1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/06 20:03:27 $
7 Version: $Revision: 1.161 $
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 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 Doc 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(std::string const &value,
552 ValEntry *valEntry = 0;
553 DocEntry *currentEntry = GetDocEntryByNumber( group, elem);
557 valEntry = dynamic_cast< ValEntry* >(currentEntry);
561 if( valEntry->GetVR()!=vr )
564 // if currentEntry doesn't correspond to the requested valEntry
567 if (!RemoveEntry(currentEntry))
569 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
570 " of previous DocEntry failed.");
577 // Create a new valEntry if necessary
580 valEntry = NewValEntryByNumber(group, elem, vr);
582 if ( !AddEntry(valEntry))
584 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
585 " failed allthough this is a creation.");
592 // Set the binEntry value
593 SetEntry(value, valEntry);
598 * \brief Modifies the value of a given Header Entry (Dicom Element)
599 * when it exists. Create it with the given value when unexistant.
600 * A copy of the binArea is made to be kept in the Document.
601 * @param binArea (binary) value to be set
602 * @param Group Group number of the Entry
603 * @param Elem Element number of the Entry
604 * @param vr V(alue) R(epresentation) of the Entry -if private Entry-
605 * \return pointer to the modified/created Header Entry (NULL when creation
608 BinEntry *Document::ReplaceOrCreateByNumber(uint8_t *binArea,
614 BinEntry *binEntry = 0;
615 DocEntry *currentEntry = GetDocEntryByNumber( group, elem);
617 // Verify the currentEntry
620 binEntry = dynamic_cast< BinEntry* >(currentEntry);
624 if( binEntry->GetVR()!=vr )
627 // if currentEntry doesn't correspond to the requested valEntry
630 if (!RemoveEntry(currentEntry))
632 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
633 " of previous DocEntry failed.");
640 // Create a new binEntry if necessary
643 binEntry = NewBinEntryByNumber(group, elem, vr);
645 if ( !AddEntry(binEntry))
647 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
648 " failed allthough this is a creation.");
655 // Set the binEntry value
657 if (lgth>0 && binArea)
659 tmpArea = new uint8_t[lgth];
660 memcpy(tmpArea,binArea,lgth);
666 if (!SetEntry(tmpArea,lgth,binEntry))
678 * \brief Modifies the value of a given Header Entry (Dicom Element)
679 * when it exists. Create it when unexistant.
680 * @param Group Group number of the Entry
681 * @param Elem Element number of the Entry
682 * \return pointer to the modified/created SeqEntry (NULL when creation
685 SeqEntry *Document::ReplaceOrCreateByNumber( uint16_t group, uint16_t elem)
687 SeqEntry *seqEntry = 0;
688 DocEntry *currentEntry = GetDocEntryByNumber( group, elem);
690 // Verify the currentEntry
693 seqEntry = dynamic_cast< SeqEntry* >(currentEntry);
697 if( seqEntry->GetVR()!="SQ" )
700 // if currentEntry doesn't correspond to the requested valEntry
703 if (!RemoveEntry(currentEntry))
705 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: removal"
706 " of previous DocEntry failed.");
713 // Create a new seqEntry if necessary
716 seqEntry = NewSeqEntryByNumber(group, elem);
718 if ( !AddEntry(seqEntry))
720 dbg.Verbose(0, "Document::ReplaceOrCreateByNumber: AddEntry"
721 " failed allthough this is a creation.");
732 * \brief Set a new value if the invoked element exists
733 * Seems to be useless !!!
734 * @param value new element value
735 * @param group group number of the Entry
736 * @param elem element number of the Entry
739 bool Document::ReplaceIfExistByNumber(std::string const &value,
740 uint16_t group, uint16_t elem )
742 SetEntryByNumber(value, group, elem);
747 std::string Document::GetTransferSyntaxValue(TransferSyntaxType type)
749 return TransferSyntaxStrings[type];
752 //-----------------------------------------------------------------------------
756 * \brief Checks if a given Dicom Element exists within the H table
757 * @param group Group number of the searched Dicom Element
758 * @param element Element number of the searched Dicom Element
759 * @return true is found
761 bool Document::CheckIfEntryExistByNumber(uint16_t group, uint16_t element )
763 const std::string &key = DictEntry::TranslateToKey(group, element );
764 return TagHT.count(key) != 0;
768 * \brief Searches within Header Entries (Dicom Elements) parsed with
769 * the public and private dictionaries
770 * for the element value of a given tag.
771 * \warning Don't use any longer : use GetPubEntryByName
772 * @param tagName name of the searched element.
773 * @return Corresponding element value when it exists,
774 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
776 std::string Document::GetEntryByName(TagName const &tagName)
778 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
784 return GetEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
788 * \brief Searches within Header Entries (Dicom Elements) parsed with
789 * the public and private dictionaries
790 * for the element value representation of a given tag.
792 * Obtaining the VR (Value Representation) might be needed by caller
793 * to convert the string typed content to caller's native type
794 * (think of C++ vs Python). The VR is actually of a higher level
795 * of semantics than just the native C++ type.
796 * @param tagName name of the searched element.
797 * @return Corresponding element value representation when it exists,
798 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
800 std::string Document::GetEntryVRByName(TagName const& tagName)
802 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
803 if( dictEntry == NULL)
808 DocEntry *elem = GetDocEntryByNumber(dictEntry->GetGroup(),
809 dictEntry->GetElement());
810 return elem->GetVR();
814 * \brief Searches within Header Entries (Dicom Elements) parsed with
815 * the public and private dictionaries
816 * for the element value representation of a given tag.
817 * @param group Group number of the searched tag.
818 * @param element Element number of the searched tag.
819 * @return Corresponding element value representation when it exists,
820 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
822 std::string Document::GetEntryByNumber(uint16_t group, uint16_t element)
824 TagKey key = DictEntry::TranslateToKey(group, element);
825 if ( !TagHT.count(key))
830 return ((ValEntry *)TagHT.find(key)->second)->GetValue();
834 * \brief Searches within Header Entries (Dicom Elements) parsed with
835 * the public and private dictionaries
836 * for the element value representation of a given tag..
838 * Obtaining the VR (Value Representation) might be needed by caller
839 * to convert the string typed content to caller's native type
840 * (think of C++ vs Python). The VR is actually of a higher level
841 * of semantics than just the native C++ type.
842 * @param group Group number of the searched tag.
843 * @param element Element number of the searched tag.
844 * @return Corresponding element value representation when it exists,
845 * and the string GDCM_UNFOUND ("gdcm::Unfound") otherwise.
847 std::string Document::GetEntryVRByNumber(uint16_t group, uint16_t element)
849 DocEntry *elem = GetDocEntryByNumber(group, element);
854 return elem->GetVR();
858 * \brief Searches within Header Entries (Dicom Elements) parsed with
859 * the public and private dictionaries
860 * for the value length of a given tag..
861 * @param group Group number of the searched tag.
862 * @param element Element number of the searched tag.
863 * @return Corresponding element length; -2 if not found
865 int Document::GetEntryLengthByNumber(uint16_t group, uint16_t element)
867 DocEntry *elem = GetDocEntryByNumber(group, element);
870 return -2; //magic number
872 return elem->GetLength();
875 * \brief Sets the value (string) of the Header Entry (Dicom Element)
876 * @param content string value of the Dicom Element
877 * @param tagName name of the searched Dicom Element.
878 * @return true when found
880 bool Document::SetEntryByName( std::string const &content,
881 TagName const &tagName)
883 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
889 return SetEntryByNumber(content,dictEntry->GetGroup(),
890 dictEntry->GetElement());
894 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
895 * through it's (group, element) and modifies it's content with
897 * @param content new value (string) to substitute with
898 * @param group group number of the Dicom Element to modify
899 * @param element element number of the Dicom Element to modify
901 bool Document::SetEntryByNumber(std::string const& content,
902 uint16_t group, uint16_t element)
904 ValEntry *entry = GetValEntryByNumber(group, element);
907 dbg.Verbose(0, "Document::SetEntryByNumber: no corresponding",
908 " ValEntry (try promotion first).");
911 return SetEntry(content,entry);
915 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
916 * through it's (group, element) and modifies it's content with
918 * @param content new value (void* -> uint8_t*) to substitute with
919 * @param lgth new value length
920 * @param group group number of the Dicom Element to modify
921 * @param element element number of the Dicom Element to modify
923 bool Document::SetEntryByNumber(uint8_t*content, int lgth,
924 uint16_t group, uint16_t element)
926 BinEntry *entry = GetBinEntryByNumber(group, element);
929 dbg.Verbose(0, "Document::SetEntryByNumber: no corresponding",
930 " ValEntry (try promotion first).");
934 return SetEntry(content,lgth,entry);
938 * \brief Accesses an existing DocEntry (i.e. a Dicom Element)
939 * and modifies it's content with the given value.
940 * @param content new value (string) to substitute with
941 * @param entry Entry to be modified
943 bool Document::SetEntry(std::string const &content,ValEntry *entry)
947 entry->SetValue(content);
954 * \brief Accesses an existing BinEntry (i.e. a Dicom Element)
955 * and modifies it's content with the given value.
956 * @param content new value (void* -> uint8_t*) to substitute with
957 * @param entry Entry to be modified
958 * @param lgth new value length
960 bool Document::SetEntry(uint8_t *content, int lgth, BinEntry *entry)
964 // Hope Binary field length is *never* wrong
965 /*if(lgth%2) // Non even length are padded with a space (020H).
968 //content = content + '\0'; // fing a trick to enlarge a binary field?
971 entry->SetBinArea(content);
972 entry->SetLength(lgth);
973 entry->SetValue(GDCM_BINLOADED);
980 * \brief Gets (from Header) a 'non string' element value
981 * (LoadElementValues has already be executed)
982 * @param group group number of the Entry
983 * @param elem element number of the Entry
984 * @return Pointer to the 'non string' area
986 void *Document::GetEntryBinAreaByNumber(uint16_t group, uint16_t elem)
988 DocEntry *entry = GetDocEntryByNumber(group, elem);
991 dbg.Verbose(1, "Document::GetDocEntryByNumber: no entry");
994 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(entry) )
996 return binEntry->GetBinArea();
1003 * \brief Loads (from disk) the element content
1004 * when a string is not suitable
1005 * @param group group number of the Entry
1006 * @param elem element number of the Entry
1008 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
1010 // Search the corresponding DocEntry
1011 DocEntry *docElement = GetDocEntryByNumber(group, elem);
1015 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
1019 LoadEntryBinArea(binElement);
1023 * \brief Loads (from disk) the element content
1024 * when a string is not suitable
1025 * @param element Entry whose binArea is going to be loaded
1027 void Document::LoadEntryBinArea(BinEntry *element)
1029 if(element->GetBinArea())
1032 bool openFile = !Fp;
1036 size_t o =(size_t)element->GetOffset();
1037 Fp->seekg(o, std::ios::beg);
1039 size_t l = element->GetLength();
1040 uint8_t *a = new uint8_t[l];
1043 dbg.Verbose(0, "Document::LoadEntryBinArea cannot allocate a");
1047 /// \todo check the result
1048 Fp->read((char*)a, l);
1049 if( Fp->fail() || Fp->eof()) //Fp->gcount() == 1
1055 element->SetBinArea(a);
1062 * \brief Sets a 'non string' value to a given Dicom Element
1063 * @param area area containing the 'non string' value
1064 * @param group Group number of the searched Dicom Element
1065 * @param element Element number of the searched Dicom Element
1068 /*bool Document::SetEntryBinAreaByNumber(uint8_t *area,
1069 uint16_t group, uint16_t element)
1071 DocEntry *currentEntry = GetDocEntryByNumber(group, element);
1072 if ( !currentEntry )
1077 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(currentEntry) )
1079 binEntry->SetBinArea( area );
1087 * \brief Searches within the Header Entries for a Dicom Element of
1089 * @param tagName name of the searched Dicom Element.
1090 * @return Corresponding Dicom Element when it exists, and NULL
1093 DocEntry *Document::GetDocEntryByName(TagName const &tagName)
1095 DictEntry *dictEntry = RefPubDict->GetDictEntryByName(tagName);
1101 return GetDocEntryByNumber(dictEntry->GetGroup(),dictEntry->GetElement());
1105 * \brief retrieves a Dicom Element (the first one) using (group, element)
1106 * \warning (group, element) IS NOT an identifier inside the Dicom Header
1107 * if you think it's NOT UNIQUE, check the count number
1108 * and use iterators to retrieve ALL the Dicoms Elements within
1109 * a given couple (group, element)
1110 * @param group Group number of the searched Dicom Element
1111 * @param element Element number of the searched Dicom Element
1114 DocEntry *Document::GetDocEntryByNumber(uint16_t group, uint16_t element)
1116 TagKey key = DictEntry::TranslateToKey(group, element);
1117 if ( !TagHT.count(key))
1121 return TagHT.find(key)->second;
1125 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1126 * returns a result when the corresponding entry is of type
1128 * @return When present, the corresponding ValEntry.
1130 ValEntry *Document::GetValEntryByNumber(uint16_t group, uint16_t element)
1132 DocEntry *currentEntry = GetDocEntryByNumber(group, element);
1133 if ( !currentEntry )
1137 if ( ValEntry *entry = dynamic_cast<ValEntry*>(currentEntry) )
1141 dbg.Verbose(0, "Document::GetValEntryByNumber: unfound ValEntry.");
1147 * \brief Same as \ref Document::GetDocEntryByNumber except it only
1148 * returns a result when the corresponding entry is of type
1150 * @return When present, the corresponding BinEntry.
1152 BinEntry *Document::GetBinEntryByNumber(uint16_t group, uint16_t element)
1154 DocEntry *currentEntry = GetDocEntryByNumber(group, element);
1155 if ( !currentEntry )
1159 if ( BinEntry *entry = dynamic_cast<BinEntry*>(currentEntry) )
1163 dbg.Verbose(0, "Document::GetBinEntryByNumber: unfound BinEntry.");
1169 * \brief Loads the element while preserving the current
1170 * underlying file position indicator as opposed to
1171 * to LoadDocEntry that modifies it.
1172 * @param entry Header Entry whose value shall be loaded.
1175 void Document::LoadDocEntrySafe(DocEntry *entry)
1179 long PositionOnEntry = Fp->tellg();
1180 LoadDocEntry(entry);
1181 Fp->seekg(PositionOnEntry, std::ios::beg);
1186 * \brief Swaps back the bytes of 4-byte long integer accordingly to
1188 * @return The properly swaped 32 bits integer.
1190 uint32_t Document::SwapLong(uint32_t a)
1197 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
1198 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
1202 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
1206 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
1209 //std::cout << "swapCode= " << SwapCode << std::endl;
1210 dbg.Error(" Document::SwapLong : unset swap code");
1217 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
1219 * @return The properly unswaped 32 bits integer.
1221 uint32_t Document::UnswapLong(uint32_t a)
1227 * \brief Swaps the bytes so they agree with the processor order
1228 * @return The properly swaped 16 bits integer.
1230 uint16_t Document::SwapShort(uint16_t a)
1232 if ( SwapCode == 4321 || SwapCode == 2143 )
1234 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
1240 * \brief Unswaps the bytes so they agree with the processor order
1241 * @return The properly unswaped 16 bits integer.
1243 uint16_t Document::UnswapShort(uint16_t a)
1245 return SwapShort(a);
1248 //-----------------------------------------------------------------------------
1252 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1253 * @return length of the parsed set.
1255 void Document::ParseDES(DocEntrySet *set, long offset,
1256 long l_max, bool delim_mode)
1258 DocEntry *newDocEntry = 0;
1259 ValEntry *newValEntry;
1260 BinEntry *newBinEntry;
1261 SeqEntry *newSeqEntry;
1264 long offsetEntry,readEntry;
1268 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1274 newDocEntry = ReadNextDocEntry( );
1280 vr = newDocEntry->GetVR();
1281 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
1282 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
1283 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1285 if ( newValEntry || newBinEntry )
1287 offsetEntry=newDocEntry->GetOffset();
1288 readEntry=newDocEntry->GetReadLength();
1292 if ( ! Global::GetVR()->IsVROfBinaryRepresentable(vr) )
1294 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1295 dbg.Verbose(0, "Document::ParseDES: neither Valentry, "
1296 "nor BinEntry. Probably unknown VR.");
1299 //////////////////// BinEntry or UNKOWN VR:
1300 // When "this" is a Document the Key is simply of the
1301 // form ( group, elem )...
1302 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1305 newBinEntry->SetKey( newBinEntry->GetKey() );
1307 // but when "this" is a SQItem, we are inserting this new
1308 // valEntry in a sequence item, and the kay has the
1309 // generalized form (refer to \ref BaseTagKey):
1310 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1312 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1313 + newBinEntry->GetKey() );
1316 LoadDocEntry( newBinEntry );
1317 if( !set->AddEntry( newBinEntry ) )
1319 //Expect big troubles if here
1320 //delete newBinEntry;
1326 /////////////////////// ValEntry
1327 // When "set" is a Document, then we are at the top of the
1328 // hierarchy and the Key is simply of the form ( group, elem )...
1329 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1332 newValEntry->SetKey( newValEntry->GetKey() );
1334 // ...but when "set" is a SQItem, we are inserting this new
1335 // valEntry in a sequence item. Hence the key has the
1336 // generalized form (refer to \ref BaseTagKey):
1337 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1339 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1340 + newValEntry->GetKey() );
1343 LoadDocEntry( newValEntry );
1344 bool delimitor=newValEntry->IsItemDelimitor();
1345 if( !set->AddEntry( newValEntry ) )
1347 // If here expect big troubles
1348 //delete newValEntry; //otherwise mem leak
1358 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1366 if ( ( newDocEntry->GetGroup() == 0x7fe0 )
1367 && ( newDocEntry->GetElement() == 0x0010 ) )
1369 TransferSyntaxType ts = GetTransferSyntax();
1370 if ( ts == RLELossless )
1372 long positionOnEntry = Fp->tellg();
1373 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1375 Fp->seekg( positionOnEntry, std::ios::beg );
1377 else if ( IsJPEG() )
1379 long positionOnEntry = Fp->tellg();
1380 Fp->seekg( newDocEntry->GetOffset(), std::ios::beg );
1381 ComputeJPEGFragmentInfo();
1382 Fp->seekg( positionOnEntry, std::ios::beg );
1386 // Just to make sure we are at the beginning of next entry.
1387 SkipToNextDocEntry(offsetEntry,readEntry);
1392 unsigned long l = newDocEntry->GetReadLength();
1393 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
1395 if ( l == 0xffffffff )
1404 // no other way to create it ...
1405 newSeqEntry->SetDelimitorMode( delim_mode );
1407 // At the top of the hierarchy, stands a Document. When "set"
1408 // is a Document, then we are building the first depth level.
1409 // Hence the SeqEntry we are building simply has a depth
1411 if (Document *dummy = dynamic_cast< Document* > ( set ) )
1414 newSeqEntry->SetDepthLevel( 1 );
1415 newSeqEntry->SetKey( newSeqEntry->GetKey() );
1417 // But when "set" is allready a SQItem, we are building a nested
1418 // sequence, and hence the depth level of the new SeqEntry
1419 // we are building, is one level deeper:
1420 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1422 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1423 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1424 + newSeqEntry->GetKey() );
1428 { // Don't try to parse zero-length sequences
1429 ParseSQ( newSeqEntry,
1430 newDocEntry->GetOffset(),
1433 set->AddEntry( newSeqEntry );
1434 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1446 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1447 * @return parsed length for this level
1449 void Document::ParseSQ( SeqEntry *seqEntry,
1450 long offset, long l_max, bool delim_mode)
1452 int SQItemNumber = 0;
1457 DocEntry *newDocEntry = ReadNextDocEntry();
1460 // FIXME Should warn user
1465 if ( newDocEntry->IsSequenceDelimitor() )
1467 seqEntry->SetSequenceDelimitationItem( newDocEntry );
1471 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1477 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1478 std::ostringstream newBase;
1479 newBase << seqEntry->GetKey()
1483 itemSQ->SetBaseTagKey( newBase.str() );
1484 unsigned int l = newDocEntry->GetReadLength();
1486 if ( l == 0xffffffff )
1495 ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
1498 seqEntry->AddEntry( itemSQ, SQItemNumber );
1500 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1508 * \brief Loads the element content if its length doesn't exceed
1509 * the value specified with Document::SetMaxSizeLoadEntry()
1510 * @param entry Header Entry (Dicom Element) to be dealt with
1512 void Document::LoadDocEntry(DocEntry *entry)
1514 uint16_t group = entry->GetGroup();
1515 std::string vr = entry->GetVR();
1516 uint32_t length = entry->GetLength();
1518 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1520 // A SeQuence "contains" a set of Elements.
1521 // (fffe e000) tells us an Element is beginning
1522 // (fffe e00d) tells us an Element just ended
1523 // (fffe e0dd) tells us the current SeQuence just ended
1524 if( group == 0xfffe )
1526 // NO more value field for SQ !
1530 // When the length is zero things are easy:
1533 ((ValEntry *)entry)->SetValue("");
1537 // The elements whose length is bigger than the specified upper bound
1538 // are not loaded. Instead we leave a short notice of the offset of
1539 // the element content and it's length.
1541 std::ostringstream s;
1542 if (length > MaxSizeLoadEntry)
1544 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1546 //s << "gdcm::NotLoaded (BinEntry)";
1547 s << GDCM_NOTLOADED;
1548 s << " Address:" << (long)entry->GetOffset();
1549 s << " Length:" << entry->GetLength();
1550 s << " x(" << std::hex << entry->GetLength() << ")";
1551 binEntryPtr->SetValue(s.str());
1553 // Be carefull : a BinEntry IS_A ValEntry ...
1554 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1556 // s << "gdcm::NotLoaded. (ValEntry)";
1557 s << GDCM_NOTLOADED;
1558 s << " Address:" << (long)entry->GetOffset();
1559 s << " Length:" << entry->GetLength();
1560 s << " x(" << std::hex << entry->GetLength() << ")";
1561 valEntryPtr->SetValue(s.str());
1566 std::cout<< "MaxSizeLoadEntry exceeded, neither a BinEntry "
1567 << "nor a ValEntry ?! Should never print that !" << std::endl;
1570 // to be sure we are at the end of the value ...
1571 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1576 // When we find a BinEntry not very much can be done :
1577 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1579 s << GDCM_BINLOADED;
1580 binEntryPtr->SetValue(s.str());
1581 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1585 /// \todo Any compacter code suggested (?)
1586 if ( IsDocEntryAnInteger(entry) )
1590 // When short integer(s) are expected, read and convert the following
1591 // n *two characters properly i.e. consider them as short integers as
1592 // opposed to strings.
1593 // Elements with Value Multiplicity > 1
1594 // contain a set of integers (not a single one)
1595 if (vr == "US" || vr == "SS")
1598 NewInt = ReadInt16();
1602 for (int i=1; i < nbInt; i++)
1605 NewInt = ReadInt16();
1610 // See above comment on multiple integers (mutatis mutandis).
1611 else if (vr == "UL" || vr == "SL")
1614 NewInt = ReadInt32();
1618 for (int i=1; i < nbInt; i++)
1621 NewInt = ReadInt32();
1626 #ifdef GDCM_NO_ANSI_STRING_STREAM
1627 s << std::ends; // to avoid oddities on Solaris
1628 #endif //GDCM_NO_ANSI_STRING_STREAM
1630 ((ValEntry *)entry)->SetValue(s.str());
1634 // FIXME: We need an additional byte for storing \0 that is not on disk
1635 char *str = new char[length+1];
1636 Fp->read(str, (size_t)length);
1637 str[length] = '\0'; //this is only useful when length is odd
1638 // Special DicomString call to properly handle \0 and even length
1639 std::string newValue;
1642 newValue = Util::DicomString(str, length+1);
1643 //dbg.Verbose(0, "Warning: bad length: ", length );
1644 dbg.Verbose(0, "For string :", newValue.c_str());
1645 // Since we change the length of string update it length
1646 entry->SetReadLength(length+1);
1650 newValue = Util::DicomString(str, length);
1654 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1656 if ( Fp->fail() || Fp->eof())//Fp->gcount() == 1
1658 dbg.Verbose(1, "Document::LoadDocEntry",
1659 "unread element value");
1660 valEntry->SetValue(GDCM_UNREAD);
1666 // Because of correspondance with the VR dic
1667 valEntry->SetValue(newValue);
1671 valEntry->SetValue(newValue);
1676 dbg.Error(true, "Document::LoadDocEntry"
1677 "Should have a ValEntry, here !");
1683 * \brief Find the value Length of the passed Header Entry
1684 * @param entry Header Entry whose length of the value shall be loaded.
1686 void Document::FindDocEntryLength( DocEntry *entry )
1687 throw ( FormatError )
1689 uint16_t element = entry->GetElement();
1690 std::string vr = entry->GetVR();
1693 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1695 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1697 // The following reserved two bytes (see PS 3.5-2003, section
1698 // "7.1.2 Data element structure with explicit vr", p 27) must be
1699 // skipped before proceeding on reading the length on 4 bytes.
1700 Fp->seekg( 2L, std::ios::cur);
1701 uint32_t length32 = ReadInt32();
1703 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1708 /// \todo rename that to FindDocEntryLengthOBOrOW since
1709 /// the above test is on both OB and OW...
1710 lengthOB = FindDocEntryLengthOB();
1712 catch ( FormatUnexpected )
1714 // Computing the length failed (this happens with broken
1715 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1716 // chance to get the pixels by deciding the element goes
1717 // until the end of the file. Hence we artificially fix the
1718 // the length and proceed.
1719 long currentPosition = Fp->tellg();
1720 Fp->seekg(0L,std::ios::end);
1721 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1722 Fp->seekg(currentPosition, std::ios::beg);
1723 entry->SetLength(lengthUntilEOF);
1726 entry->SetLength(lengthOB);
1729 FixDocEntryFoundLength(entry, length32);
1733 // Length is encoded on 2 bytes.
1734 length16 = ReadInt16();
1736 // We can tell the current file is encoded in big endian (like
1737 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1738 // and it's value is the one of the encoding of a big endian file.
1739 // In order to deal with such big endian encoded files, we have
1740 // (at least) two strategies:
1741 // * when we load the "Transfer Syntax" tag with value of big endian
1742 // encoding, we raise the proper flags. Then we wait for the end
1743 // of the META group (0x0002) among which is "Transfer Syntax",
1744 // before switching the swap code to big endian. We have to postpone
1745 // the switching of the swap code since the META group is fully encoded
1746 // in little endian, and big endian coding only starts at the next
1747 // group. The corresponding code can be hard to analyse and adds
1748 // many additional unnecessary tests for regular tags.
1749 // * the second strategy consists in waiting for trouble, that shall
1750 // appear when we find the first group with big endian encoding. This
1751 // is easy to detect since the length of a "Group Length" tag (the
1752 // ones with zero as element number) has to be of 4 (0x0004). When we
1753 // encounter 1024 (0x0400) chances are the encoding changed and we
1754 // found a group with big endian encoding.
1755 // We shall use this second strategy. In order to make sure that we
1756 // can interpret the presence of an apparently big endian encoded
1757 // length of a "Group Length" without committing a big mistake, we
1758 // add an additional check: we look in the already parsed elements
1759 // for the presence of a "Transfer Syntax" whose value has to be "big
1760 // endian encoding". When this is the case, chances are we have got our
1761 // hands on a big endian encoded file: we switch the swap code to
1762 // big endian and proceed...
1763 if ( element == 0x0000 && length16 == 0x0400 )
1765 TransferSyntaxType ts = GetTransferSyntax();
1766 if ( ts != ExplicitVRBigEndian )
1768 throw FormatError( "Document::FindDocEntryLength()",
1769 " not explicit VR." );
1773 SwitchSwapToBigEndian();
1774 // Restore the unproperly loaded values i.e. the group, the element
1775 // and the dictionary entry depending on them.
1776 uint16_t correctGroup = SwapShort( entry->GetGroup() );
1777 uint16_t correctElem = SwapShort( entry->GetElement() );
1778 DictEntry *newTag = GetDictEntryByNumber( correctGroup,
1782 // This correct tag is not in the dictionary. Create a new one.
1783 newTag = NewVirtualDictEntry(correctGroup, correctElem);
1785 // FIXME this can create a memory leaks on the old entry that be
1786 // left unreferenced.
1787 entry->SetDictEntry( newTag );
1790 // Heuristic: well, some files are really ill-formed.
1791 if ( length16 == 0xffff)
1793 // 0xffff means that we deal with 'Unknown Length' Sequence
1796 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1801 // Either implicit VR or a non DICOM conformal (see note below) explicit
1802 // VR that ommited the VR of (at least) this element. Farts happen.
1803 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1804 // on Data elements "Implicit and Explicit VR Data Elements shall
1805 // not coexist in a Data Set and Data Sets nested within it".]
1806 // Length is on 4 bytes.
1808 FixDocEntryFoundLength( entry, ReadInt32() );
1814 * \brief Find the Value Representation of the current Dicom Element.
1815 * @return Value Representation of the current Entry
1817 std::string Document::FindDocEntryVR()
1819 if ( Filetype != ExplicitVR )
1820 return(GDCM_UNKNOWN);
1822 long positionOnEntry = Fp->tellg();
1823 // Warning: we believe this is explicit VR (Value Representation) because
1824 // we used a heuristic that found "UL" in the first tag. Alas this
1825 // doesn't guarantee that all the tags will be in explicit VR. In some
1826 // cases (see e-film filtered files) one finds implicit VR tags mixed
1827 // within an explicit VR file. Hence we make sure the present tag
1828 // is in explicit VR and try to fix things if it happens not to be
1832 Fp->read (vr, (size_t)2);
1835 if( !CheckDocEntryVR(vr) )
1837 Fp->seekg(positionOnEntry, std::ios::beg);
1838 return(GDCM_UNKNOWN);
1844 * \brief Check the correspondance between the VR of the header entry
1845 * and the taken VR. If they are different, the header entry is
1846 * updated with the new VR.
1847 * @param vr Dicom Value Representation
1848 * @return false if the VR is incorrect of if the VR isn't referenced
1849 * otherwise, it returns true
1851 bool Document::CheckDocEntryVR(VRKey vr)
1853 // CLEANME searching the dicom_vr at each occurence is expensive.
1854 // PostPone this test in an optional integrity check at the end
1855 // of parsing or only in debug mode.
1856 if ( !Global::GetVR()->IsValidVR(vr) )
1863 * \brief Get the transformed value of the header entry. The VR value
1864 * is used to define the transformation to operate on the value
1865 * \warning NOT end user intended method !
1866 * @param entry entry to tranform
1867 * @return Transformed entry value
1869 std::string Document::GetDocEntryValue(DocEntry *entry)
1871 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1873 std::string val = ((ValEntry *)entry)->GetValue();
1874 std::string vr = entry->GetVR();
1875 uint32_t length = entry->GetLength();
1876 std::ostringstream s;
1879 // When short integer(s) are expected, read and convert the following
1880 // n * 2 bytes properly i.e. as a multivaluated strings
1881 // (each single value is separated fromthe next one by '\'
1882 // as usual for standard multivaluated filels
1883 // Elements with Value Multiplicity > 1
1884 // contain a set of short integers (not a single one)
1886 if( vr == "US" || vr == "SS" )
1891 for (int i=0; i < nbInt; i++)
1897 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1898 newInt16 = SwapShort( newInt16 );
1903 // When integer(s) are expected, read and convert the following
1904 // n * 4 bytes properly i.e. as a multivaluated strings
1905 // (each single value is separated fromthe next one by '\'
1906 // as usual for standard multivaluated filels
1907 // Elements with Value Multiplicity > 1
1908 // contain a set of integers (not a single one)
1909 else if( vr == "UL" || vr == "SL" )
1914 for (int i=0; i < nbInt; i++)
1920 newInt32 = ( val[4*i+0] & 0xFF )
1921 + (( val[4*i+1] & 0xFF ) << 8 )
1922 + (( val[4*i+2] & 0xFF ) << 16 )
1923 + (( val[4*i+3] & 0xFF ) << 24 );
1924 newInt32 = SwapLong( newInt32 );
1928 #ifdef GDCM_NO_ANSI_STRING_STREAM
1929 s << std::ends; // to avoid oddities on Solaris
1930 #endif //GDCM_NO_ANSI_STRING_STREAM
1934 return ((ValEntry *)entry)->GetValue();
1938 * \brief Get the reverse transformed value of the header entry. The VR
1939 * value is used to define the reverse transformation to operate on
1941 * \warning NOT end user intended method !
1942 * @param entry Entry to reverse transform
1943 * @return Reverse transformed entry value
1945 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1947 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1949 std::string vr = entry->GetVR();
1950 std::vector<std::string> tokens;
1951 std::ostringstream s;
1953 if ( vr == "US" || vr == "SS" )
1957 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1958 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1959 for (unsigned int i=0; i<tokens.size(); i++)
1961 newInt16 = atoi(tokens[i].c_str());
1962 s << ( newInt16 & 0xFF )
1963 << (( newInt16 >> 8 ) & 0xFF );
1967 if ( vr == "UL" || vr == "SL")
1971 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1972 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1973 for (unsigned int i=0; i<tokens.size();i++)
1975 newInt32 = atoi(tokens[i].c_str());
1976 s << (char)( newInt32 & 0xFF )
1977 << (char)(( newInt32 >> 8 ) & 0xFF )
1978 << (char)(( newInt32 >> 16 ) & 0xFF )
1979 << (char)(( newInt32 >> 24 ) & 0xFF );
1984 #ifdef GDCM_NO_ANSI_STRING_STREAM
1985 s << std::ends; // to avoid oddities on Solaris
1986 #endif //GDCM_NO_ANSI_STRING_STREAM
1990 return ((ValEntry *)entry)->GetValue();
1994 * \brief Skip a given Header Entry
1995 * \warning NOT end user intended method !
1996 * @param entry entry to skip
1998 void Document::SkipDocEntry(DocEntry *entry)
2000 SkipBytes(entry->GetLength());
2004 * \brief Skips to the begining of the next Header Entry
2005 * \warning NOT end user intended method !
2006 * @param offset start of skipping
2007 * @param readLgth length to skip
2010 void Document::SkipToNextDocEntry(long offset,long readLgth)
2012 Fp->seekg((long)(offset), std::ios::beg);
2013 Fp->seekg( (long)(readLgth), std::ios::cur);
2017 * \brief When the length of an element value is obviously wrong (because
2018 * the parser went Jabberwocky) one can hope improving things by
2019 * applying some heuristics.
2020 * @param entry entry to check
2021 * @param foundLength fist assumption about length
2023 void Document::FixDocEntryFoundLength(DocEntry *entry,
2024 uint32_t foundLength)
2026 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
2027 if ( foundLength == 0xffffffff)
2032 uint16_t gr = entry->GetGroup();
2033 uint16_t el = entry->GetElement();
2035 if ( foundLength % 2)
2037 std::ostringstream s;
2038 s << "Warning : Tag with uneven length "
2040 << " in x(" << std::hex << gr << "," << el <<")" << std::dec;
2041 dbg.Verbose(0, s.str().c_str());
2044 //////// Fix for some naughty General Electric images.
2045 // Allthough not recent many such GE corrupted images are still present
2046 // on Creatis hard disks. Hence this fix shall remain when such images
2047 // are no longer in user (we are talking a few years, here)...
2048 // Note: XMedCom probably uses such a trick since it is able to read
2049 // those pesky GE images ...
2050 if ( foundLength == 13)
2052 // Only happens for this length !
2053 if ( entry->GetGroup() != 0x0008
2054 || ( entry->GetElement() != 0x0070
2055 && entry->GetElement() != 0x0080 ) )
2058 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
2062 //////// Fix for some brain-dead 'Leonardo' Siemens images.
2063 // Occurence of such images is quite low (unless one leaves close to a
2064 // 'Leonardo' source. Hence, one might consider commenting out the
2065 // following fix on efficiency reasons.
2066 else if ( entry->GetGroup() == 0x0009
2067 && ( entry->GetElement() == 0x1113
2068 || entry->GetElement() == 0x1114 ) )
2071 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
2074 else if ( entry->GetVR() == "SQ" )
2076 foundLength = 0; // ReadLength is unchanged
2079 //////// We encountered a 'delimiter' element i.e. a tag of the form
2080 // "fffe|xxxx" which is just a marker. Delimiters length should not be
2081 // taken into account.
2082 else if( entry->GetGroup() == 0xfffe )
2084 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
2085 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
2086 // causes extra troubles...
2087 if( entry->GetElement() != 0x0000 )
2093 entry->SetUsableLength(foundLength);
2097 * \brief Apply some heuristics to predict whether the considered
2098 * element value contains/represents an integer or not.
2099 * @param entry The element value on which to apply the predicate.
2100 * @return The result of the heuristical predicate.
2102 bool Document::IsDocEntryAnInteger(DocEntry *entry)
2104 uint16_t element = entry->GetElement();
2105 uint16_t group = entry->GetGroup();
2106 const std::string &vr = entry->GetVR();
2107 uint32_t length = entry->GetLength();
2109 // When we have some semantics on the element we just read, and if we
2110 // a priori know we are dealing with an integer, then we shall be
2111 // able to swap it's element value properly.
2112 if ( element == 0 ) // This is the group length of the group
2120 // Allthough this should never happen, still some images have a
2121 // corrupted group length [e.g. have a glance at offset x(8336) of
2122 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
2123 // Since for dicom compliant and well behaved headers, the present
2124 // test is useless (and might even look a bit paranoid), when we
2125 // encounter such an ill-formed image, we simply display a warning
2126 // message and proceed on parsing (while crossing fingers).
2127 std::ostringstream s;
2128 long filePosition = Fp->tellg();
2129 s << "Erroneous Group Length element length on : (" \
2130 << std::hex << group << " , " << element
2131 << ") -before- position x(" << filePosition << ")"
2132 << "lgt : " << length;
2133 dbg.Verbose(0, "Document::IsDocEntryAnInteger", s.str().c_str() );
2137 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2145 * \brief Find the Length till the next sequence delimiter
2146 * \warning NOT end user intended method !
2150 uint32_t Document::FindDocEntryLengthOB()
2151 throw( FormatUnexpected )
2153 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
2154 long positionOnEntry = Fp->tellg();
2155 bool foundSequenceDelimiter = false;
2156 uint32_t totalLength = 0;
2158 while ( !foundSequenceDelimiter )
2164 group = ReadInt16();
2167 catch ( FormatError )
2169 throw FormatError("Document::FindDocEntryLengthOB()",
2170 " group or element not present.");
2173 // We have to decount the group and element we just read
2176 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
2178 dbg.Verbose(1, "Document::FindDocEntryLengthOB: neither an Item "
2179 "tag nor a Sequence delimiter tag.");
2180 Fp->seekg(positionOnEntry, std::ios::beg);
2181 throw FormatUnexpected("Document::FindDocEntryLengthOB()",
2182 "Neither an Item tag nor a Sequence "
2186 if ( elem == 0xe0dd )
2188 foundSequenceDelimiter = true;
2191 uint32_t itemLength = ReadInt32();
2192 // We add 4 bytes since we just read the ItemLength with ReadInt32
2193 totalLength += itemLength + 4;
2194 SkipBytes(itemLength);
2196 if ( foundSequenceDelimiter )
2201 Fp->seekg( positionOnEntry, std::ios::beg);
2206 * \brief Reads a supposed to be 16 Bits integer
2207 * (swaps it depending on processor endianity)
2208 * @return read value
2210 uint16_t Document::ReadInt16()
2211 throw( FormatError )
2214 Fp->read ((char*)&g, (size_t)2);
2217 throw FormatError( "Document::ReadInt16()", " file error." );
2221 throw FormatError( "Document::ReadInt16()", "EOF." );
2228 * \brief Reads a supposed to be 32 Bits integer
2229 * (swaps it depending on processor endianity)
2230 * @return read value
2232 uint32_t Document::ReadInt32()
2233 throw( FormatError )
2236 Fp->read ((char*)&g, (size_t)4);
2239 throw FormatError( "Document::ReadInt32()", " file error." );
2243 throw FormatError( "Document::ReadInt32()", "EOF." );
2250 * \brief skips bytes inside the source file
2251 * \warning NOT end user intended method !
2254 void Document::SkipBytes(uint32_t nBytes)
2256 //FIXME don't dump the returned value
2257 Fp->seekg((long)nBytes, std::ios::cur);
2261 * \brief Loads all the needed Dictionaries
2262 * \warning NOT end user intended method !
2264 void Document::Initialise()
2266 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
2268 RLEInfo = new RLEFramesInfo;
2269 JPEGInfo = new JPEGFragmentsInfo;
2274 * \brief Discover what the swap code is (among little endian, big endian,
2275 * bad little endian, bad big endian).
2277 * @return false when we are absolutely sure
2278 * it's neither ACR-NEMA nor DICOM
2279 * true when we hope ours assuptions are OK
2281 bool Document::CheckSwap()
2283 // The only guaranted way of finding the swap code is to find a
2284 // group tag since we know it's length has to be of four bytes i.e.
2285 // 0x00000004. Finding the swap code in then straigthforward. Trouble
2286 // occurs when we can't find such group...
2288 uint32_t x = 4; // x : for ntohs
2289 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
2295 // First, compare HostByteOrder and NetworkByteOrder in order to
2296 // determine if we shall need to swap bytes (i.e. the Endian type).
2297 if ( x == ntohs(x) )
2306 // The easiest case is the one of a DICOM header, since it possesses a
2307 // file preamble where it suffice to look for the string "DICM".
2310 char *entCur = deb + 128;
2311 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
2313 dbg.Verbose(1, "Document::CheckSwap:", "looks like DICOM Version3");
2315 // Next, determine the value representation (VR). Let's skip to the
2316 // first element (0002, 0000) and check there if we find "UL"
2317 // - or "OB" if the 1st one is (0002,0001) -,
2318 // in which case we (almost) know it is explicit VR.
2319 // WARNING: if it happens to be implicit VR then what we will read
2320 // is the length of the group. If this ascii representation of this
2321 // length happens to be "UL" then we shall believe it is explicit VR.
2322 // FIXME: in order to fix the above warning, we could read the next
2323 // element value (or a couple of elements values) in order to make
2324 // sure we are not commiting a big mistake.
2325 // We need to skip :
2326 // * the 128 bytes of File Preamble (often padded with zeroes),
2327 // * the 4 bytes of "DICM" string,
2328 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2329 // i.e. a total of 136 bytes.
2333 // Sometimes (see : gdcmData/icone.dcm) group 0x0002 *is* Explicit VR,
2334 // but elem 0002,0010 (Transfert Syntax) tells us the file is
2335 // *Implicit* VR. -and it is !-
2337 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
2338 memcmp(entCur, "OB", (size_t)2) == 0 ||
2339 memcmp(entCur, "UI", (size_t)2) == 0 ||
2340 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2341 // when Write DCM *adds*
2343 // Use Document::dicom_vr to test all the possibilities
2344 // instead of just checking for UL, OB and UI !? group 0000
2346 Filetype = ExplicitVR;
2347 dbg.Verbose(1, "Document::CheckSwap:",
2348 "explicit Value Representation");
2352 Filetype = ImplicitVR;
2353 dbg.Verbose(1, "Document::CheckSwap:",
2354 "not an explicit Value Representation");
2360 dbg.Verbose(1, "Document::CheckSwap:",
2361 "HostByteOrder != NetworkByteOrder");
2366 dbg.Verbose(1, "Document::CheckSwap:",
2367 "HostByteOrder = NetworkByteOrder");
2370 // Position the file position indicator at first tag (i.e.
2371 // after the file preamble and the "DICM" string).
2372 Fp->seekg(0, std::ios::beg);
2373 Fp->seekg ( 132L, std::ios::beg);
2377 // Alas, this is not a DicomV3 file and whatever happens there is no file
2378 // preamble. We can reset the file position indicator to where the data
2379 // is (i.e. the beginning of the file).
2380 dbg.Verbose(1, "Document::CheckSwap:", "not a DICOM Version3 file");
2381 Fp->seekg(0, std::ios::beg);
2383 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2384 // By clean we mean that the length of the first tag is written down.
2385 // If this is the case and since the length of the first group HAS to be
2386 // four (bytes), then determining the proper swap code is straightforward.
2389 // We assume the array of char we are considering contains the binary
2390 // representation of a 32 bits integer. Hence the following dirty
2392 s32 = *((uint32_t *)(entCur));
2413 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2414 // It is time for despaired wild guesses.
2415 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2416 // i.e. the 'group length' element is not present :
2418 // check the supposed-to-be 'group number'
2419 // in ( 0x0001 .. 0x0008 )
2420 // to determine ' SwapCode' value .
2421 // Only 0 or 4321 will be possible
2422 // (no oportunity to check for the formerly well known
2423 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2424 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2425 // the file IS NOT ACR-NEMA nor DICOM V3
2426 // Find a trick to tell it the caller...
2428 s16 = *((uint16_t *)(deb));
2455 dbg.Verbose(0, "Document::CheckSwap:",
2456 "ACR/NEMA unfound swap info (Really hopeless !)");
2460 // Then the only info we have is the net2host one.
2472 * \brief Restore the unproperly loaded values i.e. the group, the element
2473 * and the dictionary entry depending on them.
2475 void Document::SwitchSwapToBigEndian()
2477 dbg.Verbose(1, "Document::SwitchSwapToBigEndian",
2478 "Switching to BigEndian mode.");
2479 if ( SwapCode == 0 )
2483 else if ( SwapCode == 4321 )
2487 else if ( SwapCode == 3412 )
2491 else if ( SwapCode == 2143 )
2498 * \brief during parsing, Header Elements too long are not loaded in memory
2501 void Document::SetMaxSizeLoadEntry(long newSize)
2507 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2509 MaxSizeLoadEntry = 0xffffffff;
2512 MaxSizeLoadEntry = newSize;
2517 * \brief Header Elements too long will not be printed
2518 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
2521 void Document::SetMaxSizePrintEntry(long newSize)
2523 //DOH !! This is exactly SetMaxSizeLoadEntry FIXME FIXME
2528 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2530 MaxSizePrintEntry = 0xffffffff;
2533 MaxSizePrintEntry = newSize;
2539 * \brief Handle broken private tag from Philips NTSCAN
2540 * where the endianess is being switch to BigEndian for no
2544 void Document::HandleBrokenEndian(uint16_t group, uint16_t elem)
2546 // Endian reversion. Some files contain groups of tags with reversed endianess.
2547 static int reversedEndian = 0;
2548 // try to fix endian switching in the middle of headers
2549 if ((group == 0xfeff) && (elem == 0x00e0))
2551 // start endian swap mark for group found
2553 SwitchSwapToBigEndian();
2558 else if ((group == 0xfffe) && (elem == 0xe00d) && reversedEndian)
2560 // end of reversed endian group
2562 SwitchSwapToBigEndian();
2567 * \brief Read the next tag but WITHOUT loading it's value
2568 * (read the 'Group Number', the 'Element Number',
2569 * gets the Dict Entry
2570 * gets the VR, gets the length, gets the offset value)
2571 * @return On succes the newly created DocEntry, NULL on failure.
2573 DocEntry *Document::ReadNextDocEntry()
2580 group = ReadInt16();
2583 catch ( FormatError e )
2585 // We reached the EOF (or an error occured) therefore
2586 // header parsing has to be considered as finished.
2591 HandleBrokenEndian(group, elem);
2592 std::string vr = FindDocEntryVR();
2593 std::string realVR = vr;
2595 if( vr == GDCM_UNKNOWN)
2597 DictEntry *dictEntry = GetDictEntryByNumber(group,elem);
2599 realVR = dictEntry->GetVR();
2603 if( Global::GetVR()->IsVROfSequence(realVR) )
2604 newEntry = NewSeqEntryByNumber(group, elem);
2605 else if( Global::GetVR()->IsVROfStringRepresentable(realVR) )
2606 newEntry = NewValEntryByNumber(group, elem,vr);
2608 newEntry = NewBinEntryByNumber(group, elem,vr);
2610 if( vr == GDCM_UNKNOWN )
2612 if( Filetype == ExplicitVR )
2614 // We thought this was explicit VR, but we end up with an
2615 // implicit VR tag. Let's backtrack.
2617 msg = Util::Format("Falsely explicit vr file (%04x,%04x)\n",
2618 newEntry->GetGroup(), newEntry->GetElement());
2619 dbg.Verbose(1, "Document::FindVR: ", msg.c_str());
2621 newEntry->SetImplicitVR();
2626 FindDocEntryLength(newEntry);
2628 catch ( FormatError e )
2636 newEntry->SetOffset(Fp->tellg());
2643 * \brief Generate a free TagKey i.e. a TagKey that is not present
2644 * in the TagHt dictionary.
2645 * @param group The generated tag must belong to this group.
2646 * @return The element of tag with given group which is fee.
2648 uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2650 for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2652 TagKey key = DictEntry::TranslateToKey(group, elem);
2653 if (TagHT.count(key) == 0)
2662 * \brief Assuming the internal file pointer \ref Document::Fp
2663 * is placed at the beginning of a tag check whether this
2664 * tag is (TestGroup, TestElement).
2665 * \warning On success the internal file pointer \ref Document::Fp
2666 * is modified to point after the tag.
2667 * On failure (i.e. when the tag wasn't the expected tag
2668 * (TestGroup, TestElement) the internal file pointer
2669 * \ref Document::Fp is restored to it's original position.
2670 * @param testGroup The expected group of the tag.
2671 * @param testElement The expected Element of the tag.
2672 * @return True on success, false otherwise.
2674 bool Document::ReadTag(uint16_t testGroup, uint16_t testElement)
2676 long positionOnEntry = Fp->tellg();
2677 long currentPosition = Fp->tellg(); // On debugging purposes
2679 //// Read the Item Tag group and element, and make
2680 // sure they are what we expected:
2681 uint16_t itemTagGroup;
2682 uint16_t itemTagElement;
2685 itemTagGroup = ReadInt16();
2686 itemTagElement = ReadInt16();
2688 catch ( FormatError e )
2690 //std::cerr << e << std::endl;
2693 if ( itemTagGroup != testGroup || itemTagElement != testElement )
2695 std::ostringstream s;
2696 s << " We should have found tag (";
2697 s << std::hex << testGroup << "," << testElement << ")" << std::endl;
2698 s << " but instead we encountered tag (";
2699 s << std::hex << itemTagGroup << "," << itemTagElement << ")"
2701 s << " at address: " << (unsigned)currentPosition << std::endl;
2702 dbg.Verbose(0, "Document::ReadItemTagLength: wrong Item Tag found:");
2703 dbg.Verbose(0, s.str().c_str());
2704 Fp->seekg(positionOnEntry, std::ios::beg);
2712 * \brief Assuming the internal file pointer \ref Document::Fp
2713 * is placed at the beginning of a tag (TestGroup, TestElement),
2714 * read the length associated to the Tag.
2715 * \warning On success the internal file pointer \ref Document::Fp
2716 * is modified to point after the tag and it's length.
2717 * On failure (i.e. when the tag wasn't the expected tag
2718 * (TestGroup, TestElement) the internal file pointer
2719 * \ref Document::Fp is restored to it's original position.
2720 * @param testGroup The expected group of the tag.
2721 * @param testElement The expected Element of the tag.
2722 * @return On success returns the length associated to the tag. On failure
2725 uint32_t Document::ReadTagLength(uint16_t testGroup, uint16_t testElement)
2727 long positionOnEntry = Fp->tellg();
2728 (void)positionOnEntry;
2730 if ( !ReadTag(testGroup, testElement) )
2735 //// Then read the associated Item Length
2736 long currentPosition = Fp->tellg();
2737 uint32_t itemLength = ReadInt32();
2739 std::ostringstream s;
2740 s << "Basic Item Length is: "
2741 << itemLength << std::endl;
2742 s << " at address: " << (unsigned)currentPosition << std::endl;
2743 dbg.Verbose(0, "Document::ReadItemTagLength: ", s.str().c_str());
2749 * \brief When parsing the Pixel Data of an encapsulated file, read
2750 * the basic offset table (when present, and BTW dump it).
2752 void Document::ReadAndSkipEncapsulatedBasicOffsetTable()
2754 //// Read the Basic Offset Table Item Tag length...
2755 uint32_t itemLength = ReadTagLength(0xfffe, 0xe000);
2757 // When present, read the basic offset table itself.
2758 // Notes: - since the presence of this basic offset table is optional
2759 // we can't rely on it for the implementation, and we will simply
2760 // trash it's content (when present).
2761 // - still, when present, we could add some further checks on the
2762 // lengths, but we won't bother with such fuses for the time being.
2763 if ( itemLength != 0 )
2765 char *basicOffsetTableItemValue = new char[itemLength + 1];
2766 Fp->read(basicOffsetTableItemValue, itemLength);
2769 for (unsigned int i=0; i < itemLength; i += 4 )
2771 uint32_t individualLength = str2num( &basicOffsetTableItemValue[i],
2773 std::ostringstream s;
2774 s << " Read one length: ";
2775 s << std::hex << individualLength << std::endl;
2777 "Document::ReadAndSkipEncapsulatedBasicOffsetTable: ",
2782 delete[] basicOffsetTableItemValue;
2787 * \brief Parse pixel data from disk of [multi-]fragment RLE encoding.
2788 * Compute the RLE extra information and store it in \ref RLEInfo
2789 * for later pixel retrieval usage.
2791 void Document::ComputeRLEInfo()
2793 TransferSyntaxType ts = GetTransferSyntax();
2794 if ( ts != RLELossless )
2799 // Encoded pixel data: for the time being we are only concerned with
2800 // Jpeg or RLE Pixel data encodings.
2801 // As stated in PS 3.5-2003, section 8.2 p44:
2802 // "If sent in Encapsulated Format (i.e. other than the Native Format) the
2803 // value representation OB is used".
2804 // Hence we expect an OB value representation. Concerning OB VR,
2805 // the section PS 3.5-2003, section A.4.c p 58-59, states:
2806 // "For the Value Representations OB and OW, the encoding shall meet the
2807 // following specifications depending on the Data element tag:"
2809 // - the first item in the sequence of items before the encoded pixel
2810 // data stream shall be basic offset table item. The basic offset table
2811 // item value, however, is not required to be present"
2813 ReadAndSkipEncapsulatedBasicOffsetTable();
2815 // Encapsulated RLE Compressed Images (see PS 3.5-2003, Annex G)
2816 // Loop on the individual frame[s] and store the information
2817 // on the RLE fragments in a RLEFramesInfo.
2818 // Note: - when only a single frame is present, this is a
2820 // - when more than one frame are present, then we are in
2821 // the case of a multi-frame image.
2823 while ( (frameLength = ReadTagLength(0xfffe, 0xe000)) )
2825 // Parse the RLE Header and store the corresponding RLE Segment
2826 // Offset Table information on fragments of this current Frame.
2827 // Note that the fragment pixels themselves are not loaded
2828 // (but just skipped).
2829 long frameOffset = Fp->tellg();
2831 uint32_t nbRleSegments = ReadInt32();
2832 if ( nbRleSegments > 16 )
2834 // There should be at most 15 segments (refer to RLEFrame class)
2835 dbg.Verbose(0, "Document::ComputeRLEInfo: too many segments.");
2838 uint32_t rleSegmentOffsetTable[16];
2839 for( int k = 1; k <= 15; k++ )
2841 rleSegmentOffsetTable[k] = ReadInt32();
2844 // Deduce from both the RLE Header and the frameLength the
2845 // fragment length, and again store this info in a
2847 long rleSegmentLength[15];
2848 // skipping (not reading) RLE Segments
2849 if ( nbRleSegments > 1)
2851 for(unsigned int k = 1; k <= nbRleSegments-1; k++)
2853 rleSegmentLength[k] = rleSegmentOffsetTable[k+1]
2854 - rleSegmentOffsetTable[k];
2855 SkipBytes(rleSegmentLength[k]);
2859 rleSegmentLength[nbRleSegments] = frameLength
2860 - rleSegmentOffsetTable[nbRleSegments];
2861 SkipBytes(rleSegmentLength[nbRleSegments]);
2863 // Store the collected info
2864 RLEFrame *newFrameInfo = new RLEFrame;
2865 newFrameInfo->NumberFragments = nbRleSegments;
2866 for( unsigned int uk = 1; uk <= nbRleSegments; uk++ )
2868 newFrameInfo->Offset[uk] = frameOffset + rleSegmentOffsetTable[uk];
2869 newFrameInfo->Length[uk] = rleSegmentLength[uk];
2871 RLEInfo->Frames.push_back( newFrameInfo );
2874 // Make sure that at the end of the item we encounter a 'Sequence
2876 if ( !ReadTag(0xfffe, 0xe0dd) )
2878 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2879 dbg.Verbose(0, " item at end of RLE item sequence");
2884 * \brief Parse pixel data from disk of [multi-]fragment Jpeg encoding.
2885 * Compute the jpeg extra information (fragment[s] offset[s] and
2886 * length) and store it[them] in \ref JPEGInfo for later pixel
2889 void Document::ComputeJPEGFragmentInfo()
2891 // If you need to, look for comments of ComputeRLEInfo().
2897 ReadAndSkipEncapsulatedBasicOffsetTable();
2899 // Loop on the fragments[s] and store the parsed information in a
2901 long fragmentLength;
2902 while ( (fragmentLength = ReadTagLength(0xfffe, 0xe000)) )
2904 long fragmentOffset = Fp->tellg();
2906 // Store the collected info
2907 JPEGFragment *newFragment = new JPEGFragment;
2908 newFragment->Offset = fragmentOffset;
2909 newFragment->Length = fragmentLength;
2910 JPEGInfo->Fragments.push_back( newFragment );
2912 SkipBytes( fragmentLength );
2915 // Make sure that at the end of the item we encounter a 'Sequence
2917 if ( !ReadTag(0xfffe, 0xe0dd) )
2919 dbg.Verbose(0, "Document::ComputeRLEInfo: no sequence delimiter ");
2920 dbg.Verbose(0, " item at end of JPEG item sequence");
2925 * \brief Walk recursively the given \ref DocEntrySet, and feed
2926 * the given hash table (\ref TagDocEntryHT) with all the
2927 * \ref DocEntry (Dicom entries) encountered.
2928 * This method does the job for \ref BuildFlatHashTable.
2929 * @param builtHT Where to collect all the \ref DocEntry encountered
2930 * when recursively walking the given set.
2931 * @param set The structure to be traversed (recursively).
2933 void Document::BuildFlatHashTableRecurse( TagDocEntryHT &builtHT,
2936 if (ElementSet *elementSet = dynamic_cast< ElementSet* > ( set ) )
2938 TagDocEntryHT const ¤tHT = elementSet->GetTagHT();
2939 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2940 i != currentHT.end();
2943 DocEntry *entry = i->second;
2944 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2946 const ListSQItem& items = seqEntry->GetSQItems();
2947 for( ListSQItem::const_iterator item = items.begin();
2948 item != items.end();
2951 BuildFlatHashTableRecurse( builtHT, *item );
2955 builtHT[entry->GetKey()] = entry;
2960 if (SQItem *SQItemSet = dynamic_cast< SQItem* > ( set ) )
2962 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
2963 for (ListDocEntry::const_iterator i = currentList.begin();
2964 i != currentList.end();
2967 DocEntry *entry = *i;
2968 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2970 const ListSQItem& items = seqEntry->GetSQItems();
2971 for( ListSQItem::const_iterator item = items.begin();
2972 item != items.end();
2975 BuildFlatHashTableRecurse( builtHT, *item );
2979 builtHT[entry->GetKey()] = entry;
2986 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2989 * The structure used by a Document (through \ref ElementSet),
2990 * in order to hold the parsed entries of a Dicom header, is a recursive
2991 * one. This is due to the fact that the sequences (when present)
2992 * can be nested. Additionaly, the sequence items (represented in
2993 * gdcm as \ref SQItem) add an extra complexity to the data
2994 * structure. Hence, a gdcm user whishing to visit all the entries of
2995 * a Dicom header will need to dig in the gdcm internals (which
2996 * implies exposing all the internal data structures to the API).
2997 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2998 * recursively builds a temporary hash table, which holds all the
2999 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
3001 * \warning Of course there is NO integrity constrain between the
3002 * returned \ref TagDocEntryHT and the \ref ElementSet used
3003 * to build it. Hence if the underlying \ref ElementSet is
3004 * altered, then it is the caller responsability to invoke
3005 * \ref BuildFlatHashTable again...
3006 * @return The flat std::map<> we juste build.
3008 TagDocEntryHT *Document::BuildFlatHashTable()
3010 TagDocEntryHT *FlatHT = new TagDocEntryHT;
3011 BuildFlatHashTableRecurse( *FlatHT, this );
3018 * \brief Compares two documents, according to \ref DicomDir rules
3019 * \warning Does NOT work with ACR-NEMA files
3020 * \todo Find a trick to solve the pb (use RET fields ?)
3022 * @return true if 'smaller'
3024 bool Document::operator<(Document &document)
3027 std::string s1 = GetEntryByNumber(0x0010,0x0010);
3028 std::string s2 = document.GetEntryByNumber(0x0010,0x0010);
3040 s1 = GetEntryByNumber(0x0010,0x0020);
3041 s2 = document.GetEntryByNumber(0x0010,0x0020);
3052 // Study Instance UID
3053 s1 = GetEntryByNumber(0x0020,0x000d);
3054 s2 = document.GetEntryByNumber(0x0020,0x000d);
3065 // Serie Instance UID
3066 s1 = GetEntryByNumber(0x0020,0x000e);
3067 s2 = document.GetEntryByNumber(0x0020,0x000e);
3084 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
3085 * @param filetype Type of the File to be written
3087 int Document::ComputeGroup0002Length( FileType filetype )
3092 int groupLength = 0;
3093 bool found0002 = false;
3095 // for each zero-level Tag in the DCM Header
3099 entry = GetNextEntry();
3102 gr = entry->GetGroup();
3108 el = entry->GetElement();
3109 vr = entry->GetVR();
3111 if (filetype == ExplicitVR)
3113 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
3115 groupLength += 4; // explicit VR AND OB, OW, SQ : 4 more bytes
3118 groupLength += 2 + 2 + 4 + entry->GetLength();
3120 else if (found0002 )
3123 entry = GetNextEntry();
3128 } // end namespace gdcm
3130 //-----------------------------------------------------------------------------