdcmBinEntry.
126 /// In order to fix things "Quick and Dirty" the dictionnary was
127 /// altered on PURPOUS but now contains a WRONG value.
128 /// In order to fix things and restore the dictionary to its
129 /// correct value, one needs to decided of the semantics by deciding
130 /// wether the following tags are either:
131 /// - multivaluated US, and hence loaded as gdcmValEntry, but afterwards
132 /// also used as gdcmBinEntry, which requires the proper conversion,
133 /// - OW, and hence loaded as gdcmBinEntry, but afterwards also used
134 /// as gdcmValEntry, which requires the proper conversion.
135 LoadEntryVoidArea(0x0028,0x1201); // R LUT
136 LoadEntryVoidArea(0x0028,0x1202); // G LUT
137 LoadEntryVoidArea(0x0028,0x1203); // B LUT
139 // Segmented Red Palette Color LUT Data
140 LoadEntryVoidArea(0x0028,0x1221);
141 // Segmented Green Palette Color LUT Data
142 LoadEntryVoidArea(0x0028,0x1222);
143 // Segmented Blue Palette Color LUT Data
144 LoadEntryVoidArea(0x0028,0x1223);
146 //FIXME later : how to use it?
147 LoadEntryVoidArea(0x0028,0x3006); //LUT Data (CTX dependent)
151 // --------------------------------------------------------------
152 // Specific code to allow gdcm to read ACR-LibIDO formated images
153 // Note: ACR-LibIDO is an extension of the ACR standard that was
154 // used at CREATIS. For the time being (say a couple years)
155 // we keep this kludge to allow a smooth move to gdcm for
156 // CREATIS developpers (sorry folks).
158 // if recognition code tells us we deal with a LibIDO image
159 // we switch lineNumber and columnNumber
162 RecCode = GetEntryByNumber(0x0008, 0x0010); // recognition code
163 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
164 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
165 // with "little-endian strings"
167 Filetype = gdcmACR_LIBIDO;
168 std::string rows = GetEntryByNumber(0x0028, 0x0010);
169 std::string columns = GetEntryByNumber(0x0028, 0x0011);
170 SetEntryByNumber(columns, 0x0028, 0x0010);
171 SetEntryByNumber(rows , 0x0028, 0x0011);
173 // ----------------- End of ACR-LibIDO kludge ------------------
175 PrintLevel = 1; // 'Medium' print level by default
179 * \brief This default constructor doesn't parse the file. You should
180 * then invoke \ref gdcmDocument::SetFileName and then the parsing.
182 gdcmDocument::gdcmDocument()
185 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
187 PrintLevel = 1; // 'Medium' print level by default
191 * \brief Canonical destructor.
193 gdcmDocument::~gdcmDocument ()
198 // Recursive clean up of sequences
199 for (TagDocEntryHT::const_iterator it = TagHT.begin();
200 it != TagHT.end(); ++it )
202 //delete it->second; //temp remove
207 //-----------------------------------------------------------------------------
211 * \brief Prints The Dict Entries of THE public Dicom Dictionary
214 void gdcmDocument::PrintPubDict(std::ostream & os)
216 RefPubDict->Print(os);
220 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
223 void gdcmDocument::PrintShaDict(std::ostream & os)
225 RefShaDict->Print(os);
228 //-----------------------------------------------------------------------------
231 * \brief Get the public dictionary used
233 gdcmDict *gdcmDocument::GetPubDict()
239 * \brief Get the shadow dictionary used
241 gdcmDict *gdcmDocument::GetShaDict()
247 * \brief Set the shadow dictionary used
248 * \param dict dictionary to use in shadow
250 bool gdcmDocument::SetShaDict(gdcmDict *dict)
257 * \brief Set the shadow dictionary used
258 * \param dictName name of the dictionary to use in shadow
260 bool gdcmDocument::SetShaDict(DictKey const & dictName)
262 RefShaDict = gdcmGlobal::GetDicts()->GetDict(dictName);
267 * \brief This predicate, based on hopefully reasonable heuristics,
268 * decides whether or not the current gdcmDocument was properly parsed
269 * and contains the mandatory information for being considered as
270 * a well formed and usable Dicom/Acr File.
271 * @return true when gdcmDocument is the one of a reasonable Dicom/Acr file,
274 bool gdcmDocument::IsReadable()
276 if( Filetype == gdcmUnknown)
278 dbg.Verbose(0, "gdcmDocument::IsReadable: wrong filetype");
284 dbg.Verbose(0, "gdcmDocument::IsReadable: no tags in internal"
294 * \brief Internal function that checks whether the Transfer Syntax given
295 * as argument is the one present in the current document.
296 * @param syntaxToCheck The transfert syntax we need to check against.
297 * @return True when SyntaxToCheck corresponds to the Transfer Syntax of
298 * the current document. False either when the document contains
299 * no Transfer Syntax, or when the Tranfer Syntaxes doesn't match.
301 bool gdcmDocument::IsGivenTransferSyntax(std::string const & syntaxToCheck)
303 gdcmDocEntry *entry = GetDocEntryByNumber(0x0002, 0x0010);
309 // The entry might be present but not loaded (parsing and loading
310 // happen at different stages): try loading and proceed with check...
311 LoadDocEntrySafe(entry);
312 if (gdcmValEntry* valEntry = dynamic_cast< gdcmValEntry* >(entry) )
314 std::string transfer = valEntry->GetValue();
315 // The actual transfer (as read from disk) might be padded. We
316 // first need to remove the potential padding. We can make the
317 // weak assumption that padding was not executed with digits...
318 if ( transfer.length() == 0 ) { // for brain damaged headers
321 while ( ! isdigit(transfer[transfer.length()-1]) )
323 transfer.erase(transfer.length()-1, 1);
325 if ( transfer == syntaxToCheck )
334 * \brief Determines if the Transfer Syntax of the present document
335 * corresponds to a Implicit Value Representation of
337 * \sa \ref gdcmDocument::IsGivenTransferSyntax.
338 * @return True when ImplicitVRLittleEndian found. False in all other cases.
340 bool gdcmDocument::IsImplicitVRLittleEndianTransferSyntax()
342 return IsGivenTransferSyntax(UI1_2_840_10008_1_2);
346 * \brief Determines if the Transfer Syntax was already encountered
347 * and if it corresponds to a ExplicitVRLittleEndian one.
348 * @return True when ExplicitVRLittleEndian found. False in all other cases.
350 bool gdcmDocument::IsExplicitVRLittleEndianTransferSyntax()
352 return IsGivenTransferSyntax(UI1_2_840_10008_1_2_1);
356 * \brief Determines if the Transfer Syntax was already encountered
357 * and if it corresponds to a DeflatedExplicitVRLittleEndian one.
358 * @return True when DeflatedExplicitVRLittleEndian found. False in all other cases.
360 bool gdcmDocument::IsDeflatedExplicitVRLittleEndianTransferSyntax()
362 return IsGivenTransferSyntax(UI1_2_840_10008_1_2_1_99);
366 * \brief Determines if the Transfer Syntax was already encountered
367 * and if it corresponds to a Explicit VR Big Endian one.
368 * @return True when big endian found. False in all other cases.
370 bool gdcmDocument::IsExplicitVRBigEndianTransferSyntax()
372 return IsGivenTransferSyntax(UI1_2_840_10008_1_2_2);
376 * \brief Determines if the Transfer Syntax was already encountered
377 * and if it corresponds to a JPEGBaseLineProcess1 one.
378 * @return True when JPEGBaseLineProcess1found. False in all other cases.
380 bool gdcmDocument::IsJPEGBaseLineProcess1TransferSyntax()
382 return IsGivenTransferSyntax(UI1_2_840_10008_1_2_4_50);
386 * \brief Determines if the Transfer Syntax was already encountered
387 * and if it corresponds to a JPEGExtendedProcess2-4 one.
388 * @return True when JPEGExtendedProcess2-4 found. False in all other cases.
390 bool gdcmDocument::IsJPEGExtendedProcess2_4TransferSyntax()
392 return IsGivenTransferSyntax(UI1_2_840_10008_1_2_4_51);
396 * \brief Determines if the Transfer Syntax was already encountered
397 * and if it corresponds to a JPEGExtendeProcess3-5 one.
398 * @return True when JPEGExtendedProcess3-5 found. False in all other cases.
400 bool gdcmDocument::IsJPEGExtendedProcess3_5TransferSyntax()
402 return IsGivenTransferSyntax(UI1_2_840_10008_1_2_4_52);
406 * \brief Determines if the Transfer Syntax was already encountered
407 * and if it corresponds to a JPEGSpectralSelectionProcess6-8 one.
408 * @return True when JPEGSpectralSelectionProcess6-8 found. False in all
411 bool gdcmDocument::IsJPEGSpectralSelectionProcess6_8TransferSyntax()
413 return IsGivenTransferSyntax(UI1_2_840_10008_1_2_4_53);
417 * \brief Determines if the Transfer Syntax was already encountered
418 * and if it corresponds to a RLE Lossless one.
419 * @return True when RLE Lossless found. False in all
422 bool gdcmDocument::IsRLELossLessTransferSyntax()
424 return IsGivenTransferSyntax(UI1_2_840_10008_1_2_5);
428 * \brief Determines if Transfer Syntax was already encountered
429 * and if it corresponds to a JPEG Lossless one.
430 * @return True when RLE Lossless found. False in all
434 bool gdcmDocument::IsJPEGLossless()
436 return ( IsGivenTransferSyntax(UI1_2_840_10008_1_2_4_55)
437 || IsGivenTransferSyntax(UI1_2_840_10008_1_2_4_57)
438 || IsGivenTransferSyntax(UI1_2_840_10008_1_2_4_70) );
442 * \brief Determines if the Transfer Syntax was already encountered
443 * and if it corresponds to a JPEG2000 one
444 * @return True when JPEG2000 (Lossly or LossLess) found. False in all
447 bool gdcmDocument::IsJPEG2000()
449 return ( IsGivenTransferSyntax(UI1_2_840_10008_1_2_4_90)
450 || IsGivenTransferSyntax(UI1_2_840_10008_1_2_4_91) );
454 * \brief Predicate for dicom version 3 file.
455 * @return True when the file is a dicom version 3.
457 bool gdcmDocument::IsDicomV3()
459 // Checking if Transfert Syntax exists is enough
460 // Anyway, it's to late check if the 'Preamble' was found ...
461 // And ... would it be a rich idea to check ?
462 // (some 'no Preamble' DICOM images exist !)
463 return GetDocEntryByNumber(0x0002, 0x0010) != NULL;
467 * \brief returns the File Type
468 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
469 * @return the FileType code
471 FileType gdcmDocument::GetFileType()
477 * \brief Tries to open the file \ref gdcmDocument::Filename and
478 * checks the preamble when existing.
479 * @return The FILE pointer on success.
481 FILE* gdcmDocument::OpenFile()
483 Fp = fopen(Filename.c_str(),"rb");
488 "gdcmDocument::OpenFile cannot open file: ",
494 fread(&zero, (size_t)2, (size_t)1, Fp);
496 //ACR -- or DICOM with no Preamble --
497 if( zero == 0x0008 || zero == 0x0800 || zero == 0x0002 || zero == 0x0200 )
503 fseek(Fp, 126L, SEEK_CUR);
505 fread(dicm, (size_t)4, (size_t)1, Fp);
506 if( memcmp(dicm, "DICM", 4) == 0 )
513 "gdcmDocument::OpenFile not DICOM/ACR (missing preamble)",
520 * \brief closes the file
521 * @return TRUE if the close was successfull
523 bool gdcmDocument::CloseFile()
525 int closed = fclose(Fp);
532 * \brief Writes in a file all the Header Entries (Dicom Elements)
533 * @param fp file pointer on an already open file
534 * @param filetype Type of the File to be written
535 * (ACR-NEMA, ExplicitVR, ImplicitVR)
536 * \return Always true.
538 void gdcmDocument::Write(FILE* fp,FileType filetype)
540 /// \todo move the following lines (and a lot of others, to be written)
541 /// to a future function CheckAndCorrectHeader
542 /// (necessary if user wants to write a DICOM V3 file
543 /// starting from an ACR-NEMA (V2) gdcmHeader
545 if (filetype == gdcmImplicitVR)
547 std::string implicitVRTransfertSyntax = UI1_2_840_10008_1_2;
548 ReplaceOrCreateByNumber(implicitVRTransfertSyntax,0x0002, 0x0010);
550 /// \todo Refer to standards on page 21, chapter 6.2
551 /// "Value representation": values with a VR of UI shall be
552 /// padded with a single trailing null
553 /// in the following case we have to padd manually with a 0
555 SetEntryLengthByNumber(18, 0x0002, 0x0010);
558 if (filetype == gdcmExplicitVR)
560 std::string explicitVRTransfertSyntax = UI1_2_840_10008_1_2_1;
561 ReplaceOrCreateByNumber(explicitVRTransfertSyntax,0x0002, 0x0010);
563 /// \todo Refer to standards on page 21, chapter 6.2
564 /// "Value representation": values with a VR of UI shall be
565 /// padded with a single trailing null
566 /// Dans le cas suivant on doit pader manuellement avec un 0
568 SetEntryLengthByNumber(20, 0x0002, 0x0010);
572 * \todo rewrite later, if really usefull
573 * - 'Group Length' element is optional in DICOM
574 * - but un-updated odd groups lengthes can causes pb
577 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
578 * UpdateGroupLength(false,filetype);
579 * if ( filetype == ACR)
580 * UpdateGroupLength(true,ACR);
583 gdcmElementSet::Write(fp, filetype); // This one is recursive
588 * \brief Modifies the value of a given Header Entry (Dicom Element)
589 * when it exists. Create it with the given value when unexistant.
590 * @param value (string) Value to be set
591 * @param group Group number of the Entry
592 * @param elem Element number of the Entry
593 * @param VR V(alue) R(epresentation) of the Entry -if private Entry-
594 * \return pointer to the modified/created Header Entry (NULL when creation
597 gdcmValEntry * gdcmDocument::ReplaceOrCreateByNumber(
598 std::string const & value,
601 std::string const & VR )
603 gdcmValEntry* valEntry = 0;
604 gdcmDocEntry* currentEntry = GetDocEntryByNumber( group, elem);
608 // check if (group,element) DictEntry exists
609 // if it doesn't, create an entry in gdcmDictSet::VirtualEntry
612 // Find out if the tag we received is in the dictionaries:
613 gdcmDict *pubDict = gdcmGlobal::GetDicts()->GetDefaultPubDict();
614 gdcmDictEntry *dictEntry = pubDict->GetDictEntryByNumber(group, elem);
617 currentEntry = NewDocEntryByNumber(group, elem,VR);
621 currentEntry = NewDocEntryByNumber(group, elem);
626 dbg.Verbose(0, "gdcmDocument::ReplaceOrCreateByNumber: call to"
627 " NewDocEntryByNumber failed.");
630 valEntry = new gdcmValEntry(currentEntry);
631 if ( !AddEntry(valEntry))
633 dbg.Verbose(0, "gdcmDocument::ReplaceOrCreateByNumber: AddEntry"
634 " failed allthough this is a creation.");
639 valEntry = dynamic_cast< gdcmValEntry* >(currentEntry);
640 if ( !valEntry ) // Euuuuh? It wasn't a ValEntry
641 // then we change it to a ValEntry ?
642 // Shouldn't it be considered as an error ?
644 // We need to promote the gdcmDocEntry to a gdcmValEntry:
645 valEntry = new gdcmValEntry(currentEntry);
646 if (!RemoveEntry(currentEntry))
648 dbg.Verbose(0, "gdcmDocument::ReplaceOrCreateByNumber: removal"
649 " of previous DocEntry failed.");
652 if ( !AddEntry(valEntry))
654 dbg.Verbose(0, "gdcmDocument::ReplaceOrCreateByNumber: adding"
655 " promoted ValEntry failed.");
661 SetEntryByNumber(value, group, elem);
667 * \brief Modifies the value of a given Header Entry (Dicom Element)
668 * when it exists. Create it with the given value when unexistant.
669 * @param voidArea (binary) value to be set
670 * @param Group Group number of the Entry
671 * @param Elem Element number of the Entry
672 * \return pointer to the modified/created Header Entry (NULL when creation
675 gdcmBinEntry * gdcmDocument::ReplaceOrCreateByNumber(
680 std::string const & VR )
682 gdcmBinEntry* binEntry = 0;
683 gdcmDocEntry* currentEntry = GetDocEntryByNumber( group, elem);
687 // check if (group,element) DictEntry exists
688 // if it doesn't, create an entry in gdcmDictSet::VirtualEntry
691 // Find out if the tag we received is in the dictionaries:
692 gdcmDict *pubDict = gdcmGlobal::GetDicts()->GetDefaultPubDict();
693 gdcmDictEntry *dictEntry = pubDict->GetDictEntryByNumber(group, elem);
697 currentEntry = NewDocEntryByNumber(group, elem, VR);
701 currentEntry = NewDocEntryByNumber(group, elem);
705 dbg.Verbose(0, "gdcmDocument::ReplaceOrCreateByNumber: call to"
706 " NewDocEntryByNumber failed.");
709 binEntry = new gdcmBinEntry(currentEntry);
710 if ( !AddEntry(binEntry))
712 dbg.Verbose(0, "gdcmDocument::ReplaceOrCreateByNumber: AddEntry"
713 " failed allthough this is a creation.");
718 binEntry = dynamic_cast< gdcmBinEntry* >(currentEntry);
719 if ( !binEntry ) // Euuuuh? It wasn't a BinEntry
720 // then we change it to a BinEntry ?
721 // Shouldn't it be considered as an error ?
723 // We need to promote the gdcmDocEntry to a gdcmBinEntry:
724 binEntry = new gdcmBinEntry(currentEntry);
725 if (!RemoveEntry(currentEntry))
727 dbg.Verbose(0, "gdcmDocument::ReplaceOrCreateByNumber: removal"
728 " of previous DocEntry failed.");
731 if ( !AddEntry(binEntry))
733 dbg.Verbose(0, "gdcmDocument::ReplaceOrCreateByNumber: adding"
734 " promoted BinEntry failed.");
740 SetEntryByNumber(voidArea, lgth, group, elem);
747 * \brief Modifies the value of a given Header Entry (Dicom Element)
748 * when it exists. Create it when unexistant.
749 * @param Group Group number of the Entry
750 * @param Elem Element number of the Entry
751 * \return pointer to the modified/created SeqEntry (NULL when creation
754 gdcmSeqEntry * gdcmDocument::ReplaceOrCreateByNumber(
759 gdcmDocEntry* a = GetDocEntryByNumber( group, elem);
762 a = NewSeqEntryByNumber(group, elem);
768 b = new gdcmSeqEntry(a, 1); // FIXME : 1 (Depth)
775 * \brief Set a new value if the invoked element exists
776 * Seems to be useless !!!
777 * @param value new element value
778 * @param group group number of the Entry
779 * @param elem element number of the Entry
782 bool gdcmDocument::ReplaceIfExistByNumber(std::string const & value,
783 uint16_t group, uint16_t elem )
785 SetEntryByNumber(value, group, elem);
790 //-----------------------------------------------------------------------------
794 * \brief Checks if a given Dicom Element exists within the H table
795 * @param group Group number of the searched Dicom Element
796 * @param element Element number of the searched Dicom Element
797 * @return true is found
799 bool gdcmDocument::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
801 const std::string &key = gdcmDictEntry::TranslateToKey(group, element );
802 return TagHT.count(key);
806 * \brief Searches within Header Entries (Dicom Elements) parsed with
807 * the public and private dictionaries
808 * for the element value of a given tag.
809 * \warning Don't use any longer : use GetPubEntryByName
810 * @param tagName name of the searched element.
811 * @return Corresponding element value when it exists,
812 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
814 std::string gdcmDocument::GetEntryByName(TagName const & tagName)
816 gdcmDictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
822 return GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
826 * \brief Searches within Header Entries (Dicom Elements) parsed with
827 * the public and private dictionaries
828 * for the element value representation of a given tag.
830 * Obtaining the VR (Value Representation) might be needed by caller
831 * to convert the string typed content to caller's native type
832 * (think of C++ vs Python). The VR is actually of a higher level
833 * of semantics than just the native C++ type.
834 * @param tagName name of the searched element.
835 * @return Corresponding element value representation when it exists,
836 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
838 std::string gdcmDocument::GetEntryVRByName(TagName const & tagName)
840 gdcmDictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
841 if( dictEntry == NULL)
846 gdcmDocEntry* elem = GetDocEntryByNumber(dictEntry->GetGroup(),
847 dictEntry->GetElement());
848 return elem->GetVR();
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.
855 * @param group Group number of the searched tag.
856 * @param element Element number of the searched tag.
857 * @return Corresponding element value representation when it exists,
858 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
860 std::string gdcmDocument::GetEntryByNumber(uint16_t group, uint16_t element)
862 gdcmTagKey key = gdcmDictEntry::TranslateToKey(group, element);
863 /// \todo use map methods, instead of multimap JPR
864 if ( !TagHT.count(key))
869 return ((gdcmValEntry *)TagHT.find(key)->second)->GetValue();
873 * \brief Searches within Header Entries (Dicom Elements) parsed with
874 * the public and private dictionaries
875 * for the element value representation of a given tag..
877 * Obtaining the VR (Value Representation) might be needed by caller
878 * to convert the string typed content to caller's native type
879 * (think of C++ vs Python). The VR is actually of a higher level
880 * of semantics than just the native C++ type.
881 * @param group Group number of the searched tag.
882 * @param element Element number of the searched tag.
883 * @return Corresponding element value representation when it exists,
884 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
886 std::string gdcmDocument::GetEntryVRByNumber(uint16_t group, uint16_t element)
888 gdcmDocEntry* elem = GetDocEntryByNumber(group, element);
893 return elem->GetVR();
897 * \brief Searches within Header Entries (Dicom Elements) parsed with
898 * the public and private dictionaries
899 * for the value length of a given tag..
900 * @param group Group number of the searched tag.
901 * @param element Element number of the searched tag.
902 * @return Corresponding element length; -2 if not found
904 int gdcmDocument::GetEntryLengthByNumber(uint16_t group, uint16_t element)
906 gdcmDocEntry* elem = GetDocEntryByNumber(group, element);
909 return -2; //magic number
911 return elem->GetLength();
914 * \brief Sets the value (string) of the Header Entry (Dicom Element)
915 * @param content string value of the Dicom Element
916 * @param tagName name of the searched Dicom Element.
917 * @return true when found
919 bool gdcmDocument::SetEntryByName(std::string const & content,std::string const & tagName)
921 gdcmDictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
927 return SetEntryByNumber(content,dictEntry->GetGroup(),
928 dictEntry->GetElement());
932 * \brief Accesses an existing gdcmDocEntry (i.e. a Dicom Element)
933 * through it's (group, element) and modifies it's content with
935 * @param content new value (string) to substitute with
936 * @param group group number of the Dicom Element to modify
937 * @param element element number of the Dicom Element to modify
939 bool gdcmDocument::SetEntryByNumber(std::string const & content,
946 gdcmValEntry* valEntry = GetValEntryByNumber(group, element);
949 dbg.Verbose(0, "gdcmDocument::SetEntryByNumber: no corresponding",
950 " ValEntry (try promotion first).");
953 // Non even content must be padded with a space (020H)...
954 std::string finalContent = content;
955 if( finalContent.length() % 2 )
957 finalContent += '\0'; // ... therefore we padd with (000H) .!?!
959 valEntry->SetValue(finalContent);
961 // Integers have a special treatement for their length:
963 l = finalContent.length();
964 if ( l != 0) // To avoid to be cheated by 'zero length' integers
966 gdcmVRKey vr = valEntry->GetVR();
967 if( vr == "US" || vr == "SS" )
969 c = CountSubstring(content, "\\") + 1; // for multivaluated items
972 else if( vr == "UL" || vr == "SL" )
974 c = CountSubstring(content, "\\") + 1; // for multivaluated items
978 valEntry->SetLength(l);
983 * \brief Accesses an existing gdcmDocEntry (i.e. a Dicom Element)
984 * through it's (group, element) and modifies it's content with
986 * @param content new value (void *) to substitute with
987 * @param lgth new value length
988 * @param group group number of the Dicom Element to modify
989 * @param element element number of the Dicom Element to modify
991 bool gdcmDocument::SetEntryByNumber(void *content,
996 (void)lgth; //not used
997 gdcmTagKey key = gdcmDictEntry::TranslateToKey(group, element);
998 if ( !TagHT.count(key))
1003 /* Hope Binary field length is *never* wrong
1004 if(lgth%2) // Non even length are padded with a space (020H).
1007 //content = content + '\0'; // fing a trick to enlarge a binary field?
1010 gdcmBinEntry* a = (gdcmBinEntry *)TagHT[key];
1011 a->SetVoidArea(content);
1013 a->SetValue(GDCM_BINLOADED);
1019 * \brief Accesses an existing gdcmDocEntry (i.e. a Dicom Element)
1020 * in the PubDocEntrySet of this instance
1021 * through it's (group, element) and modifies it's length with
1023 * \warning Use with extreme caution.
1024 * @param l new length to substitute with
1025 * @param group group number of the Entry to modify
1026 * @param element element number of the Entry to modify
1027 * @return true on success, false otherwise.
1029 bool gdcmDocument::SetEntryLengthByNumber(uint32_t l,
1033 /// \todo use map methods, instead of multimap JPR
1034 gdcmTagKey key = gdcmDictEntry::TranslateToKey(group, element);
1035 if ( !TagHT.count(key) )
1041 l++; // length must be even
1043 ( ((TagHT.equal_range(key)).first)->second )->SetLength(l);
1049 * \brief Gets (from Header) the offset of a 'non string' element value
1050 * (LoadElementValues has already be executed)
1051 * @param group group number of the Entry
1052 * @param elem element number of the Entry
1053 * @return File Offset of the Element Value
1055 size_t gdcmDocument::GetEntryOffsetByNumber(uint16_t group, uint16_t elem)
1057 gdcmDocEntry* entry = GetDocEntryByNumber(group, elem);
1060 dbg.Verbose(1, "gdcmDocument::GetDocEntryByNumber: no entry present.");
1063 return entry->GetOffset();
1067 * \brief Gets (from Header) a 'non string' element value
1068 * (LoadElementValues has already be executed)
1069 * @param group group number of the Entry
1070 * @param elem element number of the Entry
1071 * @return Pointer to the 'non string' area
1073 void * gdcmDocument::GetEntryVoidAreaByNumber(uint16_t group, uint16_t elem)
1075 gdcmDocEntry* entry = GetDocEntryByNumber(group, elem);
1078 dbg.Verbose(1, "gdcmDocument::GetDocEntryByNumber: no entry");
1081 return ((gdcmBinEntry *)entry)->GetVoidArea();
1085 * \brief Loads (from disk) the element content
1086 * when a string is not suitable
1087 * @param group group number of the Entry
1088 * @param elem element number of the Entry
1090 void* gdcmDocument::LoadEntryVoidArea(uint16_t group, uint16_t elem)
1092 gdcmDocEntry *docElement = GetDocEntryByNumber(group, elem);
1097 size_t o =(size_t)docElement->GetOffset();
1098 fseek(Fp, o, SEEK_SET);
1099 size_t l = docElement->GetLength();
1100 char* a = new char[l];
1103 dbg.Verbose(0, "gdcmDocument::LoadEntryVoidArea cannot allocate a");
1106 size_t l2 = fread(a, 1, l , Fp);
1112 /// \todo Drop any already existing void area! JPR
1113 if( !SetEntryVoidAreaByNumber( a, group, elem ) );
1115 dbg.Verbose(0, "gdcmDocument::LoadEntryVoidArea setting failed.");
1120 * \brief Loads (from disk) the element content
1121 * when a string is not suitable
1122 * @param element Entry whose voidArea is going to be loaded
1124 void *gdcmDocument::LoadEntryVoidArea(gdcmBinEntry *element)
1126 size_t o =(size_t)element->GetOffset();
1127 fseek(Fp, o, SEEK_SET);
1128 size_t l = element->GetLength();
1129 char* a = new char[l];
1132 dbg.Verbose(0, "gdcmDocument::LoadEntryVoidArea cannot allocate a");
1135 element->SetVoidArea((void *)a);
1136 /// \todo check the result
1137 size_t l2 = fread(a, 1, l , Fp);
1148 * \brief Sets a 'non string' value to a given Dicom Element
1149 * @param area area containing the 'non string' value
1150 * @param group Group number of the searched Dicom Element
1151 * @param element Element number of the searched Dicom Element
1154 bool gdcmDocument::SetEntryVoidAreaByNumber(void * area,
1158 gdcmDocEntry* currentEntry = GetDocEntryByNumber(group, element);
1159 if ( !currentEntry )
1163 if ( gdcmBinEntry* binEntry = dynamic_cast<gdcmBinEntry*>(currentEntry) )
1165 binEntry->SetVoidArea( area );
1172 * \brief Update the entries with the shadow dictionary.
1173 * Only non even entries are analyzed
1175 void gdcmDocument::UpdateShaEntries()
1177 //gdcmDictEntry *entry;
1180 /// \todo TODO : still any use to explore recursively the whole structure?
1182 for(ListTag::iterator it=listEntries.begin();
1183 it!=listEntries.end();
1186 // Odd group => from public dictionary
1187 if((*it)->GetGroup()%2==0)
1190 // Peer group => search the corresponding dict entry
1192 entry=RefShaDict->GetDictEntryByNumber((*it)->GetGroup(),(*it)->GetElement());
1196 if((*it)->IsImplicitVR())
1201 (*it)->SetValue(GetDocEntryUnvalue(*it)); // to go on compiling
1203 // Set the new entry and the new value
1204 (*it)->SetDictEntry(entry);
1205 CheckDocEntryVR(*it,vr);
1207 (*it)->SetValue(GetDocEntryValue(*it)); // to go on compiling
1212 // Remove precedent value transformation
1213 (*it)->SetDictEntry(NewVirtualDictEntry((*it)->GetGroup(),(*it)->GetElement(),vr));
1220 * \brief Searches within the Header Entries for a Dicom Element of
1222 * @param tagName name of the searched Dicom Element.
1223 * @return Corresponding Dicom Element when it exists, and NULL
1226 gdcmDocEntry* gdcmDocument::GetDocEntryByName(std::string const & tagName)
1228 gdcmDictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
1234 return GetDocEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
1238 * \brief retrieves a Dicom Element (the first one) using (group, element)
1239 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1240 * if you think it's NOT UNIQUE, check the count number
1241 * and use iterators to retrieve ALL the Dicoms Elements within
1242 * a given couple (group, element)
1243 * @param group Group number of the searched Dicom Element
1244 * @param element Element number of the searched Dicom Element
1247 gdcmDocEntry* gdcmDocument::GetDocEntryByNumber(uint16_t group,
1250 gdcmTagKey key = gdcmDictEntry::TranslateToKey(group, element);
1251 if ( !TagHT.count(key))
1255 return TagHT.find(key)->second;
1259 * \brief Same as \ref gdcmDocument::GetDocEntryByNumber except it only
1260 * returns a result when the corresponding entry is of type
1262 * @return When present, the corresponding ValEntry.
1264 gdcmValEntry* gdcmDocument::GetValEntryByNumber(uint16_t group,
1267 gdcmDocEntry* currentEntry = GetDocEntryByNumber(group, element);
1268 if ( !currentEntry )
1272 if ( gdcmValEntry* valEntry = dynamic_cast<gdcmValEntry*>(currentEntry) )
1276 dbg.Verbose(0, "gdcmDocument::GetValEntryByNumber: unfound ValEntry.");
1282 * \brief Loads the element while preserving the current
1283 * underlying file position indicator as opposed to
1284 * to LoadDocEntry that modifies it.
1285 * @param entry Header Entry whose value shall be loaded.
1288 void gdcmDocument::LoadDocEntrySafe(gdcmDocEntry * entry)
1290 long PositionOnEntry = ftell(Fp);
1291 LoadDocEntry(entry);
1292 fseek(Fp, PositionOnEntry, SEEK_SET);
1296 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1298 * @return The properly swaped 32 bits integer.
1300 uint32_t gdcmDocument::SwapLong(uint32_t a)
1307 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1308 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1312 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1316 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1319 //std::cout << "swapCode= " << SwapCode << std::endl;
1320 dbg.Error(" gdcmDocument::SwapLong : unset swap code");
1327 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1329 * @return The properly unswaped 32 bits integer.
1331 uint32_t gdcmDocument::UnswapLong(uint32_t a)
1337 * \brief Swaps the bytes so they agree with the processor order
1338 * @return The properly swaped 16 bits integer.
1340 uint16_t gdcmDocument::SwapShort(uint16_t a)
1342 if ( SwapCode == 4321 || SwapCode == 2143 )
1344 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1350 * \brief Unswaps the bytes so they agree with the processor order
1351 * @return The properly unswaped 16 bits integer.
1353 uint16_t gdcmDocument::UnswapShort(uint16_t a)
1355 return SwapShort(a);
1358 //-----------------------------------------------------------------------------
1362 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1363 * @return length of the parsed set.
1366 long gdcmDocument::ParseDES(gdcmDocEntrySet *set,
1371 gdcmDocEntry *newDocEntry = 0;
1372 unsigned long l = 0;
1376 if ( !delim_mode && (ftell(Fp)-offset) >= l_max)
1380 newDocEntry = ReadNextDocEntry( );
1386 gdcmVRKey vr = newDocEntry->GetVR();
1390 if ( gdcmGlobal::GetVR()->IsVROfGdcmStringRepresentable(vr) )
1392 /////////////////////// ValEntry
1393 gdcmValEntry* newValEntry =
1394 new gdcmValEntry( newDocEntry->GetDictEntry() );
1395 newValEntry->Copy( newDocEntry );
1397 // When "set" is a gdcmDocument, then we are at the top of the
1398 // hierarchy and the Key is simply of the form ( group, elem )...
1399 if (gdcmDocument* dummy = dynamic_cast< gdcmDocument* > ( set ) )
1402 newValEntry->SetKey( newValEntry->GetKey() );
1404 // ...but when "set" is a gdcmSQItem, we are inserting this new
1405 // valEntry in a sequence item. Hence the key has the
1406 // generalized form (refer to \ref gdcmBaseTagKey):
1407 if (gdcmSQItem* parentSQItem = dynamic_cast< gdcmSQItem* > ( set ) )
1409 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1410 + newValEntry->GetKey() );
1413 set->AddEntry( newValEntry );
1414 LoadDocEntry( newValEntry );
1415 if (newValEntry->IsItemDelimitor())
1419 if ( !delim_mode && (ftell(Fp)-offset) >= l_max)
1426 if ( ! gdcmGlobal::GetVR()->IsVROfGdcmBinaryRepresentable(vr) )
1428 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1429 dbg.Verbose(0, "gdcmDocument::ParseDES: neither Valentry, "
1430 "nor BinEntry. Probably unknown VR.");
1433 //////////////////// BinEntry or UNKOWN VR:
1434 gdcmBinEntry* newBinEntry =
1435 new gdcmBinEntry( newDocEntry->GetDictEntry() );
1436 newBinEntry->Copy( newDocEntry );
1438 // When "this" is a gdcmDocument the Key is simply of the
1439 // form ( group, elem )...
1440 if (gdcmDocument* dummy = dynamic_cast< gdcmDocument* > ( set ) )
1443 newBinEntry->SetKey( newBinEntry->GetKey() );
1445 // but when "this" is a SQItem, we are inserting this new
1446 // valEntry in a sequence item, and the kay has the
1447 // generalized form (refer to \ref gdcmBaseTagKey):
1448 if (gdcmSQItem* parentSQItem = dynamic_cast< gdcmSQItem* > ( set ) )
1450 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1451 + newBinEntry->GetKey() );
1454 set->AddEntry( newBinEntry );
1455 LoadDocEntry( newBinEntry );
1458 if (newDocEntry->GetGroup() == 0x7fe0 &&
1459 newDocEntry->GetElement() == 0x0010 )
1461 if (newDocEntry->GetReadLength()==0xffffffff)
1463 // Broken US.3405.1.dcm
1464 Parse7FE0(); // to skip the pixels
1465 // (multipart JPEG/RLE are trouble makers)
1469 SkipToNextDocEntry(newDocEntry);
1470 l = newDocEntry->GetFullLength();
1475 // to be sure we are at the beginning
1476 SkipToNextDocEntry(newDocEntry);
1477 l = newDocEntry->GetFullLength();
1483 l = newDocEntry->GetReadLength();
1484 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1486 if ( l == 0xffffffff )
1495 // no other way to create it ...
1496 gdcmSeqEntry* newSeqEntry =
1497 new gdcmSeqEntry( newDocEntry->GetDictEntry() );
1498 newSeqEntry->Copy( newDocEntry );
1499 newSeqEntry->SetDelimitorMode( delim_mode );
1501 // At the top of the hierarchy, stands a gdcmDocument. When "set"
1502 // is a gdcmDocument, then we are building the first depth level.
1503 // Hence the gdcmSeqEntry we are building simply has a depth
1505 if (gdcmDocument* dummy = dynamic_cast< gdcmDocument* > ( set ) )
1508 newSeqEntry->SetDepthLevel( 1 );
1509 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1511 // But when "set" is allready a SQItem, we are building a nested
1512 // sequence, and hence the depth level of the new gdcmSeqEntry
1513 // we are building, is one level deeper:
1514 if (gdcmSQItem* parentSQItem = dynamic_cast< gdcmSQItem* > ( set ) )
1516 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1517 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1518 + newSeqEntry->GetKey() );
1522 { // Don't try to parse zero-length sequences
1523 (void)ParseSQ( newSeqEntry,
1524 newDocEntry->GetOffset(),
1527 set->AddEntry( newSeqEntry );
1528 if ( !delim_mode && (ftell(Fp)-offset) >= l_max)
1535 return l; // Probably useless
1539 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1540 * @return parsed length for this level
1542 long gdcmDocument::ParseSQ( gdcmSeqEntry* seqEntry,
1543 long offset, long l_max, bool delim_mode)
1545 int SQItemNumber = 0;
1550 gdcmDocEntry* newDocEntry = ReadNextDocEntry();
1553 // FIXME Should warn user
1558 if ( newDocEntry->IsSequenceDelimitor() )
1560 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1564 if ( !delim_mode && (ftell(Fp)-offset) >= l_max)
1569 gdcmSQItem *itemSQ = new gdcmSQItem( seqEntry->GetDepthLevel() );
1570 std::ostringstream newBase;
1571 newBase << seqEntry->GetKey()
1575 itemSQ->SetBaseTagKey( newBase.str() );
1576 unsigned int l = newDocEntry->GetReadLength();
1578 if ( l == 0xffffffff )
1587 (void)ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1589 seqEntry->AddEntry( itemSQ, SQItemNumber );
1591 if ( !delim_mode && ( ftell(Fp) - offset ) >= l_max )
1597 int lgth = ftell(Fp) - offset;
1602 * \brief Loads the element content if its length doesn't exceed
1603 * the value specified with gdcmDocument::SetMaxSizeLoadEntry()
1604 * @param entry Header Entry (Dicom Element) to be dealt with
1606 void gdcmDocument::LoadDocEntry(gdcmDocEntry* entry)
1609 uint16_t group = entry->GetGroup();
1610 std::string vr = entry->GetVR();
1611 uint32_t length = entry->GetLength();
1613 fseek(Fp, (long)entry->GetOffset(), SEEK_SET);
1615 // A SeQuence "contains" a set of Elements.
1616 // (fffe e000) tells us an Element is beginning
1617 // (fffe e00d) tells us an Element just ended
1618 // (fffe e0dd) tells us the current SeQuence just ended
1619 if( group == 0xfffe )
1621 // NO more value field for SQ !
1625 // When the length is zero things are easy:
1628 ((gdcmValEntry *)entry)->SetValue("");
1632 // The elements whose length is bigger than the specified upper bound
1633 // are not loaded. Instead we leave a short notice of the offset of
1634 // the element content and it's length.
1636 std::ostringstream s;
1637 if (length > MaxSizeLoadEntry)
1639 if (gdcmBinEntry* binEntryPtr = dynamic_cast< gdcmBinEntry* >(entry) )
1641 //s << "gdcm::NotLoaded (BinEntry)";
1642 s << GDCM_NOTLOADED;
1643 s << " Address:" << (long)entry->GetOffset();
1644 s << " Length:" << entry->GetLength();
1645 s << " x(" << std::hex << entry->GetLength() << ")";
1646 binEntryPtr->SetValue(s.str());
1648 // Be carefull : a BinEntry IS_A ValEntry ...
1649 else if (gdcmValEntry* valEntryPtr = dynamic_cast< gdcmValEntry* >(entry) )
1651 // s << "gdcm::NotLoaded. (ValEntry)";
1652 s << GDCM_NOTLOADED;
1653 s << " Address:" << (long)entry->GetOffset();
1654 s << " Length:" << entry->GetLength();
1655 s << " x(" << std::hex << entry->GetLength() << ")";
1656 valEntryPtr->SetValue(s.str());
1661 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1662 << "nor a ValEntry ?! Should never print that !" << std::endl;
1665 // to be sure we are at the end of the value ...
1666 fseek(Fp,(long)entry->GetOffset()+(long)entry->GetLength(),SEEK_SET);
1670 // When we find a BinEntry not very much can be done :
1671 if (gdcmBinEntry* binEntryPtr = dynamic_cast< gdcmBinEntry* >(entry) )
1673 s << GDCM_BINLOADED;
1674 binEntryPtr->SetValue(s.str());
1675 LoadEntryVoidArea(binEntryPtr); // last one, not to erase length !
1679 /// \todo Any compacter code suggested (?)
1680 if ( IsDocEntryAnInteger(entry) )
1684 // When short integer(s) are expected, read and convert the following
1685 // n *two characters properly i.e. consider them as short integers as
1686 // opposed to strings.
1687 // Elements with Value Multiplicity > 1
1688 // contain a set of integers (not a single one)
1689 if (vr == "US" || vr == "SS")
1692 NewInt = ReadInt16();
1696 for (int i=1; i < nbInt; i++)
1699 NewInt = ReadInt16();
1704 // See above comment on multiple integers (mutatis mutandis).
1705 else if (vr == "UL" || vr == "SL")
1708 NewInt = ReadInt32();
1712 for (int i=1; i < nbInt; i++)
1715 NewInt = ReadInt32();
1720 #ifdef GDCM_NO_ANSI_STRING_STREAM
1721 s << std::ends; // to avoid oddities on Solaris
1722 #endif //GDCM_NO_ANSI_STRING_STREAM
1724 ((gdcmValEntry *)entry)->SetValue(s.str());
1728 // We need an additional byte for storing \0 that is not on disk
1729 //std::string newValue(length,0);
1730 //item_read = fread(&(newValue[0]), (size_t)length, (size_t)1, Fp);
1731 //rah !! I can't believe it could work, normally this is a const char* !!!
1732 char *str = new char[length+1];
1733 item_read = fread(str, (size_t)length, (size_t)1, Fp);
1735 std::string newValue = str;
1737 if ( gdcmValEntry* valEntry = dynamic_cast<gdcmValEntry* >(entry) )
1739 if ( item_read != 1 )
1741 dbg.Verbose(1, "gdcmDocument::LoadDocEntry",
1742 "unread element value");
1743 valEntry->SetValue(GDCM_UNREAD);
1749 // Because of correspondance with the VR dic
1750 valEntry->SetValue(newValue);
1754 valEntry->SetValue(newValue);
1759 dbg.Error(true, "gdcmDocument::LoadDocEntry"
1760 "Should have a ValEntry, here !");
1766 * \brief Find the value Length of the passed Header Entry
1767 * @param entry Header Entry whose length of the value shall be loaded.
1769 void gdcmDocument::FindDocEntryLength( gdcmDocEntry *entry )
1770 throw ( gdcmFormatError )
1772 uint16_t element = entry->GetElement();
1773 std::string vr = entry->GetVR();
1776 if ( Filetype == gdcmExplicitVR && !entry->IsImplicitVR() )
1778 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1780 // The following reserved two bytes (see PS 3.5-2001, section
1781 // 7.1.2 Data element structure with explicit vr p27) must be
1782 // skipped before proceeding on reading the length on 4 bytes.
1783 fseek(Fp, 2L, SEEK_CUR);
1784 uint32_t length32 = ReadInt32();
1786 if ( vr == "OB" && length32 == 0xffffffff )
1791 lengthOB = FindDocEntryLengthOB();
1793 catch ( gdcmFormatUnexpected )
1795 // Computing the length failed (this happens with broken
1796 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1797 // chance to get the pixels by deciding the element goes
1798 // until the end of the file. Hence we artificially fix the
1799 // the length and proceed.
1800 long currentPosition = ftell(Fp);
1801 fseek(Fp,0L,SEEK_END);
1802 long lengthUntilEOF = ftell(Fp) - currentPosition;
1803 fseek(Fp, currentPosition, SEEK_SET);
1804 entry->SetLength(lengthUntilEOF);
1807 entry->SetLength(lengthOB);
1810 FixDocEntryFoundLength(entry, length32);
1814 // Length is encoded on 2 bytes.
1815 length16 = ReadInt16();
1817 // We can tell the current file is encoded in big endian (like
1818 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1819 // and it's value is the one of the encoding of a big endian file.
1820 // In order to deal with such big endian encoded files, we have
1821 // (at least) two strategies:
1822 // * when we load the "Transfer Syntax" tag with value of big endian
1823 // encoding, we raise the proper flags. Then we wait for the end
1824 // of the META group (0x0002) among which is "Transfer Syntax",
1825 // before switching the swap code to big endian. We have to postpone
1826 // the switching of the swap code since the META group is fully encoded
1827 // in little endian, and big endian coding only starts at the next
1828 // group. The corresponding code can be hard to analyse and adds
1829 // many additional unnecessary tests for regular tags.
1830 // * the second strategy consists in waiting for trouble, that shall
1831 // appear when we find the first group with big endian encoding. This
1832 // is easy to detect since the length of a "Group Length" tag (the
1833 // ones with zero as element number) has to be of 4 (0x0004). When we
1834 // encounter 1024 (0x0400) chances are the encoding changed and we
1835 // found a group with big endian encoding.
1836 // We shall use this second strategy. In order to make sure that we
1837 // can interpret the presence of an apparently big endian encoded
1838 // length of a "Group Length" without committing a big mistake, we
1839 // add an additional check: we look in the already parsed elements
1840 // for the presence of a "Transfer Syntax" whose value has to be "big
1841 // endian encoding". When this is the case, chances are we have got our
1842 // hands on a big endian encoded file: we switch the swap code to
1843 // big endian and proceed...
1844 if ( element == 0x0000 && length16 == 0x0400 )
1846 if ( !IsExplicitVRBigEndianTransferSyntax() )
1848 throw gdcmFormatError( "gdcmDocument::FindDocEntryLength()",
1849 " not explicit VR." );
1853 SwitchSwapToBigEndian();
1854 // Restore the unproperly loaded values i.e. the group, the element
1855 // and the dictionary entry depending on them.
1856 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1857 uint16_t correctElem = SwapShort( entry->GetElement() );
1858 gdcmDictEntry* newTag = GetDictEntryByNumber( correctGroup,
1862 // This correct tag is not in the dictionary. Create a new one.
1863 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1865 // FIXME this can create a memory leaks on the old entry that be
1866 // left unreferenced.
1867 entry->SetDictEntry( newTag );
1870 // Heuristic: well, some files are really ill-formed.
1871 if ( length16 == 0xffff)
1873 // 0xffff means that we deal with 'Unknown Length' Sequence
1876 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1881 // Either implicit VR or a non DICOM conformal (see note below) explicit
1882 // VR that ommited the VR of (at least) this element. Farts happen.
1883 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1884 // on Data elements "Implicit and Explicit VR Data Elements shall
1885 // not coexist in a Data Set and Data Sets nested within it".]
1886 // Length is on 4 bytes.
1888 FixDocEntryFoundLength( entry, ReadInt32() );
1894 * \brief Find the Value Representation of the current Dicom Element.
1897 void gdcmDocument::FindDocEntryVR( gdcmDocEntry *entry )
1899 if ( Filetype != gdcmExplicitVR )
1906 long positionOnEntry = ftell(Fp);
1907 // Warning: we believe this is explicit VR (Value Representation) because
1908 // we used a heuristic that found "UL" in the first tag. Alas this
1909 // doesn't guarantee that all the tags will be in explicit VR. In some
1910 // cases (see e-film filtered files) one finds implicit VR tags mixed
1911 // within an explicit VR file. Hence we make sure the present tag
1912 // is in explicit VR and try to fix things if it happens not to be
1915 fread (vr, (size_t)2,(size_t)1, Fp);
1918 if( !CheckDocEntryVR(entry, vr) )
1920 fseek(Fp, positionOnEntry, SEEK_SET);
1921 // When this element is known in the dictionary we shall use, e.g. for
1922 // the semantics (see the usage of IsAnInteger), the VR proposed by the
1923 // dictionary entry. Still we have to flag the element as implicit since
1924 // we know now our assumption on expliciteness is not furfilled.
1926 if ( entry->IsVRUnknown() )
1928 entry->SetVR("Implicit");
1930 entry->SetImplicitVR();
1935 * \brief Check the correspondance between the VR of the header entry
1936 * and the taken VR. If they are different, the header entry is
1937 * updated with the new VR.
1938 * @param entry Header Entry to check
1939 * @param vr Dicom Value Representation
1940 * @return false if the VR is incorrect of if the VR isn't referenced
1941 * otherwise, it returns true
1943 bool gdcmDocument::CheckDocEntryVR(gdcmDocEntry *entry, gdcmVRKey vr)
1945 char msg[100]; // for sprintf
1946 bool realExplicit = true;
1948 // Assume we are reading a falsely explicit VR file i.e. we reached
1949 // a tag where we expect reading a VR but are in fact we read the
1950 // first to bytes of the length. Then we will interogate (through find)
1951 // the dicom_vr dictionary with oddities like "\004\0" which crashes
1952 // both GCC and VC++ implementations of the STL map. Hence when the
1953 // expected VR read happens to be non-ascii characters we consider
1954 // we hit falsely explicit VR tag.
1956 if ( !isalpha(vr[0]) && !isalpha(vr[1]) )
1958 realExplicit = false;
1961 // CLEANME searching the dicom_vr at each occurence is expensive.
1962 // PostPone this test in an optional integrity check at the end
1963 // of parsing or only in debug mode.
1964 if ( realExplicit && !gdcmGlobal::GetVR()->Count(vr) )
1966 realExplicit = false;
1969 if ( !realExplicit )
1971 // We thought this was explicit VR, but we end up with an
1972 // implicit VR tag. Let's backtrack.
1973 sprintf(msg,"Falsely explicit vr file (%04x,%04x)\n",
1974 entry->GetGroup(), entry->GetElement());
1975 dbg.Verbose(1, "gdcmDocument::FindVR: ",msg);
1977 if( entry->GetGroup() % 2 && entry->GetElement() == 0x0000)
1979 // Group length is UL !
1980 gdcmDictEntry* newEntry = NewVirtualDictEntry(
1981 entry->GetGroup(), entry->GetElement(),
1982 "UL", "FIXME", "Group Length");
1983 entry->SetDictEntry( newEntry );
1988 if ( entry->IsVRUnknown() )
1990 // When not a dictionary entry, we can safely overwrite the VR.
1991 if( entry->GetElement() == 0x0000 )
1993 // Group length is UL !
2001 else if ( entry->GetVR() != vr )
2003 // The VR present in the file and the dictionary disagree. We assume
2004 // the file writer knew best and use the VR of the file. Since it would
2005 // be unwise to overwrite the VR of a dictionary (since it would
2006 // compromise it's next user), we need to clone the actual DictEntry
2007 // and change the VR for the read one.
2008 gdcmDictEntry* newEntry = NewVirtualDictEntry(
2009 entry->GetGroup(), entry->GetElement(),
2010 vr, "FIXME", entry->GetName());
2011 entry->SetDictEntry(newEntry);
2018 * \brief Get the transformed value of the header entry. The VR value
2019 * is used to define the transformation to operate on the value
2020 * \warning NOT end user intended method !
2021 * @param entry entry to tranform
2022 * @return Transformed entry value
2024 std::string gdcmDocument::GetDocEntryValue(gdcmDocEntry *entry)
2026 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2028 std::string val = ((gdcmValEntry *)entry)->GetValue();
2029 std::string vr = entry->GetVR();
2030 uint32_t length = entry->GetLength();
2031 std::ostringstream s;
2034 // When short integer(s) are expected, read and convert the following
2035 // n * 2 bytes properly i.e. as a multivaluated strings
2036 // (each single value is separated fromthe next one by '\'
2037 // as usual for standard multivaluated filels
2038 // Elements with Value Multiplicity > 1
2039 // contain a set of short integers (not a single one)
2041 if( vr == "US" || vr == "SS" )
2046 for (int i=0; i < nbInt; i++)
2052 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
2053 newInt16 = SwapShort( newInt16 );
2058 // When integer(s) are expected, read and convert the following
2059 // n * 4 bytes properly i.e. as a multivaluated strings
2060 // (each single value is separated fromthe next one by '\'
2061 // as usual for standard multivaluated filels
2062 // Elements with Value Multiplicity > 1
2063 // contain a set of integers (not a single one)
2064 else if( vr == "UL" || vr == "SL" )
2069 for (int i=0; i < nbInt; i++)
2075 newInt32 = ( val[4*i+0] & 0xFF )
2076 + (( val[4*i+1] & 0xFF ) << 8 )
2077 + (( val[4*i+2] & 0xFF ) << 16 )
2078 + (( val[4*i+3] & 0xFF ) << 24 );
2079 newInt32 = SwapLong( newInt32 );
2083 #ifdef GDCM_NO_ANSI_STRING_STREAM
2084 s << std::ends; // to avoid oddities on Solaris
2085 #endif //GDCM_NO_ANSI_STRING_STREAM
2089 return ((gdcmValEntry *)entry)->GetValue();
2093 * \brief Get the reverse transformed value of the header entry. The VR
2094 * value is used to define the reverse transformation to operate on
2096 * \warning NOT end user intended method !
2097 * @param entry Entry to reverse transform
2098 * @return Reverse transformed entry value
2100 std::string gdcmDocument::GetDocEntryUnvalue(gdcmDocEntry* entry)
2102 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2104 std::string vr = entry->GetVR();
2105 std::vector<std::string> tokens;
2106 std::ostringstream s;
2108 if ( vr == "US" || vr == "SS" )
2112 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
2113 Tokenize (((gdcmValEntry *)entry)->GetValue(), tokens, "\\");
2114 for (unsigned int i=0; i<tokens.size(); i++)
2116 newInt16 = atoi(tokens[i].c_str());
2117 s << ( newInt16 & 0xFF )
2118 << (( newInt16 >> 8 ) & 0xFF );
2122 if ( vr == "UL" || vr == "SL")
2126 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
2127 Tokenize (((gdcmValEntry *)entry)->GetValue(), tokens, "\\");
2128 for (unsigned int i=0; i<tokens.size();i++)
2130 newInt32 = atoi(tokens[i].c_str());
2131 s << (char)( newInt32 & 0xFF )
2132 << (char)(( newInt32 >> 8 ) & 0xFF )
2133 << (char)(( newInt32 >> 16 ) & 0xFF )
2134 << (char)(( newInt32 >> 24 ) & 0xFF );
2139 #ifdef GDCM_NO_ANSI_STRING_STREAM
2140 s << std::ends; // to avoid oddities on Solaris
2141 #endif //GDCM_NO_ANSI_STRING_STREAM
2145 return ((gdcmValEntry *)entry)->GetValue();
2149 * \brief Skip a given Header Entry
2150 * \warning NOT end user intended method !
2151 * @param entry entry to skip
2153 void gdcmDocument::SkipDocEntry(gdcmDocEntry *entry)
2155 SkipBytes(entry->GetLength());
2159 * \brief Skips to the begining of the next Header Entry
2160 * \warning NOT end user intended method !
2161 * @param entry entry to skip
2163 void gdcmDocument::SkipToNextDocEntry(gdcmDocEntry *entry)
2165 fseek(Fp, (long)(entry->GetOffset()), SEEK_SET);
2166 fseek(Fp, (long)(entry->GetReadLength()), SEEK_CUR);
2170 * \brief When the length of an element value is obviously wrong (because
2171 * the parser went Jabberwocky) one can hope improving things by
2172 * applying some heuristics.
2173 * @param entry entry to check
2174 * @param foundLength fist assumption about length
2176 void gdcmDocument::FixDocEntryFoundLength(gdcmDocEntry *entry,
2177 uint32_t foundLength)
2179 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
2180 if ( foundLength == 0xffffffff)
2185 uint16_t gr = entry->GetGroup();
2186 uint16_t el = entry->GetElement();
2188 if ( foundLength % 2)
2190 std::ostringstream s;
2191 s << "Warning : Tag with uneven length "
2193 << " in x(" << std::hex << gr << "," << el <<")" << std::dec;
2194 dbg.Verbose(0, s.str().c_str());
2197 //////// Fix for some naughty General Electric images.
2198 // Allthough not recent many such GE corrupted images are still present
2199 // on Creatis hard disks. Hence this fix shall remain when such images
2200 // are no longer in user (we are talking a few years, here)...
2201 // Note: XMedCom probably uses such a trick since it is able to read
2202 // those pesky GE images ...
2203 if ( foundLength == 13)
2205 // Only happens for this length !
2206 if ( entry->GetGroup() != 0x0008
2207 || ( entry->GetElement() != 0x0070
2208 && entry->GetElement() != 0x0080 ) )
2211 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
2215 //////// Fix for some brain-dead 'Leonardo' Siemens images.
2216 // Occurence of such images is quite low (unless one leaves close to a
2217 // 'Leonardo' source. Hence, one might consider commenting out the
2218 // following fix on efficiency reasons.
2219 else if ( entry->GetGroup() == 0x0009
2220 && ( entry->GetElement() == 0x1113
2221 || entry->GetElement() == 0x1114 ) )
2224 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
2227 else if ( entry->GetVR() == "SQ" )
2229 foundLength = 0; // ReadLength is unchanged
2232 //////// We encountered a 'delimiter' element i.e. a tag of the form
2233 // "fffe|xxxx" which is just a marker. Delimiters length should not be
2234 // taken into account.
2235 else if( entry->GetGroup() == 0xfffe )
2237 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
2238 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
2239 // causes extra troubles...
2240 if( entry->GetElement() != 0x0000 )
2246 entry->SetUsableLength(foundLength);
2250 * \brief Apply some heuristics to predict whether the considered
2251 * element value contains/represents an integer or not.
2252 * @param entry The element value on which to apply the predicate.
2253 * @return The result of the heuristical predicate.
2255 bool gdcmDocument::IsDocEntryAnInteger(gdcmDocEntry *entry)
2257 uint16_t element = entry->GetElement();
2258 uint16_t group = entry->GetGroup();
2259 std::string vr = entry->GetVR();
2260 uint32_t length = entry->GetLength();
2262 // When we have some semantics on the element we just read, and if we
2263 // a priori know we are dealing with an integer, then we shall be
2264 // able to swap it's element value properly.
2265 if ( element == 0 ) // This is the group length of the group
2273 // Allthough this should never happen, still some images have a
2274 // corrupted group length [e.g. have a glance at offset x(8336) of
2275 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2276 // Since for dicom compliant and well behaved headers, the present
2277 // test is useless (and might even look a bit paranoid), when we
2278 // encounter such an ill-formed image, we simply display a warning
2279 // message and proceed on parsing (while crossing fingers).
2280 std::ostringstream s;
2281 int filePosition = ftell(Fp);
2282 s << "Erroneous Group Length element length on : (" \
2283 << std::hex << group << " , " << element
2284 << ") -before- position x(" << filePosition << ")"
2285 << "lgt : " << length;
2286 dbg.Verbose(0, "gdcmDocument::IsDocEntryAnInteger", s.str().c_str() );
2290 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2299 * \brief Find the Length till the next sequence delimiter
2300 * \warning NOT end user intended method !
2304 uint32_t gdcmDocument::FindDocEntryLengthOB()
2305 throw( gdcmFormatUnexpected )
2307 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2308 long positionOnEntry = ftell(Fp);
2309 bool foundSequenceDelimiter = false;
2310 uint32_t totalLength = 0;
2312 while ( !foundSequenceDelimiter )
2318 group = ReadInt16();
2321 catch ( gdcmFormatError )
2323 throw gdcmFormatError("gdcmDocument::FindDocEntryLengthOB()",
2324 " group or element not present.");
2327 // We have to decount the group and element we just read
2330 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2332 dbg.Verbose(1, "gdcmDocument::FindDocEntryLengthOB: neither an Item "
2333 "tag nor a Sequence delimiter tag.");
2334 fseek(Fp, positionOnEntry, SEEK_SET);
2335 throw gdcmFormatUnexpected("gdcmDocument::FindDocEntryLengthOB()",
2336 "Neither an Item tag nor a Sequence "
2340 if ( elem == 0xe0dd )
2342 foundSequenceDelimiter = true;
2345 uint32_t itemLength = ReadInt32();
2346 // We add 4 bytes since we just read the ItemLength with ReadInt32
2347 totalLength += itemLength + 4;
2348 SkipBytes(itemLength);
2350 if ( foundSequenceDelimiter )
2355 fseek(Fp, positionOnEntry, SEEK_SET);
2360 * \brief Reads a supposed to be 16 Bits integer
2361 * (swaps it depending on processor endianity)
2362 * @return read value
2364 uint16_t gdcmDocument::ReadInt16()
2365 throw( gdcmFormatError )
2368 size_t item_read = fread (&g, (size_t)2,(size_t)1, Fp);
2369 if ( item_read != 1 )
2373 throw gdcmFormatError( "gdcmDocument::ReadInt16()", " file error." );
2375 throw gdcmFormatError( "gdcmDocument::ReadInt16()", "EOF." );
2382 * \brief Reads a supposed to be 32 Bits integer
2383 * (swaps it depending on processor endianity)
2384 * @return read value
2386 uint32_t gdcmDocument::ReadInt32()
2387 throw( gdcmFormatError )
2390 size_t item_read = fread (&g, (size_t)4,(size_t)1, Fp);
2391 if ( item_read != 1 )
2395 throw gdcmFormatError( "gdcmDocument::ReadInt16()", " file error." );
2397 throw gdcmFormatError( "gdcmDocument::ReadInt32()", "EOF." );
2404 * \brief skips bytes inside the source file
2405 * \warning NOT end user intended method !
2408 void gdcmDocument::SkipBytes(uint32_t nBytes)
2410 //FIXME don't dump the returned value
2411 (void)fseek(Fp, (long)nBytes, SEEK_CUR);
2415 * \brief Loads all the needed Dictionaries
2416 * \warning NOT end user intended method !
2418 void gdcmDocument::Initialise()
2420 RefPubDict = gdcmGlobal::GetDicts()->GetDefaultPubDict();
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 gdcmDocument::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
2444 char deb[HEADER_LENGTH_TO_READ];
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".
2459 int lgrLue = fread(deb, 1, HEADER_LENGTH_TO_READ, Fp);
2460 (void)lgrLue; //FIXME not used
2462 char *entCur = deb + 128;
2463 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2465 dbg.Verbose(1, "gdcmDocument::CheckSwap:", "looks like DICOM Version3");
2467 // Next, determine the value representation (VR). Let's skip to the
2468 // first element (0002, 0000) and check there if we find "UL"
2469 // - or "OB" if the 1st one is (0002,0001) -,
2470 // in which case we (almost) know it is explicit VR.
2471 // WARNING: if it happens to be implicit VR then what we will read
2472 // is the length of the group. If this ascii representation of this
2473 // length happens to be "UL" then we shall believe it is explicit VR.
2474 // FIXME: in order to fix the above warning, we could read the next
2475 // element value (or a couple of elements values) in order to make
2476 // sure we are not commiting a big mistake.
2477 // We need to skip :
2478 // * the 128 bytes of File Preamble (often padded with zeroes),
2479 // * the 4 bytes of "DICM" string,
2480 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2481 // i.e. a total of 136 bytes.
2485 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2486 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2487 // *Implicit* VR. -and it is !-
2489 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2490 memcmp(entCur, "OB", (size_t)2) == 0 ||
2491 memcmp(entCur, "UI", (size_t)2) == 0 ||
2492 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2493 // when Write DCM *adds*
2495 // Use gdcmDocument::dicom_vr to test all the possibilities
2496 // instead of just checking for UL, OB and UI !? group 0000
2498 Filetype = gdcmExplicitVR;
2499 dbg.Verbose(1, "gdcmDocument::CheckSwap:",
2500 "explicit Value Representation");
2504 Filetype = gdcmImplicitVR;
2505 dbg.Verbose(1, "gdcmDocument::CheckSwap:",
2506 "not an explicit Value Representation");
2512 dbg.Verbose(1, "gdcmDocument::CheckSwap:",
2513 "HostByteOrder != NetworkByteOrder");
2518 dbg.Verbose(1, "gdcmDocument::CheckSwap:",
2519 "HostByteOrder = NetworkByteOrder");
2522 // Position the file position indicator at first tag (i.e.
2523 // after the file preamble and the "DICM" string).
2525 fseek (Fp, 132L, SEEK_SET);
2529 // Alas, this is not a DicomV3 file and whatever happens there is no file
2530 // preamble. We can reset the file position indicator to where the data
2531 // is (i.e. the beginning of the file).
2532 dbg.Verbose(1, "gdcmDocument::CheckSwap:", "not a DICOM Version3 file");
2535 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2536 // By clean we mean that the length of the first tag is written down.
2537 // If this is the case and since the length of the first group HAS to be
2538 // four (bytes), then determining the proper swap code is straightforward.
2541 // We assume the array of char we are considering contains the binary
2542 // representation of a 32 bits integer. Hence the following dirty
2544 s32 = *((uint32_t *)(entCur));
2565 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2566 // It is time for despaired wild guesses.
2567 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2568 // i.e. the 'group length' element is not present :
2570 // check the supposed to be 'group number'
2571 // 0x0002 or 0x0004 or 0x0008
2572 // to determine ' SwapCode' value .
2573 // Only 0 or 4321 will be possible
2574 // (no oportunity to check for the formerly well known
2575 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2576 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -4, 8-)
2577 // the file IS NOT ACR-NEMA nor DICOM V3
2578 // Find a trick to tell it the caller...
2580 s16 = *((uint16_t *)(deb));
2597 dbg.Verbose(0, "gdcmDocument::CheckSwap:",
2598 "ACR/NEMA unfound swap info (Really hopeless !)");
2599 Filetype = gdcmUnknown;
2602 // Then the only info we have is the net2host one.
2612 * \brief Restore the unproperly loaded values i.e. the group, the element
2613 * and the dictionary entry depending on them.
2615 void gdcmDocument::SwitchSwapToBigEndian()
2617 dbg.Verbose(1, "gdcmDocument::SwitchSwapToBigEndian",
2618 "Switching to BigEndian mode.");
2619 if ( SwapCode == 0 )
2623 else if ( SwapCode == 4321 )
2627 else if ( SwapCode == 3412 )
2631 else if ( SwapCode == 2143 )
2638 * \brief during parsing, Header Elements too long are not loaded in memory
2641 void gdcmDocument::SetMaxSizeLoadEntry(long newSize)
2647 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2649 MaxSizeLoadEntry = 0xffffffff;
2652 MaxSizeLoadEntry = newSize;
2657 * \brief Header Elements too long will not be printed
2658 * \todo See comments of \ref gdcmDocument::MAX_SIZE_PRINT_ELEMENT_VALUE
2661 void gdcmDocument::SetMaxSizePrintEntry(long newSize)
2663 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2668 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2670 MaxSizePrintEntry = 0xffffffff;
2673 MaxSizePrintEntry = newSize;
2679 * \brief Read the next tag but WITHOUT loading it's value
2680 * (read the 'Group Number', the 'Element Number',
2681 * gets the Dict Entry
2682 * gets the VR, gets the length, gets the offset value)
2683 * @return On succes the newly created DocEntry, NULL on failure.
2685 gdcmDocEntry* gdcmDocument::ReadNextDocEntry()
2692 group = ReadInt16();
2695 catch ( gdcmFormatError e )
2697 // We reached the EOF (or an error occured) therefore
2698 // header parsing has to be considered as finished.
2703 gdcmDocEntry *newEntry = NewDocEntryByNumber(group, elem);
2704 FindDocEntryVR(newEntry);
2708 FindDocEntryLength(newEntry);
2710 catch ( gdcmFormatError e )
2718 newEntry->SetOffset(ftell(Fp));
2725 * \brief Generate a free gdcmTagKey i.e. a gdcmTagKey that is not present
2726 * in the TagHt dictionary.
2727 * @param group The generated tag must belong to this group.
2728 * @return The element of tag with given group which is fee.
2730 uint32_t gdcmDocument::GenerateFreeTagKeyInGroup(uint16_t group)
2732 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2734 gdcmTagKey key = gdcmDictEntry::TranslateToKey(group, elem);
2735 if (TagHT.count(key) == 0)
2744 * \brief Assuming the internal file pointer \ref gdcmDocument::Fp
2745 * is placed at the beginning of a tag check whether this
2746 * tag is (TestGroup, TestElement).
2747 * \warning On success the internal file pointer \ref gdcmDocument::Fp
2748 * is modified to point after the tag.
2749 * On failure (i.e. when the tag wasn't the expected tag
2750 * (TestGroup, TestElement) the internal file pointer
2751 * \ref gdcmDocument::Fp is restored to it's original position.
2752 * @param testGroup The expected group of the tag.
2753 * @param testElement The expected Element of the tag.
2754 * @return True on success, false otherwise.
2756 bool gdcmDocument::ReadTag(uint16_t testGroup, uint16_t testElement)
2758 long positionOnEntry = ftell(Fp);
2759 long currentPosition = ftell(Fp); // On debugging purposes
2761 //// Read the Item Tag group and element, and make
2762 // sure they are what we expected:
2763 uint16_t itemTagGroup = ReadInt16();
2764 uint16_t itemTagElement = ReadInt16();
2765 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2767 std::ostringstream s;
2768 s << " We should have found tag (";
2769 s << std::hex << testGroup << "," << testElement << ")" << std::endl;
2770 s << " but instead we encountered tag (";
2771 s << std::hex << itemTagGroup << "," << itemTagElement << ")"
2773 s << " at address: " << (unsigned)currentPosition << std::endl;
2774 dbg.Verbose(0, "gdcmDocument::ReadItemTagLength: wrong Item Tag found:");
2775 dbg.Verbose(0, s.str().c_str());
2776 fseek(Fp, positionOnEntry, SEEK_SET);
2784 * \brief Assuming the internal file pointer \ref gdcmDocument::Fp
2785 * is placed at the beginning of a tag (TestGroup, TestElement),
2786 * read the length associated to the Tag.
2787 * \warning On success the internal file pointer \ref gdcmDocument::Fp
2788 * is modified to point after the tag and it's length.
2789 * On failure (i.e. when the tag wasn't the expected tag
2790 * (TestGroup, TestElement) the internal file pointer
2791 * \ref gdcmDocument::Fp is restored to it's original position.
2792 * @param testGroup The expected group of the tag.
2793 * @param testElement The expected Element of the tag.
2794 * @return On success returns the length associated to the tag. On failure
2797 uint32_t gdcmDocument::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2799 long positionOnEntry = ftell(Fp);
2800 (void)positionOnEntry;
2802 if ( !ReadTag(testGroup, testElement) )
2807 //// Then read the associated Item Length
2808 long currentPosition = ftell(Fp);
2809 uint32_t itemLength = ReadInt32();
2811 std::ostringstream s;
2812 s << "Basic Item Length is: "
2813 << itemLength << std::endl;
2814 s << " at address: " << (unsigned)currentPosition << std::endl;
2815 dbg.Verbose(0, "gdcmDocument::ReadItemTagLength: ", s.str().c_str());
2821 * \brief Parse pixel data from disk for multi-fragment Jpeg/Rle files
2822 * No other way so 'skip' the Data
2824 void gdcmDocument::Parse7FE0 ()
2826 gdcmDocEntry* element = GetDocEntryByNumber(0x0002, 0x0010);
2829 // Should warn user FIXME
2833 if ( IsImplicitVRLittleEndianTransferSyntax()
2834 || IsExplicitVRLittleEndianTransferSyntax()
2835 || IsExplicitVRBigEndianTransferSyntax() /// \todo 1.2.2 ??? A verifier !
2836 || IsDeflatedExplicitVRLittleEndianTransferSyntax() )
2841 // ---------------- for Parsing : Position on begining of Jpeg/RLE Pixels
2843 //// Read the Basic Offset Table Item Tag length...
2844 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2846 //// ... and then read length[s] itself[themselves]. We don't use
2847 // the values read (BTW what is the purpous of those lengths ?)
2848 if ( itemLength != 0 )
2850 // BTW, what is the purpous of those length anyhow !?
2851 char* basicOffsetTableItemValue = new char[itemLength + 1];
2852 fread(basicOffsetTableItemValue, itemLength, 1, Fp);
2854 for (unsigned int i=0; i < itemLength; i += 4 )
2856 uint32_t individualLength = str2num(&basicOffsetTableItemValue[i],uint32_t);
2857 std::ostringstream s;
2858 s << " Read one length: ";
2859 s << std::hex << individualLength << std::endl;
2860 dbg.Verbose(0, "gdcmDocument::Parse7FE0: ", s.str().c_str());
2862 delete[] basicOffsetTableItemValue;
2865 if ( ! IsRLELossLessTransferSyntax() )
2869 //// We then skip (not reading them) all the fragments of images:
2870 while ( (itemLength = ReadTagLength(0xfffe, 0xe000)) )
2872 SkipBytes(itemLength);
2879 long rleSegmentLength[15], fragmentLength;
2881 // While we find some items:
2882 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2884 // Parse fragments of the current Fragment (Frame)
2885 //------------------ scanning (not reading) fragment pixels
2886 uint32_t nbRleSegments = ReadInt32();
2888 //// Reading RLE Segments Offset Table
2889 uint32_t rleSegmentOffsetTable[15];
2890 for(int k=1; k<=15; k++)
2892 ftellRes = ftell(Fp);
2893 rleSegmentOffsetTable[k] = ReadInt32();
2896 // skipping (not reading) RLE Segments
2897 if ( nbRleSegments > 1)
2899 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2901 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2902 - rleSegmentOffsetTable[k];
2903 ftellRes = ftell(Fp);
2904 SkipBytes(rleSegmentLength[k]);
2908 rleSegmentLength[nbRleSegments] = fragmentLength
2909 - rleSegmentOffsetTable[nbRleSegments];
2910 ftellRes = ftell(Fp);
2911 SkipBytes(rleSegmentLength[nbRleSegments]);
2914 // Make sure that at the end of the item we encounter a 'Sequence
2916 if ( !ReadTag(0xfffe, 0xe0dd) )
2918 dbg.Verbose(0, "gdcmDocument::Parse7FE0: no sequence delimiter item");
2919 dbg.Verbose(0, " at end of RLE item sequence");
2925 * \brief Walk recursively the given \ref gdcmDocEntrySet, and feed
2926 * the given hash table (\ref TagDocEntryHT) with all the
2927 * \ref gdcmDocEntry (Dicom entries) encountered.
2928 * This method does the job for \ref BuildFlatHashTable.
2929 * @param builtHT Where to collect all the \ref gdcmDocEntry encountered
2930 * when recursively walking the given set.
2931 * @param set The structure to be traversed (recursively).
2933 void gdcmDocument::BuildFlatHashTableRecurse( TagDocEntryHT& builtHT,
2934 gdcmDocEntrySet* set )
2936 if (gdcmElementSet* elementSet = dynamic_cast< gdcmElementSet* > ( set ) )
2938 TagDocEntryHT* currentHT = elementSet->GetTagHT();
2939 for( TagDocEntryHT::const_iterator i = currentHT->begin();
2940 i != currentHT->end();
2943 gdcmDocEntry* entry = i->second;
2944 if ( gdcmSeqEntry* seqEntry = dynamic_cast<gdcmSeqEntry*>(entry) )
2946 ListSQItem& items = seqEntry->GetSQItems();
2947 for( ListSQItem::const_iterator item = items.begin();
2948 item != items.end();
2951 BuildFlatHashTableRecurse( builtHT, *item );
2955 builtHT[entry->GetKey()] = entry;
2960 if (gdcmSQItem* SQItemSet = dynamic_cast< gdcmSQItem* > ( set ) )
2962 ListDocEntry& currentList = SQItemSet->GetDocEntries();
2963 for (ListDocEntry::iterator i = currentList.begin();
2964 i != currentList.end();
2967 gdcmDocEntry* entry = *i;
2968 if ( gdcmSeqEntry* seqEntry = dynamic_cast<gdcmSeqEntry*>(entry) )
2970 ListSQItem& items = seqEntry->GetSQItems();
2971 for( ListSQItem::const_iterator item = items.begin();
2972 item != items.end();
2975 BuildFlatHashTableRecurse( builtHT, *item );
2979 builtHT[entry->GetKey()] = entry;
2986 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2989 * The structure used by a gdcmDocument (through \ref gdcmElementSet),
2990 * in order to old the parsed entries of a Dicom header, is a recursive
2991 * one. This is due to the fact that the sequences (when present)
2992 * can be nested. Additionaly, the sequence items (represented in
2993 * gdcm as \ref gdcmSQItem) add an extra complexity to the data
2994 * structure. Hence, a gdcm user whishing to visit all the entries of
2995 * a Dicom header will need to dig in the gdcm internals (which
2996 * implies exposing all the internal data structures to the API).
2997 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2998 * recursively builds a temporary hash table, which holds all the
2999 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
3001 * \warning Of course there is NO integrity constrain between the
3002 * returned \ref TagDocEntryHT and the \ref gdcmElementSet used
3003 * to build it. Hence if the underlying \ref gdcmElementSet is
3004 * altered, then it is the caller responsability to invoke
3005 * \ref BuildFlatHashTable again...
3006 * @return The flat std::map<> we juste build.
3008 TagDocEntryHT* gdcmDocument::BuildFlatHashTable()
3010 TagDocEntryHT* FlatHT = new TagDocEntryHT;
3011 BuildFlatHashTableRecurse( *FlatHT, this );
3018 * \brief Compares two documents, according to \ref gdcmDicomDir rules
3019 * \warning Does NOT work with ACR-NEMA files
3020 * \todo Find a trick to solve the pb (use RET fields ?)
3022 * @return true if 'smaller'
3024 bool gdcmDocument::operator<(gdcmDocument &document)
3027 std::string s1 = GetEntryByNumber(0x0010,0x0010);
3028 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
3040 s1 = GetEntryByNumber(0x0010,0x0020);
3041 s2 = document.GetEntryByNumber(0x0010,0x0020);
3052 // Study Instance UID
3053 s1 = GetEntryByNumber(0x0020,0x000d);
3054 s2 = document.GetEntryByNumber(0x0020,0x000d);
3065 // Serie Instance UID
3066 s1 = GetEntryByNumber(0x0020,0x000e);
3067 s2 = document.GetEntryByNumber(0x0020,0x000e);
3083 //-----------------------------------------------------------------------------