1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/07 19:20:38 $
7 Version: $Revision: 1.165 $
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 Big 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 gdcmVerboseMacro("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 gdcmVerboseMacro("Document::IsReadable: wrong filetype");
292 gdcmVerboseMacro("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)
427 gdcmVerboseMacro( "Document::OpenFile is already opened when opening: " <<
431 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
434 gdcmVerboseMacro( "Document::OpenFile cannot open file: " <<
442 Fp->read((char*)&zero, (size_t)2);
449 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
451 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
452 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
453 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
454 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
460 Fp->seekg(126L, std::ios::cur);
462 Fp->read(dicm, (size_t)4);
468 if( memcmp(dicm, "DICM", 4) == 0 )
474 gdcmVerboseMacro( "Document::OpenFile not DICOM/ACR (missing preamble)" <<
481 * \brief closes the file
482 * @return TRUE if the close was successfull
484 bool Document::CloseFile()
493 return true; //FIXME how do we detect a non-close ifstream ?
497 * \brief Writes in a file all the Header Entries (Dicom Elements)
498 * @param fp file pointer on an already open file
499 * @param filetype Type of the File to be written
500 * (ACR-NEMA, ExplicitVR, ImplicitVR)
501 * \return Always true.
503 void Document::WriteContent(std::ofstream *fp, FileType filetype)
505 /// \todo move the following lines (and a lot of others, to be written)
506 /// to a future function CheckAndCorrectHeader
507 /// (necessary if user wants to write a DICOM V3 file
508 /// starting from an ACR-NEMA (V2) Header
510 if ( filetype == ImplicitVR || filetype == ExplicitVR )
512 // writing Dicom File Preamble
513 char filePreamble[128];
514 memset(filePreamble, 0, 128);
515 fp->write(filePreamble, 128);
516 fp->write("DICM", 4);
520 * \todo rewrite later, if really usefull
521 * - 'Group Length' element is optional in DICOM
522 * - but un-updated odd groups lengthes can causes pb
525 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
526 * UpdateGroupLength(false,filetype);
527 * if ( filetype == ACR)
528 * UpdateGroupLength(true,ACR);
531 ElementSet::WriteContent(fp, filetype); // This one is recursive
535 * \brief Modifies the value of a given Doc Entry (Dicom Element)
536 * when it exists. Create it with the given value when unexistant.
537 * @param value (string) Value to be set
538 * @param group Group number of the Entry
539 * @param elem Element number of the Entry
540 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
541 * \return pointer to the modified/created Header Entry (NULL when creation
544 ValEntry *Document::ReplaceOrCreateByNumber(std::string const &value,
549 ValEntry *valEntry = 0;
550 DocEntry *currentEntry = GetDocEntryByNumber( group, elem);
554 valEntry = dynamic_cast< ValEntry* >(currentEntry);
558 if( valEntry->GetVR()!=vr )
561 // if currentEntry doesn't correspond to the requested valEntry
564 if (!RemoveEntry(currentEntry))
566 gdcmVerboseMacro("Document::ReplaceOrCreateByNumber: removal"
567 " of previous DocEntry failed.");
574 // Create a new valEntry if necessary
577 valEntry = NewValEntryByNumber(group, elem, vr);
579 if ( !AddEntry(valEntry))
581 gdcmVerboseMacro("Document::ReplaceOrCreateByNumber: AddEntry"
582 " failed allthough this is a creation.");
589 // Set the binEntry value
590 SetEntry(value, valEntry);
595 * \brief Modifies the value of a given Header Entry (Dicom Element)
596 * when it exists. Create it with the given value when unexistant.
597 * A copy of the binArea is made to be kept in the Document.
598 * @param binArea (binary) value to be set
599 * @param Group Group number of the Entry
600 * @param Elem Element number of the Entry
601 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
602 * \return pointer to the modified/created Header Entry (NULL when creation
605 BinEntry *Document::ReplaceOrCreateByNumber(uint8_t *binArea,
611 BinEntry *binEntry = 0;
612 DocEntry *currentEntry = GetDocEntryByNumber( group, elem);
614 // Verify the currentEntry
617 binEntry = dynamic_cast< BinEntry* >(currentEntry);
621 if( binEntry->GetVR()!=vr )
624 // if currentEntry doesn't correspond to the requested valEntry
627 if (!RemoveEntry(currentEntry))
629 gdcmVerboseMacro("Document::ReplaceOrCreateByNumber: removal"
630 " of previous DocEntry failed.");
637 // Create a new binEntry if necessary
640 binEntry = NewBinEntryByNumber(group, elem, vr);
642 if ( !AddEntry(binEntry))
644 gdcmVerboseMacro("Document::ReplaceOrCreateByNumber: AddEntry"
645 " failed allthough this is a creation.");
652 // Set the binEntry value
654 if (lgth>0 && binArea)
656 tmpArea = new uint8_t[lgth];
657 memcpy(tmpArea,binArea,lgth);
663 if (!SetEntry(tmpArea,lgth,binEntry))
675 * \brief Modifies the value of a given Header Entry (Dicom Element)
676 * when it exists. Create it when unexistant.
677 * @param Group Group number of the Entry
678 * @param Elem Element number of the Entry
679 * \return pointer to the modified/created SeqEntry (NULL when creation
682 SeqEntry *Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem)
684 SeqEntry *seqEntry = 0;
685 DocEntry *currentEntry = GetDocEntryByNumber( group, elem);
687 // Verify the currentEntry
690 seqEntry = dynamic_cast< SeqEntry* >(currentEntry);
694 if( seqEntry->GetVR()!="SQ" )
697 // if currentEntry doesn't correspond to the requested valEntry
700 if (!RemoveEntry(currentEntry))
702 gdcmVerboseMacro("Document::ReplaceOrCreateByNumber: removal"
703 " of previous DocEntry failed.");
710 // Create a new seqEntry if necessary
713 seqEntry = NewSeqEntryByNumber(group, elem);
715 if ( !AddEntry(seqEntry))
717 gdcmVerboseMacro("Document::ReplaceOrCreateByNumber: AddEntry"
718 " failed allthough this is a creation.");
729 * \brief Set a new value if the invoked element exists
730 * Seems to be useless !!!
731 * @param value new element value
732 * @param group group number of the Entry
733 * @param elem element number of the Entry
736 bool Document::ReplaceIfExistByNumber(std::string const &value,
737 uint16_t group, uint16_t elem )
739 SetEntryByNumber(value, group, elem);
744 std::string Document::GetTransferSyntaxValue(TransferSyntaxType type)
746 return TransferSyntaxStrings[type];
749 //-----------------------------------------------------------------------------
753 * \brief Checks if a given Dicom Element exists within the H table
754 * @param group Group number of the searched Dicom Element
755 * @param element Element number of the searched Dicom Element
756 * @return true is found
758 bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
760 const std::string &key = DictEntry::TranslateToKey(group, element );
761 return TagHT.count(key) != 0;
766 * \brief Searches within Header Entries (Dicom Elements) parsed with
767 * the public and private dictionaries
768 * for the element value representation of a given tag.
769 * @param group Group number of the searched tag.
770 * @param element Element number of the searched tag.
771 * @return Corresponding element value representation when it exists,
772 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
774 std::string Document::GetEntryByNumber(uint16_t group, uint16_t element)
776 TagKey key = DictEntry::TranslateToKey(group, element);
777 if ( !TagHT.count(key))
782 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
786 * \brief Searches within Header Entries (Dicom Elements) parsed with
787 * the public and private dictionaries
788 * for the element value representation of a given tag..
790 * Obtaining the VR (Value Representation) might be needed by caller
791 * to convert the string typed content to caller's native type
792 * (think of C++ vs Python). The VR is actually of a higher level
793 * of semantics than just the native C++ type.
794 * @param group Group number of the searched tag.
795 * @param element Element number of the searched tag.
796 * @return Corresponding element value representation when it exists,
797 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
799 std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element)
801 DocEntry *elem = GetDocEntryByNumber(group, element);
806 return elem->GetVR();
810 * \brief Searches within Header Entries (Dicom Elements) parsed with
811 * the public and private dictionaries
812 * for the value length of a given tag..
813 * @param group Group number of the searched tag.
814 * @param element Element number of the searched tag.
815 * @return Corresponding element length; -2 if not found
817 int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element)
819 DocEntry *elem = GetDocEntryByNumber(group, element);
822 return -2; //magic number
824 return elem->GetLength();
828 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
829 * through it's (group, element) and modifies it's content with
831 * @param content new value (string) to substitute with
832 * @param group group number of the Dicom Element to modify
833 * @param element element number of the Dicom Element to modify
835 bool Document::SetEntryByNumber(std::string const& content,
836 uint16_t group, uint16_t element)
838 ValEntry *entry = GetValEntryByNumber(group, element);
841 gdcmVerboseMacro("Document::SetEntryByNumber: no corresponding"
842 " ValEntry (try promotion first).");
845 return SetEntry(content,entry);
849 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
850 * through it's (group, element) and modifies it's content with
852 * @param content new value (void* -> uint8_t*) to substitute with
853 * @param lgth new value length
854 * @param group group number of the Dicom Element to modify
855 * @param element element number of the Dicom Element to modify
857 bool Document::SetEntryByNumber(uint8_t*content, int lgth,
858 uint16_t group, uint16_t element)
860 BinEntry *entry = GetBinEntryByNumber(group, element);
863 gdcmVerboseMacro( "Document::SetEntryByNumber: no corresponding"
864 " ValEntry (try promotion first).");
868 return SetEntry(content,lgth,entry);
872 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
873 * and modifies it's content with the given value.
874 * @param content new value (string) to substitute with
875 * @param entry Entry to be modified
877 bool Document::SetEntry(std::string const &content,ValEntry *entry)
881 entry->SetValue(content);
888 * \brief Accesses an existing BinEntry (i.e. a Dicom Element)
889 * and modifies it's content with the given value.
890 * @param content new value (void* -> uint8_t*) to substitute with
891 * @param entry Entry to be modified
892 * @param lgth new value length
894 bool Document::SetEntry(uint8_t *content, int lgth, BinEntry *entry)
898 // Hope Binary field length is *never* wrong
899 /*if(lgth%2) // Non even length are padded with a space (020H).
902 //content = content + '\0'; // fing a trick to enlarge a binary field?
905 entry->SetBinArea(content);
906 entry->SetLength(lgth);
907 entry->SetValue(GDCM_BINLOADED);
914 * \brief Gets (from Header) a 'non string' element value
915 * (LoadElementValues has already be executed)
916 * @param group group number of the Entry
917 * @param elem element number of the Entry
918 * @return Pointer to the 'non string' area
920 void *Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem)
922 DocEntry *entry = GetDocEntryByNumber(group, elem);
925 gdcmVerboseMacro("Document::GetDocEntryByNumber: no entry");
928 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(entry) )
930 return binEntry->GetBinArea();
937 * \brief Loads (from disk) the element content
938 * when a string is not suitable
939 * @param group group number of the Entry
940 * @param elem element number of the Entry
942 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
944 // Search the corresponding DocEntry
945 DocEntry *docElement = GetDocEntryByNumber(group, elem);
949 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
953 LoadEntryBinArea(binElement);
957 * \brief Loads (from disk) the element content
958 * when a string is not suitable
959 * @param element Entry whose binArea is going to be loaded
961 void Document::LoadEntryBinArea(BinEntry *element)
963 if(element->GetBinArea())
970 size_t o =(size_t)element->GetOffset();
971 Fp->seekg(o, std::ios::beg);
973 size_t l = element->GetLength();
974 uint8_t *a = new uint8_t[l];
977 gdcmVerboseMacro("Document::LoadEntryBinArea cannot allocate a");
981 /// \todo check the result
982 Fp->read((char*)a, l);
983 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
989 element->SetBinArea(a);
996 * \brief Sets a 'non string' value to a given Dicom Element
997 * @param area area containing the 'non string' value
998 * @param group Group number of the searched Dicom Element
999 * @param element Element number of the searched Dicom Element
1002 /*bool Document::SetEntryBinAreaByNumber(uint8_t *area,
1003 uint16_t group, uint16_t element)
1005 DocEntry *currentEntry = GetDocEntryByNumber(group, element);
1006 if ( !currentEntry )
1011 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1013 binEntry->SetBinArea( area );
1021 * \brief retrieves a Dicom Element (the first one) using (group, element)
1022 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1023 * if you think it's NOT UNIQUE, check the count number
1024 * and use iterators to retrieve ALL the Dicoms Elements within
1025 * a given couple (group, element)
1026 * @param group Group number of the searched Dicom Element
1027 * @param element Element number of the searched Dicom Element
1030 DocEntry *Document::GetDocEntryByNumber(uint16_t group, uint16_t element)
1032 TagKey key = DictEntry::TranslateToKey(group, element);
1033 if ( !TagHT.count(key))
1037 return TagHT.find(key)->second;
1041 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1042 * returns a result when the corresponding entry is of type
1044 * @return When present, the corresponding ValEntry.
1046 ValEntry *Document::GetValEntryByNumber(uint16_t group, uint16_t element)
1048 DocEntry *currentEntry = GetDocEntryByNumber(group, element);
1049 if ( !currentEntry )
1053 if ( ValEntry *entry = dynamic_cast<ValEntry*>(currentEntry) )
1057 gdcmVerboseMacro("Document::GetValEntryByNumber: unfound ValEntry.");
1063 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1064 * returns a result when the corresponding entry is of type
1066 * @return When present, the corresponding BinEntry.
1068 BinEntry *Document::GetBinEntryByNumber(uint16_t group, uint16_t element)
1070 DocEntry *currentEntry = GetDocEntryByNumber(group, element);
1071 if ( !currentEntry )
1075 if ( BinEntry *entry = dynamic_cast<BinEntry*>(currentEntry) )
1079 gdcmVerboseMacro("Document::GetBinEntryByNumber: unfound BinEntry.");
1085 * \brief Loads the element while preserving the current
1086 * underlying file position indicator as opposed to
1087 * to LoadDocEntry that modifies it.
1088 * @param entry Header Entry whose value shall be loaded.
1091 void Document::LoadDocEntrySafe(DocEntry *entry)
1095 long PositionOnEntry = Fp->tellg();
1096 LoadDocEntry(entry);
1097 Fp->seekg(PositionOnEntry, std::ios::beg);
1102 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1104 * @return The properly swaped 32 bits integer.
1106 uint32_t Document::SwapLong(uint32_t a)
1113 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1114 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1118 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1122 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1125 //std::cout << "swapCode= " << SwapCode << std::endl;
1126 gdcmErrorMacro(" Document::SwapLong : unset swap code");
1133 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1135 * @return The properly unswaped 32 bits integer.
1137 uint32_t Document::UnswapLong(uint32_t a)
1143 * \brief Swaps the bytes so they agree with the processor order
1144 * @return The properly swaped 16 bits integer.
1146 uint16_t Document::SwapShort(uint16_t a)
1148 if ( SwapCode == 4321 || SwapCode == 2143 )
1150 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1156 * \brief Unswaps the bytes so they agree with the processor order
1157 * @return The properly unswaped 16 bits integer.
1159 uint16_t Document::UnswapShort(uint16_t a)
1161 return SwapShort(a);
1164 //-----------------------------------------------------------------------------
1168 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1169 * @return length of the parsed set.
1171 void Document::ParseDES(DocEntrySet *set, long offset,
1172 long l_max, bool delim_mode)
1174 DocEntry *newDocEntry = 0;
1175 ValEntry *newValEntry;
1176 BinEntry *newBinEntry;
1177 SeqEntry *newSeqEntry;
1183 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1189 newDocEntry = ReadNextDocEntry( );
1195 vr = newDocEntry->GetVR();
1196 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
1197 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
1198 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1200 if ( newValEntry || newBinEntry )
1204 if ( ! Global::GetVR()->IsVROfBinaryRepresentable(vr) )
1206 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1207 gdcmVerboseMacro("Document::ParseDES: neither Valentry, "
1208 "nor BinEntry. Probably unknown VR.");
1211 //////////////////// BinEntry or UNKOWN VR:
1212 // When "this" is a Document the Key is simply of the
1213 // form ( group, elem )...
1214 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1217 newBinEntry->SetKey( newBinEntry->GetKey() );
1219 // but when "this" is a SQItem, we are inserting this new
1220 // valEntry in a sequence item, and the kay has the
1221 // generalized form (refer to \ref BaseTagKey):
1222 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1224 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1225 + newBinEntry->GetKey() );
1228 LoadDocEntry( newBinEntry );
1229 if( !set->AddEntry( newBinEntry ) )
1231 //Expect big troubles if here
1232 //delete newBinEntry;
1238 /////////////////////// ValEntry
1239 // When "set" is a Document, then we are at the top of the
1240 // hierarchy and the Key is simply of the form ( group, elem )...
1241 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1244 newValEntry->SetKey( newValEntry->GetKey() );
1246 // ...but when "set" is a SQItem, we are inserting this new
1247 // valEntry in a sequence item. Hence the key has the
1248 // generalized form (refer to \ref BaseTagKey):
1249 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1251 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1252 + newValEntry->GetKey() );
1255 LoadDocEntry( newValEntry );
1256 bool delimitor=newValEntry->IsItemDelimitor();
1257 if( !set->AddEntry( newValEntry ) )
1259 // If here expect big troubles
1260 //delete newValEntry; //otherwise mem leak
1270 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1278 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1279 && ( newDocEntry->GetElement() == 0x0010 ) )
1281 TransferSyntaxType ts = GetTransferSyntax();
1282 if ( ts == RLELossless )
1284 long positionOnEntry = Fp->tellg();
1285 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1287 Fp->seekg( positionOnEntry, std::ios::beg );
1289 else if ( IsJPEG() )
1291 long positionOnEntry = Fp->tellg();
1292 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1293 ComputeJPEGFragmentInfo();
1294 Fp->seekg( positionOnEntry, std::ios::beg );
1298 // Just to make sure we are at the beginning of next entry.
1299 SkipToNextDocEntry(newDocEntry);
1304 unsigned long l = newDocEntry->GetReadLength();
1305 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1307 if ( l == 0xffffffff )
1316 // no other way to create it ...
1317 newSeqEntry->SetDelimitorMode( delim_mode );
1319 // At the top of the hierarchy, stands a Document. When "set"
1320 // is a Document, then we are building the first depth level.
1321 // Hence the SeqEntry we are building simply has a depth
1323 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1326 newSeqEntry->SetDepthLevel( 1 );
1327 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1329 // But when "set" is allready a SQItem, we are building a nested
1330 // sequence, and hence the depth level of the new SeqEntry
1331 // we are building, is one level deeper:
1332 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1334 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1335 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1336 + newSeqEntry->GetKey() );
1340 { // Don't try to parse zero-length sequences
1341 ParseSQ( newSeqEntry,
1342 newDocEntry->GetOffset(),
1345 set->AddEntry( newSeqEntry );
1346 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1358 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1359 * @return parsed length for this level
1361 void Document::ParseSQ( SeqEntry *seqEntry,
1362 long offset, long l_max, bool delim_mode)
1364 int SQItemNumber = 0;
1369 DocEntry *newDocEntry = ReadNextDocEntry();
1372 // FIXME Should warn user
1377 if ( newDocEntry->IsSequenceDelimitor() )
1379 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1383 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1389 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1390 std::ostringstream newBase;
1391 newBase << seqEntry->GetKey()
1395 itemSQ->SetBaseTagKey( newBase.str() );
1396 unsigned int l = newDocEntry->GetReadLength();
1398 if ( l == 0xffffffff )
1407 ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1410 seqEntry->AddEntry( itemSQ, SQItemNumber );
1412 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1420 * \brief Loads the element content if its length doesn't exceed
1421 * the value specified with Document::SetMaxSizeLoadEntry()
1422 * @param entry Header Entry (Dicom Element) to be dealt with
1424 void Document::LoadDocEntry(DocEntry *entry)
1426 uint16_t group = entry->GetGroup();
1427 std::string vr = entry->GetVR();
1428 uint32_t length = entry->GetLength();
1430 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1432 // A SeQuence "contains" a set of Elements.
1433 // (fffe e000) tells us an Element is beginning
1434 // (fffe e00d) tells us an Element just ended
1435 // (fffe e0dd) tells us the current SeQuence just ended
1436 if( group == 0xfffe )
1438 // NO more value field for SQ !
1442 // When the length is zero things are easy:
1445 ((ValEntry *)entry)->SetValue("");
1449 // The elements whose length is bigger than the specified upper bound
1450 // are not loaded. Instead we leave a short notice of the offset of
1451 // the element content and it's length.
1453 std::ostringstream s;
1454 if (length > MaxSizeLoadEntry)
1456 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1458 //s << "gdcm::NotLoaded (BinEntry)";
1459 s << GDCM_NOTLOADED;
1460 s << " Address:" << (long)entry->GetOffset();
1461 s << " Length:" << entry->GetLength();
1462 s << " x(" << std::hex << entry->GetLength() << ")";
1463 binEntryPtr->SetValue(s.str());
1465 // Be carefull : a BinEntry IS_A ValEntry ...
1466 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1468 // s << "gdcm::NotLoaded. (ValEntry)";
1469 s << GDCM_NOTLOADED;
1470 s << " Address:" << (long)entry->GetOffset();
1471 s << " Length:" << entry->GetLength();
1472 s << " x(" << std::hex << entry->GetLength() << ")";
1473 valEntryPtr->SetValue(s.str());
1478 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1479 << "nor a ValEntry ?! Should never print that !" << std::endl;
1482 // to be sure we are at the end of the value ...
1483 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1488 // When we find a BinEntry not very much can be done :
1489 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1491 s << GDCM_BINLOADED;
1492 binEntryPtr->SetValue(s.str());
1493 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1497 /// \todo Any compacter code suggested (?)
1498 if ( IsDocEntryAnInteger(entry) )
1502 // When short integer(s) are expected, read and convert the following
1503 // n *two characters properly i.e. consider them as short integers as
1504 // opposed to strings.
1505 // Elements with Value Multiplicity > 1
1506 // contain a set of integers (not a single one)
1507 if (vr == "US" || vr == "SS")
1510 NewInt = ReadInt16();
1514 for (int i=1; i < nbInt; i++)
1517 NewInt = ReadInt16();
1522 // See above comment on multiple integers (mutatis mutandis).
1523 else if (vr == "UL" || vr == "SL")
1526 NewInt = ReadInt32();
1530 for (int i=1; i < nbInt; i++)
1533 NewInt = ReadInt32();
1538 #ifdef GDCM_NO_ANSI_STRING_STREAM
1539 s << std::ends; // to avoid oddities on Solaris
1540 #endif //GDCM_NO_ANSI_STRING_STREAM
1542 ((ValEntry *)entry)->SetValue(s.str());
1546 // FIXME: We need an additional byte for storing \0 that is not on disk
1547 char *str = new char[length+1];
1548 Fp->read(str, (size_t)length);
1549 str[length] = '\0'; //this is only useful when length is odd
1550 // Special DicomString call to properly handle \0 and even length
1551 std::string newValue;
1554 newValue = Util::DicomString(str, length+1);
1555 gdcmVerboseMacro("Warning: bad length: " << length );
1556 gdcmVerboseMacro("For string :" << newValue.c_str());
1557 // Since we change the length of string update it length
1558 //entry->SetReadLength(length+1);
1562 newValue = Util::DicomString(str, length);
1566 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1568 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1570 gdcmVerboseMacro("Document::LoadDocEntry"
1571 "unread element value");
1572 valEntry->SetValue(GDCM_UNREAD);
1578 // Because of correspondance with the VR dic
1579 valEntry->SetValue(newValue);
1583 valEntry->SetValue(newValue);
1588 gdcmErrorMacro("Document::LoadDocEntry"
1589 "Should have a ValEntry, here !");
1595 * \brief Find the value Length of the passed Header Entry
1596 * @param entry Header Entry whose length of the value shall be loaded.
1598 void Document::FindDocEntryLength( DocEntry *entry )
1599 throw ( FormatError )
1601 uint16_t element = entry->GetElement();
1602 std::string vr = entry->GetVR();
1605 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1607 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1609 // The following reserved two bytes (see PS 3.5-2003, section
1610 // "7.1.2 Data element structure with explicit vr", p 27) must be
1611 // skipped before proceeding on reading the length on 4 bytes.
1612 Fp->seekg( 2L, std::ios::cur);
1613 uint32_t length32 = ReadInt32();
1615 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1620 lengthOB = FindDocEntryLengthOBOrOW();
1622 catch ( FormatUnexpected )
1624 // Computing the length failed (this happens with broken
1625 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1626 // chance to get the pixels by deciding the element goes
1627 // until the end of the file. Hence we artificially fix the
1628 // the length and proceed.
1629 long currentPosition = Fp->tellg();
1630 Fp->seekg(0L,std::ios::end);
1632 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1633 Fp->seekg(currentPosition, std::ios::beg);
1635 entry->SetReadLength(lengthUntilEOF);
1636 entry->SetLength(lengthUntilEOF);
1639 entry->SetReadLength(lengthOB);
1640 entry->SetLength(lengthOB);
1643 FixDocEntryFoundLength(entry, length32);
1647 // Length is encoded on 2 bytes.
1648 length16 = ReadInt16();
1650 // We can tell the current file is encoded in big endian (like
1651 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1652 // and it's value is the one of the encoding of a big endian file.
1653 // In order to deal with such big endian encoded files, we have
1654 // (at least) two strategies:
1655 // * when we load the "Transfer Syntax" tag with value of big endian
1656 // encoding, we raise the proper flags. Then we wait for the end
1657 // of the META group (0x0002) among which is "Transfer Syntax",
1658 // before switching the swap code to big endian. We have to postpone
1659 // the switching of the swap code since the META group is fully encoded
1660 // in little endian, and big endian coding only starts at the next
1661 // group. The corresponding code can be hard to analyse and adds
1662 // many additional unnecessary tests for regular tags.
1663 // * the second strategy consists in waiting for trouble, that shall
1664 // appear when we find the first group with big endian encoding. This
1665 // is easy to detect since the length of a "Group Length" tag (the
1666 // ones with zero as element number) has to be of 4 (0x0004). When we
1667 // encounter 1024 (0x0400) chances are the encoding changed and we
1668 // found a group with big endian encoding.
1669 // We shall use this second strategy. In order to make sure that we
1670 // can interpret the presence of an apparently big endian encoded
1671 // length of a "Group Length" without committing a big mistake, we
1672 // add an additional check: we look in the already parsed elements
1673 // for the presence of a "Transfer Syntax" whose value has to be "big
1674 // endian encoding". When this is the case, chances are we have got our
1675 // hands on a big endian encoded file: we switch the swap code to
1676 // big endian and proceed...
1677 if ( element == 0x0000 && length16 == 0x0400 )
1679 TransferSyntaxType ts = GetTransferSyntax();
1680 if ( ts != ExplicitVRBigEndian )
1682 throw FormatError( "Document::FindDocEntryLength()",
1683 " not explicit VR." );
1687 SwitchSwapToBigEndian();
1689 // Restore the unproperly loaded values i.e. the group, the element
1690 // and the dictionary entry depending on them.
1691 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1692 uint16_t correctElem = SwapShort( entry->GetElement() );
1693 DictEntry *newTag = GetDictEntryByNumber( correctGroup,
1697 // This correct tag is not in the dictionary. Create a new one.
1698 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1700 // FIXME this can create a memory leaks on the old entry that be
1701 // left unreferenced.
1702 entry->SetDictEntry( newTag );
1705 // Heuristic: well, some files are really ill-formed.
1706 if ( length16 == 0xffff)
1708 // 0xffff means that we deal with 'Unknown Length' Sequence
1711 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1716 // Either implicit VR or a non DICOM conformal (see note below) explicit
1717 // VR that ommited the VR of (at least) this element. Farts happen.
1718 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1719 // on Data elements "Implicit and Explicit VR Data Elements shall
1720 // not coexist in a Data Set and Data Sets nested within it".]
1721 // Length is on 4 bytes.
1723 FixDocEntryFoundLength( entry, ReadInt32() );
1729 * \brief Find the Value Representation of the current Dicom Element.
1730 * @return Value Representation of the current Entry
1732 std::string Document::FindDocEntryVR()
1734 if ( Filetype != ExplicitVR )
1735 return(GDCM_UNKNOWN);
1737 long positionOnEntry = Fp->tellg();
1738 // Warning: we believe this is explicit VR (Value Representation) because
1739 // we used a heuristic that found "UL" in the first tag. Alas this
1740 // doesn't guarantee that all the tags will be in explicit VR. In some
1741 // cases (see e-film filtered files) one finds implicit VR tags mixed
1742 // within an explicit VR file. Hence we make sure the present tag
1743 // is in explicit VR and try to fix things if it happens not to be
1747 Fp->read (vr, (size_t)2);
1750 if( !CheckDocEntryVR(vr) )
1752 Fp->seekg(positionOnEntry, std::ios::beg);
1753 return(GDCM_UNKNOWN);
1759 * \brief Check the correspondance between the VR of the header entry
1760 * and the taken VR. If they are different, the header entry is
1761 * updated with the new VR.
1762 * @param vr Dicom Value Representation
1763 * @return false if the VR is incorrect of if the VR isn't referenced
1764 * otherwise, it returns true
1766 bool Document::CheckDocEntryVR(VRKey vr)
1768 // CLEANME searching the dicom_vr at each occurence is expensive.
1769 // PostPone this test in an optional integrity check at the end
1770 // of parsing or only in debug mode.
1771 if ( !Global::GetVR()->IsValidVR(vr) )
1778 * \brief Get the transformed value of the header entry. The VR value
1779 * is used to define the transformation to operate on the value
1780 * \warning NOT end user intended method !
1781 * @param entry entry to tranform
1782 * @return Transformed entry value
1784 std::string Document::GetDocEntryValue(DocEntry *entry)
1786 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1788 std::string val = ((ValEntry *)entry)->GetValue();
1789 std::string vr = entry->GetVR();
1790 uint32_t length = entry->GetLength();
1791 std::ostringstream s;
1794 // When short integer(s) are expected, read and convert the following
1795 // n * 2 bytes properly i.e. as a multivaluated strings
1796 // (each single value is separated fromthe next one by '\'
1797 // as usual for standard multivaluated filels
1798 // Elements with Value Multiplicity > 1
1799 // contain a set of short integers (not a single one)
1801 if( vr == "US" || vr == "SS" )
1806 for (int i=0; i < nbInt; i++)
1812 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1813 newInt16 = SwapShort( newInt16 );
1818 // When integer(s) are expected, read and convert the following
1819 // n * 4 bytes properly i.e. as a multivaluated strings
1820 // (each single value is separated fromthe next one by '\'
1821 // as usual for standard multivaluated filels
1822 // Elements with Value Multiplicity > 1
1823 // contain a set of integers (not a single one)
1824 else if( vr == "UL" || vr == "SL" )
1829 for (int i=0; i < nbInt; i++)
1835 newInt32 = ( val[4*i+0] & 0xFF )
1836 + (( val[4*i+1] & 0xFF ) << 8 )
1837 + (( val[4*i+2] & 0xFF ) << 16 )
1838 + (( val[4*i+3] & 0xFF ) << 24 );
1839 newInt32 = SwapLong( newInt32 );
1843 #ifdef GDCM_NO_ANSI_STRING_STREAM
1844 s << std::ends; // to avoid oddities on Solaris
1845 #endif //GDCM_NO_ANSI_STRING_STREAM
1849 return ((ValEntry *)entry)->GetValue();
1853 * \brief Get the reverse transformed value of the header entry. The VR
1854 * value is used to define the reverse transformation to operate on
1856 * \warning NOT end user intended method !
1857 * @param entry Entry to reverse transform
1858 * @return Reverse transformed entry value
1860 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1862 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1864 std::string vr = entry->GetVR();
1865 std::vector<std::string> tokens;
1866 std::ostringstream s;
1868 if ( vr == "US" || vr == "SS" )
1872 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1873 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1874 for (unsigned int i=0; i<tokens.size(); i++)
1876 newInt16 = atoi(tokens[i].c_str());
1877 s << ( newInt16 & 0xFF )
1878 << (( newInt16 >> 8 ) & 0xFF );
1882 if ( vr == "UL" || vr == "SL")
1886 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1887 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1888 for (unsigned int i=0; i<tokens.size();i++)
1890 newInt32 = atoi(tokens[i].c_str());
1891 s << (char)( newInt32 & 0xFF )
1892 << (char)(( newInt32 >> 8 ) & 0xFF )
1893 << (char)(( newInt32 >> 16 ) & 0xFF )
1894 << (char)(( newInt32 >> 24 ) & 0xFF );
1899 #ifdef GDCM_NO_ANSI_STRING_STREAM
1900 s << std::ends; // to avoid oddities on Solaris
1901 #endif //GDCM_NO_ANSI_STRING_STREAM
1905 return ((ValEntry *)entry)->GetValue();
1909 * \brief Skip a given Header Entry
1910 * \warning NOT end user intended method !
1911 * @param entry entry to skip
1913 void Document::SkipDocEntry(DocEntry *entry)
1915 SkipBytes(entry->GetLength());
1919 * \brief Skips to the begining of the next Header Entry
1920 * \warning NOT end user intended method !
1921 * @param offset start of skipping
1922 * @param readLgth length to skip
1925 void Document::SkipToNextDocEntry(DocEntry *newDocEntry)
1927 Fp->seekg((long)(newDocEntry->GetOffset()), std::ios::beg);
1928 Fp->seekg( (long)(newDocEntry->GetReadLength()),std::ios::cur);
1932 * \brief When the length of an element value is obviously wrong (because
1933 * the parser went Jabberwocky) one can hope improving things by
1934 * applying some heuristics.
1935 * @param entry entry to check
1936 * @param foundLength fist assumption about length
1938 void Document::FixDocEntryFoundLength(DocEntry *entry,
1939 uint32_t foundLength)
1941 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1942 if ( foundLength == 0xffffffff)
1947 uint16_t gr = entry->GetGroup();
1948 uint16_t elem = entry->GetElement();
1950 if ( foundLength % 2)
1952 std::ostringstream s;
1953 s << "Warning : Tag with uneven length "
1955 << " in x(" << std::hex << gr << "," << elem <<")" << std::dec;
1956 gdcmVerboseMacro(s.str().c_str());
1959 //////// Fix for some naughty General Electric images.
1960 // Allthough not recent many such GE corrupted images are still present
1961 // on Creatis hard disks. Hence this fix shall remain when such images
1962 // are no longer in use (we are talking a few years, here)...
1963 // Note: XMedCom probably uses such a trick since it is able to read
1964 // those pesky GE images ...
1965 if ( foundLength == 13)
1967 // Only happens for this length !
1968 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1971 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
1975 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1976 // Occurence of such images is quite low (unless one leaves close to a
1977 // 'Leonardo' source. Hence, one might consider commenting out the
1978 // following fix on efficiency reasons.
1979 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1982 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
1985 else if ( entry->GetVR() == "SQ" )
1987 foundLength = 0; // ReadLength is unchanged
1990 //////// We encountered a 'delimiter' element i.e. a tag of the form
1991 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1992 // taken into account.
1993 else if( gr == 0xfffe )
1995 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1996 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1997 // causes extra troubles...
1998 if( entry->GetElement() != 0x0000 )
2004 entry->SetLength(foundLength);
2008 * \brief Apply some heuristics to predict whether the considered
2009 * element value contains/represents an integer or not.
2010 * @param entry The element value on which to apply the predicate.
2011 * @return The result of the heuristical predicate.
2013 bool Document::IsDocEntryAnInteger(DocEntry *entry)
2015 uint16_t element = entry->GetElement();
2016 uint16_t group = entry->GetGroup();
2017 const std::string &vr = entry->GetVR();
2018 uint32_t length = entry->GetLength();
2020 // When we have some semantics on the element we just read, and if we
2021 // a priori know we are dealing with an integer, then we shall be
2022 // able to swap it's element value properly.
2023 if ( element == 0 ) // This is the group length of the group
2031 // Allthough this should never happen, still some images have a
2032 // corrupted group length [e.g. have a glance at offset x(8336) of
2033 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2034 // Since for dicom compliant and well behaved headers, the present
2035 // test is useless (and might even look a bit paranoid), when we
2036 // encounter such an ill-formed image, we simply display a warning
2037 // message and proceed on parsing (while crossing fingers).
2038 std::ostringstream s;
2039 long filePosition = Fp->tellg();
2040 s << "Erroneous Group Length element length on : (" \
2041 << std::hex << group << " , " << element
2042 << ") -before- position x(" << filePosition << ")"
2043 << "lgt : " << length;
2044 gdcmVerboseMacro("Document::IsDocEntryAnInteger" << s.str().c_str() );
2048 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2056 * \brief Find the Length till the next sequence delimiter
2057 * \warning NOT end user intended method !
2061 uint32_t Document::FindDocEntryLengthOBOrOW()
2062 throw( FormatUnexpected )
2064 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2065 long positionOnEntry = Fp->tellg();
2066 bool foundSequenceDelimiter = false;
2067 uint32_t totalLength = 0;
2069 while ( !foundSequenceDelimiter )
2075 group = ReadInt16();
2078 catch ( FormatError )
2080 throw FormatError("Document::FindDocEntryLengthOBOrOW()",
2081 " group or element not present.");
2084 // We have to decount the group and element we just read
2087 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2089 gdcmVerboseMacro("Document::FindDocEntryLengthOBOrOW: neither an Item "
2090 "tag nor a Sequence delimiter tag.");
2091 Fp->seekg(positionOnEntry, std::ios::beg);
2092 throw FormatUnexpected("Document::FindDocEntryLengthOBOrOW()",
2093 "Neither an Item tag nor a Sequence "
2097 if ( elem == 0xe0dd )
2099 foundSequenceDelimiter = true;
2102 uint32_t itemLength = ReadInt32();
2103 // We add 4 bytes since we just read the ItemLength with ReadInt32
2104 totalLength += itemLength + 4;
2105 SkipBytes(itemLength);
2107 if ( foundSequenceDelimiter )
2112 Fp->seekg( positionOnEntry, std::ios::beg);
2117 * \brief Reads a supposed to be 16 Bits integer
2118 * (swaps it depending on processor endianity)
2119 * @return read value
2121 uint16_t Document::ReadInt16()
2122 throw( FormatError )
2125 Fp->read ((char*)&g, (size_t)2);
2128 throw FormatError( "Document::ReadInt16()", " file error." );
2132 throw FormatError( "Document::ReadInt16()", "EOF." );
2139 * \brief Reads a supposed to be 32 Bits integer
2140 * (swaps it depending on processor endianity)
2141 * @return read value
2143 uint32_t Document::ReadInt32()
2144 throw( FormatError )
2147 Fp->read ((char*)&g, (size_t)4);
2150 throw FormatError( "Document::ReadInt32()", " file error." );
2154 throw FormatError( "Document::ReadInt32()", "EOF." );
2161 * \brief skips bytes inside the source file
2162 * \warning NOT end user intended method !
2165 void Document::SkipBytes(uint32_t nBytes)
2167 //FIXME don't dump the returned value
2168 Fp->seekg((long)nBytes, std::ios::cur);
2172 * \brief Loads all the needed Dictionaries
2173 * \warning NOT end user intended method !
2175 void Document::Initialise()
2177 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2179 RLEInfo = new RLEFramesInfo;
2180 JPEGInfo = new JPEGFragmentsInfo;
2185 * \brief Discover what the swap code is (among little endian, big endian,
2186 * bad little endian, bad big endian).
2188 * @return false when we are absolutely sure
2189 * it's neither ACR-NEMA nor DICOM
2190 * true when we hope ours assuptions are OK
2192 bool Document::CheckSwap()
2194 // The only guaranted way of finding the swap code is to find a
2195 // group tag since we know it's length has to be of four bytes i.e.
2196 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2197 // occurs when we can't find such group...
2199 uint32_t x = 4; // x : for ntohs
2200 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2206 // First, compare HostByteOrder and NetworkByteOrder in order to
2207 // determine if we shall need to swap bytes (i.e. the Endian type).
2208 if ( x == ntohs(x) )
2217 // The easiest case is the one of a DICOM header, since it possesses a
2218 // file preamble where it suffice to look for the string "DICM".
2221 char *entCur = deb + 128;
2222 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2224 gdcmVerboseMacro("Document::CheckSwap:" "looks like DICOM Version3");
2226 // Next, determine the value representation (VR). Let's skip to the
2227 // first element (0002, 0000) and check there if we find "UL"
2228 // - or "OB" if the 1st one is (0002,0001) -,
2229 // in which case we (almost) know it is explicit VR.
2230 // WARNING: if it happens to be implicit VR then what we will read
2231 // is the length of the group. If this ascii representation of this
2232 // length happens to be "UL" then we shall believe it is explicit VR.
2233 // FIXME: in order to fix the above warning, we could read the next
2234 // element value (or a couple of elements values) in order to make
2235 // sure we are not commiting a big mistake.
2236 // We need to skip :
2237 // * the 128 bytes of File Preamble (often padded with zeroes),
2238 // * the 4 bytes of "DICM" string,
2239 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2240 // i.e. a total of 136 bytes.
2244 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2245 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2246 // *Implicit* VR. -and it is !-
2248 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2249 memcmp(entCur, "OB", (size_t)2) == 0 ||
2250 memcmp(entCur, "UI", (size_t)2) == 0 ||
2251 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2252 // when Write DCM *adds*
2254 // Use Document::dicom_vr to test all the possibilities
2255 // instead of just checking for UL, OB and UI !? group 0000
2257 Filetype = ExplicitVR;
2258 gdcmVerboseMacro( "Document::CheckSwap:"
2259 "explicit Value Representation");
2263 Filetype = ImplicitVR;
2264 gdcmVerboseMacro("Document::CheckSwap:"
2265 "not an explicit Value Representation");
2271 gdcmVerboseMacro("Document::CheckSwap:"
2272 "HostByteOrder != NetworkByteOrder");
2277 gdcmVerboseMacro("Document::CheckSwap:"
2278 "HostByteOrder = NetworkByteOrder");
2281 // Position the file position indicator at first tag (i.e.
2282 // after the file preamble and the "DICM" string).
2283 Fp->seekg(0, std::ios::beg);
2284 Fp->seekg ( 132L, std::ios::beg);
2288 // Alas, this is not a DicomV3 file and whatever happens there is no file
2289 // preamble. We can reset the file position indicator to where the data
2290 // is (i.e. the beginning of the file).
2291 gdcmVerboseMacro("Document::CheckSwap:" "not a DICOM Version3 file");
2292 Fp->seekg(0, std::ios::beg);
2294 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2295 // By clean we mean that the length of the first tag is written down.
2296 // If this is the case and since the length of the first group HAS to be
2297 // four (bytes), then determining the proper swap code is straightforward.
2300 // We assume the array of char we are considering contains the binary
2301 // representation of a 32 bits integer. Hence the following dirty
2303 s32 = *((uint32_t *)(entCur));
2324 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2325 // It is time for despaired wild guesses.
2326 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2327 // i.e. the 'group length' element is not present :
2329 // check the supposed-to-be 'group number'
2330 // in ( 0x0001 .. 0x0008 )
2331 // to determine ' SwapCode' value .
2332 // Only 0 or 4321 will be possible
2333 // (no oportunity to check for the formerly well known
2334 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2335 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2336 // the file IS NOT ACR-NEMA nor DICOM V3
2337 // Find a trick to tell it the caller...
2339 s16 = *((uint16_t *)(deb));
2366 gdcmVerboseMacro( "Document::CheckSwap:"
2367 "ACR/NEMA unfound swap info (Really hopeless !)");
2371 // Then the only info we have is the net2host one.
2383 * \brief Restore the unproperly loaded values i.e. the group, the element
2384 * and the dictionary entry depending on them.
2386 void Document::SwitchSwapToBigEndian()
2388 gdcmVerboseMacro("Document::SwitchSwapToBigEndian"
2389 "Switching to BigEndian mode.");
2390 if ( SwapCode == 0 )
2394 else if ( SwapCode == 4321 )
2398 else if ( SwapCode == 3412 )
2402 else if ( SwapCode == 2143 )
2409 * \brief during parsing, Header Elements too long are not loaded in memory
2412 void Document::SetMaxSizeLoadEntry(long newSize)
2418 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2420 MaxSizeLoadEntry = 0xffffffff;
2423 MaxSizeLoadEntry = newSize;
2428 * \brief Header Elements too long will not be printed
2429 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2432 void Document::SetMaxSizePrintEntry(long newSize)
2434 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2439 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2441 MaxSizePrintEntry = 0xffffffff;
2444 MaxSizePrintEntry = newSize;
2450 * \brief Handle broken private tag from Philips NTSCAN
2451 * where the endianess is being switch to BigEndian for no
2455 void Document::HandleBrokenEndian(uint16_t group, uint16_t elem)
2457 // Endian reversion. Some files contain groups of tags with reversed endianess.
2458 static int reversedEndian = 0;
2459 // try to fix endian switching in the middle of headers
2460 if ((group == 0xfeff) && (elem == 0x00e0))
2462 // start endian swap mark for group found
2464 SwitchSwapToBigEndian();
2469 else if ((group == 0xfffe) && (elem == 0xe00d) && reversedEndian)
2471 // end of reversed endian group
2473 SwitchSwapToBigEndian();
2478 * \brief Read the next tag but WITHOUT loading it's value
2479 * (read the 'Group Number', the 'Element Number',
2480 * gets the Dict Entry
2481 * gets the VR, gets the length, gets the offset value)
2482 * @return On succes the newly created DocEntry, NULL on failure.
2484 DocEntry *Document::ReadNextDocEntry()
2491 group = ReadInt16();
2494 catch ( FormatError e )
2496 // We reached the EOF (or an error occured) therefore
2497 // header parsing has to be considered as finished.
2502 HandleBrokenEndian(group, elem);
2503 std::string vr = FindDocEntryVR();
2504 std::string realVR = vr;
2506 if( vr == GDCM_UNKNOWN)
2508 DictEntry *dictEntry = GetDictEntryByNumber(group,elem);
2510 realVR = dictEntry->GetVR();
2514 if( Global::GetVR()->IsVROfSequence(realVR) )
2515 newEntry = NewSeqEntryByNumber(group, elem);
2516 else if( Global::GetVR()->IsVROfStringRepresentable(realVR) )
2517 newEntry = NewValEntryByNumber(group, elem,vr);
2519 newEntry = NewBinEntryByNumber(group, elem,vr);
2521 if( vr == GDCM_UNKNOWN )
2523 if( Filetype == ExplicitVR )
2525 // We thought this was explicit VR, but we end up with an
2526 // implicit VR tag. Let's backtrack.
2528 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
2529 newEntry->GetGroup(), newEntry->GetElement());
2530 gdcmVerboseMacro("Document::FindVR: " << msg.c_str());
2532 newEntry->SetImplicitVR();
2537 FindDocEntryLength(newEntry);
2539 catch ( FormatError e )
2547 newEntry->SetOffset(Fp->tellg());
2554 * \brief Generate a free TagKey i.e. a TagKey that is not present
2555 * in the TagHt dictionary.
2556 * @param group The generated tag must belong to this group.
2557 * @return The element of tag with given group which is fee.
2559 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2561 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2563 TagKey key = DictEntry::TranslateToKey(group, elem);
2564 if (TagHT.count(key) == 0)
2573 * \brief Assuming the internal file pointer \ref Document::Fp
2574 * is placed at the beginning of a tag check whether this
2575 * tag is (TestGroup, TestElement).
2576 * \warning On success the internal file pointer \ref Document::Fp
2577 * is modified to point after the tag.
2578 * On failure (i.e. when the tag wasn't the expected tag
2579 * (TestGroup, TestElement) the internal file pointer
2580 * \ref Document::Fp is restored to it's original position.
2581 * @param testGroup The expected group of the tag.
2582 * @param testElement The expected Element of the tag.
2583 * @return True on success, false otherwise.
2585 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2587 long positionOnEntry = Fp->tellg();
2588 long currentPosition = Fp->tellg(); // On debugging purposes
2590 //// Read the Item Tag group and element, and make
2591 // sure they are what we expected:
2592 uint16_t itemTagGroup;
2593 uint16_t itemTagElement;
2596 itemTagGroup = ReadInt16();
2597 itemTagElement = ReadInt16();
2599 catch ( FormatError e )
2601 //std::cerr << e << std::endl;
2604 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2606 std::ostringstream s;
2607 s << " We should have found tag (";
2608 s << std::hex << testGroup << "," << testElement << ")" << std::endl;
2609 s << " but instead we encountered tag (";
2610 s << std::hex << itemTagGroup << "," << itemTagElement << ")"
2612 s << " at address: " << (unsigned)currentPosition << std::endl;
2613 gdcmVerboseMacro("Document::ReadItemTagLength: wrong Item Tag found:"
2614 << s.str().c_str());
2615 Fp->seekg(positionOnEntry, std::ios::beg);
2623 * \brief Assuming the internal file pointer \ref Document::Fp
2624 * is placed at the beginning of a tag (TestGroup, TestElement),
2625 * read the length associated to the Tag.
2626 * \warning On success the internal file pointer \ref Document::Fp
2627 * is modified to point after the tag and it's length.
2628 * On failure (i.e. when the tag wasn't the expected tag
2629 * (TestGroup, TestElement) the internal file pointer
2630 * \ref Document::Fp is restored to it's original position.
2631 * @param testGroup The expected group of the tag.
2632 * @param testElement The expected Element of the tag.
2633 * @return On success returns the length associated to the tag. On failure
2636 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2638 long positionOnEntry = Fp->tellg();
2639 (void)positionOnEntry;
2641 if ( !ReadTag(testGroup, testElement) )
2646 //// Then read the associated Item Length
2647 long currentPosition = Fp->tellg();
2648 uint32_t itemLength = ReadInt32();
2650 std::ostringstream s;
2651 s << "Basic Item Length is: "
2652 << itemLength << std::endl;
2653 s << " at address: " << (unsigned)currentPosition << std::endl;
2654 gdcmVerboseMacro("Document::ReadItemTagLength: " << s.str().c_str());
2660 * \brief When parsing the Pixel Data of an encapsulated file, read
2661 * the basic offset table (when present, and BTW dump it).
2663 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2665 //// Read the Basic Offset Table Item Tag length...
2666 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2668 // When present, read the basic offset table itself.
2669 // Notes: - since the presence of this basic offset table is optional
2670 // we can't rely on it for the implementation, and we will simply
2671 // trash it's content (when present).
2672 // - still, when present, we could add some further checks on the
2673 // lengths, but we won't bother with such fuses for the time being.
2674 if ( itemLength != 0 )
2676 char *basicOffsetTableItemValue = new char[itemLength + 1];
2677 Fp->read(basicOffsetTableItemValue, itemLength);
2680 for (unsigned int i=0; i < itemLength; i += 4 )
2682 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2684 std::ostringstream s;
2685 s << " Read one length: ";
2686 s << std::hex << individualLength << std::endl;
2688 "Document::ReadAndSkipEncapsulatedBasicOffsetTable: ",
2693 delete[] basicOffsetTableItemValue;
2698 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2699 * Compute the RLE extra information and store it in \ref RLEInfo
2700 * for later pixel retrieval usage.
2702 void Document::ComputeRLEInfo()
2704 TransferSyntaxType ts = GetTransferSyntax();
2705 if ( ts != RLELossless )
2710 // Encoded pixel data: for the time being we are only concerned with
2711 // Jpeg or RLE Pixel data encodings.
2712 // As stated in PS 3.5-2003, section 8.2 p44:
2713 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2714 // value representation OB is used".
2715 // Hence we expect an OB value representation. Concerning OB VR,
2716 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2717 // "For the Value Representations OB and OW, the encoding shall meet the
2718 // following specifications depending on the Data element tag:"
2720 // - the first item in the sequence of items before the encoded pixel
2721 // data stream shall be basic offset table item. The basic offset table
2722 // item value, however, is not required to be present"
2724 ReadAndSkipEncapsulatedBasicOffsetTable();
2726 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2727 // Loop on the individual frame[s] and store the information
2728 // on the RLE fragments in a RLEFramesInfo.
2729 // Note: - when only a single frame is present, this is a
2731 // - when more than one frame are present, then we are in
2732 // the case of a multi-frame image.
2734 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2736 // Parse the RLE Header and store the corresponding RLE Segment
2737 // Offset Table information on fragments of this current Frame.
2738 // Note that the fragment pixels themselves are not loaded
2739 // (but just skipped).
2740 long frameOffset = Fp->tellg();
2742 uint32_t nbRleSegments = ReadInt32();
2743 if ( nbRleSegments > 16 )
2745 // There should be at most 15 segments (refer to RLEFrame class)
2746 gdcmVerboseMacro("Document::ComputeRLEInfo: too many segments.");
2749 uint32_t rleSegmentOffsetTable[16];
2750 for( int k = 1; k <= 15; k++ )
2752 rleSegmentOffsetTable[k] = ReadInt32();
2755 // Deduce from both the RLE Header and the frameLength the
2756 // fragment length, and again store this info in a
2758 long rleSegmentLength[15];
2759 // skipping (not reading) RLE Segments
2760 if ( nbRleSegments > 1)
2762 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2764 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2765 - rleSegmentOffsetTable[k];
2766 SkipBytes(rleSegmentLength[k]);
2770 rleSegmentLength[nbRleSegments] = frameLength
2771 - rleSegmentOffsetTable[nbRleSegments];
2772 SkipBytes(rleSegmentLength[nbRleSegments]);
2774 // Store the collected info
2775 RLEFrame *newFrameInfo = new RLEFrame;
2776 newFrameInfo->NumberFragments = nbRleSegments;
2777 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2779 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2780 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2782 RLEInfo->Frames.push_back( newFrameInfo );
2785 // Make sure that at the end of the item we encounter a 'Sequence
2787 if ( !ReadTag(0xfffe, 0xe0dd) )
2789 gdcmVerboseMacro("Document::ComputeRLEInfo: no sequence delimiter "
2790 " item at end of RLE item sequence");
2795 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2796 * Compute the jpeg extra information (fragment[s] offset[s] and
2797 * length) and store it[them] in \ref JPEGInfo for later pixel
2800 void Document::ComputeJPEGFragmentInfo()
2802 // If you need to, look for comments of ComputeRLEInfo().
2808 ReadAndSkipEncapsulatedBasicOffsetTable();
2810 // Loop on the fragments[s] and store the parsed information in a
2812 long fragmentLength;
2813 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2815 long fragmentOffset = Fp->tellg();
2817 // Store the collected info
2818 JPEGFragment *newFragment = new JPEGFragment;
2819 newFragment->Offset = fragmentOffset;
2820 newFragment->Length = fragmentLength;
2821 JPEGInfo->Fragments.push_back( newFragment );
2823 SkipBytes( fragmentLength );
2826 // Make sure that at the end of the item we encounter a 'Sequence
2828 if ( !ReadTag(0xfffe, 0xe0dd) )
2830 gdcmVerboseMacro("Document::ComputeRLEInfo: no sequence delimiter "
2831 " item at end of JPEG item sequence");
2836 * \brief Walk recursively the given \ref DocEntrySet, and feed
2837 * the given hash table (\ref TagDocEntryHT) with all the
2838 * \ref DocEntry (Dicom entries) encountered.
2839 * This method does the job for \ref BuildFlatHashTable.
2840 * @param builtHT Where to collect all the \ref DocEntry encountered
2841 * when recursively walking the given set.
2842 * @param set The structure to be traversed (recursively).
2844 void Document::BuildFlatHashTableRecurse( TagDocEntryHT &builtHT,
2847 if (ElementSet *elementSet = dynamic_cast< ElementSet* > ( set ) )
2849 TagDocEntryHT const ¤tHT = elementSet->GetTagHT();
2850 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2851 i != currentHT.end();
2854 DocEntry *entry = i->second;
2855 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2857 const ListSQItem& items = seqEntry->GetSQItems();
2858 for( ListSQItem::const_iterator item = items.begin();
2859 item != items.end();
2862 BuildFlatHashTableRecurse( builtHT, *item );
2866 builtHT[entry->GetKey()] = entry;
2871 if (SQItem *SQItemSet = dynamic_cast< SQItem* > ( set ) )
2873 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
2874 for (ListDocEntry::const_iterator i = currentList.begin();
2875 i != currentList.end();
2878 DocEntry *entry = *i;
2879 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2881 const ListSQItem& items = seqEntry->GetSQItems();
2882 for( ListSQItem::const_iterator item = items.begin();
2883 item != items.end();
2886 BuildFlatHashTableRecurse( builtHT, *item );
2890 builtHT[entry->GetKey()] = entry;
2897 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2900 * The structure used by a Document (through \ref ElementSet),
2901 * in order to hold the parsed entries of a Dicom header, is a recursive
2902 * one. This is due to the fact that the sequences (when present)
2903 * can be nested. Additionaly, the sequence items (represented in
2904 * gdcm as \ref SQItem) add an extra complexity to the data
2905 * structure. Hence, a gdcm user whishing to visit all the entries of
2906 * a Dicom header will need to dig in the gdcm internals (which
2907 * implies exposing all the internal data structures to the API).
2908 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2909 * recursively builds a temporary hash table, which holds all the
2910 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
2912 * \warning Of course there is NO integrity constrain between the
2913 * returned \ref TagDocEntryHT and the \ref ElementSet used
2914 * to build it. Hence if the underlying \ref ElementSet is
2915 * altered, then it is the caller responsability to invoke
2916 * \ref BuildFlatHashTable again...
2917 * @return The flat std::map<> we juste build.
2919 TagDocEntryHT *Document::BuildFlatHashTable()
2921 TagDocEntryHT *FlatHT = new TagDocEntryHT;
2922 BuildFlatHashTableRecurse( *FlatHT, this );
2929 * \brief Compares two documents, according to \ref DicomDir rules
2930 * \warning Does NOT work with ACR-NEMA files
2931 * \todo Find a trick to solve the pb (use RET fields ?)
2933 * @return true if 'smaller'
2935 bool Document::operator<(Document &document)
2938 std::string s1 = GetEntryByNumber(0x0010,0x0010);
2939 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
2951 s1 = GetEntryByNumber(0x0010,0x0020);
2952 s2 = document.GetEntryByNumber(0x0010,0x0020);
2963 // Study Instance UID
2964 s1 = GetEntryByNumber(0x0020,0x000d);
2965 s2 = document.GetEntryByNumber(0x0020,0x000d);
2976 // Serie Instance UID
2977 s1 = GetEntryByNumber(0x0020,0x000e);
2978 s2 = document.GetEntryByNumber(0x0020,0x000e);
2995 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
2996 * @param filetype Type of the File to be written
2998 int Document::ComputeGroup0002Length( FileType filetype )
3003 int groupLength = 0;
3004 bool found0002 = false;
3006 // for each zero-level Tag in the DCM Header
3010 entry = GetNextEntry();
3013 gr = entry->GetGroup();
3019 el = entry->GetElement();
3020 vr = entry->GetVR();
3022 if (filetype == ExplicitVR)
3024 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
3026 groupLength += 4; // explicit VR AND OB, OW, SQ : 4 more bytes
3029 groupLength += 2 + 2 + 4 + entry->GetLength();
3031 else if (found0002 )
3034 entry = GetNextEntry();
3039 } // end namespace gdcm
3041 //-----------------------------------------------------------------------------