1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/05 16:53:23 $
7 Version: $Revision: 1.152 $
9 Copyright (c) CREATIS (Centre de Recherche et d'Applications en Traitement de
10 l'Image). All rights reserved. See Doc/License.txt or
11 http://www.creatis.insa-lyon.fr/Public/Gdcm/License.html for details.
13 This software is distributed WITHOUT ANY WARRANTY; without even
14 the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
15 PURPOSE. See the above copyright notices for more information.
17 =========================================================================*/
19 #include "gdcmDocument.h"
20 #include "gdcmValEntry.h"
21 #include "gdcmBinEntry.h"
22 #include "gdcmSeqEntry.h"
23 #include "gdcmGlobal.h"
25 #include "gdcmDebug.h"
27 #include "gdcmException.h"
28 #include "gdcmDictSet.h"
29 #include "gdcmRLEFramesInfo.h"
30 #include "gdcmJPEGFragmentsInfo.h"
31 #include "gdcmDocEntrySet.h"
32 #include "gdcmSQItem.h"
38 #if defined(_MSC_VER) || defined(__BORLANDC__)
41 #include <netinet/in.h>
46 //-----------------------------------------------------------------------------
47 static const char *TransferSyntaxStrings[] = {
48 // Implicit VR Little Endian
50 // Implicit VR Little Endian DLX G.E?
52 // Explicit VR Little Endian
53 "1.2.840.10008.1.2.1",
54 // Deflated Explicit VR Little Endian
55 "1.2.840.10008.1.2.1.99",
56 // Explicit VR Big Endian
57 "1.2.840.10008.1.2.2",
58 // JPEG Baseline (Process 1)
59 "1.2.840.10008.1.2.4.50",
60 // JPEG Extended (Process 2 & 4)
61 "1.2.840.10008.1.2.4.51",
62 // JPEG Extended (Process 3 & 5)
63 "1.2.840.10008.1.2.4.52",
64 // JPEG Spectral Selection, Non-Hierarchical (Process 6 & 8)
65 "1.2.840.10008.1.2.4.53",
66 // JPEG Full Progression, Non-Hierarchical (Process 10 & 12)
67 "1.2.840.10008.1.2.4.55",
68 // JPEG Lossless, Non-Hierarchical (Process 14)
69 "1.2.840.10008.1.2.4.57",
70 // JPEG Lossless, Hierarchical, First-Order Prediction (Process 14, [Selection Value 1])
71 "1.2.840.10008.1.2.4.70",
73 "1.2.840.10008.1.2.4.90",
75 "1.2.840.10008.1.2.4.91",
77 "1.2.840.10008.1.2.5",
79 "Unknown Transfer Syntax"
82 //-----------------------------------------------------------------------------
83 // Refer to Document::CheckSwap()
84 //const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
86 // Refer to Document::SetMaxSizeLoadEntry()
87 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
88 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
90 //-----------------------------------------------------------------------------
91 // Constructor / Destructor
95 * @param filename file to be opened for parsing
97 Document::Document( std::string const & filename ) : ElementSet(-1)
99 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
109 dbg.Verbose(0, "Document::Document: starting parsing of file: ",
111 Fp->seekg( 0, std::ios::beg);
113 Fp->seekg(0, std::ios::end);
114 long lgt = Fp->tellg();
116 Fp->seekg( 0, std::ios::beg);
118 long beg = Fp->tellg();
121 ParseDES( this, beg, lgt, false); // le Load sera fait a la volee
123 Fp->seekg( 0, std::ios::beg);
125 // Load 'non string' values
127 std::string PhotometricInterpretation = GetEntryByNumber(0x0028,0x0004);
128 if( PhotometricInterpretation == "PALETTE COLOR " )
130 LoadEntryBinArea(0x0028,0x1200); // gray LUT
131 /// FIXME FIXME FIXME
132 /// The tags refered by the three following lines used to be CORRECTLY
133 /// defined as having an US Value Representation in the public
134 /// dictionnary. BUT the semantics implied by the three following
135 /// lines state that the corresponding tag contents are in fact
136 /// the ones of a BinEntry.
137 /// In order to fix things "Quick and Dirty" the dictionnary was
138 /// altered on PURPOUS but now contains a WRONG value.
139 /// In order to fix things and restore the dictionary to its
140 /// correct value, one needs to decided of the semantics by deciding
141 /// wether the following tags are either:
142 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
143 /// also used as BinEntry, which requires the proper conversion,
144 /// - OW, and hence loaded as BinEntry, but afterwards also used
145 /// as ValEntry, which requires the proper conversion.
146 LoadEntryBinArea(0x0028,0x1201); // R LUT
147 LoadEntryBinArea(0x0028,0x1202); // G LUT
148 LoadEntryBinArea(0x0028,0x1203); // B LUT
150 // Segmented Red Palette Color LUT Data
151 LoadEntryBinArea(0x0028,0x1221);
152 // Segmented Green Palette Color LUT Data
153 LoadEntryBinArea(0x0028,0x1222);
154 // Segmented Blue Palette Color LUT Data
155 LoadEntryBinArea(0x0028,0x1223);
157 //FIXME later : how to use it?
158 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
162 // --------------------------------------------------------------
163 // Specific code to allow gdcm to read ACR-LibIDO formated images
164 // Note: ACR-LibIDO is an extension of the ACR standard that was
165 // used at CREATIS. For the time being (say a couple years)
166 // we keep this kludge to allow a smooth move to gdcm for
167 // CREATIS developpers (sorry folks).
169 // if recognition code tells us we deal with a LibIDO image
170 // we switch lineNumber and columnNumber
173 RecCode = GetEntryByNumber(0x0008, 0x0010); // recognition code
174 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
175 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
176 // with "little-endian strings"
178 Filetype = ACR_LIBIDO;
179 std::string rows = GetEntryByNumber(0x0028, 0x0010);
180 std::string columns = GetEntryByNumber(0x0028, 0x0011);
181 SetEntryByNumber(columns, 0x0028, 0x0010);
182 SetEntryByNumber(rows , 0x0028, 0x0011);
184 // ----------------- End of ACR-LibIDO kludge ------------------
188 * \brief This default constructor doesn't parse the file. You should
189 * then invoke \ref Document::SetFileName and then the parsing.
191 Document::Document() : ElementSet(-1)
195 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
198 Filetype = ExplicitVR;
202 * \brief Canonical destructor.
204 Document::~Document ()
213 //-----------------------------------------------------------------------------
217 * \brief Prints The Dict Entries of THE public Dicom Dictionary
220 void Document::PrintPubDict(std::ostream & os)
222 RefPubDict->SetPrintLevel(PrintLevel);
223 RefPubDict->Print(os);
227 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
230 void Document::PrintShaDict(std::ostream & os)
232 RefShaDict->SetPrintLevel(PrintLevel);
233 RefShaDict->Print(os);
236 //-----------------------------------------------------------------------------
239 * \brief Get the public dictionary used
241 Dict* Document::GetPubDict()
247 * \brief Get the shadow dictionary used
249 Dict* Document::GetShaDict()
255 * \brief Set the shadow dictionary used
256 * \param dict dictionary to use in shadow
258 bool Document::SetShaDict(Dict *dict)
265 * \brief Set the shadow dictionary used
266 * \param dictName name of the dictionary to use in shadow
268 bool Document::SetShaDict(DictKey const & dictName)
270 RefShaDict = Global::GetDicts()->GetDict(dictName);
275 * \brief This predicate, based on hopefully reasonable heuristics,
276 * decides whether or not the current Document was properly parsed
277 * and contains the mandatory information for being considered as
278 * a well formed and usable Dicom/Acr File.
279 * @return true when Document is the one of a reasonable Dicom/Acr file,
282 bool Document::IsReadable()
284 if( Filetype == Unknown)
286 dbg.Verbose(0, "Document::IsReadable: wrong filetype");
292 dbg.Verbose(0, "Document::IsReadable: no tags in internal"
301 * \brief Accessor to the Transfer Syntax (when present) of the
302 * current document (it internally handles reading the
303 * value from disk when only parsing occured).
304 * @return The encountered Transfer Syntax of the current document.
306 TransferSyntaxType Document::GetTransferSyntax()
308 DocEntry *entry = GetDocEntryByNumber(0x0002, 0x0010);
314 // The entry might be present but not loaded (parsing and loading
315 // happen at different stages): try loading and proceed with check...
316 LoadDocEntrySafe(entry);
317 if (ValEntry* valEntry = dynamic_cast< ValEntry* >(entry) )
319 std::string transfer = valEntry->GetValue();
320 // The actual transfer (as read from disk) might be padded. We
321 // first need to remove the potential padding. We can make the
322 // weak assumption that padding was not executed with digits...
323 if ( transfer.length() == 0 )
325 // for brain damaged headers
328 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
330 transfer.erase(transfer.length()-1, 1);
332 for (int i = 0; TransferSyntaxStrings[i] != NULL; i++)
334 if ( TransferSyntaxStrings[i] == transfer )
336 return TransferSyntaxType(i);
343 bool Document::IsJPEGLossless()
345 TransferSyntaxType r = GetTransferSyntax();
346 return r == JPEGFullProgressionProcess10_12
347 || r == JPEGLosslessProcess14
348 || r == JPEGLosslessProcess14_1;
352 * \brief Determines if the Transfer Syntax was already encountered
353 * and if it corresponds to a JPEG2000 one
354 * @return True when JPEG2000 (Lossly or LossLess) found. False in all
357 bool Document::IsJPEG2000()
359 TransferSyntaxType r = GetTransferSyntax();
360 return r == JPEG2000Lossless || r == JPEG2000;
364 * \brief Determines if the Transfer Syntax corresponds to any form
365 * of Jpeg encoded Pixel data.
366 * @return True when any form of JPEG found. False otherwise.
368 bool Document::IsJPEG()
370 TransferSyntaxType r = GetTransferSyntax();
371 return r == JPEGBaselineProcess1
372 || r == JPEGExtendedProcess2_4
373 || r == JPEGExtendedProcess3_5
374 || r == JPEGSpectralSelectionProcess6_8
380 * \brief Determines if the Transfer Syntax corresponds to encapsulated
381 * of encoded Pixel Data (as opposed to native).
382 * @return True when encapsulated. False when native.
384 bool Document::IsEncapsulate()
386 TransferSyntaxType r = GetTransferSyntax();
387 return IsJPEG() || r == RLELossless;
391 * \brief Predicate for dicom version 3 file.
392 * @return True when the file is a dicom version 3.
394 bool Document::IsDicomV3()
396 // Checking if Transfert Syntax exists is enough
397 // Anyway, it's to late check if the 'Preamble' was found ...
398 // And ... would it be a rich idea to check ?
399 // (some 'no Preamble' DICOM images exist !)
400 return GetDocEntryByNumber(0x0002, 0x0010) != NULL;
404 * \brief returns the File Type
405 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
406 * @return the FileType code
408 FileType Document::GetFileType()
414 * \brief Tries to open the file \ref Document::Filename and
415 * checks the preamble when existing.
416 * @return The FILE pointer on success.
418 std::ifstream* Document::OpenFile()
420 if (Filename.length() == 0)
428 "Document::OpenFile is already opened when opening: ",
432 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
436 "Document::OpenFile cannot open file: ",
444 Fp->read((char*)&zero, (size_t)2);
451 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
453 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
454 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
455 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
456 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
462 Fp->seekg(126L, std::ios::cur);
464 Fp->read(dicm, (size_t)4);
470 if( memcmp(dicm, "DICM", 4) == 0 )
477 "Document::OpenFile not DICOM/ACR (missing preamble)",
484 * \brief closes the file
485 * @return TRUE if the close was successfull
487 bool Document::CloseFile()
496 return true; //FIXME how do we detect a non-close ifstream ?
500 * \brief Writes in a file all the Header Entries (Dicom Elements)
501 * @param fp file pointer on an already open file
502 * @param filetype Type of the File to be written
503 * (ACR-NEMA, ExplicitVR, ImplicitVR)
504 * \return Always true.
506 void Document::WriteContent(std::ofstream* fp, FileType filetype)
508 /// \todo move the following lines (and a lot of others, to be written)
509 /// to a future function CheckAndCorrectHeader
510 /// (necessary if user wants to write a DICOM V3 file
511 /// starting from an ACR-NEMA (V2) Header
513 if ( filetype == ImplicitVR || filetype == ExplicitVR )
515 // writing Dicom File Preamble
516 char filePreamble[128];
517 memset(filePreamble, 0, 128);
518 fp->write(filePreamble, 128);
519 fp->write("DICM", 4);
523 * \todo rewrite later, if really usefull
524 * - 'Group Length' element is optional in DICOM
525 * - but un-updated odd groups lengthes can causes pb
528 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
529 * UpdateGroupLength(false,filetype);
530 * if ( filetype == ACR)
531 * UpdateGroupLength(true,ACR);
534 ElementSet::WriteContent(fp, filetype); // This one is recursive
538 * \brief Modifies the value of a given Header Entry (Dicom Element)
539 * when it exists. Create it with the given value when unexistant.
540 * @param value (string) Value to be set
541 * @param group Group number of the Entry
542 * @param elem Element number of the Entry
543 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
544 * \return pointer to the modified/created Header Entry (NULL when creation
547 ValEntry* Document::ReplaceOrCreateByNumber(
548 std::string const & value,
553 ValEntry* valEntry = 0;
554 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
558 // check if (group,element) DictEntry exists
559 // if it doesn't, create an entry in DictSet::VirtualEntry
562 // Find out if the tag we received is in the dictionaries:
563 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
564 DictEntry* dictEntry = pubDict->GetDictEntryByNumber(group, elem);
567 currentEntry = NewDocEntryByNumber(group, elem, vr);
571 currentEntry = NewDocEntryByNumber(group, elem);
576 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
577 " NewDocEntryByNumber failed.");
581 valEntry = new ValEntry(currentEntry);
584 if ( !AddEntry(valEntry))
587 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
588 " failed allthough this is a creation.");
594 valEntry = dynamic_cast< ValEntry* >(currentEntry);
595 if ( !valEntry ) // Euuuuh? It wasn't a ValEntry
596 // then we change it to a ValEntry ?
597 // Shouldn't it be considered as an error ?
599 // We need to promote the DocEntry to a ValEntry:
600 valEntry = new ValEntry(currentEntry);
601 if (!RemoveEntry(currentEntry))
604 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
605 " of previous DocEntry failed.");
608 if ( !AddEntry(valEntry))
611 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
612 " promoted ValEntry failed.");
618 SetEntryByNumber(value, group, elem);
624 * \brief Modifies the value of a given Header Entry (Dicom Element)
625 * when it exists. Create it with the given value when unexistant.
626 * A copy of the binArea is made to be kept in the Document.
627 * @param binArea (binary) value to be set
628 * @param Group Group number of the Entry
629 * @param Elem Element number of the Entry
630 * \return pointer to the modified/created Header Entry (NULL when creation
633 BinEntry* Document::ReplaceOrCreateByNumber(
640 BinEntry* binEntry = 0;
641 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
645 // check if (group,element) DictEntry exists
646 // if it doesn't, create an entry in DictSet::VirtualEntry
649 // Find out if the tag we received is in the dictionaries:
650 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
651 DictEntry *dictEntry = pubDict->GetDictEntryByNumber(group, elem);
655 currentEntry = NewDocEntryByNumber(group, elem, vr);
659 currentEntry = NewDocEntryByNumber(group, elem);
663 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
664 " NewDocEntryByNumber failed.");
667 binEntry = new BinEntry(currentEntry);
668 if ( !AddEntry(binEntry))
670 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
671 " failed allthough this is a creation.");
677 binEntry = dynamic_cast< BinEntry* >(currentEntry);
678 if ( !binEntry ) // Euuuuh? It wasn't a BinEntry
679 // then we change it to a BinEntry ?
680 // Shouldn't it be considered as an error ?
682 // We need to promote the DocEntry to a BinEntry:
683 binEntry = new BinEntry(currentEntry);
684 if (!RemoveEntry(currentEntry))
686 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
687 " of previous DocEntry failed.");
690 if ( !AddEntry(binEntry))
692 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
693 " promoted BinEntry failed.");
700 if (lgth>0 && binArea)
702 tmpArea = new uint8_t[lgth];
703 memcpy(tmpArea,binArea,lgth);
709 if (!SetEntryByNumber(tmpArea, lgth, group, elem))
722 * \brief Modifies the value of a given Header Entry (Dicom Element)
723 * when it exists. Create it when unexistant.
724 * @param Group Group number of the Entry
725 * @param Elem Element number of the Entry
726 * \return pointer to the modified/created SeqEntry (NULL when creation
729 SeqEntry* Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem)
732 DocEntry* a = GetDocEntryByNumber( group, elem);
735 a = NewSeqEntryByNumber(group, elem);
741 b = new SeqEntry(a, 1); // FIXME : 1 (Depth)
748 * \brief Set a new value if the invoked element exists
749 * Seems to be useless !!!
750 * @param value new element value
751 * @param group group number of the Entry
752 * @param elem element number of the Entry
755 bool Document::ReplaceIfExistByNumber(std::string const & value,
756 uint16_t group, uint16_t elem )
758 SetEntryByNumber(value, group, elem);
763 std::string Document::GetTransferSyntaxValue(TransferSyntaxType type)
765 return TransferSyntaxStrings[type];
768 //-----------------------------------------------------------------------------
772 * \brief Checks if a given Dicom Element exists within the H table
773 * @param group Group number of the searched Dicom Element
774 * @param element Element number of the searched Dicom Element
775 * @return true is found
777 bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
779 const std::string &key = DictEntry::TranslateToKey(group, element );
780 return TagHT.count(key) != 0;
784 * \brief Searches within Header Entries (Dicom Elements) parsed with
785 * the public and private dictionaries
786 * for the element value of a given tag.
787 * \warning Don't use any longer : use GetPubEntryByName
788 * @param tagName name of the searched element.
789 * @return Corresponding element value when it exists,
790 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
792 std::string Document::GetEntryByName(TagName const & tagName)
794 DictEntry* dictEntry = RefPubDict->GetDictEntryByName(tagName);
800 return GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
804 * \brief Searches within Header Entries (Dicom Elements) parsed with
805 * the public and private dictionaries
806 * for the element value representation of a given tag.
808 * Obtaining the VR (Value Representation) might be needed by caller
809 * to convert the string typed content to caller's native type
810 * (think of C++ vs Python). The VR is actually of a higher level
811 * of semantics than just the native C++ type.
812 * @param tagName name of the searched element.
813 * @return Corresponding element value representation when it exists,
814 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
816 std::string Document::GetEntryVRByName(TagName const& tagName)
818 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
819 if( dictEntry == NULL)
824 DocEntry* elem = GetDocEntryByNumber(dictEntry->GetGroup(),
825 dictEntry->GetElement());
826 return elem->GetVR();
830 * \brief Searches within Header Entries (Dicom Elements) parsed with
831 * the public and private dictionaries
832 * for the element value representation of a given tag.
833 * @param group Group number of the searched tag.
834 * @param element Element number of the searched tag.
835 * @return Corresponding element value representation when it exists,
836 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
838 std::string Document::GetEntryByNumber(uint16_t group, uint16_t element)
840 TagKey key = DictEntry::TranslateToKey(group, element);
841 if ( !TagHT.count(key))
846 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
850 * \brief Searches within Header Entries (Dicom Elements) parsed with
851 * the public and private dictionaries
852 * for the element value representation of a given tag..
854 * Obtaining the VR (Value Representation) might be needed by caller
855 * to convert the string typed content to caller's native type
856 * (think of C++ vs Python). The VR is actually of a higher level
857 * of semantics than just the native C++ type.
858 * @param group Group number of the searched tag.
859 * @param element Element number of the searched tag.
860 * @return Corresponding element value representation when it exists,
861 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
863 std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element)
865 DocEntry* elem = GetDocEntryByNumber(group, element);
870 return elem->GetVR();
874 * \brief Searches within Header Entries (Dicom Elements) parsed with
875 * the public and private dictionaries
876 * for the value length of a given tag..
877 * @param group Group number of the searched tag.
878 * @param element Element number of the searched tag.
879 * @return Corresponding element length; -2 if not found
881 int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element)
883 DocEntry* elem = GetDocEntryByNumber(group, element);
886 return -2; //magic number
888 return elem->GetLength();
891 * \brief Sets the value (string) of the Header Entry (Dicom Element)
892 * @param content string value of the Dicom Element
893 * @param tagName name of the searched Dicom Element.
894 * @return true when found
896 bool Document::SetEntryByName( std::string const & content,
897 TagName const & tagName)
899 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
905 return SetEntryByNumber(content,dictEntry->GetGroup(),
906 dictEntry->GetElement());
910 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
911 * through it's (group, element) and modifies it's content with
913 * @param content new value (string) to substitute with
914 * @param group group number of the Dicom Element to modify
915 * @param element element number of the Dicom Element to modify
917 bool Document::SetEntryByNumber(std::string const& content,
918 uint16_t group, uint16_t element)
920 ValEntry* valEntry = GetValEntryByNumber(group, element);
923 dbg.Verbose(0, "Document::SetEntryByNumber: no corresponding",
924 " ValEntry (try promotion first).");
927 // Non even content must be padded with a space (020H)...
928 valEntry->SetValue(content);
933 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
934 * through it's (group, element) and modifies it's content with
936 * @param content new value (void* -> uint8_t*) to substitute with
937 * @param lgth new value length
938 * @param group group number of the Dicom Element to modify
939 * @param element element number of the Dicom Element to modify
941 bool Document::SetEntryByNumber(uint8_t*content, int lgth,
942 uint16_t group, uint16_t element)
944 (void)lgth; //not used
945 TagKey key = DictEntry::TranslateToKey(group, element);
946 if ( !TagHT.count(key))
951 /* Hope Binary field length is *never* wrong
952 if(lgth%2) // Non even length are padded with a space (020H).
955 //content = content + '\0'; // fing a trick to enlarge a binary field?
958 BinEntry* entry = (BinEntry *)TagHT[key];
959 entry->SetBinArea(content);
960 entry->SetLength(lgth);
961 entry->SetValue(GDCM_BINLOADED);
967 * \brief Gets (from Header) a 'non string' element value
968 * (LoadElementValues has already be executed)
969 * @param group group number of the Entry
970 * @param elem element number of the Entry
971 * @return Pointer to the 'non string' area
973 void* Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem)
975 DocEntry* entry = GetDocEntryByNumber(group, elem);
978 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry");
981 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(entry) )
983 return binEntry->GetBinArea();
990 * \brief Loads (from disk) the element content
991 * when a string is not suitable
992 * @param group group number of the Entry
993 * @param elem element number of the Entry
995 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
997 // Search the corresponding DocEntry
998 DocEntry *docElement = GetDocEntryByNumber(group, elem);
1002 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
1006 LoadEntryBinArea(binElement);
1010 * \brief Loads (from disk) the element content
1011 * when a string is not suitable
1012 * @param element Entry whose binArea is going to be loaded
1014 void Document::LoadEntryBinArea(BinEntry* element)
1016 if(element->GetBinArea())
1019 bool openFile = !Fp;
1023 size_t o =(size_t)element->GetOffset();
1024 Fp->seekg(o, std::ios::beg);
1026 size_t l = element->GetLength();
1027 uint8_t* a = new uint8_t[l];
1030 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1034 /// \todo check the result
1035 Fp->read((char*)a, l);
1036 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
1042 element->SetBinArea(a);
1049 * \brief Sets a 'non string' value to a given Dicom Element
1050 * @param area area containing the 'non string' value
1051 * @param group Group number of the searched Dicom Element
1052 * @param element Element number of the searched Dicom Element
1055 /*bool Document::SetEntryBinAreaByNumber(uint8_t* area,
1056 uint16_t group, uint16_t element)
1058 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1059 if ( !currentEntry )
1064 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1066 binEntry->SetBinArea( area );
1074 * \brief Update the entries with the shadow dictionary.
1075 * Only non even entries are analyzed
1077 void Document::UpdateShaEntries()
1082 /// \todo TODO : still any use to explore recursively the whole structure?
1084 for(ListTag::iterator it=listEntries.begin();
1085 it!=listEntries.end();
1088 // Odd group => from public dictionary
1089 if((*it)->GetGroup()%2==0)
1092 // Peer group => search the corresponding dict entry
1094 entry=RefShaDict->GetDictEntryByNumber((*it)->GetGroup(),(*it)->GetElement());
1098 if((*it)->IsImplicitVR())
1103 (*it)->SetValue(GetDocEntryUnvalue(*it)); // to go on compiling
1105 // Set the new entry and the new value
1106 (*it)->SetDictEntry(entry);
1107 CheckDocEntryVR(*it,vr);
1109 (*it)->SetValue(GetDocEntryValue(*it)); // to go on compiling
1114 // Remove precedent value transformation
1115 (*it)->SetDictEntry(NewVirtualDictEntry((*it)->GetGroup(),(*it)->GetElement(),vr));
1122 * \brief Searches within the Header Entries for a Dicom Element of
1124 * @param tagName name of the searched Dicom Element.
1125 * @return Corresponding Dicom Element when it exists, and NULL
1128 DocEntry* Document::GetDocEntryByName(TagName const & tagName)
1130 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
1136 return GetDocEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
1140 * \brief retrieves a Dicom Element (the first one) using (group, element)
1141 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1142 * if you think it's NOT UNIQUE, check the count number
1143 * and use iterators to retrieve ALL the Dicoms Elements within
1144 * a given couple (group, element)
1145 * @param group Group number of the searched Dicom Element
1146 * @param element Element number of the searched Dicom Element
1149 DocEntry* Document::GetDocEntryByNumber(uint16_t group, uint16_t element)
1151 TagKey key = DictEntry::TranslateToKey(group, element);
1152 if ( !TagHT.count(key))
1156 return TagHT.find(key)->second;
1160 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1161 * returns a result when the corresponding entry is of type
1163 * @return When present, the corresponding ValEntry.
1165 ValEntry* Document::GetValEntryByNumber(uint16_t group, uint16_t element)
1167 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1168 if ( !currentEntry )
1172 if ( ValEntry* valEntry = dynamic_cast<ValEntry*>(currentEntry) )
1176 dbg.Verbose(0, "Document::GetValEntryByNumber: unfound ValEntry.");
1182 * \brief Loads the element while preserving the current
1183 * underlying file position indicator as opposed to
1184 * to LoadDocEntry that modifies it.
1185 * @param entry Header Entry whose value shall be loaded.
1188 void Document::LoadDocEntrySafe(DocEntry * entry)
1192 long PositionOnEntry = Fp->tellg();
1193 LoadDocEntry(entry);
1194 Fp->seekg(PositionOnEntry, std::ios::beg);
1199 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1201 * @return The properly swaped 32 bits integer.
1203 uint32_t Document::SwapLong(uint32_t a)
1210 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1211 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1215 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1219 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1222 //std::cout << "swapCode= " << SwapCode << std::endl;
1223 dbg.Error(" Document::SwapLong : unset swap code");
1230 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1232 * @return The properly unswaped 32 bits integer.
1234 uint32_t Document::UnswapLong(uint32_t a)
1240 * \brief Swaps the bytes so they agree with the processor order
1241 * @return The properly swaped 16 bits integer.
1243 uint16_t Document::SwapShort(uint16_t a)
1245 if ( SwapCode == 4321 || SwapCode == 2143 )
1247 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1253 * \brief Unswaps the bytes so they agree with the processor order
1254 * @return The properly unswaped 16 bits integer.
1256 uint16_t Document::UnswapShort(uint16_t a)
1258 return SwapShort(a);
1261 //-----------------------------------------------------------------------------
1265 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1266 * @return length of the parsed set.
1268 void Document::ParseDES(DocEntrySet *set, long offset,
1269 long l_max, bool delim_mode)
1271 DocEntry *newDocEntry = 0;
1275 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1279 newDocEntry = ReadNextDocEntry( );
1285 VRKey vr = newDocEntry->GetVR();
1289 if ( Global::GetVR()->IsVROfGdcmStringRepresentable(vr) )
1291 /////////////////////// ValEntry
1292 ValEntry* newValEntry =
1293 new ValEntry( newDocEntry->GetDictEntry() ); //LEAK
1294 newValEntry->Copy( newDocEntry );
1296 // When "set" is a Document, then we are at the top of the
1297 // hierarchy and the Key is simply of the form ( group, elem )...
1298 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1301 newValEntry->SetKey( newValEntry->GetKey() );
1303 // ...but when "set" is a SQItem, we are inserting this new
1304 // valEntry in a sequence item. Hence the key has the
1305 // generalized form (refer to \ref BaseTagKey):
1306 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1308 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1309 + newValEntry->GetKey() );
1312 LoadDocEntry( newValEntry );
1313 bool delimitor=newValEntry->IsItemDelimitor();
1314 if( !set->AddEntry( newValEntry ) )
1316 // If here expect big troubles
1317 delete newValEntry; //otherwise mem leak
1325 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1333 if ( ! Global::GetVR()->IsVROfGdcmBinaryRepresentable(vr) )
1335 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1336 dbg.Verbose(0, "Document::ParseDES: neither Valentry, "
1337 "nor BinEntry. Probably unknown VR.");
1340 //////////////////// BinEntry or UNKOWN VR:
1341 BinEntry* newBinEntry = new BinEntry( newDocEntry ); //LEAK
1343 // When "this" is a Document the Key is simply of the
1344 // form ( group, elem )...
1345 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1348 newBinEntry->SetKey( newBinEntry->GetKey() );
1350 // but when "this" is a SQItem, we are inserting this new
1351 // valEntry in a sequence item, and the kay has the
1352 // generalized form (refer to \ref BaseTagKey):
1353 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1355 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1356 + newBinEntry->GetKey() );
1359 LoadDocEntry( newBinEntry );
1360 if( !set->AddEntry( newBinEntry ) )
1362 //Expect big troubles if here
1367 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1368 && ( newDocEntry->GetElement() == 0x0010 ) )
1370 TransferSyntaxType ts = GetTransferSyntax();
1371 if ( ts == RLELossless )
1373 long PositionOnEntry = Fp->tellg();
1374 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1376 Fp->seekg( PositionOnEntry, std::ios::beg );
1378 else if ( IsJPEG() )
1380 long PositionOnEntry = Fp->tellg();
1381 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1382 ComputeJPEGFragmentInfo();
1383 Fp->seekg( PositionOnEntry, std::ios::beg );
1387 // Just to make sure we are at the beginning of next entry.
1388 SkipToNextDocEntry(newDocEntry);
1389 //delete newDocEntry;
1394 unsigned long l = newDocEntry->GetReadLength();
1395 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1397 if ( l == 0xffffffff )
1406 // no other way to create it ...
1407 SeqEntry* newSeqEntry =
1408 new SeqEntry( newDocEntry->GetDictEntry() );
1409 newSeqEntry->Copy( newDocEntry );
1410 newSeqEntry->SetDelimitorMode( delim_mode );
1412 // At the top of the hierarchy, stands a Document. When "set"
1413 // is a Document, then we are building the first depth level.
1414 // Hence the SeqEntry we are building simply has a depth
1416 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1419 newSeqEntry->SetDepthLevel( 1 );
1420 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1422 // But when "set" is allready a SQItem, we are building a nested
1423 // sequence, and hence the depth level of the new SeqEntry
1424 // we are building, is one level deeper:
1425 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1427 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1428 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1429 + newSeqEntry->GetKey() );
1433 { // Don't try to parse zero-length sequences
1434 ParseSQ( newSeqEntry,
1435 newDocEntry->GetOffset(),
1438 set->AddEntry( newSeqEntry );
1439 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1450 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1451 * @return parsed length for this level
1453 void Document::ParseSQ( SeqEntry* seqEntry,
1454 long offset, long l_max, bool delim_mode)
1456 int SQItemNumber = 0;
1461 DocEntry* newDocEntry = ReadNextDocEntry();
1464 // FIXME Should warn user
1469 if ( newDocEntry->IsSequenceDelimitor() )
1471 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1475 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1481 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1482 std::ostringstream newBase;
1483 newBase << seqEntry->GetKey()
1487 itemSQ->SetBaseTagKey( newBase.str() );
1488 unsigned int l = newDocEntry->GetReadLength();
1490 if ( l == 0xffffffff )
1499 ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1502 seqEntry->AddEntry( itemSQ, SQItemNumber );
1504 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1512 * \brief Loads the element content if its length doesn't exceed
1513 * the value specified with Document::SetMaxSizeLoadEntry()
1514 * @param entry Header Entry (Dicom Element) to be dealt with
1516 void Document::LoadDocEntry(DocEntry* entry)
1518 uint16_t group = entry->GetGroup();
1519 std::string vr = entry->GetVR();
1520 uint32_t length = entry->GetLength();
1522 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1524 // A SeQuence "contains" a set of Elements.
1525 // (fffe e000) tells us an Element is beginning
1526 // (fffe e00d) tells us an Element just ended
1527 // (fffe e0dd) tells us the current SeQuence just ended
1528 if( group == 0xfffe )
1530 // NO more value field for SQ !
1534 // When the length is zero things are easy:
1537 ((ValEntry *)entry)->SetValue("");
1541 // The elements whose length is bigger than the specified upper bound
1542 // are not loaded. Instead we leave a short notice of the offset of
1543 // the element content and it's length.
1545 std::ostringstream s;
1546 if (length > MaxSizeLoadEntry)
1548 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1550 //s << "gdcm::NotLoaded (BinEntry)";
1551 s << GDCM_NOTLOADED;
1552 s << " Address:" << (long)entry->GetOffset();
1553 s << " Length:" << entry->GetLength();
1554 s << " x(" << std::hex << entry->GetLength() << ")";
1555 binEntryPtr->SetValue(s.str());
1557 // Be carefull : a BinEntry IS_A ValEntry ...
1558 else if (ValEntry* valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1560 // s << "gdcm::NotLoaded. (ValEntry)";
1561 s << GDCM_NOTLOADED;
1562 s << " Address:" << (long)entry->GetOffset();
1563 s << " Length:" << entry->GetLength();
1564 s << " x(" << std::hex << entry->GetLength() << ")";
1565 valEntryPtr->SetValue(s.str());
1570 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1571 << "nor a ValEntry ?! Should never print that !" << std::endl;
1574 // to be sure we are at the end of the value ...
1575 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1580 // When we find a BinEntry not very much can be done :
1581 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1583 s << GDCM_BINLOADED;
1584 binEntryPtr->SetValue(s.str());
1585 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1589 /// \todo Any compacter code suggested (?)
1590 if ( IsDocEntryAnInteger(entry) )
1594 // When short integer(s) are expected, read and convert the following
1595 // n *two characters properly i.e. consider them as short integers as
1596 // opposed to strings.
1597 // Elements with Value Multiplicity > 1
1598 // contain a set of integers (not a single one)
1599 if (vr == "US" || vr == "SS")
1602 NewInt = ReadInt16();
1606 for (int i=1; i < nbInt; i++)
1609 NewInt = ReadInt16();
1614 // See above comment on multiple integers (mutatis mutandis).
1615 else if (vr == "UL" || vr == "SL")
1618 NewInt = ReadInt32();
1622 for (int i=1; i < nbInt; i++)
1625 NewInt = ReadInt32();
1630 #ifdef GDCM_NO_ANSI_STRING_STREAM
1631 s << std::ends; // to avoid oddities on Solaris
1632 #endif //GDCM_NO_ANSI_STRING_STREAM
1634 ((ValEntry *)entry)->SetValue(s.str());
1638 // FIXME: We need an additional byte for storing \0 that is not on disk
1639 char *str = new char[length+1];
1640 Fp->read(str, (size_t)length);
1641 str[length] = '\0'; //this is only useful when length is odd
1642 // Special DicomString call to properly handle \0 and even length
1643 std::string newValue;
1646 newValue = Util::DicomString(str, length+1);
1647 //dbg.Verbose(0, "Warning: bad length: ", length );
1648 dbg.Verbose(0, "For string :", newValue.c_str());
1649 // Since we change the length of string update it length
1650 entry->SetReadLength(length+1);
1654 newValue = Util::DicomString(str, length);
1658 if ( ValEntry* valEntry = dynamic_cast<ValEntry* >(entry) )
1660 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1662 dbg.Verbose(1, "Document::LoadDocEntry",
1663 "unread element value");
1664 valEntry->SetValue(GDCM_UNREAD);
1670 // Because of correspondance with the VR dic
1671 valEntry->SetValue(newValue);
1675 valEntry->SetValue(newValue);
1680 dbg.Error(true, "Document::LoadDocEntry"
1681 "Should have a ValEntry, here !");
1687 * \brief Find the value Length of the passed Header Entry
1688 * @param entry Header Entry whose length of the value shall be loaded.
1690 void Document::FindDocEntryLength( DocEntry *entry )
1691 throw ( FormatError )
1693 uint16_t element = entry->GetElement();
1694 std::string vr = entry->GetVR();
1697 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1699 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1701 // The following reserved two bytes (see PS 3.5-2003, section
1702 // "7.1.2 Data element structure with explicit vr", p 27) must be
1703 // skipped before proceeding on reading the length on 4 bytes.
1704 Fp->seekg( 2L, std::ios::cur);
1705 uint32_t length32 = ReadInt32();
1707 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1712 /// \todo rename that to FindDocEntryLengthOBOrOW since
1713 /// the above test is on both OB and OW...
1714 lengthOB = FindDocEntryLengthOB();
1716 catch ( FormatUnexpected )
1718 // Computing the length failed (this happens with broken
1719 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1720 // chance to get the pixels by deciding the element goes
1721 // until the end of the file. Hence we artificially fix the
1722 // the length and proceed.
1723 long currentPosition = Fp->tellg();
1724 Fp->seekg(0L,std::ios::end);
1725 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1726 Fp->seekg(currentPosition, std::ios::beg);
1727 entry->SetLength(lengthUntilEOF);
1730 entry->SetLength(lengthOB);
1733 FixDocEntryFoundLength(entry, length32);
1737 // Length is encoded on 2 bytes.
1738 length16 = ReadInt16();
1740 // We can tell the current file is encoded in big endian (like
1741 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1742 // and it's value is the one of the encoding of a big endian file.
1743 // In order to deal with such big endian encoded files, we have
1744 // (at least) two strategies:
1745 // * when we load the "Transfer Syntax" tag with value of big endian
1746 // encoding, we raise the proper flags. Then we wait for the end
1747 // of the META group (0x0002) among which is "Transfer Syntax",
1748 // before switching the swap code to big endian. We have to postpone
1749 // the switching of the swap code since the META group is fully encoded
1750 // in little endian, and big endian coding only starts at the next
1751 // group. The corresponding code can be hard to analyse and adds
1752 // many additional unnecessary tests for regular tags.
1753 // * the second strategy consists in waiting for trouble, that shall
1754 // appear when we find the first group with big endian encoding. This
1755 // is easy to detect since the length of a "Group Length" tag (the
1756 // ones with zero as element number) has to be of 4 (0x0004). When we
1757 // encounter 1024 (0x0400) chances are the encoding changed and we
1758 // found a group with big endian encoding.
1759 // We shall use this second strategy. In order to make sure that we
1760 // can interpret the presence of an apparently big endian encoded
1761 // length of a "Group Length" without committing a big mistake, we
1762 // add an additional check: we look in the already parsed elements
1763 // for the presence of a "Transfer Syntax" whose value has to be "big
1764 // endian encoding". When this is the case, chances are we have got our
1765 // hands on a big endian encoded file: we switch the swap code to
1766 // big endian and proceed...
1767 if ( element == 0x0000 && length16 == 0x0400 )
1769 TransferSyntaxType ts = GetTransferSyntax();
1770 if ( ts != ExplicitVRBigEndian )
1772 throw FormatError( "Document::FindDocEntryLength()",
1773 " not explicit VR." );
1777 SwitchSwapToBigEndian();
1778 // Restore the unproperly loaded values i.e. the group, the element
1779 // and the dictionary entry depending on them.
1780 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1781 uint16_t correctElem = SwapShort( entry->GetElement() );
1782 DictEntry* newTag = GetDictEntryByNumber( correctGroup,
1786 // This correct tag is not in the dictionary. Create a new one.
1787 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1789 // FIXME this can create a memory leaks on the old entry that be
1790 // left unreferenced.
1791 entry->SetDictEntry( newTag );
1794 // Heuristic: well, some files are really ill-formed.
1795 if ( length16 == 0xffff)
1797 // 0xffff means that we deal with 'Unknown Length' Sequence
1800 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1805 // Either implicit VR or a non DICOM conformal (see note below) explicit
1806 // VR that ommited the VR of (at least) this element. Farts happen.
1807 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1808 // on Data elements "Implicit and Explicit VR Data Elements shall
1809 // not coexist in a Data Set and Data Sets nested within it".]
1810 // Length is on 4 bytes.
1812 FixDocEntryFoundLength( entry, ReadInt32() );
1818 * \brief Find the Value Representation of the current Dicom Element.
1821 void Document::FindDocEntryVR( DocEntry *entry )
1823 if ( Filetype != ExplicitVR )
1830 long positionOnEntry = Fp->tellg();
1831 // Warning: we believe this is explicit VR (Value Representation) because
1832 // we used a heuristic that found "UL" in the first tag. Alas this
1833 // doesn't guarantee that all the tags will be in explicit VR. In some
1834 // cases (see e-film filtered files) one finds implicit VR tags mixed
1835 // within an explicit VR file. Hence we make sure the present tag
1836 // is in explicit VR and try to fix things if it happens not to be
1838 Fp->read (vr, (size_t)2);
1841 if( !CheckDocEntryVR(entry, vr) )
1843 Fp->seekg(positionOnEntry, std::ios::beg);
1844 // When this element is known in the dictionary we shall use, e.g. for
1845 // the semantics (see the usage of IsAnInteger), the VR proposed by the
1846 // dictionary entry. Still we have to flag the element as implicit since
1847 // we know now our assumption on expliciteness is not furfilled.
1849 if ( entry->IsVRUnknown() )
1851 entry->SetVR("Implicit");
1853 entry->SetImplicitVR();
1858 * \brief Check the correspondance between the VR of the header entry
1859 * and the taken VR. If they are different, the header entry is
1860 * updated with the new VR.
1861 * @param entry Header Entry to check
1862 * @param vr Dicom Value Representation
1863 * @return false if the VR is incorrect of if the VR isn't referenced
1864 * otherwise, it returns true
1866 bool Document::CheckDocEntryVR(DocEntry *entry, VRKey vr)
1869 bool realExplicit = true;
1871 // Assume we are reading a falsely explicit VR file i.e. we reached
1872 // a tag where we expect reading a VR but are in fact we read the
1873 // first to bytes of the length. Then we will interogate (through find)
1874 // the dicom_vr dictionary with oddities like "\004\0" which crashes
1875 // both GCC and VC++ implementations of the STL map. Hence when the
1876 // expected VR read happens to be non-ascii characters we consider
1877 // we hit falsely explicit VR tag.
1879 if ( !isalpha((unsigned char)vr[0]) && !isalpha((unsigned char)vr[1]) )
1881 realExplicit = false;
1884 // CLEANME searching the dicom_vr at each occurence is expensive.
1885 // PostPone this test in an optional integrity check at the end
1886 // of parsing or only in debug mode.
1887 if ( realExplicit && !Global::GetVR()->Count(vr) )
1889 realExplicit = false;
1892 if ( !realExplicit )
1894 // We thought this was explicit VR, but we end up with an
1895 // implicit VR tag. Let's backtrack.
1896 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
1897 entry->GetGroup(), entry->GetElement());
1898 dbg.Verbose(1, "Document::FindVR: ", msg.c_str());
1900 if( entry->GetGroup() % 2 && entry->GetElement() == 0x0000)
1902 // Group length is UL !
1903 DictEntry* newEntry = NewVirtualDictEntry(
1904 entry->GetGroup(), entry->GetElement(),
1905 "UL", "FIXME", "Group Length");
1906 entry->SetDictEntry( newEntry );
1911 if ( entry->IsVRUnknown() )
1913 // When not a dictionary entry, we can safely overwrite the VR.
1914 if( entry->GetElement() == 0x0000 )
1916 // Group length is UL !
1924 else if ( entry->GetVR() != vr )
1926 // The VR present in the file and the dictionary disagree. We assume
1927 // the file writer knew best and use the VR of the file. Since it would
1928 // be unwise to overwrite the VR of a dictionary (since it would
1929 // compromise it's next user), we need to clone the actual DictEntry
1930 // and change the VR for the read one.
1931 DictEntry* newEntry = NewVirtualDictEntry(
1932 entry->GetGroup(), entry->GetElement(),
1933 vr, "FIXME", entry->GetName());
1934 entry->SetDictEntry(newEntry);
1941 * \brief Get the transformed value of the header entry. The VR value
1942 * is used to define the transformation to operate on the value
1943 * \warning NOT end user intended method !
1944 * @param entry entry to tranform
1945 * @return Transformed entry value
1947 std::string Document::GetDocEntryValue(DocEntry *entry)
1949 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1951 std::string val = ((ValEntry *)entry)->GetValue();
1952 std::string vr = entry->GetVR();
1953 uint32_t length = entry->GetLength();
1954 std::ostringstream s;
1957 // When short integer(s) are expected, read and convert the following
1958 // n * 2 bytes properly i.e. as a multivaluated strings
1959 // (each single value is separated fromthe next one by '\'
1960 // as usual for standard multivaluated filels
1961 // Elements with Value Multiplicity > 1
1962 // contain a set of short integers (not a single one)
1964 if( vr == "US" || vr == "SS" )
1969 for (int i=0; i < nbInt; i++)
1975 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1976 newInt16 = SwapShort( newInt16 );
1981 // When integer(s) are expected, read and convert the following
1982 // n * 4 bytes properly i.e. as a multivaluated strings
1983 // (each single value is separated fromthe next one by '\'
1984 // as usual for standard multivaluated filels
1985 // Elements with Value Multiplicity > 1
1986 // contain a set of integers (not a single one)
1987 else if( vr == "UL" || vr == "SL" )
1992 for (int i=0; i < nbInt; i++)
1998 newInt32 = ( val[4*i+0] & 0xFF )
1999 + (( val[4*i+1] & 0xFF ) << 8 )
2000 + (( val[4*i+2] & 0xFF ) << 16 )
2001 + (( val[4*i+3] & 0xFF ) << 24 );
2002 newInt32 = SwapLong( newInt32 );
2006 #ifdef GDCM_NO_ANSI_STRING_STREAM
2007 s << std::ends; // to avoid oddities on Solaris
2008 #endif //GDCM_NO_ANSI_STRING_STREAM
2012 return ((ValEntry *)entry)->GetValue();
2016 * \brief Get the reverse transformed value of the header entry. The VR
2017 * value is used to define the reverse transformation to operate on
2019 * \warning NOT end user intended method !
2020 * @param entry Entry to reverse transform
2021 * @return Reverse transformed entry value
2023 std::string Document::GetDocEntryUnvalue(DocEntry* entry)
2025 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2027 std::string vr = entry->GetVR();
2028 std::vector<std::string> tokens;
2029 std::ostringstream s;
2031 if ( vr == "US" || vr == "SS" )
2035 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
2036 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2037 for (unsigned int i=0; i<tokens.size(); i++)
2039 newInt16 = atoi(tokens[i].c_str());
2040 s << ( newInt16 & 0xFF )
2041 << (( newInt16 >> 8 ) & 0xFF );
2045 if ( vr == "UL" || vr == "SL")
2049 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
2050 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2051 for (unsigned int i=0; i<tokens.size();i++)
2053 newInt32 = atoi(tokens[i].c_str());
2054 s << (char)( newInt32 & 0xFF )
2055 << (char)(( newInt32 >> 8 ) & 0xFF )
2056 << (char)(( newInt32 >> 16 ) & 0xFF )
2057 << (char)(( newInt32 >> 24 ) & 0xFF );
2062 #ifdef GDCM_NO_ANSI_STRING_STREAM
2063 s << std::ends; // to avoid oddities on Solaris
2064 #endif //GDCM_NO_ANSI_STRING_STREAM
2068 return ((ValEntry *)entry)->GetValue();
2072 * \brief Skip a given Header Entry
2073 * \warning NOT end user intended method !
2074 * @param entry entry to skip
2076 void Document::SkipDocEntry(DocEntry *entry)
2078 SkipBytes(entry->GetLength());
2082 * \brief Skips to the begining of the next Header Entry
2083 * \warning NOT end user intended method !
2084 * @param entry entry to skip
2086 void Document::SkipToNextDocEntry(DocEntry *entry)
2088 Fp->seekg((long)(entry->GetOffset()), std::ios::beg);
2089 Fp->seekg( (long)(entry->GetReadLength()), std::ios::cur);
2093 * \brief When the length of an element value is obviously wrong (because
2094 * the parser went Jabberwocky) one can hope improving things by
2095 * applying some heuristics.
2096 * @param entry entry to check
2097 * @param foundLength fist assumption about length
2099 void Document::FixDocEntryFoundLength(DocEntry *entry,
2100 uint32_t foundLength)
2102 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
2103 if ( foundLength == 0xffffffff)
2108 uint16_t gr = entry->GetGroup();
2109 uint16_t el = entry->GetElement();
2111 if ( foundLength % 2)
2113 std::ostringstream s;
2114 s << "Warning : Tag with uneven length "
2116 << " in x(" << std::hex << gr << "," << el <<")" << std::dec;
2117 dbg.Verbose(0, s.str().c_str());
2120 //////// Fix for some naughty General Electric images.
2121 // Allthough not recent many such GE corrupted images are still present
2122 // on Creatis hard disks. Hence this fix shall remain when such images
2123 // are no longer in user (we are talking a few years, here)...
2124 // Note: XMedCom probably uses such a trick since it is able to read
2125 // those pesky GE images ...
2126 if ( foundLength == 13)
2128 // Only happens for this length !
2129 if ( entry->GetGroup() != 0x0008
2130 || ( entry->GetElement() != 0x0070
2131 && entry->GetElement() != 0x0080 ) )
2134 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
2138 //////// Fix for some brain-dead 'Leonardo' Siemens images.
2139 // Occurence of such images is quite low (unless one leaves close to a
2140 // 'Leonardo' source. Hence, one might consider commenting out the
2141 // following fix on efficiency reasons.
2142 else if ( entry->GetGroup() == 0x0009
2143 && ( entry->GetElement() == 0x1113
2144 || entry->GetElement() == 0x1114 ) )
2147 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
2150 else if ( entry->GetVR() == "SQ" )
2152 foundLength = 0; // ReadLength is unchanged
2155 //////// We encountered a 'delimiter' element i.e. a tag of the form
2156 // "fffe|xxxx" which is just a marker. Delimiters length should not be
2157 // taken into account.
2158 else if( entry->GetGroup() == 0xfffe )
2160 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
2161 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
2162 // causes extra troubles...
2163 if( entry->GetElement() != 0x0000 )
2169 entry->SetUsableLength(foundLength);
2173 * \brief Apply some heuristics to predict whether the considered
2174 * element value contains/represents an integer or not.
2175 * @param entry The element value on which to apply the predicate.
2176 * @return The result of the heuristical predicate.
2178 bool Document::IsDocEntryAnInteger(DocEntry *entry)
2180 uint16_t element = entry->GetElement();
2181 uint16_t group = entry->GetGroup();
2182 const std::string & vr = entry->GetVR();
2183 uint32_t length = entry->GetLength();
2185 // When we have some semantics on the element we just read, and if we
2186 // a priori know we are dealing with an integer, then we shall be
2187 // able to swap it's element value properly.
2188 if ( element == 0 ) // This is the group length of the group
2196 // Allthough this should never happen, still some images have a
2197 // corrupted group length [e.g. have a glance at offset x(8336) of
2198 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2199 // Since for dicom compliant and well behaved headers, the present
2200 // test is useless (and might even look a bit paranoid), when we
2201 // encounter such an ill-formed image, we simply display a warning
2202 // message and proceed on parsing (while crossing fingers).
2203 std::ostringstream s;
2204 long filePosition = Fp->tellg();
2205 s << "Erroneous Group Length element length on : (" \
2206 << std::hex << group << " , " << element
2207 << ") -before- position x(" << filePosition << ")"
2208 << "lgt : " << length;
2209 dbg.Verbose(0, "Document::IsDocEntryAnInteger", s.str().c_str() );
2213 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2221 * \brief Find the Length till the next sequence delimiter
2222 * \warning NOT end user intended method !
2226 uint32_t Document::FindDocEntryLengthOB()
2227 throw( FormatUnexpected )
2229 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2230 long positionOnEntry = Fp->tellg();
2231 bool foundSequenceDelimiter = false;
2232 uint32_t totalLength = 0;
2234 while ( !foundSequenceDelimiter )
2240 group = ReadInt16();
2243 catch ( FormatError )
2245 throw FormatError("Document::FindDocEntryLengthOB()",
2246 " group or element not present.");
2249 // We have to decount the group and element we just read
2252 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2254 dbg.Verbose(1, "Document::FindDocEntryLengthOB: neither an Item "
2255 "tag nor a Sequence delimiter tag.");
2256 Fp->seekg(positionOnEntry, std::ios::beg);
2257 throw FormatUnexpected("Document::FindDocEntryLengthOB()",
2258 "Neither an Item tag nor a Sequence "
2262 if ( elem == 0xe0dd )
2264 foundSequenceDelimiter = true;
2267 uint32_t itemLength = ReadInt32();
2268 // We add 4 bytes since we just read the ItemLength with ReadInt32
2269 totalLength += itemLength + 4;
2270 SkipBytes(itemLength);
2272 if ( foundSequenceDelimiter )
2277 Fp->seekg( positionOnEntry, std::ios::beg);
2282 * \brief Reads a supposed to be 16 Bits integer
2283 * (swaps it depending on processor endianity)
2284 * @return read value
2286 uint16_t Document::ReadInt16()
2287 throw( FormatError )
2290 Fp->read ((char*)&g, (size_t)2);
2293 throw FormatError( "Document::ReadInt16()", " file error." );
2297 throw FormatError( "Document::ReadInt16()", "EOF." );
2304 * \brief Reads a supposed to be 32 Bits integer
2305 * (swaps it depending on processor endianity)
2306 * @return read value
2308 uint32_t Document::ReadInt32()
2309 throw( FormatError )
2312 Fp->read ((char*)&g, (size_t)4);
2315 throw FormatError( "Document::ReadInt32()", " file error." );
2319 throw FormatError( "Document::ReadInt32()", "EOF." );
2326 * \brief skips bytes inside the source file
2327 * \warning NOT end user intended method !
2330 void Document::SkipBytes(uint32_t nBytes)
2332 //FIXME don't dump the returned value
2333 Fp->seekg((long)nBytes, std::ios::cur);
2337 * \brief Loads all the needed Dictionaries
2338 * \warning NOT end user intended method !
2340 void Document::Initialise()
2342 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2344 RLEInfo = new RLEFramesInfo;
2345 JPEGInfo = new JPEGFragmentsInfo;
2350 * \brief Discover what the swap code is (among little endian, big endian,
2351 * bad little endian, bad big endian).
2353 * @return false when we are absolutely sure
2354 * it's neither ACR-NEMA nor DICOM
2355 * true when we hope ours assuptions are OK
2357 bool Document::CheckSwap()
2359 // The only guaranted way of finding the swap code is to find a
2360 // group tag since we know it's length has to be of four bytes i.e.
2361 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2362 // occurs when we can't find such group...
2364 uint32_t x = 4; // x : for ntohs
2365 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2371 // First, compare HostByteOrder and NetworkByteOrder in order to
2372 // determine if we shall need to swap bytes (i.e. the Endian type).
2373 if ( x == ntohs(x) )
2382 // The easiest case is the one of a DICOM header, since it possesses a
2383 // file preamble where it suffice to look for the string "DICM".
2386 char *entCur = deb + 128;
2387 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2389 dbg.Verbose(1, "Document::CheckSwap:", "looks like DICOM Version3");
2391 // Next, determine the value representation (VR). Let's skip to the
2392 // first element (0002, 0000) and check there if we find "UL"
2393 // - or "OB" if the 1st one is (0002,0001) -,
2394 // in which case we (almost) know it is explicit VR.
2395 // WARNING: if it happens to be implicit VR then what we will read
2396 // is the length of the group. If this ascii representation of this
2397 // length happens to be "UL" then we shall believe it is explicit VR.
2398 // FIXME: in order to fix the above warning, we could read the next
2399 // element value (or a couple of elements values) in order to make
2400 // sure we are not commiting a big mistake.
2401 // We need to skip :
2402 // * the 128 bytes of File Preamble (often padded with zeroes),
2403 // * the 4 bytes of "DICM" string,
2404 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2405 // i.e. a total of 136 bytes.
2409 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2410 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2411 // *Implicit* VR. -and it is !-
2413 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2414 memcmp(entCur, "OB", (size_t)2) == 0 ||
2415 memcmp(entCur, "UI", (size_t)2) == 0 ||
2416 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2417 // when Write DCM *adds*
2419 // Use Document::dicom_vr to test all the possibilities
2420 // instead of just checking for UL, OB and UI !? group 0000
2422 Filetype = ExplicitVR;
2423 dbg.Verbose(1, "Document::CheckSwap:",
2424 "explicit Value Representation");
2428 Filetype = ImplicitVR;
2429 dbg.Verbose(1, "Document::CheckSwap:",
2430 "not an explicit Value Representation");
2436 dbg.Verbose(1, "Document::CheckSwap:",
2437 "HostByteOrder != NetworkByteOrder");
2442 dbg.Verbose(1, "Document::CheckSwap:",
2443 "HostByteOrder = NetworkByteOrder");
2446 // Position the file position indicator at first tag (i.e.
2447 // after the file preamble and the "DICM" string).
2448 Fp->seekg(0, std::ios::beg);
2449 Fp->seekg ( 132L, std::ios::beg);
2453 // Alas, this is not a DicomV3 file and whatever happens there is no file
2454 // preamble. We can reset the file position indicator to where the data
2455 // is (i.e. the beginning of the file).
2456 dbg.Verbose(1, "Document::CheckSwap:", "not a DICOM Version3 file");
2457 Fp->seekg(0, std::ios::beg);
2459 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2460 // By clean we mean that the length of the first tag is written down.
2461 // If this is the case and since the length of the first group HAS to be
2462 // four (bytes), then determining the proper swap code is straightforward.
2465 // We assume the array of char we are considering contains the binary
2466 // representation of a 32 bits integer. Hence the following dirty
2468 s32 = *((uint32_t *)(entCur));
2489 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2490 // It is time for despaired wild guesses.
2491 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2492 // i.e. the 'group length' element is not present :
2494 // check the supposed-to-be 'group number'
2495 // in ( 0x0001 .. 0x0008 )
2496 // to determine ' SwapCode' value .
2497 // Only 0 or 4321 will be possible
2498 // (no oportunity to check for the formerly well known
2499 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2500 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2501 // the file IS NOT ACR-NEMA nor DICOM V3
2502 // Find a trick to tell it the caller...
2504 s16 = *((uint16_t *)(deb));
2531 dbg.Verbose(0, "Document::CheckSwap:",
2532 "ACR/NEMA unfound swap info (Really hopeless !)");
2536 // Then the only info we have is the net2host one.
2548 * \brief Restore the unproperly loaded values i.e. the group, the element
2549 * and the dictionary entry depending on them.
2551 void Document::SwitchSwapToBigEndian()
2553 dbg.Verbose(1, "Document::SwitchSwapToBigEndian",
2554 "Switching to BigEndian mode.");
2555 if ( SwapCode == 0 )
2559 else if ( SwapCode == 4321 )
2563 else if ( SwapCode == 3412 )
2567 else if ( SwapCode == 2143 )
2574 * \brief during parsing, Header Elements too long are not loaded in memory
2577 void Document::SetMaxSizeLoadEntry(long newSize)
2583 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2585 MaxSizeLoadEntry = 0xffffffff;
2588 MaxSizeLoadEntry = newSize;
2593 * \brief Header Elements too long will not be printed
2594 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2597 void Document::SetMaxSizePrintEntry(long newSize)
2599 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2604 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2606 MaxSizePrintEntry = 0xffffffff;
2609 MaxSizePrintEntry = newSize;
2615 * \brief Handle broken private tag from Philips NTSCAN
2616 * where the endianess is being switch to BigEndian for no
2620 void Document::HandleBrokenEndian(uint16_t group, uint16_t elem)
2622 // Endian reversion. Some files contain groups of tags with reversed endianess.
2623 static int reversedEndian = 0;
2624 // try to fix endian switching in the middle of headers
2625 if ((group == 0xfeff) && (elem == 0x00e0))
2627 // start endian swap mark for group found
2629 SwitchSwapToBigEndian();
2634 else if ((group == 0xfffe) && (elem == 0xe00d) && reversedEndian)
2636 // end of reversed endian group
2638 SwitchSwapToBigEndian();
2643 * \brief Read the next tag but WITHOUT loading it's value
2644 * (read the 'Group Number', the 'Element Number',
2645 * gets the Dict Entry
2646 * gets the VR, gets the length, gets the offset value)
2647 * @return On succes the newly created DocEntry, NULL on failure.
2649 DocEntry* Document::ReadNextDocEntry()
2656 group = ReadInt16();
2659 catch ( FormatError e )
2661 // We reached the EOF (or an error occured) therefore
2662 // header parsing has to be considered as finished.
2667 HandleBrokenEndian(group, elem);
2668 DocEntry *newEntry = NewDocEntryByNumber(group, elem);
2669 FindDocEntryVR(newEntry);
2673 FindDocEntryLength(newEntry);
2675 catch ( FormatError e )
2683 newEntry->SetOffset(Fp->tellg());
2690 * \brief Generate a free TagKey i.e. a TagKey that is not present
2691 * in the TagHt dictionary.
2692 * @param group The generated tag must belong to this group.
2693 * @return The element of tag with given group which is fee.
2695 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2697 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2699 TagKey key = DictEntry::TranslateToKey(group, elem);
2700 if (TagHT.count(key) == 0)
2709 * \brief Assuming the internal file pointer \ref Document::Fp
2710 * is placed at the beginning of a tag check whether this
2711 * tag is (TestGroup, TestElement).
2712 * \warning On success the internal file pointer \ref Document::Fp
2713 * is modified to point after the tag.
2714 * On failure (i.e. when the tag wasn't the expected tag
2715 * (TestGroup, TestElement) the internal file pointer
2716 * \ref Document::Fp is restored to it's original position.
2717 * @param testGroup The expected group of the tag.
2718 * @param testElement The expected Element of the tag.
2719 * @return True on success, false otherwise.
2721 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2723 long positionOnEntry = Fp->tellg();
2724 long currentPosition = Fp->tellg(); // On debugging purposes
2726 //// Read the Item Tag group and element, and make
2727 // sure they are what we expected:
2728 uint16_t itemTagGroup;
2729 uint16_t itemTagElement;
2732 itemTagGroup = ReadInt16();
2733 itemTagElement = ReadInt16();
2735 catch ( FormatError e )
2737 //std::cerr << e << std::endl;
2740 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2742 std::ostringstream s;
2743 s << " We should have found tag (";
2744 s << std::hex << testGroup << "," << testElement << ")" << std::endl;
2745 s << " but instead we encountered tag (";
2746 s << std::hex << itemTagGroup << "," << itemTagElement << ")"
2748 s << " at address: " << (unsigned)currentPosition << std::endl;
2749 dbg.Verbose(0, "Document::ReadItemTagLength: wrong Item Tag found:");
2750 dbg.Verbose(0, s.str().c_str());
2751 Fp->seekg(positionOnEntry, std::ios::beg);
2759 * \brief Assuming the internal file pointer \ref Document::Fp
2760 * is placed at the beginning of a tag (TestGroup, TestElement),
2761 * read the length associated to the Tag.
2762 * \warning On success the internal file pointer \ref Document::Fp
2763 * is modified to point after the tag and it's length.
2764 * On failure (i.e. when the tag wasn't the expected tag
2765 * (TestGroup, TestElement) the internal file pointer
2766 * \ref Document::Fp is restored to it's original position.
2767 * @param testGroup The expected group of the tag.
2768 * @param testElement The expected Element of the tag.
2769 * @return On success returns the length associated to the tag. On failure
2772 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2774 long positionOnEntry = Fp->tellg();
2775 (void)positionOnEntry;
2777 if ( !ReadTag(testGroup, testElement) )
2782 //// Then read the associated Item Length
2783 long currentPosition = Fp->tellg();
2784 uint32_t itemLength = ReadInt32();
2786 std::ostringstream s;
2787 s << "Basic Item Length is: "
2788 << itemLength << std::endl;
2789 s << " at address: " << (unsigned)currentPosition << std::endl;
2790 dbg.Verbose(0, "Document::ReadItemTagLength: ", s.str().c_str());
2796 * \brief When parsing the Pixel Data of an encapsulated file, read
2797 * the basic offset table (when present, and BTW dump it).
2799 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2801 //// Read the Basic Offset Table Item Tag length...
2802 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2804 // When present, read the basic offset table itself.
2805 // Notes: - since the presence of this basic offset table is optional
2806 // we can't rely on it for the implementation, and we will simply
2807 // trash it's content (when present).
2808 // - still, when present, we could add some further checks on the
2809 // lengths, but we won't bother with such fuses for the time being.
2810 if ( itemLength != 0 )
2812 char* basicOffsetTableItemValue = new char[itemLength + 1];
2813 Fp->read(basicOffsetTableItemValue, itemLength);
2816 for (unsigned int i=0; i < itemLength; i += 4 )
2818 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2820 std::ostringstream s;
2821 s << " Read one length: ";
2822 s << std::hex << individualLength << std::endl;
2824 "Document::ReadAndSkipEncapsulatedBasicOffsetTable: ",
2829 delete[] basicOffsetTableItemValue;
2834 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2835 * Compute the RLE extra information and store it in \ref RLEInfo
2836 * for later pixel retrieval usage.
2838 void Document::ComputeRLEInfo()
2840 TransferSyntaxType ts = GetTransferSyntax();
2841 if ( ts != RLELossless )
2846 // Encoded pixel data: for the time being we are only concerned with
2847 // Jpeg or RLE Pixel data encodings.
2848 // As stated in PS 3.5-2003, section 8.2 p44:
2849 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2850 // value representation OB is used".
2851 // Hence we expect an OB value representation. Concerning OB VR,
2852 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2853 // "For the Value Representations OB and OW, the encoding shall meet the
2854 // following specifications depending on the Data element tag:"
2856 // - the first item in the sequence of items before the encoded pixel
2857 // data stream shall be basic offset table item. The basic offset table
2858 // item value, however, is not required to be present"
2860 ReadAndSkipEncapsulatedBasicOffsetTable();
2862 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2863 // Loop on the individual frame[s] and store the information
2864 // on the RLE fragments in a RLEFramesInfo.
2865 // Note: - when only a single frame is present, this is a
2867 // - when more than one frame are present, then we are in
2868 // the case of a multi-frame image.
2870 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2872 // Parse the RLE Header and store the corresponding RLE Segment
2873 // Offset Table information on fragments of this current Frame.
2874 // Note that the fragment pixels themselves are not loaded
2875 // (but just skipped).
2876 long frameOffset = Fp->tellg();
2878 uint32_t nbRleSegments = ReadInt32();
2879 if ( nbRleSegments > 16 )
2881 // There should be at most 15 segments (refer to RLEFrame class)
2882 dbg.Verbose(0, "Document::ComputeRLEInfo: too many segments.");
2885 uint32_t rleSegmentOffsetTable[16];
2886 for( int k = 1; k <= 15; k++ )
2888 rleSegmentOffsetTable[k] = ReadInt32();
2891 // Deduce from both the RLE Header and the frameLength the
2892 // fragment length, and again store this info in a
2894 long rleSegmentLength[15];
2895 // skipping (not reading) RLE Segments
2896 if ( nbRleSegments > 1)
2898 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2900 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2901 - rleSegmentOffsetTable[k];
2902 SkipBytes(rleSegmentLength[k]);
2906 rleSegmentLength[nbRleSegments] = frameLength
2907 - rleSegmentOffsetTable[nbRleSegments];
2908 SkipBytes(rleSegmentLength[nbRleSegments]);
2910 // Store the collected info
2911 RLEFrame* newFrameInfo = new RLEFrame;
2912 newFrameInfo->NumberFragments = nbRleSegments;
2913 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2915 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2916 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2918 RLEInfo->Frames.push_back( newFrameInfo );
2921 // Make sure that at the end of the item we encounter a 'Sequence
2923 if ( !ReadTag(0xfffe, 0xe0dd) )
2925 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2926 dbg.Verbose(0, " item at end of RLE item sequence");
2931 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2932 * Compute the jpeg extra information (fragment[s] offset[s] and
2933 * length) and store it[them] in \ref JPEGInfo for later pixel
2936 void Document::ComputeJPEGFragmentInfo()
2938 // If you need to, look for comments of ComputeRLEInfo().
2944 ReadAndSkipEncapsulatedBasicOffsetTable();
2946 // Loop on the fragments[s] and store the parsed information in a
2948 long fragmentLength;
2949 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2951 long fragmentOffset = Fp->tellg();
2953 // Store the collected info
2954 JPEGFragment* newFragment = new JPEGFragment;
2955 newFragment->Offset = fragmentOffset;
2956 newFragment->Length = fragmentLength;
2957 JPEGInfo->Fragments.push_back( newFragment );
2959 SkipBytes( fragmentLength );
2962 // Make sure that at the end of the item we encounter a 'Sequence
2964 if ( !ReadTag(0xfffe, 0xe0dd) )
2966 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2967 dbg.Verbose(0, " item at end of JPEG item sequence");
2972 * \brief Walk recursively the given \ref DocEntrySet, and feed
2973 * the given hash table (\ref TagDocEntryHT) with all the
2974 * \ref DocEntry (Dicom entries) encountered.
2975 * This method does the job for \ref BuildFlatHashTable.
2976 * @param builtHT Where to collect all the \ref DocEntry encountered
2977 * when recursively walking the given set.
2978 * @param set The structure to be traversed (recursively).
2980 void Document::BuildFlatHashTableRecurse( TagDocEntryHT& builtHT,
2983 if (ElementSet* elementSet = dynamic_cast< ElementSet* > ( set ) )
2985 TagDocEntryHT const & currentHT = elementSet->GetTagHT();
2986 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2987 i != currentHT.end();
2990 DocEntry* entry = i->second;
2991 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
2993 const ListSQItem& items = seqEntry->GetSQItems();
2994 for( ListSQItem::const_iterator item = items.begin();
2995 item != items.end();
2998 BuildFlatHashTableRecurse( builtHT, *item );
3002 builtHT[entry->GetKey()] = entry;
3007 if (SQItem* SQItemSet = dynamic_cast< SQItem* > ( set ) )
3009 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
3010 for (ListDocEntry::const_iterator i = currentList.begin();
3011 i != currentList.end();
3014 DocEntry* entry = *i;
3015 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3017 const ListSQItem& items = seqEntry->GetSQItems();
3018 for( ListSQItem::const_iterator item = items.begin();
3019 item != items.end();
3022 BuildFlatHashTableRecurse( builtHT, *item );
3026 builtHT[entry->GetKey()] = entry;
3033 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
3036 * The structure used by a Document (through \ref ElementSet),
3037 * in order to hold the parsed entries of a Dicom header, is a recursive
3038 * one. This is due to the fact that the sequences (when present)
3039 * can be nested. Additionaly, the sequence items (represented in
3040 * gdcm as \ref SQItem) add an extra complexity to the data
3041 * structure. Hence, a gdcm user whishing to visit all the entries of
3042 * a Dicom header will need to dig in the gdcm internals (which
3043 * implies exposing all the internal data structures to the API).
3044 * In order to avoid this burden to the user, \ref BuildFlatHashTable
3045 * recursively builds a temporary hash table, which holds all the
3046 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
3048 * \warning Of course there is NO integrity constrain between the
3049 * returned \ref TagDocEntryHT and the \ref ElementSet used
3050 * to build it. Hence if the underlying \ref ElementSet is
3051 * altered, then it is the caller responsability to invoke
3052 * \ref BuildFlatHashTable again...
3053 * @return The flat std::map<> we juste build.
3055 TagDocEntryHT* Document::BuildFlatHashTable()
3057 TagDocEntryHT* FlatHT = new TagDocEntryHT;
3058 BuildFlatHashTableRecurse( *FlatHT, this );
3065 * \brief Compares two documents, according to \ref DicomDir rules
3066 * \warning Does NOT work with ACR-NEMA files
3067 * \todo Find a trick to solve the pb (use RET fields ?)
3069 * @return true if 'smaller'
3071 bool Document::operator<(Document &document)
3074 std::string s1 = GetEntryByNumber(0x0010,0x0010);
3075 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
3087 s1 = GetEntryByNumber(0x0010,0x0020);
3088 s2 = document.GetEntryByNumber(0x0010,0x0020);
3099 // Study Instance UID
3100 s1 = GetEntryByNumber(0x0020,0x000d);
3101 s2 = document.GetEntryByNumber(0x0020,0x000d);
3112 // Serie Instance UID
3113 s1 = GetEntryByNumber(0x0020,0x000e);
3114 s2 = document.GetEntryByNumber(0x0020,0x000e);
3131 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
3132 * @param filetype Type of the File to be written
3134 int Document::ComputeGroup0002Length( FileType filetype )
3139 int groupLength = 0;
3140 bool found0002 = false;
3142 // for each zero-level Tag in the DCM Header
3149 entry = GetNextEntry();
3154 gr = entry->GetGroup();
3164 el = entry->GetElement();
3165 vr = entry->GetVR();
3167 if (filetype == ExplicitVR)
3169 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
3171 groupLength += 4; // explicit VR AND OB, OW, SQ : 4 more bytes
3174 groupLength += 2 + 2 + 4 + entry->GetLength();
3179 } // end namespace gdcm
3181 //-----------------------------------------------------------------------------