1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2004/12/02 15:14:17 $
7 Version: $Revision: 1.145 $
9 Copyright (c) CREATIS (Centre de Recherche et d'Applications en Traitement de
10 l'Image). All rights reserved. See Doc/License.txt or
11 http://www.creatis.insa-lyon.fr/Public/Gdcm/License.html for details.
13 This software is distributed WITHOUT ANY WARRANTY; without even
14 the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
15 PURPOSE. See the above copyright notices for more information.
17 =========================================================================*/
19 #include "gdcmDocument.h"
20 #include "gdcmValEntry.h"
21 #include "gdcmBinEntry.h"
22 #include "gdcmSeqEntry.h"
23 #include "gdcmGlobal.h"
25 #include "gdcmDebug.h"
31 #if defined(_MSC_VER) || defined(__BORLANDC__)
34 #include <netinet/in.h>
39 //-----------------------------------------------------------------------------
40 static const char *TransferSyntaxStrings[] = {
41 // Implicit VR Little Endian
43 // Implicit VR Little Endian DLX G.E?
45 // Explicit VR Little Endian
46 "1.2.840.10008.1.2.1",
47 // Deflated Explicit VR Little Endian
48 "1.2.840.10008.1.2.1.99",
49 // Explicit VR Big Endian
50 "1.2.840.10008.1.2.2",
51 // JPEG Baseline (Process 1)
52 "1.2.840.10008.1.2.4.50",
53 // JPEG Extended (Process 2 & 4)
54 "1.2.840.10008.1.2.4.51",
55 // JPEG Extended (Process 3 & 5)
56 "1.2.840.10008.1.2.4.52",
57 // JPEG Spectral Selection, Non-Hierarchical (Process 6 & 8)
58 "1.2.840.10008.1.2.4.53",
59 // JPEG Full Progression, Non-Hierarchical (Process 10 & 12)
60 "1.2.840.10008.1.2.4.55",
61 // JPEG Lossless, Non-Hierarchical (Process 14)
62 "1.2.840.10008.1.2.4.57",
63 // JPEG Lossless, Hierarchical, First-Order Prediction (Process 14, [Selection Value 1])
64 "1.2.840.10008.1.2.4.70",
66 "1.2.840.10008.1.2.4.90",
68 "1.2.840.10008.1.2.4.91",
70 "1.2.840.10008.1.2.5",
72 "Unknown Transfer Syntax"
75 //-----------------------------------------------------------------------------
76 // Refer to Document::CheckSwap()
77 //const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
79 // Refer to Document::SetMaxSizeLoadEntry()
80 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
81 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
83 //-----------------------------------------------------------------------------
84 // Constructor / Destructor
88 * @param filename file to be opened for parsing
90 Document::Document( std::string const & filename ) : ElementSet(-1)
92 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
102 dbg.Verbose(0, "Document::Document: starting parsing of file: ",
104 Fp->seekg( 0, std::ios::beg);
106 Fp->seekg(0, std::ios::end);
107 long lgt = Fp->tellg();
109 Fp->seekg( 0, std::ios::beg);
111 long beg = Fp->tellg();
114 ParseDES( this, beg, lgt, false); // le Load sera fait a la volee
116 Fp->seekg( 0, std::ios::beg);
118 // Load 'non string' values
120 std::string PhotometricInterpretation = GetEntryByNumber(0x0028,0x0004);
121 if( PhotometricInterpretation == "PALETTE COLOR " )
123 LoadEntryBinArea(0x0028,0x1200); // gray LUT
124 /// FIXME FIXME FIXME
125 /// The tags refered by the three following lines used to be CORRECTLY
126 /// defined as having an US Value Representation in the public
127 /// dictionnary. BUT the semantics implied by the three following
128 /// lines state that the corresponding tag contents are in fact
129 /// the ones of a BinEntry.
130 /// In order to fix things "Quick and Dirty" the dictionnary was
131 /// altered on PURPOUS but now contains a WRONG value.
132 /// In order to fix things and restore the dictionary to its
133 /// correct value, one needs to decided of the semantics by deciding
134 /// wether the following tags are either:
135 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
136 /// also used as BinEntry, which requires the proper conversion,
137 /// - OW, and hence loaded as BinEntry, but afterwards also used
138 /// as ValEntry, which requires the proper conversion.
139 LoadEntryBinArea(0x0028,0x1201); // R LUT
140 LoadEntryBinArea(0x0028,0x1202); // G LUT
141 LoadEntryBinArea(0x0028,0x1203); // B LUT
143 // Segmented Red Palette Color LUT Data
144 LoadEntryBinArea(0x0028,0x1221);
145 // Segmented Green Palette Color LUT Data
146 LoadEntryBinArea(0x0028,0x1222);
147 // Segmented Blue Palette Color LUT Data
148 LoadEntryBinArea(0x0028,0x1223);
150 //FIXME later : how to use it?
151 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
155 // --------------------------------------------------------------
156 // Specific code to allow gdcm to read ACR-LibIDO formated images
157 // Note: ACR-LibIDO is an extension of the ACR standard that was
158 // used at CREATIS. For the time being (say a couple years)
159 // we keep this kludge to allow a smooth move to gdcm for
160 // CREATIS developpers (sorry folks).
162 // if recognition code tells us we deal with a LibIDO image
163 // we switch lineNumber and columnNumber
166 RecCode = GetEntryByNumber(0x0008, 0x0010); // recognition code
167 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
168 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
169 // with "little-endian strings"
171 Filetype = ACR_LIBIDO;
172 std::string rows = GetEntryByNumber(0x0028, 0x0010);
173 std::string columns = GetEntryByNumber(0x0028, 0x0011);
174 SetEntryByNumber(columns, 0x0028, 0x0010);
175 SetEntryByNumber(rows , 0x0028, 0x0011);
177 // ----------------- End of ACR-LibIDO kludge ------------------
179 PrintLevel = 1; // 'Medium' print level by default
183 * \brief This default constructor doesn't parse the file. You should
184 * then invoke \ref Document::SetFileName and then the parsing.
186 Document::Document() : ElementSet(-1)
190 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
193 Filetype = ExplicitVR;
194 PrintLevel = 1; // 'Medium' print level by default
198 * \brief Canonical destructor.
200 Document::~Document ()
209 //-----------------------------------------------------------------------------
213 * \brief Prints The Dict Entries of THE public Dicom Dictionary
216 void Document::PrintPubDict(std::ostream & os)
218 RefPubDict->Print(os);
222 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
225 void Document::PrintShaDict(std::ostream & os)
227 RefShaDict->Print(os);
230 //-----------------------------------------------------------------------------
233 * \brief Get the public dictionary used
235 Dict* Document::GetPubDict()
241 * \brief Get the shadow dictionary used
243 Dict* Document::GetShaDict()
249 * \brief Set the shadow dictionary used
250 * \param dict dictionary to use in shadow
252 bool Document::SetShaDict(Dict *dict)
259 * \brief Set the shadow dictionary used
260 * \param dictName name of the dictionary to use in shadow
262 bool Document::SetShaDict(DictKey const & dictName)
264 RefShaDict = Global::GetDicts()->GetDict(dictName);
269 * \brief This predicate, based on hopefully reasonable heuristics,
270 * decides whether or not the current Document was properly parsed
271 * and contains the mandatory information for being considered as
272 * a well formed and usable Dicom/Acr File.
273 * @return true when Document is the one of a reasonable Dicom/Acr file,
276 bool Document::IsReadable()
278 if( Filetype == Unknown)
280 dbg.Verbose(0, "Document::IsReadable: wrong filetype");
286 dbg.Verbose(0, "Document::IsReadable: no tags in internal"
295 * \brief Accessor to the Transfer Syntax (when present) of the
296 * current document (it internally handles reading the
297 * value from disk when only parsing occured).
298 * @return The encountered Transfer Syntax of the current document.
300 TransferSyntaxType Document::GetTransferSyntax()
302 DocEntry *entry = GetDocEntryByNumber(0x0002, 0x0010);
308 // The entry might be present but not loaded (parsing and loading
309 // happen at different stages): try loading and proceed with check...
310 LoadDocEntrySafe(entry);
311 if (ValEntry* valEntry = dynamic_cast< ValEntry* >(entry) )
313 std::string transfer = valEntry->GetValue();
314 // The actual transfer (as read from disk) might be padded. We
315 // first need to remove the potential padding. We can make the
316 // weak assumption that padding was not executed with digits...
317 if ( transfer.length() == 0 )
319 // for brain damaged headers
322 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
324 transfer.erase(transfer.length()-1, 1);
326 for (int i = 0; TransferSyntaxStrings[i] != NULL; i++)
328 if ( TransferSyntaxStrings[i] == transfer )
330 return TransferSyntaxType(i);
337 bool Document::IsJPEGLossless()
339 TransferSyntaxType r = GetTransferSyntax();
340 return r == JPEGFullProgressionProcess10_12
341 || r == JPEGLosslessProcess14
342 || r == JPEGLosslessProcess14_1;
346 * \brief Determines if the Transfer Syntax was already encountered
347 * and if it corresponds to a JPEG2000 one
348 * @return True when JPEG2000 (Lossly or LossLess) found. False in all
351 bool Document::IsJPEG2000()
353 TransferSyntaxType r = GetTransferSyntax();
354 return r == JPEG2000Lossless || r == JPEG2000;
358 * \brief Determines if the Transfer Syntax corresponds to any form
359 * of Jpeg encoded Pixel data.
360 * @return True when any form of JPEG found. False otherwise.
362 bool Document::IsJPEG()
364 TransferSyntaxType r = GetTransferSyntax();
365 return r == JPEGBaselineProcess1
366 || r == JPEGExtendedProcess2_4
367 || r == JPEGExtendedProcess3_5
368 || r == JPEGSpectralSelectionProcess6_8
374 * \brief Determines if the Transfer Syntax corresponds to encapsulated
375 * of encoded Pixel Data (as opposed to native).
376 * @return True when encapsulated. False when native.
378 bool Document::IsEncapsulate()
380 TransferSyntaxType r = GetTransferSyntax();
381 return IsJPEG() || r == RLELossless;
385 * \brief Predicate for dicom version 3 file.
386 * @return True when the file is a dicom version 3.
388 bool Document::IsDicomV3()
390 // Checking if Transfert Syntax exists is enough
391 // Anyway, it's to late check if the 'Preamble' was found ...
392 // And ... would it be a rich idea to check ?
393 // (some 'no Preamble' DICOM images exist !)
394 return GetDocEntryByNumber(0x0002, 0x0010) != NULL;
398 * \brief returns the File Type
399 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
400 * @return the FileType code
402 FileType Document::GetFileType()
408 * \brief Tries to open the file \ref Document::Filename and
409 * checks the preamble when existing.
410 * @return The FILE pointer on success.
412 std::ifstream* Document::OpenFile()
414 if (Filename.length() == 0)
422 "Document::OpenFile is already opened when opening: ",
426 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
430 "Document::OpenFile cannot open file: ",
438 Fp->read((char*)&zero, (size_t)2 );
440 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
442 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
443 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
444 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
445 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
451 Fp->seekg(126L, std::ios::cur);
453 Fp->read(dicm, (size_t)4);
454 if( memcmp(dicm, "DICM", 4) == 0 )
461 "Document::OpenFile not DICOM/ACR (missing preamble)",
468 * \brief closes the file
469 * @return TRUE if the close was successfull
471 bool Document::CloseFile()
480 return true; //FIXME how do we detect a non-close ifstream ?
484 * \brief Writes in a file all the Header Entries (Dicom Elements)
485 * @param fp file pointer on an already open file
486 * @param filetype Type of the File to be written
487 * (ACR-NEMA, ExplicitVR, ImplicitVR)
488 * \return Always true.
490 void Document::WriteContent(std::ofstream* fp, FileType filetype)
492 /// \todo move the following lines (and a lot of others, to be written)
493 /// to a future function CheckAndCorrectHeader
494 /// (necessary if user wants to write a DICOM V3 file
495 /// starting from an ACR-NEMA (V2) Header
497 if ( filetype == ImplicitVR || filetype == ExplicitVR )
499 // writing Dicom File Preamble
500 char filePreamble[128];
501 memset(filePreamble, 0, 128);
502 fp->write(filePreamble, 128);
503 fp->write("DICM", 4);
507 * \todo rewrite later, if really usefull
508 * - 'Group Length' element is optional in DICOM
509 * - but un-updated odd groups lengthes can causes pb
512 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
513 * UpdateGroupLength(false,filetype);
514 * if ( filetype == ACR)
515 * UpdateGroupLength(true,ACR);
518 ElementSet::WriteContent(fp, filetype); // This one is recursive
522 * \brief Modifies the value of a given Header Entry (Dicom Element)
523 * when it exists. Create it with the given value when unexistant.
524 * @param value (string) Value to be set
525 * @param group Group number of the Entry
526 * @param elem Element number of the Entry
527 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
528 * \return pointer to the modified/created Header Entry (NULL when creation
531 ValEntry* Document::ReplaceOrCreateByNumber(
532 std::string const & value,
537 ValEntry* valEntry = 0;
538 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
542 // check if (group,element) DictEntry exists
543 // if it doesn't, create an entry in DictSet::VirtualEntry
546 // Find out if the tag we received is in the dictionaries:
547 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
548 DictEntry* dictEntry = pubDict->GetDictEntryByNumber(group, elem);
551 currentEntry = NewDocEntryByNumber(group, elem, vr);
555 currentEntry = NewDocEntryByNumber(group, elem);
560 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
561 " NewDocEntryByNumber failed.");
565 valEntry = new ValEntry(currentEntry);
568 if ( !AddEntry(valEntry))
571 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
572 " failed allthough this is a creation.");
578 valEntry = dynamic_cast< ValEntry* >(currentEntry);
579 if ( !valEntry ) // Euuuuh? It wasn't a ValEntry
580 // then we change it to a ValEntry ?
581 // Shouldn't it be considered as an error ?
583 // We need to promote the DocEntry to a ValEntry:
584 valEntry = new ValEntry(currentEntry);
585 if (!RemoveEntry(currentEntry))
588 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
589 " of previous DocEntry failed.");
592 if ( !AddEntry(valEntry))
595 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
596 " promoted ValEntry failed.");
602 SetEntryByNumber(value, group, elem);
608 * \brief Modifies the value of a given Header Entry (Dicom Element)
609 * when it exists. Create it with the given value when unexistant.
610 * A copy of the binArea is made to be kept in the Document.
611 * @param binArea (binary) value to be set
612 * @param Group Group number of the Entry
613 * @param Elem Element number of the Entry
614 * \return pointer to the modified/created Header Entry (NULL when creation
617 BinEntry* Document::ReplaceOrCreateByNumber(
624 BinEntry* binEntry = 0;
625 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
629 // check if (group,element) DictEntry exists
630 // if it doesn't, create an entry in DictSet::VirtualEntry
633 // Find out if the tag we received is in the dictionaries:
634 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
635 DictEntry *dictEntry = pubDict->GetDictEntryByNumber(group, elem);
639 currentEntry = NewDocEntryByNumber(group, elem, vr);
643 currentEntry = NewDocEntryByNumber(group, elem);
647 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
648 " NewDocEntryByNumber failed.");
651 binEntry = new BinEntry(currentEntry);
652 if ( !AddEntry(binEntry))
654 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
655 " failed allthough this is a creation.");
661 binEntry = dynamic_cast< BinEntry* >(currentEntry);
662 if ( !binEntry ) // Euuuuh? It wasn't a BinEntry
663 // then we change it to a BinEntry ?
664 // Shouldn't it be considered as an error ?
666 // We need to promote the DocEntry to a BinEntry:
667 binEntry = new BinEntry(currentEntry);
668 if (!RemoveEntry(currentEntry))
670 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
671 " of previous DocEntry failed.");
674 if ( !AddEntry(binEntry))
676 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
677 " promoted BinEntry failed.");
684 if (lgth>0 && binArea)
686 tmpArea = new uint8_t[lgth];
687 memcpy(tmpArea,binArea,lgth);
693 if (!SetEntryByNumber(tmpArea, lgth, group, elem))
706 * \brief Modifies the value of a given Header Entry (Dicom Element)
707 * when it exists. Create it when unexistant.
708 * @param Group Group number of the Entry
709 * @param Elem Element number of the Entry
710 * \return pointer to the modified/created SeqEntry (NULL when creation
713 SeqEntry* Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem)
716 DocEntry* a = GetDocEntryByNumber( group, elem);
719 a = NewSeqEntryByNumber(group, elem);
725 b = new SeqEntry(a, 1); // FIXME : 1 (Depth)
732 * \brief Set a new value if the invoked element exists
733 * Seems to be useless !!!
734 * @param value new element value
735 * @param group group number of the Entry
736 * @param elem element number of the Entry
739 bool Document::ReplaceIfExistByNumber(std::string const & value,
740 uint16_t group, uint16_t elem )
742 SetEntryByNumber(value, group, elem);
747 std::string Document::GetTransferSyntaxValue(TransferSyntaxType type)
749 return TransferSyntaxStrings[type];
752 //-----------------------------------------------------------------------------
756 * \brief Checks if a given Dicom Element exists within the H table
757 * @param group Group number of the searched Dicom Element
758 * @param element Element number of the searched Dicom Element
759 * @return true is found
761 bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
763 const std::string &key = DictEntry::TranslateToKey(group, element );
764 return TagHT.count(key) != 0;
768 * \brief Searches within Header Entries (Dicom Elements) parsed with
769 * the public and private dictionaries
770 * for the element value of a given tag.
771 * \warning Don't use any longer : use GetPubEntryByName
772 * @param tagName name of the searched element.
773 * @return Corresponding element value when it exists,
774 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
776 std::string Document::GetEntryByName(TagName const & tagName)
778 DictEntry* dictEntry = RefPubDict->GetDictEntryByName(tagName);
784 return GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
788 * \brief Searches within Header Entries (Dicom Elements) parsed with
789 * the public and private dictionaries
790 * for the element value representation of a given tag.
792 * Obtaining the VR (Value Representation) might be needed by caller
793 * to convert the string typed content to caller's native type
794 * (think of C++ vs Python). The VR is actually of a higher level
795 * of semantics than just the native C++ type.
796 * @param tagName name of the searched element.
797 * @return Corresponding element value representation when it exists,
798 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
800 std::string Document::GetEntryVRByName(TagName const& tagName)
802 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
803 if( dictEntry == NULL)
808 DocEntry* elem = GetDocEntryByNumber(dictEntry->GetGroup(),
809 dictEntry->GetElement());
810 return elem->GetVR();
814 * \brief Searches within Header Entries (Dicom Elements) parsed with
815 * the public and private dictionaries
816 * for the element value representation of a given tag.
817 * @param group Group number of the searched tag.
818 * @param element Element number of the searched tag.
819 * @return Corresponding element value representation when it exists,
820 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
822 std::string Document::GetEntryByNumber(uint16_t group, uint16_t element)
824 TagKey key = DictEntry::TranslateToKey(group, element);
825 /// \todo use map methods, instead of multimap JPR
826 if ( !TagHT.count(key))
831 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
835 * \brief Searches within Header Entries (Dicom Elements) parsed with
836 * the public and private dictionaries
837 * for the element value representation of a given tag..
839 * Obtaining the VR (Value Representation) might be needed by caller
840 * to convert the string typed content to caller's native type
841 * (think of C++ vs Python). The VR is actually of a higher level
842 * of semantics than just the native C++ type.
843 * @param group Group number of the searched tag.
844 * @param element Element number of the searched tag.
845 * @return Corresponding element value representation when it exists,
846 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
848 std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element)
850 DocEntry* elem = GetDocEntryByNumber(group, element);
855 return elem->GetVR();
859 * \brief Searches within Header Entries (Dicom Elements) parsed with
860 * the public and private dictionaries
861 * for the value length of a given tag..
862 * @param group Group number of the searched tag.
863 * @param element Element number of the searched tag.
864 * @return Corresponding element length; -2 if not found
866 int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element)
868 DocEntry* elem = GetDocEntryByNumber(group, element);
871 return -2; //magic number
873 return elem->GetLength();
876 * \brief Sets the value (string) of the Header Entry (Dicom Element)
877 * @param content string value of the Dicom Element
878 * @param tagName name of the searched Dicom Element.
879 * @return true when found
881 bool Document::SetEntryByName( std::string const & content,
882 TagName const & tagName)
884 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
890 return SetEntryByNumber(content,dictEntry->GetGroup(),
891 dictEntry->GetElement());
895 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
896 * through it's (group, element) and modifies it's content with
898 * @param content new value (string) to substitute with
899 * @param group group number of the Dicom Element to modify
900 * @param element element number of the Dicom Element to modify
902 bool Document::SetEntryByNumber(std::string const& content,
903 uint16_t group, uint16_t element)
908 ValEntry* valEntry = GetValEntryByNumber(group, element);
911 dbg.Verbose(0, "Document::SetEntryByNumber: no corresponding",
912 " ValEntry (try promotion first).");
915 // Non even content must be padded with a space (020H)...
916 std::string finalContent = Util::DicomString( content.c_str() );
917 assert( !(finalContent.size() % 2) );
918 valEntry->SetValue(finalContent);
920 // Integers have a special treatement for their length:
922 l = finalContent.length();
923 if ( l != 0) // To avoid to be cheated by 'zero length' integers
925 VRKey vr = valEntry->GetVR();
926 if( vr == "US" || vr == "SS" )
928 // for multivaluated items
929 c = Util::CountSubstring(content, "\\") + 1;
932 else if( vr == "UL" || vr == "SL" )
934 // for multivaluated items
935 c = Util::CountSubstring(content, "\\") + 1;
939 valEntry->SetLength(l);
944 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
945 * through it's (group, element) and modifies it's content with
947 * @param content new value (void* -> uint8_t*) to substitute with
948 * @param lgth new value length
949 * @param group group number of the Dicom Element to modify
950 * @param element element number of the Dicom Element to modify
952 bool Document::SetEntryByNumber(uint8_t*content, int lgth,
953 uint16_t group, uint16_t element)
955 (void)lgth; //not used
956 TagKey key = DictEntry::TranslateToKey(group, element);
957 if ( !TagHT.count(key))
962 /* Hope Binary field length is *never* wrong
963 if(lgth%2) // Non even length are padded with a space (020H).
966 //content = content + '\0'; // fing a trick to enlarge a binary field?
969 BinEntry* entry = (BinEntry *)TagHT[key];
970 entry->SetBinArea(content);
971 entry->SetLength(lgth);
972 entry->SetValue(GDCM_BINLOADED);
978 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
979 * in the PubDocEntrySet of this instance
980 * through it's (group, element) and modifies it's length with
982 * \warning Use with extreme caution.
983 * @param l new length to substitute with
984 * @param group group number of the Entry to modify
985 * @param element element number of the Entry to modify
986 * @return true on success, false otherwise.
988 /*bool Document::SetEntryLengthByNumber(uint32_t l,
989 uint16_t group, uint16_t element)
991 /// \todo use map methods, instead of multimap JPR
992 TagKey key = DictEntry::TranslateToKey(group, element);
993 if ( !TagHT.count(key) )
999 l++; // length must be even
1001 ( ((TagHT.equal_range(key)).first)->second )->SetLength(l);
1007 * \brief Gets (from Header) the offset of a 'non string' element value
1008 * (LoadElementValues has already be executed)
1009 * @param group group number of the Entry
1010 * @param elem element number of the Entry
1011 * @return File Offset of the Element Value
1013 /*size_t Document::GetEntryOffsetByNumber(uint16_t group, uint16_t elem)
1015 DocEntry* entry = GetDocEntryByNumber(group, elem);
1018 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry present.");
1021 return entry->GetOffset();
1025 * \brief Gets (from Header) a 'non string' element value
1026 * (LoadElementValues has already be executed)
1027 * @param group group number of the Entry
1028 * @param elem element number of the Entry
1029 * @return Pointer to the 'non string' area
1031 void* Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem)
1033 DocEntry* entry = GetDocEntryByNumber(group, elem);
1036 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry");
1039 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(entry) )
1041 return binEntry->GetBinArea();
1048 * \brief Loads (from disk) the element content
1049 * when a string is not suitable
1050 * @param group group number of the Entry
1051 * @param elem element number of the Entry
1053 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
1055 // Search the corresponding DocEntry
1056 DocEntry *docElement = GetDocEntryByNumber(group, elem);
1060 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
1064 LoadEntryBinArea(binElement);
1068 * \brief Loads (from disk) the element content
1069 * when a string is not suitable
1070 * @param element Entry whose binArea is going to be loaded
1072 void Document::LoadEntryBinArea(BinEntry* element)
1074 if(element->GetBinArea())
1077 bool openFile = !Fp;
1081 size_t o =(size_t)element->GetOffset();
1082 Fp->seekg(o, std::ios::beg);
1084 size_t l = element->GetLength();
1085 uint8_t* a = new uint8_t[l];
1088 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1092 /// \todo check the result
1093 Fp->read((char*)a, l);
1094 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
1100 element->SetBinArea(a);
1107 * \brief Sets a 'non string' value to a given Dicom Element
1108 * @param area area containing the 'non string' value
1109 * @param group Group number of the searched Dicom Element
1110 * @param element Element number of the searched Dicom Element
1113 /*bool Document::SetEntryBinAreaByNumber(uint8_t* area,
1114 uint16_t group, uint16_t element)
1116 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1117 if ( !currentEntry )
1122 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1124 binEntry->SetBinArea( area );
1132 * \brief Update the entries with the shadow dictionary.
1133 * Only non even entries are analyzed
1135 void Document::UpdateShaEntries()
1140 /// \todo TODO : still any use to explore recursively the whole structure?
1142 for(ListTag::iterator it=listEntries.begin();
1143 it!=listEntries.end();
1146 // Odd group => from public dictionary
1147 if((*it)->GetGroup()%2==0)
1150 // Peer group => search the corresponding dict entry
1152 entry=RefShaDict->GetDictEntryByNumber((*it)->GetGroup(),(*it)->GetElement());
1156 if((*it)->IsImplicitVR())
1161 (*it)->SetValue(GetDocEntryUnvalue(*it)); // to go on compiling
1163 // Set the new entry and the new value
1164 (*it)->SetDictEntry(entry);
1165 CheckDocEntryVR(*it,vr);
1167 (*it)->SetValue(GetDocEntryValue(*it)); // to go on compiling
1172 // Remove precedent value transformation
1173 (*it)->SetDictEntry(NewVirtualDictEntry((*it)->GetGroup(),(*it)->GetElement(),vr));
1180 * \brief Searches within the Header Entries for a Dicom Element of
1182 * @param tagName name of the searched Dicom Element.
1183 * @return Corresponding Dicom Element when it exists, and NULL
1186 DocEntry* Document::GetDocEntryByName(TagName const & tagName)
1188 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
1194 return GetDocEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
1198 * \brief retrieves a Dicom Element (the first one) using (group, element)
1199 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1200 * if you think it's NOT UNIQUE, check the count number
1201 * and use iterators to retrieve ALL the Dicoms Elements within
1202 * a given couple (group, element)
1203 * @param group Group number of the searched Dicom Element
1204 * @param element Element number of the searched Dicom Element
1207 DocEntry* Document::GetDocEntryByNumber(uint16_t group, uint16_t element)
1209 TagKey key = DictEntry::TranslateToKey(group, element);
1210 if ( !TagHT.count(key))
1214 return TagHT.find(key)->second;
1218 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1219 * returns a result when the corresponding entry is of type
1221 * @return When present, the corresponding ValEntry.
1223 ValEntry* Document::GetValEntryByNumber(uint16_t group, uint16_t element)
1225 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1226 if ( !currentEntry )
1230 if ( ValEntry* valEntry = dynamic_cast<ValEntry*>(currentEntry) )
1234 dbg.Verbose(0, "Document::GetValEntryByNumber: unfound ValEntry.");
1240 * \brief Loads the element while preserving the current
1241 * underlying file position indicator as opposed to
1242 * to LoadDocEntry that modifies it.
1243 * @param entry Header Entry whose value shall be loaded.
1246 void Document::LoadDocEntrySafe(DocEntry * entry)
1250 long PositionOnEntry = Fp->tellg();
1251 LoadDocEntry(entry);
1252 Fp->seekg(PositionOnEntry, std::ios::beg);
1257 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1259 * @return The properly swaped 32 bits integer.
1261 uint32_t Document::SwapLong(uint32_t a)
1268 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1269 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1273 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1277 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1280 //std::cout << "swapCode= " << SwapCode << std::endl;
1281 dbg.Error(" Document::SwapLong : unset swap code");
1288 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1290 * @return The properly unswaped 32 bits integer.
1292 uint32_t Document::UnswapLong(uint32_t a)
1298 * \brief Swaps the bytes so they agree with the processor order
1299 * @return The properly swaped 16 bits integer.
1301 uint16_t Document::SwapShort(uint16_t a)
1303 if ( SwapCode == 4321 || SwapCode == 2143 )
1305 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1311 * \brief Unswaps the bytes so they agree with the processor order
1312 * @return The properly unswaped 16 bits integer.
1314 uint16_t Document::UnswapShort(uint16_t a)
1316 return SwapShort(a);
1319 //-----------------------------------------------------------------------------
1323 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1324 * @return length of the parsed set.
1326 void Document::ParseDES(DocEntrySet *set, long offset,
1327 long l_max, bool delim_mode)
1329 DocEntry *newDocEntry = 0;
1333 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1337 newDocEntry = ReadNextDocEntry( );
1343 VRKey vr = newDocEntry->GetVR();
1347 if ( Global::GetVR()->IsVROfGdcmStringRepresentable(vr) )
1349 /////////////////////// ValEntry
1350 ValEntry* newValEntry =
1351 new ValEntry( newDocEntry->GetDictEntry() ); //LEAK
1352 newValEntry->Copy( newDocEntry );
1354 // When "set" is a Document, then we are at the top of the
1355 // hierarchy and the Key is simply of the form ( group, elem )...
1356 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1359 newValEntry->SetKey( newValEntry->GetKey() );
1361 // ...but when "set" is a SQItem, we are inserting this new
1362 // valEntry in a sequence item. Hence the key has the
1363 // generalized form (refer to \ref BaseTagKey):
1364 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1366 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1367 + newValEntry->GetKey() );
1370 LoadDocEntry( newValEntry );
1371 bool delimitor=newValEntry->IsItemDelimitor();
1372 if( !set->AddEntry( newValEntry ) )
1374 // If here expect big troubles
1375 delete newValEntry; //otherwise mem leak
1383 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1391 if ( ! Global::GetVR()->IsVROfGdcmBinaryRepresentable(vr) )
1393 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1394 dbg.Verbose(0, "Document::ParseDES: neither Valentry, "
1395 "nor BinEntry. Probably unknown VR.");
1398 //////////////////// BinEntry or UNKOWN VR:
1399 BinEntry* newBinEntry = new BinEntry( newDocEntry ); //LEAK
1401 // When "this" is a Document the Key is simply of the
1402 // form ( group, elem )...
1403 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1406 newBinEntry->SetKey( newBinEntry->GetKey() );
1408 // but when "this" is a SQItem, we are inserting this new
1409 // valEntry in a sequence item, and the kay has the
1410 // generalized form (refer to \ref BaseTagKey):
1411 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1413 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1414 + newBinEntry->GetKey() );
1417 LoadDocEntry( newBinEntry );
1418 if( !set->AddEntry( newBinEntry ) )
1420 //Expect big troubles if here
1425 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1426 && ( newDocEntry->GetElement() == 0x0010 ) )
1428 TransferSyntaxType ts = GetTransferSyntax();
1429 if ( ts == RLELossless )
1431 long PositionOnEntry = Fp->tellg();
1432 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1434 Fp->seekg( PositionOnEntry, std::ios::beg );
1436 else if ( IsJPEG() )
1438 long PositionOnEntry = Fp->tellg();
1439 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1440 ComputeJPEGFragmentInfo();
1441 Fp->seekg( PositionOnEntry, std::ios::beg );
1445 // Just to make sure we are at the beginning of next entry.
1446 SkipToNextDocEntry(newDocEntry);
1447 //delete newDocEntry;
1452 unsigned long l = newDocEntry->GetReadLength();
1453 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1455 if ( l == 0xffffffff )
1464 // no other way to create it ...
1465 SeqEntry* newSeqEntry =
1466 new SeqEntry( newDocEntry->GetDictEntry() );
1467 newSeqEntry->Copy( newDocEntry );
1468 newSeqEntry->SetDelimitorMode( delim_mode );
1470 // At the top of the hierarchy, stands a Document. When "set"
1471 // is a Document, then we are building the first depth level.
1472 // Hence the SeqEntry we are building simply has a depth
1474 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1477 newSeqEntry->SetDepthLevel( 1 );
1478 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1480 // But when "set" is allready a SQItem, we are building a nested
1481 // sequence, and hence the depth level of the new SeqEntry
1482 // we are building, is one level deeper:
1483 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1485 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1486 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1487 + newSeqEntry->GetKey() );
1491 { // Don't try to parse zero-length sequences
1492 ParseSQ( newSeqEntry,
1493 newDocEntry->GetOffset(),
1496 set->AddEntry( newSeqEntry );
1497 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1508 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1509 * @return parsed length for this level
1511 void Document::ParseSQ( SeqEntry* seqEntry,
1512 long offset, long l_max, bool delim_mode)
1514 int SQItemNumber = 0;
1519 DocEntry* newDocEntry = ReadNextDocEntry();
1522 // FIXME Should warn user
1527 if ( newDocEntry->IsSequenceDelimitor() )
1529 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1533 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1539 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1540 std::ostringstream newBase;
1541 newBase << seqEntry->GetKey()
1545 itemSQ->SetBaseTagKey( newBase.str() );
1546 unsigned int l = newDocEntry->GetReadLength();
1548 if ( l == 0xffffffff )
1557 ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1560 seqEntry->AddEntry( itemSQ, SQItemNumber );
1562 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1570 * \brief Loads the element content if its length doesn't exceed
1571 * the value specified with Document::SetMaxSizeLoadEntry()
1572 * @param entry Header Entry (Dicom Element) to be dealt with
1574 void Document::LoadDocEntry(DocEntry* entry)
1576 uint16_t group = entry->GetGroup();
1577 std::string vr = entry->GetVR();
1578 uint32_t length = entry->GetLength();
1580 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1582 // A SeQuence "contains" a set of Elements.
1583 // (fffe e000) tells us an Element is beginning
1584 // (fffe e00d) tells us an Element just ended
1585 // (fffe e0dd) tells us the current SeQuence just ended
1586 if( group == 0xfffe )
1588 // NO more value field for SQ !
1592 // When the length is zero things are easy:
1595 ((ValEntry *)entry)->SetValue("");
1599 // The elements whose length is bigger than the specified upper bound
1600 // are not loaded. Instead we leave a short notice of the offset of
1601 // the element content and it's length.
1603 std::ostringstream s;
1604 if (length > MaxSizeLoadEntry)
1606 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1608 //s << "gdcm::NotLoaded (BinEntry)";
1609 s << GDCM_NOTLOADED;
1610 s << " Address:" << (long)entry->GetOffset();
1611 s << " Length:" << entry->GetLength();
1612 s << " x(" << std::hex << entry->GetLength() << ")";
1613 binEntryPtr->SetValue(s.str());
1615 // Be carefull : a BinEntry IS_A ValEntry ...
1616 else if (ValEntry* valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1618 // s << "gdcm::NotLoaded. (ValEntry)";
1619 s << GDCM_NOTLOADED;
1620 s << " Address:" << (long)entry->GetOffset();
1621 s << " Length:" << entry->GetLength();
1622 s << " x(" << std::hex << entry->GetLength() << ")";
1623 valEntryPtr->SetValue(s.str());
1628 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1629 << "nor a ValEntry ?! Should never print that !" << std::endl;
1632 // to be sure we are at the end of the value ...
1633 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1638 // When we find a BinEntry not very much can be done :
1639 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1641 s << GDCM_BINLOADED;
1642 binEntryPtr->SetValue(s.str());
1643 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1647 /// \todo Any compacter code suggested (?)
1648 if ( IsDocEntryAnInteger(entry) )
1652 // When short integer(s) are expected, read and convert the following
1653 // n *two characters properly i.e. consider them as short integers as
1654 // opposed to strings.
1655 // Elements with Value Multiplicity > 1
1656 // contain a set of integers (not a single one)
1657 if (vr == "US" || vr == "SS")
1660 NewInt = ReadInt16();
1664 for (int i=1; i < nbInt; i++)
1667 NewInt = ReadInt16();
1672 // See above comment on multiple integers (mutatis mutandis).
1673 else if (vr == "UL" || vr == "SL")
1676 NewInt = ReadInt32();
1680 for (int i=1; i < nbInt; i++)
1683 NewInt = ReadInt32();
1688 #ifdef GDCM_NO_ANSI_STRING_STREAM
1689 s << std::ends; // to avoid oddities on Solaris
1690 #endif //GDCM_NO_ANSI_STRING_STREAM
1692 ((ValEntry *)entry)->SetValue(s.str());
1696 // FIXME: We need an additional byte for storing \0 that is not on disk
1697 char *str = new char[length+1];
1698 Fp->read(str, (size_t)length);
1699 str[length] = '\0'; //this is only useful when length is odd
1700 // Special DicomString call to properly handle \0 and even length
1701 std::string newValue;
1704 newValue = Util::DicomString(str, length+1);
1705 //dbg.Verbose(0, "Warning: bad length: ", length );
1706 dbg.Verbose(0, "For string :", newValue.c_str());
1707 // Since we change the length of string update it length
1708 entry->SetReadLength(length+1);
1712 newValue = Util::DicomString(str, length);
1716 if ( ValEntry* valEntry = dynamic_cast<ValEntry* >(entry) )
1718 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1720 dbg.Verbose(1, "Document::LoadDocEntry",
1721 "unread element value");
1722 valEntry->SetValue(GDCM_UNREAD);
1728 // Because of correspondance with the VR dic
1729 valEntry->SetValue(newValue);
1733 valEntry->SetValue(newValue);
1738 dbg.Error(true, "Document::LoadDocEntry"
1739 "Should have a ValEntry, here !");
1745 * \brief Find the value Length of the passed Header Entry
1746 * @param entry Header Entry whose length of the value shall be loaded.
1748 void Document::FindDocEntryLength( DocEntry *entry )
1749 throw ( FormatError )
1751 uint16_t element = entry->GetElement();
1752 std::string vr = entry->GetVR();
1755 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1757 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1759 // The following reserved two bytes (see PS 3.5-2003, section
1760 // "7.1.2 Data element structure with explicit vr", p 27) must be
1761 // skipped before proceeding on reading the length on 4 bytes.
1762 Fp->seekg( 2L, std::ios::cur);
1763 uint32_t length32 = ReadInt32();
1765 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1770 /// \todo rename that to FindDocEntryLengthOBOrOW since
1771 /// the above test is on both OB and OW...
1772 lengthOB = FindDocEntryLengthOB();
1774 catch ( FormatUnexpected )
1776 // Computing the length failed (this happens with broken
1777 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1778 // chance to get the pixels by deciding the element goes
1779 // until the end of the file. Hence we artificially fix the
1780 // the length and proceed.
1781 long currentPosition = Fp->tellg();
1782 Fp->seekg(0L,std::ios::end);
1783 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1784 Fp->seekg(currentPosition, std::ios::beg);
1785 entry->SetLength(lengthUntilEOF);
1788 entry->SetLength(lengthOB);
1791 FixDocEntryFoundLength(entry, length32);
1795 // Length is encoded on 2 bytes.
1796 length16 = ReadInt16();
1798 // We can tell the current file is encoded in big endian (like
1799 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1800 // and it's value is the one of the encoding of a big endian file.
1801 // In order to deal with such big endian encoded files, we have
1802 // (at least) two strategies:
1803 // * when we load the "Transfer Syntax" tag with value of big endian
1804 // encoding, we raise the proper flags. Then we wait for the end
1805 // of the META group (0x0002) among which is "Transfer Syntax",
1806 // before switching the swap code to big endian. We have to postpone
1807 // the switching of the swap code since the META group is fully encoded
1808 // in little endian, and big endian coding only starts at the next
1809 // group. The corresponding code can be hard to analyse and adds
1810 // many additional unnecessary tests for regular tags.
1811 // * the second strategy consists in waiting for trouble, that shall
1812 // appear when we find the first group with big endian encoding. This
1813 // is easy to detect since the length of a "Group Length" tag (the
1814 // ones with zero as element number) has to be of 4 (0x0004). When we
1815 // encounter 1024 (0x0400) chances are the encoding changed and we
1816 // found a group with big endian encoding.
1817 // We shall use this second strategy. In order to make sure that we
1818 // can interpret the presence of an apparently big endian encoded
1819 // length of a "Group Length" without committing a big mistake, we
1820 // add an additional check: we look in the already parsed elements
1821 // for the presence of a "Transfer Syntax" whose value has to be "big
1822 // endian encoding". When this is the case, chances are we have got our
1823 // hands on a big endian encoded file: we switch the swap code to
1824 // big endian and proceed...
1825 if ( element == 0x0000 && length16 == 0x0400 )
1827 TransferSyntaxType ts = GetTransferSyntax();
1828 if ( ts != ExplicitVRBigEndian )
1830 throw FormatError( "Document::FindDocEntryLength()",
1831 " not explicit VR." );
1835 SwitchSwapToBigEndian();
1836 // Restore the unproperly loaded values i.e. the group, the element
1837 // and the dictionary entry depending on them.
1838 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1839 uint16_t correctElem = SwapShort( entry->GetElement() );
1840 DictEntry* newTag = GetDictEntryByNumber( correctGroup,
1844 // This correct tag is not in the dictionary. Create a new one.
1845 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1847 // FIXME this can create a memory leaks on the old entry that be
1848 // left unreferenced.
1849 entry->SetDictEntry( newTag );
1852 // Heuristic: well, some files are really ill-formed.
1853 if ( length16 == 0xffff)
1855 // 0xffff means that we deal with 'Unknown Length' Sequence
1858 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1863 // Either implicit VR or a non DICOM conformal (see note below) explicit
1864 // VR that ommited the VR of (at least) this element. Farts happen.
1865 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1866 // on Data elements "Implicit and Explicit VR Data Elements shall
1867 // not coexist in a Data Set and Data Sets nested within it".]
1868 // Length is on 4 bytes.
1870 FixDocEntryFoundLength( entry, ReadInt32() );
1876 * \brief Find the Value Representation of the current Dicom Element.
1879 void Document::FindDocEntryVR( DocEntry *entry )
1881 if ( Filetype != ExplicitVR )
1888 long positionOnEntry = Fp->tellg();
1889 // Warning: we believe this is explicit VR (Value Representation) because
1890 // we used a heuristic that found "UL" in the first tag. Alas this
1891 // doesn't guarantee that all the tags will be in explicit VR. In some
1892 // cases (see e-film filtered files) one finds implicit VR tags mixed
1893 // within an explicit VR file. Hence we make sure the present tag
1894 // is in explicit VR and try to fix things if it happens not to be
1898 Fp->read (vr, (size_t)2);
1901 if( !CheckDocEntryVR(entry, vr) )
1903 Fp->seekg(positionOnEntry, std::ios::beg);
1904 // When this element is known in the dictionary we shall use, e.g. for
1905 // the semantics (see the usage of IsAnInteger), the VR proposed by the
1906 // dictionary entry. Still we have to flag the element as implicit since
1907 // we know now our assumption on expliciteness is not furfilled.
1909 if ( entry->IsVRUnknown() )
1911 entry->SetVR("Implicit");
1913 entry->SetImplicitVR();
1918 * \brief Check the correspondance between the VR of the header entry
1919 * and the taken VR. If they are different, the header entry is
1920 * updated with the new VR.
1921 * @param entry Header Entry to check
1922 * @param vr Dicom Value Representation
1923 * @return false if the VR is incorrect of if the VR isn't referenced
1924 * otherwise, it returns true
1926 bool Document::CheckDocEntryVR(DocEntry *entry, VRKey vr)
1929 bool realExplicit = true;
1931 // Assume we are reading a falsely explicit VR file i.e. we reached
1932 // a tag where we expect reading a VR but are in fact we read the
1933 // first to bytes of the length. Then we will interogate (through find)
1934 // the dicom_vr dictionary with oddities like "\004\0" which crashes
1935 // both GCC and VC++ implementations of the STL map. Hence when the
1936 // expected VR read happens to be non-ascii characters we consider
1937 // we hit falsely explicit VR tag.
1939 if ( !isalpha((unsigned char)vr[0]) && !isalpha((unsigned char)vr[1]) )
1941 realExplicit = false;
1944 // CLEANME searching the dicom_vr at each occurence is expensive.
1945 // PostPone this test in an optional integrity check at the end
1946 // of parsing or only in debug mode.
1947 if ( realExplicit && !Global::GetVR()->Count(vr) )
1949 realExplicit = false;
1952 if ( !realExplicit )
1954 // We thought this was explicit VR, but we end up with an
1955 // implicit VR tag. Let's backtrack.
1956 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
1957 entry->GetGroup(), entry->GetElement());
1958 dbg.Verbose(1, "Document::FindVR: ", msg.c_str());
1960 if( entry->GetGroup() % 2 && entry->GetElement() == 0x0000)
1962 // Group length is UL !
1963 DictEntry* newEntry = NewVirtualDictEntry(
1964 entry->GetGroup(), entry->GetElement(),
1965 "UL", "FIXME", "Group Length");
1966 entry->SetDictEntry( newEntry );
1971 if ( entry->IsVRUnknown() )
1973 // When not a dictionary entry, we can safely overwrite the VR.
1974 if( entry->GetElement() == 0x0000 )
1976 // Group length is UL !
1984 else if ( entry->GetVR() != vr )
1986 // The VR present in the file and the dictionary disagree. We assume
1987 // the file writer knew best and use the VR of the file. Since it would
1988 // be unwise to overwrite the VR of a dictionary (since it would
1989 // compromise it's next user), we need to clone the actual DictEntry
1990 // and change the VR for the read one.
1991 DictEntry* newEntry = NewVirtualDictEntry(
1992 entry->GetGroup(), entry->GetElement(),
1993 vr, "FIXME", entry->GetName());
1994 entry->SetDictEntry(newEntry);
2001 * \brief Get the transformed value of the header entry. The VR value
2002 * is used to define the transformation to operate on the value
2003 * \warning NOT end user intended method !
2004 * @param entry entry to tranform
2005 * @return Transformed entry value
2007 std::string Document::GetDocEntryValue(DocEntry *entry)
2009 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2011 std::string val = ((ValEntry *)entry)->GetValue();
2012 std::string vr = entry->GetVR();
2013 uint32_t length = entry->GetLength();
2014 std::ostringstream s;
2017 // When short integer(s) are expected, read and convert the following
2018 // n * 2 bytes properly i.e. as a multivaluated strings
2019 // (each single value is separated fromthe next one by '\'
2020 // as usual for standard multivaluated filels
2021 // Elements with Value Multiplicity > 1
2022 // contain a set of short integers (not a single one)
2024 if( vr == "US" || vr == "SS" )
2029 for (int i=0; i < nbInt; i++)
2035 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
2036 newInt16 = SwapShort( newInt16 );
2041 // When integer(s) are expected, read and convert the following
2042 // n * 4 bytes properly i.e. as a multivaluated strings
2043 // (each single value is separated fromthe next one by '\'
2044 // as usual for standard multivaluated filels
2045 // Elements with Value Multiplicity > 1
2046 // contain a set of integers (not a single one)
2047 else if( vr == "UL" || vr == "SL" )
2052 for (int i=0; i < nbInt; i++)
2058 newInt32 = ( val[4*i+0] & 0xFF )
2059 + (( val[4*i+1] & 0xFF ) << 8 )
2060 + (( val[4*i+2] & 0xFF ) << 16 )
2061 + (( val[4*i+3] & 0xFF ) << 24 );
2062 newInt32 = SwapLong( newInt32 );
2066 #ifdef GDCM_NO_ANSI_STRING_STREAM
2067 s << std::ends; // to avoid oddities on Solaris
2068 #endif //GDCM_NO_ANSI_STRING_STREAM
2072 return ((ValEntry *)entry)->GetValue();
2076 * \brief Get the reverse transformed value of the header entry. The VR
2077 * value is used to define the reverse transformation to operate on
2079 * \warning NOT end user intended method !
2080 * @param entry Entry to reverse transform
2081 * @return Reverse transformed entry value
2083 std::string Document::GetDocEntryUnvalue(DocEntry* entry)
2085 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
2087 std::string vr = entry->GetVR();
2088 std::vector<std::string> tokens;
2089 std::ostringstream s;
2091 if ( vr == "US" || vr == "SS" )
2095 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
2096 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2097 for (unsigned int i=0; i<tokens.size(); i++)
2099 newInt16 = atoi(tokens[i].c_str());
2100 s << ( newInt16 & 0xFF )
2101 << (( newInt16 >> 8 ) & 0xFF );
2105 if ( vr == "UL" || vr == "SL")
2109 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
2110 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
2111 for (unsigned int i=0; i<tokens.size();i++)
2113 newInt32 = atoi(tokens[i].c_str());
2114 s << (char)( newInt32 & 0xFF )
2115 << (char)(( newInt32 >> 8 ) & 0xFF )
2116 << (char)(( newInt32 >> 16 ) & 0xFF )
2117 << (char)(( newInt32 >> 24 ) & 0xFF );
2122 #ifdef GDCM_NO_ANSI_STRING_STREAM
2123 s << std::ends; // to avoid oddities on Solaris
2124 #endif //GDCM_NO_ANSI_STRING_STREAM
2128 return ((ValEntry *)entry)->GetValue();
2132 * \brief Skip a given Header Entry
2133 * \warning NOT end user intended method !
2134 * @param entry entry to skip
2136 void Document::SkipDocEntry(DocEntry *entry)
2138 SkipBytes(entry->GetLength());
2142 * \brief Skips to the begining of the next Header Entry
2143 * \warning NOT end user intended method !
2144 * @param entry entry to skip
2146 void Document::SkipToNextDocEntry(DocEntry *entry)
2148 Fp->seekg((long)(entry->GetOffset()), std::ios::beg);
2149 Fp->seekg( (long)(entry->GetReadLength()), std::ios::cur);
2153 * \brief When the length of an element value is obviously wrong (because
2154 * the parser went Jabberwocky) one can hope improving things by
2155 * applying some heuristics.
2156 * @param entry entry to check
2157 * @param foundLength fist assumption about length
2159 void Document::FixDocEntryFoundLength(DocEntry *entry,
2160 uint32_t foundLength)
2162 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
2163 if ( foundLength == 0xffffffff)
2168 uint16_t gr = entry->GetGroup();
2169 uint16_t el = entry->GetElement();
2171 if ( foundLength % 2)
2173 std::ostringstream s;
2174 s << "Warning : Tag with uneven length "
2176 << " in x(" << std::hex << gr << "," << el <<")" << std::dec;
2177 dbg.Verbose(0, s.str().c_str());
2180 //////// Fix for some naughty General Electric images.
2181 // Allthough not recent many such GE corrupted images are still present
2182 // on Creatis hard disks. Hence this fix shall remain when such images
2183 // are no longer in user (we are talking a few years, here)...
2184 // Note: XMedCom probably uses such a trick since it is able to read
2185 // those pesky GE images ...
2186 if ( foundLength == 13)
2188 // Only happens for this length !
2189 if ( entry->GetGroup() != 0x0008
2190 || ( entry->GetElement() != 0x0070
2191 && entry->GetElement() != 0x0080 ) )
2194 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
2198 //////// Fix for some brain-dead 'Leonardo' Siemens images.
2199 // Occurence of such images is quite low (unless one leaves close to a
2200 // 'Leonardo' source. Hence, one might consider commenting out the
2201 // following fix on efficiency reasons.
2202 else if ( entry->GetGroup() == 0x0009
2203 && ( entry->GetElement() == 0x1113
2204 || entry->GetElement() == 0x1114 ) )
2207 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
2210 else if ( entry->GetVR() == "SQ" )
2212 foundLength = 0; // ReadLength is unchanged
2215 //////// We encountered a 'delimiter' element i.e. a tag of the form
2216 // "fffe|xxxx" which is just a marker. Delimiters length should not be
2217 // taken into account.
2218 else if( entry->GetGroup() == 0xfffe )
2220 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
2221 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
2222 // causes extra troubles...
2223 if( entry->GetElement() != 0x0000 )
2229 entry->SetUsableLength(foundLength);
2233 * \brief Apply some heuristics to predict whether the considered
2234 * element value contains/represents an integer or not.
2235 * @param entry The element value on which to apply the predicate.
2236 * @return The result of the heuristical predicate.
2238 bool Document::IsDocEntryAnInteger(DocEntry *entry)
2240 uint16_t element = entry->GetElement();
2241 uint16_t group = entry->GetGroup();
2242 const std::string & vr = entry->GetVR();
2243 uint32_t length = entry->GetLength();
2245 // When we have some semantics on the element we just read, and if we
2246 // a priori know we are dealing with an integer, then we shall be
2247 // able to swap it's element value properly.
2248 if ( element == 0 ) // This is the group length of the group
2256 // Allthough this should never happen, still some images have a
2257 // corrupted group length [e.g. have a glance at offset x(8336) of
2258 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2259 // Since for dicom compliant and well behaved headers, the present
2260 // test is useless (and might even look a bit paranoid), when we
2261 // encounter such an ill-formed image, we simply display a warning
2262 // message and proceed on parsing (while crossing fingers).
2263 std::ostringstream s;
2264 long filePosition = Fp->tellg();
2265 s << "Erroneous Group Length element length on : (" \
2266 << std::hex << group << " , " << element
2267 << ") -before- position x(" << filePosition << ")"
2268 << "lgt : " << length;
2269 dbg.Verbose(0, "Document::IsDocEntryAnInteger", s.str().c_str() );
2273 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2282 * \brief Find the Length till the next sequence delimiter
2283 * \warning NOT end user intended method !
2287 uint32_t Document::FindDocEntryLengthOB()
2288 throw( FormatUnexpected )
2290 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2291 long positionOnEntry = Fp->tellg();
2292 bool foundSequenceDelimiter = false;
2293 uint32_t totalLength = 0;
2295 while ( !foundSequenceDelimiter )
2301 group = ReadInt16();
2304 catch ( FormatError )
2306 throw FormatError("Document::FindDocEntryLengthOB()",
2307 " group or element not present.");
2310 // We have to decount the group and element we just read
2313 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2315 dbg.Verbose(1, "Document::FindDocEntryLengthOB: neither an Item "
2316 "tag nor a Sequence delimiter tag.");
2317 Fp->seekg(positionOnEntry, std::ios::beg);
2318 throw FormatUnexpected("Document::FindDocEntryLengthOB()",
2319 "Neither an Item tag nor a Sequence "
2323 if ( elem == 0xe0dd )
2325 foundSequenceDelimiter = true;
2328 uint32_t itemLength = ReadInt32();
2329 // We add 4 bytes since we just read the ItemLength with ReadInt32
2330 totalLength += itemLength + 4;
2331 SkipBytes(itemLength);
2333 if ( foundSequenceDelimiter )
2338 Fp->seekg( positionOnEntry, std::ios::beg);
2343 * \brief Reads a supposed to be 16 Bits integer
2344 * (swaps it depending on processor endianity)
2345 * @return read value
2347 uint16_t Document::ReadInt16()
2348 throw( FormatError )
2351 Fp->read ((char*)&g, (size_t)2);
2354 throw FormatError( "Document::ReadInt16()", " file error." );
2358 throw FormatError( "Document::ReadInt16()", "EOF." );
2365 * \brief Reads a supposed to be 32 Bits integer
2366 * (swaps it depending on processor endianity)
2367 * @return read value
2369 uint32_t Document::ReadInt32()
2370 throw( FormatError )
2373 Fp->read ((char*)&g, (size_t)4);
2376 throw FormatError( "Document::ReadInt32()", " file error." );
2380 throw FormatError( "Document::ReadInt32()", "EOF." );
2387 * \brief skips bytes inside the source file
2388 * \warning NOT end user intended method !
2391 void Document::SkipBytes(uint32_t nBytes)
2393 //FIXME don't dump the returned value
2394 Fp->seekg((long)nBytes, std::ios::cur);
2398 * \brief Loads all the needed Dictionaries
2399 * \warning NOT end user intended method !
2401 void Document::Initialise()
2403 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2405 RLEInfo = new RLEFramesInfo;
2406 JPEGInfo = new JPEGFragmentsInfo;
2411 * \brief Discover what the swap code is (among little endian, big endian,
2412 * bad little endian, bad big endian).
2414 * @return false when we are absolutely sure
2415 * it's neither ACR-NEMA nor DICOM
2416 * true when we hope ours assuptions are OK
2418 bool Document::CheckSwap()
2420 // The only guaranted way of finding the swap code is to find a
2421 // group tag since we know it's length has to be of four bytes i.e.
2422 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2423 // occurs when we can't find such group...
2425 uint32_t x = 4; // x : for ntohs
2426 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2432 // First, compare HostByteOrder and NetworkByteOrder in order to
2433 // determine if we shall need to swap bytes (i.e. the Endian type).
2434 if ( x == ntohs(x) )
2443 // The easiest case is the one of a DICOM header, since it possesses a
2444 // file preamble where it suffice to look for the string "DICM".
2447 char *entCur = deb + 128;
2448 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2450 dbg.Verbose(1, "Document::CheckSwap:", "looks like DICOM Version3");
2452 // Next, determine the value representation (VR). Let's skip to the
2453 // first element (0002, 0000) and check there if we find "UL"
2454 // - or "OB" if the 1st one is (0002,0001) -,
2455 // in which case we (almost) know it is explicit VR.
2456 // WARNING: if it happens to be implicit VR then what we will read
2457 // is the length of the group. If this ascii representation of this
2458 // length happens to be "UL" then we shall believe it is explicit VR.
2459 // FIXME: in order to fix the above warning, we could read the next
2460 // element value (or a couple of elements values) in order to make
2461 // sure we are not commiting a big mistake.
2462 // We need to skip :
2463 // * the 128 bytes of File Preamble (often padded with zeroes),
2464 // * the 4 bytes of "DICM" string,
2465 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2466 // i.e. a total of 136 bytes.
2470 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2471 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2472 // *Implicit* VR. -and it is !-
2474 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2475 memcmp(entCur, "OB", (size_t)2) == 0 ||
2476 memcmp(entCur, "UI", (size_t)2) == 0 ||
2477 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2478 // when Write DCM *adds*
2480 // Use Document::dicom_vr to test all the possibilities
2481 // instead of just checking for UL, OB and UI !? group 0000
2483 Filetype = ExplicitVR;
2484 dbg.Verbose(1, "Document::CheckSwap:",
2485 "explicit Value Representation");
2489 Filetype = ImplicitVR;
2490 dbg.Verbose(1, "Document::CheckSwap:",
2491 "not an explicit Value Representation");
2497 dbg.Verbose(1, "Document::CheckSwap:",
2498 "HostByteOrder != NetworkByteOrder");
2503 dbg.Verbose(1, "Document::CheckSwap:",
2504 "HostByteOrder = NetworkByteOrder");
2507 // Position the file position indicator at first tag (i.e.
2508 // after the file preamble and the "DICM" string).
2509 Fp->seekg(0, std::ios::beg);
2510 Fp->seekg ( 132L, std::ios::beg);
2514 // Alas, this is not a DicomV3 file and whatever happens there is no file
2515 // preamble. We can reset the file position indicator to where the data
2516 // is (i.e. the beginning of the file).
2517 dbg.Verbose(1, "Document::CheckSwap:", "not a DICOM Version3 file");
2518 Fp->seekg(0, std::ios::beg);
2520 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2521 // By clean we mean that the length of the first tag is written down.
2522 // If this is the case and since the length of the first group HAS to be
2523 // four (bytes), then determining the proper swap code is straightforward.
2526 // We assume the array of char we are considering contains the binary
2527 // representation of a 32 bits integer. Hence the following dirty
2529 s32 = *((uint32_t *)(entCur));
2550 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2551 // It is time for despaired wild guesses.
2552 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2553 // i.e. the 'group length' element is not present :
2555 // check the supposed-to-be 'group number'
2556 // in ( 0x0001 .. 0x0008 )
2557 // to determine ' SwapCode' value .
2558 // Only 0 or 4321 will be possible
2559 // (no oportunity to check for the formerly well known
2560 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2561 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2562 // the file IS NOT ACR-NEMA nor DICOM V3
2563 // Find a trick to tell it the caller...
2565 s16 = *((uint16_t *)(deb));
2592 dbg.Verbose(0, "Document::CheckSwap:",
2593 "ACR/NEMA unfound swap info (Really hopeless !)");
2597 // Then the only info we have is the net2host one.
2609 * \brief Restore the unproperly loaded values i.e. the group, the element
2610 * and the dictionary entry depending on them.
2612 void Document::SwitchSwapToBigEndian()
2614 dbg.Verbose(1, "Document::SwitchSwapToBigEndian",
2615 "Switching to BigEndian mode.");
2616 if ( SwapCode == 0 )
2620 else if ( SwapCode == 4321 )
2624 else if ( SwapCode == 3412 )
2628 else if ( SwapCode == 2143 )
2635 * \brief during parsing, Header Elements too long are not loaded in memory
2638 void Document::SetMaxSizeLoadEntry(long newSize)
2644 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2646 MaxSizeLoadEntry = 0xffffffff;
2649 MaxSizeLoadEntry = newSize;
2654 * \brief Header Elements too long will not be printed
2655 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2658 void Document::SetMaxSizePrintEntry(long newSize)
2660 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2665 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2667 MaxSizePrintEntry = 0xffffffff;
2670 MaxSizePrintEntry = newSize;
2676 * \brief Handle broken private tag from Philips NTSCAN
2677 * where the endianess is being switch to BigEndian for no
2681 void Document::HandleBrokenEndian(uint16_t group, uint16_t elem)
2683 // Endian reversion. Some files contain groups of tags with reversed endianess.
2684 static int reversedEndian = 0;
2685 // try to fix endian switching in the middle of headers
2686 if ((group == 0xfeff) && (elem == 0x00e0))
2688 // start endian swap mark for group found
2690 SwitchSwapToBigEndian();
2695 else if ((group == 0xfffe) && (elem == 0xe00d) && reversedEndian)
2697 // end of reversed endian group
2699 SwitchSwapToBigEndian();
2705 * \brief Read the next tag but WITHOUT loading it's value
2706 * (read the 'Group Number', the 'Element Number',
2707 * gets the Dict Entry
2708 * gets the VR, gets the length, gets the offset value)
2709 * @return On succes the newly created DocEntry, NULL on failure.
2711 DocEntry* Document::ReadNextDocEntry()
2718 group = ReadInt16();
2721 catch ( FormatError e )
2723 // We reached the EOF (or an error occured) therefore
2724 // header parsing has to be considered as finished.
2729 HandleBrokenEndian(group, elem);
2730 DocEntry *newEntry = NewDocEntryByNumber(group, elem);
2731 FindDocEntryVR(newEntry);
2735 FindDocEntryLength(newEntry);
2737 catch ( FormatError e )
2745 newEntry->SetOffset(Fp->tellg());
2752 * \brief Generate a free TagKey i.e. a TagKey that is not present
2753 * in the TagHt dictionary.
2754 * @param group The generated tag must belong to this group.
2755 * @return The element of tag with given group which is fee.
2757 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2759 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2761 TagKey key = DictEntry::TranslateToKey(group, elem);
2762 if (TagHT.count(key) == 0)
2771 * \brief Assuming the internal file pointer \ref Document::Fp
2772 * is placed at the beginning of a tag check whether this
2773 * tag is (TestGroup, TestElement).
2774 * \warning On success the internal file pointer \ref Document::Fp
2775 * is modified to point after the tag.
2776 * On failure (i.e. when the tag wasn't the expected tag
2777 * (TestGroup, TestElement) the internal file pointer
2778 * \ref Document::Fp is restored to it's original position.
2779 * @param testGroup The expected group of the tag.
2780 * @param testElement The expected Element of the tag.
2781 * @return True on success, false otherwise.
2783 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2785 long positionOnEntry = Fp->tellg();
2786 long currentPosition = Fp->tellg(); // On debugging purposes
2788 //// Read the Item Tag group and element, and make
2789 // sure they are what we expected:
2790 uint16_t itemTagGroup;
2791 uint16_t itemTagElement;
2794 itemTagGroup = ReadInt16();
2795 itemTagElement = ReadInt16();
2797 catch ( FormatError e )
2799 //std::cerr << e << std::endl;
2802 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2804 std::ostringstream s;
2805 s << " We should have found tag (";
2806 s << std::hex << testGroup << "," << testElement << ")" << std::endl;
2807 s << " but instead we encountered tag (";
2808 s << std::hex << itemTagGroup << "," << itemTagElement << ")"
2810 s << " at address: " << (unsigned)currentPosition << std::endl;
2811 dbg.Verbose(0, "Document::ReadItemTagLength: wrong Item Tag found:");
2812 dbg.Verbose(0, s.str().c_str());
2813 Fp->seekg(positionOnEntry, std::ios::beg);
2821 * \brief Assuming the internal file pointer \ref Document::Fp
2822 * is placed at the beginning of a tag (TestGroup, TestElement),
2823 * read the length associated to the Tag.
2824 * \warning On success the internal file pointer \ref Document::Fp
2825 * is modified to point after the tag and it's length.
2826 * On failure (i.e. when the tag wasn't the expected tag
2827 * (TestGroup, TestElement) the internal file pointer
2828 * \ref Document::Fp is restored to it's original position.
2829 * @param testGroup The expected group of the tag.
2830 * @param testElement The expected Element of the tag.
2831 * @return On success returns the length associated to the tag. On failure
2834 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2836 long positionOnEntry = Fp->tellg();
2837 (void)positionOnEntry;
2839 if ( !ReadTag(testGroup, testElement) )
2844 //// Then read the associated Item Length
2845 long currentPosition = Fp->tellg();
2846 uint32_t itemLength = ReadInt32();
2848 std::ostringstream s;
2849 s << "Basic Item Length is: "
2850 << itemLength << std::endl;
2851 s << " at address: " << (unsigned)currentPosition << std::endl;
2852 dbg.Verbose(0, "Document::ReadItemTagLength: ", s.str().c_str());
2858 * \brief When parsing the Pixel Data of an encapsulated file, read
2859 * the basic offset table (when present, and BTW dump it).
2861 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2863 //// Read the Basic Offset Table Item Tag length...
2864 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2866 // When present, read the basic offset table itself.
2867 // Notes: - since the presence of this basic offset table is optional
2868 // we can't rely on it for the implementation, and we will simply
2869 // trash it's content (when present).
2870 // - still, when present, we could add some further checks on the
2871 // lengths, but we won't bother with such fuses for the time being.
2872 if ( itemLength != 0 )
2874 char* basicOffsetTableItemValue = new char[itemLength + 1];
2875 Fp->read(basicOffsetTableItemValue, itemLength);
2878 for (unsigned int i=0; i < itemLength; i += 4 )
2880 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2882 std::ostringstream s;
2883 s << " Read one length: ";
2884 s << std::hex << individualLength << std::endl;
2886 "Document::ReadAndSkipEncapsulatedBasicOffsetTable: ",
2891 delete[] basicOffsetTableItemValue;
2896 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2897 * Compute the RLE extra information and store it in \ref RLEInfo
2898 * for later pixel retrieval usage.
2900 void Document::ComputeRLEInfo()
2902 TransferSyntaxType ts = GetTransferSyntax();
2903 if ( ts != RLELossless )
2908 // Encoded pixel data: for the time being we are only concerned with
2909 // Jpeg or RLE Pixel data encodings.
2910 // As stated in PS 3.5-2003, section 8.2 p44:
2911 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2912 // value representation OB is used".
2913 // Hence we expect an OB value representation. Concerning OB VR,
2914 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2915 // "For the Value Representations OB and OW, the encoding shall meet the
2916 // following specifications depending on the Data element tag:"
2918 // - the first item in the sequence of items before the encoded pixel
2919 // data stream shall be basic offset table item. The basic offset table
2920 // item value, however, is not required to be present"
2922 ReadAndSkipEncapsulatedBasicOffsetTable();
2924 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2925 // Loop on the individual frame[s] and store the information
2926 // on the RLE fragments in a RLEFramesInfo.
2927 // Note: - when only a single frame is present, this is a
2929 // - when more than one frame are present, then we are in
2930 // the case of a multi-frame image.
2932 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2934 // Parse the RLE Header and store the corresponding RLE Segment
2935 // Offset Table information on fragments of this current Frame.
2936 // Note that the fragment pixels themselves are not loaded
2937 // (but just skipped).
2938 long frameOffset = Fp->tellg();
2940 uint32_t nbRleSegments = ReadInt32();
2941 if ( nbRleSegments > 16 )
2943 // There should be at most 15 segments (refer to RLEFrame class)
2944 dbg.Verbose(0, "Document::ComputeRLEInfo: too many segments.");
2947 uint32_t rleSegmentOffsetTable[16];
2948 for( int k = 1; k <= 15; k++ )
2950 rleSegmentOffsetTable[k] = ReadInt32();
2953 // Deduce from both the RLE Header and the frameLength the
2954 // fragment length, and again store this info in a
2956 long rleSegmentLength[15];
2957 // skipping (not reading) RLE Segments
2958 if ( nbRleSegments > 1)
2960 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2962 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2963 - rleSegmentOffsetTable[k];
2964 SkipBytes(rleSegmentLength[k]);
2968 rleSegmentLength[nbRleSegments] = frameLength
2969 - rleSegmentOffsetTable[nbRleSegments];
2970 SkipBytes(rleSegmentLength[nbRleSegments]);
2972 // Store the collected info
2973 RLEFrame* newFrameInfo = new RLEFrame;
2974 newFrameInfo->NumberFragments = nbRleSegments;
2975 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2977 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2978 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2980 RLEInfo->Frames.push_back( newFrameInfo );
2983 // Make sure that at the end of the item we encounter a 'Sequence
2985 if ( !ReadTag(0xfffe, 0xe0dd) )
2987 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2988 dbg.Verbose(0, " item at end of RLE item sequence");
2993 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2994 * Compute the jpeg extra information (fragment[s] offset[s] and
2995 * length) and store it[them] in \ref JPEGInfo for later pixel
2998 void Document::ComputeJPEGFragmentInfo()
3000 // If you need to, look for comments of ComputeRLEInfo().
3006 ReadAndSkipEncapsulatedBasicOffsetTable();
3008 // Loop on the fragments[s] and store the parsed information in a
3010 long fragmentLength;
3011 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
3013 long fragmentOffset = Fp->tellg();
3015 // Store the collected info
3016 JPEGFragment* newFragment = new JPEGFragment;
3017 newFragment->Offset = fragmentOffset;
3018 newFragment->Length = fragmentLength;
3019 JPEGInfo->Fragments.push_back( newFragment );
3021 SkipBytes( fragmentLength );
3024 // Make sure that at the end of the item we encounter a 'Sequence
3026 if ( !ReadTag(0xfffe, 0xe0dd) )
3028 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
3029 dbg.Verbose(0, " item at end of JPEG item sequence");
3034 * \brief Walk recursively the given \ref DocEntrySet, and feed
3035 * the given hash table (\ref TagDocEntryHT) with all the
3036 * \ref DocEntry (Dicom entries) encountered.
3037 * This method does the job for \ref BuildFlatHashTable.
3038 * @param builtHT Where to collect all the \ref DocEntry encountered
3039 * when recursively walking the given set.
3040 * @param set The structure to be traversed (recursively).
3042 void Document::BuildFlatHashTableRecurse( TagDocEntryHT& builtHT,
3045 if (ElementSet* elementSet = dynamic_cast< ElementSet* > ( set ) )
3047 TagDocEntryHT const & currentHT = elementSet->GetTagHT();
3048 for( TagDocEntryHT::const_iterator i = currentHT.begin();
3049 i != currentHT.end();
3052 DocEntry* entry = i->second;
3053 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3055 const ListSQItem& items = seqEntry->GetSQItems();
3056 for( ListSQItem::const_iterator item = items.begin();
3057 item != items.end();
3060 BuildFlatHashTableRecurse( builtHT, *item );
3064 builtHT[entry->GetKey()] = entry;
3069 if (SQItem* SQItemSet = dynamic_cast< SQItem* > ( set ) )
3071 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
3072 for (ListDocEntry::const_iterator i = currentList.begin();
3073 i != currentList.end();
3076 DocEntry* entry = *i;
3077 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
3079 const ListSQItem& items = seqEntry->GetSQItems();
3080 for( ListSQItem::const_iterator item = items.begin();
3081 item != items.end();
3084 BuildFlatHashTableRecurse( builtHT, *item );
3088 builtHT[entry->GetKey()] = entry;
3095 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
3098 * The structure used by a Document (through \ref ElementSet),
3099 * in order to hold the parsed entries of a Dicom header, is a recursive
3100 * one. This is due to the fact that the sequences (when present)
3101 * can be nested. Additionaly, the sequence items (represented in
3102 * gdcm as \ref SQItem) add an extra complexity to the data
3103 * structure. Hence, a gdcm user whishing to visit all the entries of
3104 * a Dicom header will need to dig in the gdcm internals (which
3105 * implies exposing all the internal data structures to the API).
3106 * In order to avoid this burden to the user, \ref BuildFlatHashTable
3107 * recursively builds a temporary hash table, which holds all the
3108 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
3110 * \warning Of course there is NO integrity constrain between the
3111 * returned \ref TagDocEntryHT and the \ref ElementSet used
3112 * to build it. Hence if the underlying \ref ElementSet is
3113 * altered, then it is the caller responsability to invoke
3114 * \ref BuildFlatHashTable again...
3115 * @return The flat std::map<> we juste build.
3117 TagDocEntryHT* Document::BuildFlatHashTable()
3119 TagDocEntryHT* FlatHT = new TagDocEntryHT;
3120 BuildFlatHashTableRecurse( *FlatHT, this );
3127 * \brief Compares two documents, according to \ref DicomDir rules
3128 * \warning Does NOT work with ACR-NEMA files
3129 * \todo Find a trick to solve the pb (use RET fields ?)
3131 * @return true if 'smaller'
3133 bool Document::operator<(Document &document)
3136 std::string s1 = GetEntryByNumber(0x0010,0x0010);
3137 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
3149 s1 = GetEntryByNumber(0x0010,0x0020);
3150 s2 = document.GetEntryByNumber(0x0010,0x0020);
3161 // Study Instance UID
3162 s1 = GetEntryByNumber(0x0020,0x000d);
3163 s2 = document.GetEntryByNumber(0x0020,0x000d);
3174 // Serie Instance UID
3175 s1 = GetEntryByNumber(0x0020,0x000e);
3176 s2 = document.GetEntryByNumber(0x0020,0x000e);
3191 } // end namespace gdcm
3193 //-----------------------------------------------------------------------------