1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/06 13:35:38 $
7 Version: $Revision: 1.155 $
9 Copyright (c) CREATIS (Centre de Recherche et d'Applications en Traitement de
10 l'Image). All rights reserved. See Doc/License.txt or
11 http://www.creatis.insa-lyon.fr/Public/Gdcm/License.html for details.
13 This software is distributed WITHOUT ANY WARRANTY; without even
14 the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
15 PURPOSE. See the above copyright notices for more information.
17 =========================================================================*/
19 #include "gdcmDocument.h"
20 #include "gdcmValEntry.h"
21 #include "gdcmBinEntry.h"
22 #include "gdcmSeqEntry.h"
23 #include "gdcmGlobal.h"
25 #include "gdcmDebug.h"
27 #include "gdcmException.h"
28 #include "gdcmDictSet.h"
29 #include "gdcmRLEFramesInfo.h"
30 #include "gdcmJPEGFragmentsInfo.h"
31 #include "gdcmDocEntrySet.h"
32 #include "gdcmSQItem.h"
38 #if defined(_MSC_VER) || defined(__BORLANDC__)
41 #include <netinet/in.h>
46 //-----------------------------------------------------------------------------
47 static const char *TransferSyntaxStrings[] = {
48 // Implicit VR Little Endian
50 // Implicit VR Little Endian DLX G.E?
52 // Explicit VR Little Endian
53 "1.2.840.10008.1.2.1",
54 // Deflated Explicit VR Little Endian
55 "1.2.840.10008.1.2.1.99",
56 // Explicit VR Big Endian
57 "1.2.840.10008.1.2.2",
58 // JPEG Baseline (Process 1)
59 "1.2.840.10008.1.2.4.50",
60 // JPEG Extended (Process 2 & 4)
61 "1.2.840.10008.1.2.4.51",
62 // JPEG Extended (Process 3 & 5)
63 "1.2.840.10008.1.2.4.52",
64 // JPEG Spectral Selection, Non-Hierarchical (Process 6 & 8)
65 "1.2.840.10008.1.2.4.53",
66 // JPEG Full Progression, Non-Hierarchical (Process 10 & 12)
67 "1.2.840.10008.1.2.4.55",
68 // JPEG Lossless, Non-Hierarchical (Process 14)
69 "1.2.840.10008.1.2.4.57",
70 // JPEG Lossless, Hierarchical, First-Order Prediction (Process 14, [Selection Value 1])
71 "1.2.840.10008.1.2.4.70",
73 "1.2.840.10008.1.2.4.90",
75 "1.2.840.10008.1.2.4.91",
77 "1.2.840.10008.1.2.5",
79 "Unknown Transfer Syntax"
82 //-----------------------------------------------------------------------------
83 // Refer to Document::CheckSwap()
84 //const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
86 // Refer to Document::SetMaxSizeLoadEntry()
87 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
88 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
90 //-----------------------------------------------------------------------------
91 // Constructor / Destructor
95 * @param filename file to be opened for parsing
97 Document::Document( std::string const & filename ) : ElementSet(-1)
99 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
109 dbg.Verbose(0, "Document::Document: starting parsing of file: ",
111 Fp->seekg( 0, std::ios::beg);
113 Fp->seekg(0, std::ios::end);
114 long lgt = Fp->tellg();
116 Fp->seekg( 0, std::ios::beg);
118 long beg = Fp->tellg();
121 ParseDES( this, beg, lgt, false); // le Load sera fait a la volee
123 Fp->seekg( 0, std::ios::beg);
125 // Load 'non string' values
127 std::string PhotometricInterpretation = GetEntryByNumber(0x0028,0x0004);
128 if( PhotometricInterpretation == "PALETTE COLOR " )
130 LoadEntryBinArea(0x0028,0x1200); // gray LUT
131 /// FIXME FIXME FIXME
132 /// The tags refered by the three following lines used to be CORRECTLY
133 /// defined as having an US Value Representation in the public
134 /// dictionnary. BUT the semantics implied by the three following
135 /// lines state that the corresponding tag contents are in fact
136 /// the ones of a BinEntry.
137 /// In order to fix things "Quick and Dirty" the dictionnary was
138 /// altered on PURPOUS but now contains a WRONG value.
139 /// In order to fix things and restore the dictionary to its
140 /// correct value, one needs to decided of the semantics by deciding
141 /// wether the following tags are either:
142 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
143 /// also used as BinEntry, which requires the proper conversion,
144 /// - OW, and hence loaded as BinEntry, but afterwards also used
145 /// as ValEntry, which requires the proper conversion.
146 LoadEntryBinArea(0x0028,0x1201); // R LUT
147 LoadEntryBinArea(0x0028,0x1202); // G LUT
148 LoadEntryBinArea(0x0028,0x1203); // B LUT
150 // Segmented Red Palette Color LUT Data
151 LoadEntryBinArea(0x0028,0x1221);
152 // Segmented Green Palette Color LUT Data
153 LoadEntryBinArea(0x0028,0x1222);
154 // Segmented Blue Palette Color LUT Data
155 LoadEntryBinArea(0x0028,0x1223);
157 //FIXME later : how to use it?
158 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
162 // --------------------------------------------------------------
163 // Specific code to allow gdcm to read ACR-LibIDO formated images
164 // Note: ACR-LibIDO is an extension of the ACR standard that was
165 // used at CREATIS. For the time being (say a couple years)
166 // we keep this kludge to allow a smooth move to gdcm for
167 // CREATIS developpers (sorry folks).
169 // if recognition code tells us we deal with a LibIDO image
170 // we switch lineNumber and columnNumber
173 RecCode = GetEntryByNumber(0x0008, 0x0010); // recognition code
174 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
175 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
176 // with "little-endian strings"
178 Filetype = ACR_LIBIDO;
179 std::string rows = GetEntryByNumber(0x0028, 0x0010);
180 std::string columns = GetEntryByNumber(0x0028, 0x0011);
181 SetEntryByNumber(columns, 0x0028, 0x0010);
182 SetEntryByNumber(rows , 0x0028, 0x0011);
184 // ----------------- End of ACR-LibIDO kludge ------------------
188 * \brief This default constructor doesn't parse the file. You should
189 * then invoke \ref Document::SetFileName and then the parsing.
191 Document::Document() : ElementSet(-1)
195 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
198 Filetype = ExplicitVR;
202 * \brief Canonical destructor.
204 Document::~Document ()
213 //-----------------------------------------------------------------------------
217 * \brief Prints The Dict Entries of THE public Dicom Dictionary
220 void Document::PrintPubDict(std::ostream & os)
222 RefPubDict->SetPrintLevel(PrintLevel);
223 RefPubDict->Print(os);
227 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
230 void Document::PrintShaDict(std::ostream & os)
232 RefShaDict->SetPrintLevel(PrintLevel);
233 RefShaDict->Print(os);
236 //-----------------------------------------------------------------------------
239 * \brief Get the public dictionary used
241 Dict* Document::GetPubDict()
247 * \brief Get the shadow dictionary used
249 Dict* Document::GetShaDict()
255 * \brief Set the shadow dictionary used
256 * \param dict dictionary to use in shadow
258 bool Document::SetShaDict(Dict *dict)
265 * \brief Set the shadow dictionary used
266 * \param dictName name of the dictionary to use in shadow
268 bool Document::SetShaDict(DictKey const & dictName)
270 RefShaDict = Global::GetDicts()->GetDict(dictName);
275 * \brief This predicate, based on hopefully reasonable heuristics,
276 * decides whether or not the current Document was properly parsed
277 * and contains the mandatory information for being considered as
278 * a well formed and usable Dicom/Acr File.
279 * @return true when Document is the one of a reasonable Dicom/Acr file,
282 bool Document::IsReadable()
284 if( Filetype == Unknown)
286 dbg.Verbose(0, "Document::IsReadable: wrong filetype");
292 dbg.Verbose(0, "Document::IsReadable: no tags in internal"
301 * \brief Accessor to the Transfer Syntax (when present) of the
302 * current document (it internally handles reading the
303 * value from disk when only parsing occured).
304 * @return The encountered Transfer Syntax of the current document.
306 TransferSyntaxType Document::GetTransferSyntax()
308 DocEntry *entry = GetDocEntryByNumber(0x0002, 0x0010);
314 // The entry might be present but not loaded (parsing and loading
315 // happen at different stages): try loading and proceed with check...
316 LoadDocEntrySafe(entry);
317 if (ValEntry* valEntry = dynamic_cast< ValEntry* >(entry) )
319 std::string transfer = valEntry->GetValue();
320 // The actual transfer (as read from disk) might be padded. We
321 // first need to remove the potential padding. We can make the
322 // weak assumption that padding was not executed with digits...
323 if ( transfer.length() == 0 )
325 // for brain damaged headers
328 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
330 transfer.erase(transfer.length()-1, 1);
332 for (int i = 0; TransferSyntaxStrings[i] != NULL; i++)
334 if ( TransferSyntaxStrings[i] == transfer )
336 return TransferSyntaxType(i);
343 bool Document::IsJPEGLossless()
345 TransferSyntaxType r = GetTransferSyntax();
346 return r == JPEGFullProgressionProcess10_12
347 || r == JPEGLosslessProcess14
348 || r == JPEGLosslessProcess14_1;
352 * \brief Determines if the Transfer Syntax was already encountered
353 * and if it corresponds to a JPEG2000 one
354 * @return True when JPEG2000 (Lossly or LossLess) found. False in all
357 bool Document::IsJPEG2000()
359 TransferSyntaxType r = GetTransferSyntax();
360 return r == JPEG2000Lossless || r == JPEG2000;
364 * \brief Determines if the Transfer Syntax corresponds to any form
365 * of Jpeg encoded Pixel data.
366 * @return True when any form of JPEG found. False otherwise.
368 bool Document::IsJPEG()
370 TransferSyntaxType r = GetTransferSyntax();
371 return r == JPEGBaselineProcess1
372 || r == JPEGExtendedProcess2_4
373 || r == JPEGExtendedProcess3_5
374 || r == JPEGSpectralSelectionProcess6_8
380 * \brief Determines if the Transfer Syntax corresponds to encapsulated
381 * of encoded Pixel Data (as opposed to native).
382 * @return True when encapsulated. False when native.
384 bool Document::IsEncapsulate()
386 TransferSyntaxType r = GetTransferSyntax();
387 return IsJPEG() || r == RLELossless;
391 * \brief Predicate for dicom version 3 file.
392 * @return True when the file is a dicom version 3.
394 bool Document::IsDicomV3()
396 // Checking if Transfert Syntax exists is enough
397 // Anyway, it's to late check if the 'Preamble' was found ...
398 // And ... would it be a rich idea to check ?
399 // (some 'no Preamble' DICOM images exist !)
400 return GetDocEntryByNumber(0x0002, 0x0010) != NULL;
404 * \brief returns the File Type
405 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
406 * @return the FileType code
408 FileType Document::GetFileType()
414 * \brief Tries to open the file \ref Document::Filename and
415 * checks the preamble when existing.
416 * @return The FILE pointer on success.
418 std::ifstream* Document::OpenFile()
420 if (Filename.length() == 0)
428 "Document::OpenFile is already opened when opening: ",
432 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
436 "Document::OpenFile cannot open file: ",
444 Fp->read((char*)&zero, (size_t)2);
451 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
453 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
454 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
455 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
456 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
462 Fp->seekg(126L, std::ios::cur);
464 Fp->read(dicm, (size_t)4);
470 if( memcmp(dicm, "DICM", 4) == 0 )
477 "Document::OpenFile not DICOM/ACR (missing preamble)",
484 * \brief closes the file
485 * @return TRUE if the close was successfull
487 bool Document::CloseFile()
496 return true; //FIXME how do we detect a non-close ifstream ?
500 * \brief Writes in a file all the Header Entries (Dicom Elements)
501 * @param fp file pointer on an already open file
502 * @param filetype Type of the File to be written
503 * (ACR-NEMA, ExplicitVR, ImplicitVR)
504 * \return Always true.
506 void Document::WriteContent(std::ofstream* fp, FileType filetype)
508 /// \todo move the following lines (and a lot of others, to be written)
509 /// to a future function CheckAndCorrectHeader
510 /// (necessary if user wants to write a DICOM V3 file
511 /// starting from an ACR-NEMA (V2) Header
513 if ( filetype == ImplicitVR || filetype == ExplicitVR )
515 // writing Dicom File Preamble
516 char filePreamble[128];
517 memset(filePreamble, 0, 128);
518 fp->write(filePreamble, 128);
519 fp->write("DICM", 4);
523 * \todo rewrite later, if really usefull
524 * - 'Group Length' element is optional in DICOM
525 * - but un-updated odd groups lengthes can causes pb
528 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
529 * UpdateGroupLength(false,filetype);
530 * if ( filetype == ACR)
531 * UpdateGroupLength(true,ACR);
534 ElementSet::WriteContent(fp, filetype); // This one is recursive
538 * \brief Modifies the value of a given Header Entry (Dicom Element)
539 * when it exists. Create it with the given value when unexistant.
540 * @param value (string) Value to be set
541 * @param group Group number of the Entry
542 * @param elem Element number of the Entry
543 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
544 * \return pointer to the modified/created Header Entry (NULL when creation
547 ValEntry* Document::ReplaceOrCreateByNumber(
548 std::string const & value,
553 ValEntry* valEntry = 0;
554 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
558 // check if (group,element) DictEntry exists
559 // if it doesn't, create an entry in DictSet::VirtualEntry
562 // Find out if the tag we received is in the dictionaries:
563 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
564 DictEntry* dictEntry = pubDict->GetDictEntryByNumber(group, elem);
567 currentEntry = NewDocEntryByNumber(group, elem, vr);
571 currentEntry = NewDocEntryByNumber(group, elem);
576 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
577 " NewDocEntryByNumber failed.");
581 valEntry = new ValEntry(currentEntry);
584 if ( !AddEntry(valEntry))
587 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
588 " failed allthough this is a creation.");
594 valEntry = dynamic_cast< ValEntry* >(currentEntry);
595 if ( !valEntry ) // Euuuuh? It wasn't a ValEntry
596 // then we change it to a ValEntry ?
597 // Shouldn't it be considered as an error ?
599 // We need to promote the DocEntry to a ValEntry:
600 valEntry = new ValEntry(currentEntry);
601 if (!RemoveEntry(currentEntry))
604 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
605 " of previous DocEntry failed.");
608 if ( !AddEntry(valEntry))
611 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
612 " promoted ValEntry failed.");
618 SetEntryByNumber(value, group, elem);
624 * \brief Modifies the value of a given Header Entry (Dicom Element)
625 * when it exists. Create it with the given value when unexistant.
626 * A copy of the binArea is made to be kept in the Document.
627 * @param binArea (binary) value to be set
628 * @param Group Group number of the Entry
629 * @param Elem Element number of the Entry
630 * \return pointer to the modified/created Header Entry (NULL when creation
633 BinEntry* Document::ReplaceOrCreateByNumber(
640 BinEntry* binEntry = 0;
641 DocEntry* currentEntry = GetDocEntryByNumber( group, elem);
645 // check if (group,element) DictEntry exists
646 // if it doesn't, create an entry in DictSet::VirtualEntry
649 // Find out if the tag we received is in the dictionaries:
650 Dict *pubDict = Global::GetDicts()->GetDefaultPubDict();
651 DictEntry *dictEntry = pubDict->GetDictEntryByNumber(group, elem);
655 currentEntry = NewDocEntryByNumber(group, elem, vr);
659 currentEntry = NewDocEntryByNumber(group, elem);
663 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: call to"
664 " NewDocEntryByNumber failed.");
667 binEntry = new BinEntry(currentEntry);
668 if ( !AddEntry(binEntry))
670 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
671 " failed allthough this is a creation.");
677 binEntry = dynamic_cast< BinEntry* >(currentEntry);
678 if ( !binEntry ) // Euuuuh? It wasn't a BinEntry
679 // then we change it to a BinEntry ?
680 // Shouldn't it be considered as an error ?
682 // We need to promote the DocEntry to a BinEntry:
683 binEntry = new BinEntry(currentEntry);
684 if (!RemoveEntry(currentEntry))
686 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
687 " of previous DocEntry failed.");
690 if ( !AddEntry(binEntry))
692 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: adding"
693 " promoted BinEntry failed.");
700 if (lgth>0 && binArea)
702 tmpArea = new uint8_t[lgth];
703 memcpy(tmpArea,binArea,lgth);
709 if (!SetEntryByNumber(tmpArea, lgth, group, elem))
722 * \brief Modifies the value of a given Header Entry (Dicom Element)
723 * when it exists. Create it when unexistant.
724 * @param Group Group number of the Entry
725 * @param Elem Element number of the Entry
726 * \return pointer to the modified/created SeqEntry (NULL when creation
729 SeqEntry* Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem)
732 DocEntry* a = GetDocEntryByNumber( group, elem);
735 a = NewSeqEntryByNumber(group, elem);
741 b = new SeqEntry(a, 1); // FIXME : 1 (Depth)
748 * \brief Set a new value if the invoked element exists
749 * Seems to be useless !!!
750 * @param value new element value
751 * @param group group number of the Entry
752 * @param elem element number of the Entry
755 bool Document::ReplaceIfExistByNumber(std::string const & value,
756 uint16_t group, uint16_t elem )
758 SetEntryByNumber(value, group, elem);
763 std::string Document::GetTransferSyntaxValue(TransferSyntaxType type)
765 return TransferSyntaxStrings[type];
768 //-----------------------------------------------------------------------------
772 * \brief Checks if a given Dicom Element exists within the H table
773 * @param group Group number of the searched Dicom Element
774 * @param element Element number of the searched Dicom Element
775 * @return true is found
777 bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
779 const std::string &key = DictEntry::TranslateToKey(group, element );
780 return TagHT.count(key) != 0;
784 * \brief Searches within Header Entries (Dicom Elements) parsed with
785 * the public and private dictionaries
786 * for the element value of a given tag.
787 * \warning Don't use any longer : use GetPubEntryByName
788 * @param tagName name of the searched element.
789 * @return Corresponding element value when it exists,
790 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
792 std::string Document::GetEntryByName(TagName const & tagName)
794 DictEntry* dictEntry = RefPubDict->GetDictEntryByName(tagName);
800 return GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
804 * \brief Searches within Header Entries (Dicom Elements) parsed with
805 * the public and private dictionaries
806 * for the element value representation of a given tag.
808 * Obtaining the VR (Value Representation) might be needed by caller
809 * to convert the string typed content to caller's native type
810 * (think of C++ vs Python). The VR is actually of a higher level
811 * of semantics than just the native C++ type.
812 * @param tagName name of the searched element.
813 * @return Corresponding element value representation when it exists,
814 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
816 std::string Document::GetEntryVRByName(TagName const& tagName)
818 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
819 if( dictEntry == NULL)
824 DocEntry* elem = GetDocEntryByNumber(dictEntry->GetGroup(),
825 dictEntry->GetElement());
826 return elem->GetVR();
830 * \brief Searches within Header Entries (Dicom Elements) parsed with
831 * the public and private dictionaries
832 * for the element value representation of a given tag.
833 * @param group Group number of the searched tag.
834 * @param element Element number of the searched tag.
835 * @return Corresponding element value representation when it exists,
836 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
838 std::string Document::GetEntryByNumber(uint16_t group, uint16_t element)
840 TagKey key = DictEntry::TranslateToKey(group, element);
841 if ( !TagHT.count(key))
846 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
850 * \brief Searches within Header Entries (Dicom Elements) parsed with
851 * the public and private dictionaries
852 * for the element value representation of a given tag..
854 * Obtaining the VR (Value Representation) might be needed by caller
855 * to convert the string typed content to caller's native type
856 * (think of C++ vs Python). The VR is actually of a higher level
857 * of semantics than just the native C++ type.
858 * @param group Group number of the searched tag.
859 * @param element Element number of the searched tag.
860 * @return Corresponding element value representation when it exists,
861 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
863 std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element)
865 DocEntry* elem = GetDocEntryByNumber(group, element);
870 return elem->GetVR();
874 * \brief Searches within Header Entries (Dicom Elements) parsed with
875 * the public and private dictionaries
876 * for the value length of a given tag..
877 * @param group Group number of the searched tag.
878 * @param element Element number of the searched tag.
879 * @return Corresponding element length; -2 if not found
881 int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element)
883 DocEntry* elem = GetDocEntryByNumber(group, element);
886 return -2; //magic number
888 return elem->GetLength();
891 * \brief Sets the value (string) of the Header Entry (Dicom Element)
892 * @param content string value of the Dicom Element
893 * @param tagName name of the searched Dicom Element.
894 * @return true when found
896 bool Document::SetEntryByName( std::string const & content,
897 TagName const & tagName)
899 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
905 return SetEntryByNumber(content,dictEntry->GetGroup(),
906 dictEntry->GetElement());
910 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
911 * through it's (group, element) and modifies it's content with
913 * @param content new value (string) to substitute with
914 * @param group group number of the Dicom Element to modify
915 * @param element element number of the Dicom Element to modify
917 bool Document::SetEntryByNumber(std::string const& content,
918 uint16_t group, uint16_t element)
920 ValEntry* valEntry = GetValEntryByNumber(group, element);
923 dbg.Verbose(0, "Document::SetEntryByNumber: no corresponding",
924 " ValEntry (try promotion first).");
927 // Non even content must be padded with a space (020H)...
928 valEntry->SetValue(content);
933 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
934 * through it's (group, element) and modifies it's content with
936 * @param content new value (void* -> uint8_t*) to substitute with
937 * @param lgth new value length
938 * @param group group number of the Dicom Element to modify
939 * @param element element number of the Dicom Element to modify
941 bool Document::SetEntryByNumber(uint8_t*content, int lgth,
942 uint16_t group, uint16_t element)
944 (void)lgth; //not used
945 TagKey key = DictEntry::TranslateToKey(group, element);
946 if ( !TagHT.count(key))
951 /* Hope Binary field length is *never* wrong
952 if(lgth%2) // Non even length are padded with a space (020H).
955 //content = content + '\0'; // fing a trick to enlarge a binary field?
958 BinEntry* entry = (BinEntry *)TagHT[key];
959 entry->SetBinArea(content);
960 entry->SetLength(lgth);
961 entry->SetValue(GDCM_BINLOADED);
967 * \brief Gets (from Header) a 'non string' element value
968 * (LoadElementValues has already be executed)
969 * @param group group number of the Entry
970 * @param elem element number of the Entry
971 * @return Pointer to the 'non string' area
973 void* Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem)
975 DocEntry* entry = GetDocEntryByNumber(group, elem);
978 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry");
981 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(entry) )
983 return binEntry->GetBinArea();
990 * \brief Loads (from disk) the element content
991 * when a string is not suitable
992 * @param group group number of the Entry
993 * @param elem element number of the Entry
995 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
997 // Search the corresponding DocEntry
998 DocEntry *docElement = GetDocEntryByNumber(group, elem);
1002 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
1006 LoadEntryBinArea(binElement);
1010 * \brief Loads (from disk) the element content
1011 * when a string is not suitable
1012 * @param element Entry whose binArea is going to be loaded
1014 void Document::LoadEntryBinArea(BinEntry* element)
1016 if(element->GetBinArea())
1019 bool openFile = !Fp;
1023 size_t o =(size_t)element->GetOffset();
1024 Fp->seekg(o, std::ios::beg);
1026 size_t l = element->GetLength();
1027 uint8_t* a = new uint8_t[l];
1030 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1034 /// \todo check the result
1035 Fp->read((char*)a, l);
1036 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
1042 element->SetBinArea(a);
1049 * \brief Sets a 'non string' value to a given Dicom Element
1050 * @param area area containing the 'non string' value
1051 * @param group Group number of the searched Dicom Element
1052 * @param element Element number of the searched Dicom Element
1055 /*bool Document::SetEntryBinAreaByNumber(uint8_t* area,
1056 uint16_t group, uint16_t element)
1058 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1059 if ( !currentEntry )
1064 if ( BinEntry* binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1066 binEntry->SetBinArea( area );
1074 * \brief Searches within the Header Entries for a Dicom Element of
1076 * @param tagName name of the searched Dicom Element.
1077 * @return Corresponding Dicom Element when it exists, and NULL
1080 DocEntry* Document::GetDocEntryByName(TagName const & tagName)
1082 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
1088 return GetDocEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
1092 * \brief retrieves a Dicom Element (the first one) using (group, element)
1093 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1094 * if you think it's NOT UNIQUE, check the count number
1095 * and use iterators to retrieve ALL the Dicoms Elements within
1096 * a given couple (group, element)
1097 * @param group Group number of the searched Dicom Element
1098 * @param element Element number of the searched Dicom Element
1101 DocEntry* Document::GetDocEntryByNumber(uint16_t group, uint16_t element)
1103 TagKey key = DictEntry::TranslateToKey(group, element);
1104 if ( !TagHT.count(key))
1108 return TagHT.find(key)->second;
1112 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1113 * returns a result when the corresponding entry is of type
1115 * @return When present, the corresponding ValEntry.
1117 ValEntry* Document::GetValEntryByNumber(uint16_t group, uint16_t element)
1119 DocEntry* currentEntry = GetDocEntryByNumber(group, element);
1120 if ( !currentEntry )
1124 if ( ValEntry* valEntry = dynamic_cast<ValEntry*>(currentEntry) )
1128 dbg.Verbose(0, "Document::GetValEntryByNumber: unfound ValEntry.");
1134 * \brief Loads the element while preserving the current
1135 * underlying file position indicator as opposed to
1136 * to LoadDocEntry that modifies it.
1137 * @param entry Header Entry whose value shall be loaded.
1140 void Document::LoadDocEntrySafe(DocEntry * entry)
1144 long PositionOnEntry = Fp->tellg();
1145 LoadDocEntry(entry);
1146 Fp->seekg(PositionOnEntry, std::ios::beg);
1151 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1153 * @return The properly swaped 32 bits integer.
1155 uint32_t Document::SwapLong(uint32_t a)
1162 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1163 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1167 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1171 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1174 //std::cout << "swapCode= " << SwapCode << std::endl;
1175 dbg.Error(" Document::SwapLong : unset swap code");
1182 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1184 * @return The properly unswaped 32 bits integer.
1186 uint32_t Document::UnswapLong(uint32_t a)
1192 * \brief Swaps the bytes so they agree with the processor order
1193 * @return The properly swaped 16 bits integer.
1195 uint16_t Document::SwapShort(uint16_t a)
1197 if ( SwapCode == 4321 || SwapCode == 2143 )
1199 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1205 * \brief Unswaps the bytes so they agree with the processor order
1206 * @return The properly unswaped 16 bits integer.
1208 uint16_t Document::UnswapShort(uint16_t a)
1210 return SwapShort(a);
1213 //-----------------------------------------------------------------------------
1217 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1218 * @return length of the parsed set.
1220 void Document::ParseDES(DocEntrySet *set, long offset,
1221 long l_max, bool delim_mode)
1223 DocEntry *newDocEntry = 0;
1227 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1231 newDocEntry = ReadNextDocEntry( );
1237 VRKey vr = newDocEntry->GetVR();
1241 if ( Global::GetVR()->IsVROfGdcmStringRepresentable(vr) )
1243 /////////////////////// ValEntry
1244 ValEntry* newValEntry =
1245 new ValEntry( newDocEntry->GetDictEntry() ); //LEAK
1246 newValEntry->Copy( newDocEntry );
1248 // When "set" is a Document, then we are at the top of the
1249 // hierarchy and the Key is simply of the form ( group, elem )...
1250 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1253 newValEntry->SetKey( newValEntry->GetKey() );
1255 // ...but when "set" is a SQItem, we are inserting this new
1256 // valEntry in a sequence item. Hence the key has the
1257 // generalized form (refer to \ref BaseTagKey):
1258 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1260 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1261 + newValEntry->GetKey() );
1264 LoadDocEntry( newValEntry );
1265 bool delimitor=newValEntry->IsItemDelimitor();
1266 if( !set->AddEntry( newValEntry ) )
1268 // If here expect big troubles
1269 delete newValEntry; //otherwise mem leak
1277 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1285 if ( ! Global::GetVR()->IsVROfGdcmBinaryRepresentable(vr) )
1287 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1288 dbg.Verbose(0, "Document::ParseDES: neither Valentry, "
1289 "nor BinEntry. Probably unknown VR.");
1292 //////////////////// BinEntry or UNKOWN VR:
1293 BinEntry* newBinEntry = new BinEntry( newDocEntry ); //LEAK
1295 // When "this" is a Document the Key is simply of the
1296 // form ( group, elem )...
1297 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1300 newBinEntry->SetKey( newBinEntry->GetKey() );
1302 // but when "this" is a SQItem, we are inserting this new
1303 // valEntry in a sequence item, and the kay has the
1304 // generalized form (refer to \ref BaseTagKey):
1305 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1307 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1308 + newBinEntry->GetKey() );
1311 LoadDocEntry( newBinEntry );
1312 if( !set->AddEntry( newBinEntry ) )
1314 //Expect big troubles if here
1319 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1320 && ( newDocEntry->GetElement() == 0x0010 ) )
1322 TransferSyntaxType ts = GetTransferSyntax();
1323 if ( ts == RLELossless )
1325 long PositionOnEntry = Fp->tellg();
1326 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1328 Fp->seekg( PositionOnEntry, std::ios::beg );
1330 else if ( IsJPEG() )
1332 long PositionOnEntry = Fp->tellg();
1333 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1334 ComputeJPEGFragmentInfo();
1335 Fp->seekg( PositionOnEntry, std::ios::beg );
1339 // Just to make sure we are at the beginning of next entry.
1340 SkipToNextDocEntry(newDocEntry);
1341 //delete newDocEntry;
1346 unsigned long l = newDocEntry->GetReadLength();
1347 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1349 if ( l == 0xffffffff )
1358 // no other way to create it ...
1359 SeqEntry* newSeqEntry =
1360 new SeqEntry( newDocEntry->GetDictEntry() );
1361 newSeqEntry->Copy( newDocEntry );
1362 newSeqEntry->SetDelimitorMode( delim_mode );
1364 // At the top of the hierarchy, stands a Document. When "set"
1365 // is a Document, then we are building the first depth level.
1366 // Hence the SeqEntry we are building simply has a depth
1368 if (Document* dummy = dynamic_cast< Document* > ( set ) )
1371 newSeqEntry->SetDepthLevel( 1 );
1372 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1374 // But when "set" is allready a SQItem, we are building a nested
1375 // sequence, and hence the depth level of the new SeqEntry
1376 // we are building, is one level deeper:
1377 if (SQItem* parentSQItem = dynamic_cast< SQItem* > ( set ) )
1379 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1380 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1381 + newSeqEntry->GetKey() );
1385 { // Don't try to parse zero-length sequences
1386 ParseSQ( newSeqEntry,
1387 newDocEntry->GetOffset(),
1390 set->AddEntry( newSeqEntry );
1391 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1402 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1403 * @return parsed length for this level
1405 void Document::ParseSQ( SeqEntry* seqEntry,
1406 long offset, long l_max, bool delim_mode)
1408 int SQItemNumber = 0;
1413 DocEntry* newDocEntry = ReadNextDocEntry();
1416 // FIXME Should warn user
1421 if ( newDocEntry->IsSequenceDelimitor() )
1423 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1427 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1433 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1434 std::ostringstream newBase;
1435 newBase << seqEntry->GetKey()
1439 itemSQ->SetBaseTagKey( newBase.str() );
1440 unsigned int l = newDocEntry->GetReadLength();
1442 if ( l == 0xffffffff )
1451 ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1454 seqEntry->AddEntry( itemSQ, SQItemNumber );
1456 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1464 * \brief Loads the element content if its length doesn't exceed
1465 * the value specified with Document::SetMaxSizeLoadEntry()
1466 * @param entry Header Entry (Dicom Element) to be dealt with
1468 void Document::LoadDocEntry(DocEntry* entry)
1470 uint16_t group = entry->GetGroup();
1471 std::string vr = entry->GetVR();
1472 uint32_t length = entry->GetLength();
1474 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1476 // A SeQuence "contains" a set of Elements.
1477 // (fffe e000) tells us an Element is beginning
1478 // (fffe e00d) tells us an Element just ended
1479 // (fffe e0dd) tells us the current SeQuence just ended
1480 if( group == 0xfffe )
1482 // NO more value field for SQ !
1486 // When the length is zero things are easy:
1489 ((ValEntry *)entry)->SetValue("");
1493 // The elements whose length is bigger than the specified upper bound
1494 // are not loaded. Instead we leave a short notice of the offset of
1495 // the element content and it's length.
1497 std::ostringstream s;
1498 if (length > MaxSizeLoadEntry)
1500 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1502 //s << "gdcm::NotLoaded (BinEntry)";
1503 s << GDCM_NOTLOADED;
1504 s << " Address:" << (long)entry->GetOffset();
1505 s << " Length:" << entry->GetLength();
1506 s << " x(" << std::hex << entry->GetLength() << ")";
1507 binEntryPtr->SetValue(s.str());
1509 // Be carefull : a BinEntry IS_A ValEntry ...
1510 else if (ValEntry* valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1512 // s << "gdcm::NotLoaded. (ValEntry)";
1513 s << GDCM_NOTLOADED;
1514 s << " Address:" << (long)entry->GetOffset();
1515 s << " Length:" << entry->GetLength();
1516 s << " x(" << std::hex << entry->GetLength() << ")";
1517 valEntryPtr->SetValue(s.str());
1522 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1523 << "nor a ValEntry ?! Should never print that !" << std::endl;
1526 // to be sure we are at the end of the value ...
1527 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1532 // When we find a BinEntry not very much can be done :
1533 if (BinEntry* binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1535 s << GDCM_BINLOADED;
1536 binEntryPtr->SetValue(s.str());
1537 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1541 /// \todo Any compacter code suggested (?)
1542 if ( IsDocEntryAnInteger(entry) )
1546 // When short integer(s) are expected, read and convert the following
1547 // n *two characters properly i.e. consider them as short integers as
1548 // opposed to strings.
1549 // Elements with Value Multiplicity > 1
1550 // contain a set of integers (not a single one)
1551 if (vr == "US" || vr == "SS")
1554 NewInt = ReadInt16();
1558 for (int i=1; i < nbInt; i++)
1561 NewInt = ReadInt16();
1566 // See above comment on multiple integers (mutatis mutandis).
1567 else if (vr == "UL" || vr == "SL")
1570 NewInt = ReadInt32();
1574 for (int i=1; i < nbInt; i++)
1577 NewInt = ReadInt32();
1582 #ifdef GDCM_NO_ANSI_STRING_STREAM
1583 s << std::ends; // to avoid oddities on Solaris
1584 #endif //GDCM_NO_ANSI_STRING_STREAM
1586 ((ValEntry *)entry)->SetValue(s.str());
1590 // FIXME: We need an additional byte for storing \0 that is not on disk
1591 char *str = new char[length+1];
1592 Fp->read(str, (size_t)length);
1593 str[length] = '\0'; //this is only useful when length is odd
1594 // Special DicomString call to properly handle \0 and even length
1595 std::string newValue;
1598 newValue = Util::DicomString(str, length+1);
1599 //dbg.Verbose(0, "Warning: bad length: ", length );
1600 dbg.Verbose(0, "For string :", newValue.c_str());
1601 // Since we change the length of string update it length
1602 entry->SetReadLength(length+1);
1606 newValue = Util::DicomString(str, length);
1610 if ( ValEntry* valEntry = dynamic_cast<ValEntry* >(entry) )
1612 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1614 dbg.Verbose(1, "Document::LoadDocEntry",
1615 "unread element value");
1616 valEntry->SetValue(GDCM_UNREAD);
1622 // Because of correspondance with the VR dic
1623 valEntry->SetValue(newValue);
1627 valEntry->SetValue(newValue);
1632 dbg.Error(true, "Document::LoadDocEntry"
1633 "Should have a ValEntry, here !");
1639 * \brief Find the value Length of the passed Header Entry
1640 * @param entry Header Entry whose length of the value shall be loaded.
1642 void Document::FindDocEntryLength( DocEntry *entry )
1643 throw ( FormatError )
1645 uint16_t element = entry->GetElement();
1646 std::string vr = entry->GetVR();
1649 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1651 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1653 // The following reserved two bytes (see PS 3.5-2003, section
1654 // "7.1.2 Data element structure with explicit vr", p 27) must be
1655 // skipped before proceeding on reading the length on 4 bytes.
1656 Fp->seekg( 2L, std::ios::cur);
1657 uint32_t length32 = ReadInt32();
1659 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1664 /// \todo rename that to FindDocEntryLengthOBOrOW since
1665 /// the above test is on both OB and OW...
1666 lengthOB = FindDocEntryLengthOB();
1668 catch ( FormatUnexpected )
1670 // Computing the length failed (this happens with broken
1671 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1672 // chance to get the pixels by deciding the element goes
1673 // until the end of the file. Hence we artificially fix the
1674 // the length and proceed.
1675 long currentPosition = Fp->tellg();
1676 Fp->seekg(0L,std::ios::end);
1677 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1678 Fp->seekg(currentPosition, std::ios::beg);
1679 entry->SetLength(lengthUntilEOF);
1682 entry->SetLength(lengthOB);
1685 FixDocEntryFoundLength(entry, length32);
1689 // Length is encoded on 2 bytes.
1690 length16 = ReadInt16();
1692 // We can tell the current file is encoded in big endian (like
1693 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1694 // and it's value is the one of the encoding of a big endian file.
1695 // In order to deal with such big endian encoded files, we have
1696 // (at least) two strategies:
1697 // * when we load the "Transfer Syntax" tag with value of big endian
1698 // encoding, we raise the proper flags. Then we wait for the end
1699 // of the META group (0x0002) among which is "Transfer Syntax",
1700 // before switching the swap code to big endian. We have to postpone
1701 // the switching of the swap code since the META group is fully encoded
1702 // in little endian, and big endian coding only starts at the next
1703 // group. The corresponding code can be hard to analyse and adds
1704 // many additional unnecessary tests for regular tags.
1705 // * the second strategy consists in waiting for trouble, that shall
1706 // appear when we find the first group with big endian encoding. This
1707 // is easy to detect since the length of a "Group Length" tag (the
1708 // ones with zero as element number) has to be of 4 (0x0004). When we
1709 // encounter 1024 (0x0400) chances are the encoding changed and we
1710 // found a group with big endian encoding.
1711 // We shall use this second strategy. In order to make sure that we
1712 // can interpret the presence of an apparently big endian encoded
1713 // length of a "Group Length" without committing a big mistake, we
1714 // add an additional check: we look in the already parsed elements
1715 // for the presence of a "Transfer Syntax" whose value has to be "big
1716 // endian encoding". When this is the case, chances are we have got our
1717 // hands on a big endian encoded file: we switch the swap code to
1718 // big endian and proceed...
1719 if ( element == 0x0000 && length16 == 0x0400 )
1721 TransferSyntaxType ts = GetTransferSyntax();
1722 if ( ts != ExplicitVRBigEndian )
1724 throw FormatError( "Document::FindDocEntryLength()",
1725 " not explicit VR." );
1729 SwitchSwapToBigEndian();
1730 // Restore the unproperly loaded values i.e. the group, the element
1731 // and the dictionary entry depending on them.
1732 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1733 uint16_t correctElem = SwapShort( entry->GetElement() );
1734 DictEntry* newTag = GetDictEntryByNumber( correctGroup,
1738 // This correct tag is not in the dictionary. Create a new one.
1739 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1741 // FIXME this can create a memory leaks on the old entry that be
1742 // left unreferenced.
1743 entry->SetDictEntry( newTag );
1746 // Heuristic: well, some files are really ill-formed.
1747 if ( length16 == 0xffff)
1749 // 0xffff means that we deal with 'Unknown Length' Sequence
1752 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1757 // Either implicit VR or a non DICOM conformal (see note below) explicit
1758 // VR that ommited the VR of (at least) this element. Farts happen.
1759 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1760 // on Data elements "Implicit and Explicit VR Data Elements shall
1761 // not coexist in a Data Set and Data Sets nested within it".]
1762 // Length is on 4 bytes.
1764 FixDocEntryFoundLength( entry, ReadInt32() );
1770 * \brief Find the Value Representation of the current Dicom Element.
1773 std::string Document::FindDocEntryVR()
1775 if ( Filetype != ExplicitVR )
1776 return(GDCM_UNKNOWN);
1778 long positionOnEntry = Fp->tellg();
1779 // Warning: we believe this is explicit VR (Value Representation) because
1780 // we used a heuristic that found "UL" in the first tag. Alas this
1781 // doesn't guarantee that all the tags will be in explicit VR. In some
1782 // cases (see e-film filtered files) one finds implicit VR tags mixed
1783 // within an explicit VR file. Hence we make sure the present tag
1784 // is in explicit VR and try to fix things if it happens not to be
1788 Fp->read (vr, (size_t)2);
1791 if( !CheckDocEntryVR(vr) )
1793 Fp->seekg(positionOnEntry, std::ios::beg);
1794 return(GDCM_UNKNOWN);
1800 * \brief Check the correspondance between the VR of the header entry
1801 * and the taken VR. If they are different, the header entry is
1802 * updated with the new VR.
1803 * @param entry Header Entry to check
1804 * @param vr Dicom Value Representation
1805 * @return false if the VR is incorrect of if the VR isn't referenced
1806 * otherwise, it returns true
1808 bool Document::CheckDocEntryVR(VRKey vr)
1810 // CLEANME searching the dicom_vr at each occurence is expensive.
1811 // PostPone this test in an optional integrity check at the end
1812 // of parsing or only in debug mode.
1813 if ( !Global::GetVR()->IsValidVR(vr) )
1820 * \brief Get the transformed value of the header entry. The VR value
1821 * is used to define the transformation to operate on the value
1822 * \warning NOT end user intended method !
1823 * @param entry entry to tranform
1824 * @return Transformed entry value
1826 std::string Document::GetDocEntryValue(DocEntry *entry)
1828 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1830 std::string val = ((ValEntry *)entry)->GetValue();
1831 std::string vr = entry->GetVR();
1832 uint32_t length = entry->GetLength();
1833 std::ostringstream s;
1836 // When short integer(s) are expected, read and convert the following
1837 // n * 2 bytes properly i.e. as a multivaluated strings
1838 // (each single value is separated fromthe next one by '\'
1839 // as usual for standard multivaluated filels
1840 // Elements with Value Multiplicity > 1
1841 // contain a set of short integers (not a single one)
1843 if( vr == "US" || vr == "SS" )
1848 for (int i=0; i < nbInt; i++)
1854 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1855 newInt16 = SwapShort( newInt16 );
1860 // When integer(s) are expected, read and convert the following
1861 // n * 4 bytes properly i.e. as a multivaluated strings
1862 // (each single value is separated fromthe next one by '\'
1863 // as usual for standard multivaluated filels
1864 // Elements with Value Multiplicity > 1
1865 // contain a set of integers (not a single one)
1866 else if( vr == "UL" || vr == "SL" )
1871 for (int i=0; i < nbInt; i++)
1877 newInt32 = ( val[4*i+0] & 0xFF )
1878 + (( val[4*i+1] & 0xFF ) << 8 )
1879 + (( val[4*i+2] & 0xFF ) << 16 )
1880 + (( val[4*i+3] & 0xFF ) << 24 );
1881 newInt32 = SwapLong( newInt32 );
1885 #ifdef GDCM_NO_ANSI_STRING_STREAM
1886 s << std::ends; // to avoid oddities on Solaris
1887 #endif //GDCM_NO_ANSI_STRING_STREAM
1891 return ((ValEntry *)entry)->GetValue();
1895 * \brief Get the reverse transformed value of the header entry. The VR
1896 * value is used to define the reverse transformation to operate on
1898 * \warning NOT end user intended method !
1899 * @param entry Entry to reverse transform
1900 * @return Reverse transformed entry value
1902 std::string Document::GetDocEntryUnvalue(DocEntry* entry)
1904 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1906 std::string vr = entry->GetVR();
1907 std::vector<std::string> tokens;
1908 std::ostringstream s;
1910 if ( vr == "US" || vr == "SS" )
1914 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1915 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1916 for (unsigned int i=0; i<tokens.size(); i++)
1918 newInt16 = atoi(tokens[i].c_str());
1919 s << ( newInt16 & 0xFF )
1920 << (( newInt16 >> 8 ) & 0xFF );
1924 if ( vr == "UL" || vr == "SL")
1928 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1929 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1930 for (unsigned int i=0; i<tokens.size();i++)
1932 newInt32 = atoi(tokens[i].c_str());
1933 s << (char)( newInt32 & 0xFF )
1934 << (char)(( newInt32 >> 8 ) & 0xFF )
1935 << (char)(( newInt32 >> 16 ) & 0xFF )
1936 << (char)(( newInt32 >> 24 ) & 0xFF );
1941 #ifdef GDCM_NO_ANSI_STRING_STREAM
1942 s << std::ends; // to avoid oddities on Solaris
1943 #endif //GDCM_NO_ANSI_STRING_STREAM
1947 return ((ValEntry *)entry)->GetValue();
1951 * \brief Skip a given Header Entry
1952 * \warning NOT end user intended method !
1953 * @param entry entry to skip
1955 void Document::SkipDocEntry(DocEntry *entry)
1957 SkipBytes(entry->GetLength());
1961 * \brief Skips to the begining of the next Header Entry
1962 * \warning NOT end user intended method !
1963 * @param entry entry to skip
1965 void Document::SkipToNextDocEntry(DocEntry *entry)
1967 Fp->seekg((long)(entry->GetOffset()), std::ios::beg);
1968 Fp->seekg( (long)(entry->GetReadLength()), std::ios::cur);
1972 * \brief When the length of an element value is obviously wrong (because
1973 * the parser went Jabberwocky) one can hope improving things by
1974 * applying some heuristics.
1975 * @param entry entry to check
1976 * @param foundLength fist assumption about length
1978 void Document::FixDocEntryFoundLength(DocEntry *entry,
1979 uint32_t foundLength)
1981 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1982 if ( foundLength == 0xffffffff)
1987 uint16_t gr = entry->GetGroup();
1988 uint16_t el = entry->GetElement();
1990 if ( foundLength % 2)
1992 std::ostringstream s;
1993 s << "Warning : Tag with uneven length "
1995 << " in x(" << std::hex << gr << "," << el <<")" << std::dec;
1996 dbg.Verbose(0, s.str().c_str());
1999 //////// Fix for some naughty General Electric images.
2000 // Allthough not recent many such GE corrupted images are still present
2001 // on Creatis hard disks. Hence this fix shall remain when such images
2002 // are no longer in user (we are talking a few years, here)...
2003 // Note: XMedCom probably uses such a trick since it is able to read
2004 // those pesky GE images ...
2005 if ( foundLength == 13)
2007 // Only happens for this length !
2008 if ( entry->GetGroup() != 0x0008
2009 || ( entry->GetElement() != 0x0070
2010 && entry->GetElement() != 0x0080 ) )
2013 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
2017 //////// Fix for some brain-dead 'Leonardo' Siemens images.
2018 // Occurence of such images is quite low (unless one leaves close to a
2019 // 'Leonardo' source. Hence, one might consider commenting out the
2020 // following fix on efficiency reasons.
2021 else if ( entry->GetGroup() == 0x0009
2022 && ( entry->GetElement() == 0x1113
2023 || entry->GetElement() == 0x1114 ) )
2026 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
2029 else if ( entry->GetVR() == "SQ" )
2031 foundLength = 0; // ReadLength is unchanged
2034 //////// We encountered a 'delimiter' element i.e. a tag of the form
2035 // "fffe|xxxx" which is just a marker. Delimiters length should not be
2036 // taken into account.
2037 else if( entry->GetGroup() == 0xfffe )
2039 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
2040 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
2041 // causes extra troubles...
2042 if( entry->GetElement() != 0x0000 )
2048 entry->SetUsableLength(foundLength);
2052 * \brief Apply some heuristics to predict whether the considered
2053 * element value contains/represents an integer or not.
2054 * @param entry The element value on which to apply the predicate.
2055 * @return The result of the heuristical predicate.
2057 bool Document::IsDocEntryAnInteger(DocEntry *entry)
2059 uint16_t element = entry->GetElement();
2060 uint16_t group = entry->GetGroup();
2061 const std::string & vr = entry->GetVR();
2062 uint32_t length = entry->GetLength();
2064 // When we have some semantics on the element we just read, and if we
2065 // a priori know we are dealing with an integer, then we shall be
2066 // able to swap it's element value properly.
2067 if ( element == 0 ) // This is the group length of the group
2075 // Allthough this should never happen, still some images have a
2076 // corrupted group length [e.g. have a glance at offset x(8336) of
2077 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2078 // Since for dicom compliant and well behaved headers, the present
2079 // test is useless (and might even look a bit paranoid), when we
2080 // encounter such an ill-formed image, we simply display a warning
2081 // message and proceed on parsing (while crossing fingers).
2082 std::ostringstream s;
2083 long filePosition = Fp->tellg();
2084 s << "Erroneous Group Length element length on : (" \
2085 << std::hex << group << " , " << element
2086 << ") -before- position x(" << filePosition << ")"
2087 << "lgt : " << length;
2088 dbg.Verbose(0, "Document::IsDocEntryAnInteger", s.str().c_str() );
2092 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2100 * \brief Find the Length till the next sequence delimiter
2101 * \warning NOT end user intended method !
2105 uint32_t Document::FindDocEntryLengthOB()
2106 throw( FormatUnexpected )
2108 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2109 long positionOnEntry = Fp->tellg();
2110 bool foundSequenceDelimiter = false;
2111 uint32_t totalLength = 0;
2113 while ( !foundSequenceDelimiter )
2119 group = ReadInt16();
2122 catch ( FormatError )
2124 throw FormatError("Document::FindDocEntryLengthOB()",
2125 " group or element not present.");
2128 // We have to decount the group and element we just read
2131 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2133 dbg.Verbose(1, "Document::FindDocEntryLengthOB: neither an Item "
2134 "tag nor a Sequence delimiter tag.");
2135 Fp->seekg(positionOnEntry, std::ios::beg);
2136 throw FormatUnexpected("Document::FindDocEntryLengthOB()",
2137 "Neither an Item tag nor a Sequence "
2141 if ( elem == 0xe0dd )
2143 foundSequenceDelimiter = true;
2146 uint32_t itemLength = ReadInt32();
2147 // We add 4 bytes since we just read the ItemLength with ReadInt32
2148 totalLength += itemLength + 4;
2149 SkipBytes(itemLength);
2151 if ( foundSequenceDelimiter )
2156 Fp->seekg( positionOnEntry, std::ios::beg);
2161 * \brief Reads a supposed to be 16 Bits integer
2162 * (swaps it depending on processor endianity)
2163 * @return read value
2165 uint16_t Document::ReadInt16()
2166 throw( FormatError )
2169 Fp->read ((char*)&g, (size_t)2);
2172 throw FormatError( "Document::ReadInt16()", " file error." );
2176 throw FormatError( "Document::ReadInt16()", "EOF." );
2183 * \brief Reads a supposed to be 32 Bits integer
2184 * (swaps it depending on processor endianity)
2185 * @return read value
2187 uint32_t Document::ReadInt32()
2188 throw( FormatError )
2191 Fp->read ((char*)&g, (size_t)4);
2194 throw FormatError( "Document::ReadInt32()", " file error." );
2198 throw FormatError( "Document::ReadInt32()", "EOF." );
2205 * \brief skips bytes inside the source file
2206 * \warning NOT end user intended method !
2209 void Document::SkipBytes(uint32_t nBytes)
2211 //FIXME don't dump the returned value
2212 Fp->seekg((long)nBytes, std::ios::cur);
2216 * \brief Loads all the needed Dictionaries
2217 * \warning NOT end user intended method !
2219 void Document::Initialise()
2221 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2223 RLEInfo = new RLEFramesInfo;
2224 JPEGInfo = new JPEGFragmentsInfo;
2229 * \brief Discover what the swap code is (among little endian, big endian,
2230 * bad little endian, bad big endian).
2232 * @return false when we are absolutely sure
2233 * it's neither ACR-NEMA nor DICOM
2234 * true when we hope ours assuptions are OK
2236 bool Document::CheckSwap()
2238 // The only guaranted way of finding the swap code is to find a
2239 // group tag since we know it's length has to be of four bytes i.e.
2240 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2241 // occurs when we can't find such group...
2243 uint32_t x = 4; // x : for ntohs
2244 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2250 // First, compare HostByteOrder and NetworkByteOrder in order to
2251 // determine if we shall need to swap bytes (i.e. the Endian type).
2252 if ( x == ntohs(x) )
2261 // The easiest case is the one of a DICOM header, since it possesses a
2262 // file preamble where it suffice to look for the string "DICM".
2265 char *entCur = deb + 128;
2266 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2268 dbg.Verbose(1, "Document::CheckSwap:", "looks like DICOM Version3");
2270 // Next, determine the value representation (VR). Let's skip to the
2271 // first element (0002, 0000) and check there if we find "UL"
2272 // - or "OB" if the 1st one is (0002,0001) -,
2273 // in which case we (almost) know it is explicit VR.
2274 // WARNING: if it happens to be implicit VR then what we will read
2275 // is the length of the group. If this ascii representation of this
2276 // length happens to be "UL" then we shall believe it is explicit VR.
2277 // FIXME: in order to fix the above warning, we could read the next
2278 // element value (or a couple of elements values) in order to make
2279 // sure we are not commiting a big mistake.
2280 // We need to skip :
2281 // * the 128 bytes of File Preamble (often padded with zeroes),
2282 // * the 4 bytes of "DICM" string,
2283 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2284 // i.e. a total of 136 bytes.
2288 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2289 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2290 // *Implicit* VR. -and it is !-
2292 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2293 memcmp(entCur, "OB", (size_t)2) == 0 ||
2294 memcmp(entCur, "UI", (size_t)2) == 0 ||
2295 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2296 // when Write DCM *adds*
2298 // Use Document::dicom_vr to test all the possibilities
2299 // instead of just checking for UL, OB and UI !? group 0000
2301 Filetype = ExplicitVR;
2302 dbg.Verbose(1, "Document::CheckSwap:",
2303 "explicit Value Representation");
2307 Filetype = ImplicitVR;
2308 dbg.Verbose(1, "Document::CheckSwap:",
2309 "not an explicit Value Representation");
2315 dbg.Verbose(1, "Document::CheckSwap:",
2316 "HostByteOrder != NetworkByteOrder");
2321 dbg.Verbose(1, "Document::CheckSwap:",
2322 "HostByteOrder = NetworkByteOrder");
2325 // Position the file position indicator at first tag (i.e.
2326 // after the file preamble and the "DICM" string).
2327 Fp->seekg(0, std::ios::beg);
2328 Fp->seekg ( 132L, std::ios::beg);
2332 // Alas, this is not a DicomV3 file and whatever happens there is no file
2333 // preamble. We can reset the file position indicator to where the data
2334 // is (i.e. the beginning of the file).
2335 dbg.Verbose(1, "Document::CheckSwap:", "not a DICOM Version3 file");
2336 Fp->seekg(0, std::ios::beg);
2338 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2339 // By clean we mean that the length of the first tag is written down.
2340 // If this is the case and since the length of the first group HAS to be
2341 // four (bytes), then determining the proper swap code is straightforward.
2344 // We assume the array of char we are considering contains the binary
2345 // representation of a 32 bits integer. Hence the following dirty
2347 s32 = *((uint32_t *)(entCur));
2368 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2369 // It is time for despaired wild guesses.
2370 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2371 // i.e. the 'group length' element is not present :
2373 // check the supposed-to-be 'group number'
2374 // in ( 0x0001 .. 0x0008 )
2375 // to determine ' SwapCode' value .
2376 // Only 0 or 4321 will be possible
2377 // (no oportunity to check for the formerly well known
2378 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2379 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2380 // the file IS NOT ACR-NEMA nor DICOM V3
2381 // Find a trick to tell it the caller...
2383 s16 = *((uint16_t *)(deb));
2410 dbg.Verbose(0, "Document::CheckSwap:",
2411 "ACR/NEMA unfound swap info (Really hopeless !)");
2415 // Then the only info we have is the net2host one.
2427 * \brief Restore the unproperly loaded values i.e. the group, the element
2428 * and the dictionary entry depending on them.
2430 void Document::SwitchSwapToBigEndian()
2432 dbg.Verbose(1, "Document::SwitchSwapToBigEndian",
2433 "Switching to BigEndian mode.");
2434 if ( SwapCode == 0 )
2438 else if ( SwapCode == 4321 )
2442 else if ( SwapCode == 3412 )
2446 else if ( SwapCode == 2143 )
2453 * \brief during parsing, Header Elements too long are not loaded in memory
2456 void Document::SetMaxSizeLoadEntry(long newSize)
2462 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2464 MaxSizeLoadEntry = 0xffffffff;
2467 MaxSizeLoadEntry = newSize;
2472 * \brief Header Elements too long will not be printed
2473 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2476 void Document::SetMaxSizePrintEntry(long newSize)
2478 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2483 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2485 MaxSizePrintEntry = 0xffffffff;
2488 MaxSizePrintEntry = newSize;
2494 * \brief Handle broken private tag from Philips NTSCAN
2495 * where the endianess is being switch to BigEndian for no
2499 void Document::HandleBrokenEndian(uint16_t group, uint16_t elem)
2501 // Endian reversion. Some files contain groups of tags with reversed endianess.
2502 static int reversedEndian = 0;
2503 // try to fix endian switching in the middle of headers
2504 if ((group == 0xfeff) && (elem == 0x00e0))
2506 // start endian swap mark for group found
2508 SwitchSwapToBigEndian();
2513 else if ((group == 0xfffe) && (elem == 0xe00d) && reversedEndian)
2515 // end of reversed endian group
2517 SwitchSwapToBigEndian();
2522 * \brief Read the next tag but WITHOUT loading it's value
2523 * (read the 'Group Number', the 'Element Number',
2524 * gets the Dict Entry
2525 * gets the VR, gets the length, gets the offset value)
2526 * @return On succes the newly created DocEntry, NULL on failure.
2528 DocEntry* Document::ReadNextDocEntry()
2535 group = ReadInt16();
2538 catch ( FormatError e )
2540 // We reached the EOF (or an error occured) therefore
2541 // header parsing has to be considered as finished.
2546 HandleBrokenEndian(group, elem);
2547 std::string vr=FindDocEntryVR();
2549 DocEntry *newEntry = NewDocEntryByNumber(group, elem, vr);
2550 if( vr == GDCM_UNKNOWN )
2552 if( Filetype == ExplicitVR )
2554 // We thought this was explicit VR, but we end up with an
2555 // implicit VR tag. Let's backtrack.
2557 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
2558 newEntry->GetGroup(), newEntry->GetElement());
2559 dbg.Verbose(1, "Document::FindVR: ", msg.c_str());
2561 newEntry->SetImplicitVR();
2566 FindDocEntryLength(newEntry);
2568 catch ( FormatError e )
2576 newEntry->SetOffset(Fp->tellg());
2583 * \brief Generate a free TagKey i.e. a TagKey that is not present
2584 * in the TagHt dictionary.
2585 * @param group The generated tag must belong to this group.
2586 * @return The element of tag with given group which is fee.
2588 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2590 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2592 TagKey key = DictEntry::TranslateToKey(group, elem);
2593 if (TagHT.count(key) == 0)
2602 * \brief Assuming the internal file pointer \ref Document::Fp
2603 * is placed at the beginning of a tag check whether this
2604 * tag is (TestGroup, TestElement).
2605 * \warning On success the internal file pointer \ref Document::Fp
2606 * is modified to point after the tag.
2607 * On failure (i.e. when the tag wasn't the expected tag
2608 * (TestGroup, TestElement) the internal file pointer
2609 * \ref Document::Fp is restored to it's original position.
2610 * @param testGroup The expected group of the tag.
2611 * @param testElement The expected Element of the tag.
2612 * @return True on success, false otherwise.
2614 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2616 long positionOnEntry = Fp->tellg();
2617 long currentPosition = Fp->tellg(); // On debugging purposes
2619 //// Read the Item Tag group and element, and make
2620 // sure they are what we expected:
2621 uint16_t itemTagGroup;
2622 uint16_t itemTagElement;
2625 itemTagGroup = ReadInt16();
2626 itemTagElement = ReadInt16();
2628 catch ( FormatError e )
2630 //std::cerr << e << std::endl;
2633 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2635 std::ostringstream s;
2636 s << " We should have found tag (";
2637 s << std::hex << testGroup << "," << testElement << ")" << std::endl;
2638 s << " but instead we encountered tag (";
2639 s << std::hex << itemTagGroup << "," << itemTagElement << ")"
2641 s << " at address: " << (unsigned)currentPosition << std::endl;
2642 dbg.Verbose(0, "Document::ReadItemTagLength: wrong Item Tag found:");
2643 dbg.Verbose(0, s.str().c_str());
2644 Fp->seekg(positionOnEntry, std::ios::beg);
2652 * \brief Assuming the internal file pointer \ref Document::Fp
2653 * is placed at the beginning of a tag (TestGroup, TestElement),
2654 * read the length associated to the Tag.
2655 * \warning On success the internal file pointer \ref Document::Fp
2656 * is modified to point after the tag and it's length.
2657 * On failure (i.e. when the tag wasn't the expected tag
2658 * (TestGroup, TestElement) the internal file pointer
2659 * \ref Document::Fp is restored to it's original position.
2660 * @param testGroup The expected group of the tag.
2661 * @param testElement The expected Element of the tag.
2662 * @return On success returns the length associated to the tag. On failure
2665 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2667 long positionOnEntry = Fp->tellg();
2668 (void)positionOnEntry;
2670 if ( !ReadTag(testGroup, testElement) )
2675 //// Then read the associated Item Length
2676 long currentPosition = Fp->tellg();
2677 uint32_t itemLength = ReadInt32();
2679 std::ostringstream s;
2680 s << "Basic Item Length is: "
2681 << itemLength << std::endl;
2682 s << " at address: " << (unsigned)currentPosition << std::endl;
2683 dbg.Verbose(0, "Document::ReadItemTagLength: ", s.str().c_str());
2689 * \brief When parsing the Pixel Data of an encapsulated file, read
2690 * the basic offset table (when present, and BTW dump it).
2692 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2694 //// Read the Basic Offset Table Item Tag length...
2695 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2697 // When present, read the basic offset table itself.
2698 // Notes: - since the presence of this basic offset table is optional
2699 // we can't rely on it for the implementation, and we will simply
2700 // trash it's content (when present).
2701 // - still, when present, we could add some further checks on the
2702 // lengths, but we won't bother with such fuses for the time being.
2703 if ( itemLength != 0 )
2705 char* basicOffsetTableItemValue = new char[itemLength + 1];
2706 Fp->read(basicOffsetTableItemValue, itemLength);
2709 for (unsigned int i=0; i < itemLength; i += 4 )
2711 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2713 std::ostringstream s;
2714 s << " Read one length: ";
2715 s << std::hex << individualLength << std::endl;
2717 "Document::ReadAndSkipEncapsulatedBasicOffsetTable: ",
2722 delete[] basicOffsetTableItemValue;
2727 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2728 * Compute the RLE extra information and store it in \ref RLEInfo
2729 * for later pixel retrieval usage.
2731 void Document::ComputeRLEInfo()
2733 TransferSyntaxType ts = GetTransferSyntax();
2734 if ( ts != RLELossless )
2739 // Encoded pixel data: for the time being we are only concerned with
2740 // Jpeg or RLE Pixel data encodings.
2741 // As stated in PS 3.5-2003, section 8.2 p44:
2742 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2743 // value representation OB is used".
2744 // Hence we expect an OB value representation. Concerning OB VR,
2745 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2746 // "For the Value Representations OB and OW, the encoding shall meet the
2747 // following specifications depending on the Data element tag:"
2749 // - the first item in the sequence of items before the encoded pixel
2750 // data stream shall be basic offset table item. The basic offset table
2751 // item value, however, is not required to be present"
2753 ReadAndSkipEncapsulatedBasicOffsetTable();
2755 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2756 // Loop on the individual frame[s] and store the information
2757 // on the RLE fragments in a RLEFramesInfo.
2758 // Note: - when only a single frame is present, this is a
2760 // - when more than one frame are present, then we are in
2761 // the case of a multi-frame image.
2763 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2765 // Parse the RLE Header and store the corresponding RLE Segment
2766 // Offset Table information on fragments of this current Frame.
2767 // Note that the fragment pixels themselves are not loaded
2768 // (but just skipped).
2769 long frameOffset = Fp->tellg();
2771 uint32_t nbRleSegments = ReadInt32();
2772 if ( nbRleSegments > 16 )
2774 // There should be at most 15 segments (refer to RLEFrame class)
2775 dbg.Verbose(0, "Document::ComputeRLEInfo: too many segments.");
2778 uint32_t rleSegmentOffsetTable[16];
2779 for( int k = 1; k <= 15; k++ )
2781 rleSegmentOffsetTable[k] = ReadInt32();
2784 // Deduce from both the RLE Header and the frameLength the
2785 // fragment length, and again store this info in a
2787 long rleSegmentLength[15];
2788 // skipping (not reading) RLE Segments
2789 if ( nbRleSegments > 1)
2791 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2793 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2794 - rleSegmentOffsetTable[k];
2795 SkipBytes(rleSegmentLength[k]);
2799 rleSegmentLength[nbRleSegments] = frameLength
2800 - rleSegmentOffsetTable[nbRleSegments];
2801 SkipBytes(rleSegmentLength[nbRleSegments]);
2803 // Store the collected info
2804 RLEFrame* newFrameInfo = new RLEFrame;
2805 newFrameInfo->NumberFragments = nbRleSegments;
2806 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2808 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2809 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2811 RLEInfo->Frames.push_back( newFrameInfo );
2814 // Make sure that at the end of the item we encounter a 'Sequence
2816 if ( !ReadTag(0xfffe, 0xe0dd) )
2818 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2819 dbg.Verbose(0, " item at end of RLE item sequence");
2824 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2825 * Compute the jpeg extra information (fragment[s] offset[s] and
2826 * length) and store it[them] in \ref JPEGInfo for later pixel
2829 void Document::ComputeJPEGFragmentInfo()
2831 // If you need to, look for comments of ComputeRLEInfo().
2837 ReadAndSkipEncapsulatedBasicOffsetTable();
2839 // Loop on the fragments[s] and store the parsed information in a
2841 long fragmentLength;
2842 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2844 long fragmentOffset = Fp->tellg();
2846 // Store the collected info
2847 JPEGFragment* newFragment = new JPEGFragment;
2848 newFragment->Offset = fragmentOffset;
2849 newFragment->Length = fragmentLength;
2850 JPEGInfo->Fragments.push_back( newFragment );
2852 SkipBytes( fragmentLength );
2855 // Make sure that at the end of the item we encounter a 'Sequence
2857 if ( !ReadTag(0xfffe, 0xe0dd) )
2859 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2860 dbg.Verbose(0, " item at end of JPEG item sequence");
2865 * \brief Walk recursively the given \ref DocEntrySet, and feed
2866 * the given hash table (\ref TagDocEntryHT) with all the
2867 * \ref DocEntry (Dicom entries) encountered.
2868 * This method does the job for \ref BuildFlatHashTable.
2869 * @param builtHT Where to collect all the \ref DocEntry encountered
2870 * when recursively walking the given set.
2871 * @param set The structure to be traversed (recursively).
2873 void Document::BuildFlatHashTableRecurse( TagDocEntryHT& builtHT,
2876 if (ElementSet* elementSet = dynamic_cast< ElementSet* > ( set ) )
2878 TagDocEntryHT const & currentHT = elementSet->GetTagHT();
2879 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2880 i != currentHT.end();
2883 DocEntry* entry = i->second;
2884 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
2886 const ListSQItem& items = seqEntry->GetSQItems();
2887 for( ListSQItem::const_iterator item = items.begin();
2888 item != items.end();
2891 BuildFlatHashTableRecurse( builtHT, *item );
2895 builtHT[entry->GetKey()] = entry;
2900 if (SQItem* SQItemSet = dynamic_cast< SQItem* > ( set ) )
2902 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
2903 for (ListDocEntry::const_iterator i = currentList.begin();
2904 i != currentList.end();
2907 DocEntry* entry = *i;
2908 if ( SeqEntry* seqEntry = dynamic_cast<SeqEntry*>(entry) )
2910 const ListSQItem& items = seqEntry->GetSQItems();
2911 for( ListSQItem::const_iterator item = items.begin();
2912 item != items.end();
2915 BuildFlatHashTableRecurse( builtHT, *item );
2919 builtHT[entry->GetKey()] = entry;
2926 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2929 * The structure used by a Document (through \ref ElementSet),
2930 * in order to hold the parsed entries of a Dicom header, is a recursive
2931 * one. This is due to the fact that the sequences (when present)
2932 * can be nested. Additionaly, the sequence items (represented in
2933 * gdcm as \ref SQItem) add an extra complexity to the data
2934 * structure. Hence, a gdcm user whishing to visit all the entries of
2935 * a Dicom header will need to dig in the gdcm internals (which
2936 * implies exposing all the internal data structures to the API).
2937 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2938 * recursively builds a temporary hash table, which holds all the
2939 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
2941 * \warning Of course there is NO integrity constrain between the
2942 * returned \ref TagDocEntryHT and the \ref ElementSet used
2943 * to build it. Hence if the underlying \ref ElementSet is
2944 * altered, then it is the caller responsability to invoke
2945 * \ref BuildFlatHashTable again...
2946 * @return The flat std::map<> we juste build.
2948 TagDocEntryHT* Document::BuildFlatHashTable()
2950 TagDocEntryHT* FlatHT = new TagDocEntryHT;
2951 BuildFlatHashTableRecurse( *FlatHT, this );
2958 * \brief Compares two documents, according to \ref DicomDir rules
2959 * \warning Does NOT work with ACR-NEMA files
2960 * \todo Find a trick to solve the pb (use RET fields ?)
2962 * @return true if 'smaller'
2964 bool Document::operator<(Document &document)
2967 std::string s1 = GetEntryByNumber(0x0010,0x0010);
2968 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
2980 s1 = GetEntryByNumber(0x0010,0x0020);
2981 s2 = document.GetEntryByNumber(0x0010,0x0020);
2992 // Study Instance UID
2993 s1 = GetEntryByNumber(0x0020,0x000d);
2994 s2 = document.GetEntryByNumber(0x0020,0x000d);
3005 // Serie Instance UID
3006 s1 = GetEntryByNumber(0x0020,0x000e);
3007 s2 = document.GetEntryByNumber(0x0020,0x000e);
3024 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
3025 * @param filetype Type of the File to be written
3027 int Document::ComputeGroup0002Length( FileType filetype )
3032 int groupLength = 0;
3033 bool found0002 = false;
3035 // for each zero-level Tag in the DCM Header
3039 entry = GetNextEntry();
3042 gr = entry->GetGroup();
3048 el = entry->GetElement();
3049 vr = entry->GetVR();
3051 if (filetype == ExplicitVR)
3053 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
3055 groupLength += 4; // explicit VR AND OB, OW, SQ : 4 more bytes
3058 groupLength += 2 + 2 + 4 + entry->GetLength();
3060 else if (found0002 )
3063 entry = GetNextEntry();
3068 } // end namespace gdcm
3070 //-----------------------------------------------------------------------------