1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/31 12:19:34 $
7 Version: $Revision: 1.217 $
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 "gdcmDictSet.h"
28 #include "gdcmDocEntrySet.h"
29 #include "gdcmSQItem.h"
35 #if defined(_MSC_VER) || defined(__BORLANDC__) || defined(__MINGW32__)
39 #ifdef CMAKE_HAVE_NETINET_IN_H
40 #include <netinet/in.h>
46 //-----------------------------------------------------------------------------
47 // Refer to Document::CheckSwap()
48 //const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
50 // Refer to Document::SetMaxSizeLoadEntry()
51 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
52 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
54 //-----------------------------------------------------------------------------
55 // Constructor / Destructor
59 * @param filename 'Document' (File or DicomDir) to be opened for parsing
61 Document::Document( std::string const &filename ) : ElementSet(-1)
63 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
73 Group0002Parsed = false;
75 gdcmVerboseMacro( "Starting parsing of file: " << Filename.c_str());
76 // Fp->seekg( 0, std::ios::beg);
78 Fp->seekg(0, std::ios::end);
79 long lgt = Fp->tellg();
81 Fp->seekg( 0, std::ios::beg);
84 long beg = Fp->tellg();
87 ParseDES( this, beg, lgt, false); // Loading is done during parsing
89 Fp->seekg( 0, std::ios::beg);
91 // Load 'non string' values
93 std::string PhotometricInterpretation = GetEntryValue(0x0028,0x0004);
94 if( PhotometricInterpretation == "PALETTE COLOR " )
96 LoadEntryBinArea(0x0028,0x1200); // gray LUT
98 /// The tags refered by the three following lines used to be CORRECTLY
99 /// defined as having an US Value Representation in the public
100 /// dictionary. BUT the semantics implied by the three following
101 /// lines state that the corresponding tag contents are in fact
102 /// the ones of a BinEntry.
103 /// In order to fix things "Quick and Dirty" the dictionary was
104 /// altered on PURPOSE but now contains a WRONG value.
105 /// In order to fix things and restore the dictionary to its
106 /// correct value, one needs to decided of the semantics by deciding
107 /// whether the following tags are either:
108 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
109 /// also used as BinEntry, which requires the proper conversion,
110 /// - OW, and hence loaded as BinEntry, but afterwards also used
111 /// as ValEntry, which requires the proper conversion.
112 LoadEntryBinArea(0x0028,0x1201); // R LUT
113 LoadEntryBinArea(0x0028,0x1202); // G LUT
114 LoadEntryBinArea(0x0028,0x1203); // B LUT
116 // Segmented Red Palette Color LUT Data
117 LoadEntryBinArea(0x0028,0x1221);
118 // Segmented Green Palette Color LUT Data
119 LoadEntryBinArea(0x0028,0x1222);
120 // Segmented Blue Palette Color LUT Data
121 LoadEntryBinArea(0x0028,0x1223);
123 //FIXME later : how to use it?
124 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
128 // ----------------------------
129 // Specific code to allow gdcm to read ACR-LibIDO formated images
130 // Note: ACR-LibIDO is an extension of the ACR standard that was
131 // used at CREATIS. For the time being (say a couple years)
132 // we keep this kludge to allow a smooth move to gdcm for
133 // CREATIS developpers (sorry folks).
135 // if recognition code tells us we deal with a LibIDO image
136 // we switch lineNumber and columnNumber
139 RecCode = GetEntryValue(0x0008, 0x0010); // recognition code (RET)
140 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
141 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
142 // with "little-endian strings"
144 Filetype = ACR_LIBIDO;
145 std::string rows = GetEntryValue(0x0028, 0x0010);
146 std::string columns = GetEntryValue(0x0028, 0x0011);
147 SetValEntry(columns, 0x0028, 0x0010);
148 SetValEntry(rows , 0x0028, 0x0011);
150 // --- End of ACR-LibIDO kludge ---
154 * \brief This default constructor doesn't parse the file. You should
155 * then invoke \ref Document::SetFileName and then the parsing.
157 Document::Document() : ElementSet(-1)
161 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
164 Filetype = ExplicitVR;
165 Group0002Parsed = false;
169 * \brief Canonical destructor.
171 Document::~Document ()
177 //-----------------------------------------------------------------------------
181 * \brief Prints The Dict Entries of THE public Dicom Dictionary
182 * @param os ostream to print to
185 void Document::PrintPubDict(std::ostream &os)
187 RefPubDict->SetPrintLevel(PrintLevel);
188 RefPubDict->Print(os);
192 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
193 * @param os ostream to print to
196 void Document::PrintShaDict(std::ostream &os)
198 RefShaDict->SetPrintLevel(PrintLevel);
199 RefShaDict->Print(os);
202 //-----------------------------------------------------------------------------
205 * \brief Get the public dictionary used
207 Dict *Document::GetPubDict()
213 * \brief Get the shadow dictionary used
215 Dict *Document::GetShaDict()
221 * \brief Set the shadow dictionary used
222 * @param dict dictionary to use in shadow
224 bool Document::SetShaDict(Dict *dict)
231 * \brief Set the shadow dictionary used
232 * @param dictName name of the dictionary to use in shadow
234 bool Document::SetShaDict(DictKey const &dictName)
236 RefShaDict = Global::GetDicts()->GetDict(dictName);
241 * \brief This predicate, based on hopefully reasonable heuristics,
242 * decides whether or not the current Document was properly parsed
243 * and contains the mandatory information for being considered as
244 * a well formed and usable Dicom/Acr File.
245 * @return true when Document is the one of a reasonable Dicom/Acr file,
248 bool Document::IsReadable()
250 if( Filetype == Unknown)
252 gdcmVerboseMacro( "Wrong filetype");
258 gdcmVerboseMacro( "No tag in internal hash table.");
266 * \brief Predicate for dicom version 3 file.
267 * @return True when the file is a dicom version 3.
269 bool Document::IsDicomV3()
271 // Checking if Transfer Syntax exists is enough
272 // Anyway, it's to late check if the 'Preamble' was found ...
273 // And ... would it be a rich idea to check ?
274 // (some 'no Preamble' DICOM images exist !)
275 return GetDocEntry(0x0002, 0x0010) != NULL;
279 * \brief Predicate for Papyrus file
280 * Dedicated to whomsoever it may concern
281 * @return True when the file is a Papyrus file.
283 bool Document::IsPapyrus()
285 // check for Papyrus private Sequence
286 DocEntry *e = GetDocEntry(0x0041, 0x1050);
289 // check if it's actually a Sequence
290 if ( !dynamic_cast<SeqEntry*>(e) )
296 * \brief returns the File Type
297 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
298 * @return the FileType code
300 FileType Document::GetFileType()
306 * \brief Accessor to the Transfer Syntax (when present) of the
307 * current document (it internally handles reading the
308 * value from disk when only parsing occured).
309 * @return The encountered Transfer Syntax of the current document.
311 std::string Document::GetTransferSyntax()
313 DocEntry *entry = GetDocEntry(0x0002, 0x0010);
319 // The entry might be present but not loaded (parsing and loading
320 // happen at different stages): try loading and proceed with check...
321 LoadDocEntrySafe(entry);
322 if (ValEntry *valEntry = dynamic_cast< ValEntry* >(entry) )
324 std::string transfer = valEntry->GetValue();
325 // The actual transfer (as read from disk) might be padded. We
326 // first need to remove the potential padding. We can make the
327 // weak assumption that padding was not executed with digits...
328 if ( transfer.length() == 0 )
330 // for brain damaged headers
333 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
335 transfer.erase(transfer.length()-1, 1);
343 * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS
344 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
346 std::string Document::GetTransferSyntaxName()
348 // use the TS (TS : Transfer Syntax)
349 std::string transferSyntax = GetEntryValue(0x0002,0x0010);
351 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
353 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
354 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
355 return "Uncompressed ACR-NEMA";
357 if ( transferSyntax == GDCM_UNFOUND )
359 gdcmVerboseMacro( "Unfound Transfer Syntax (0002,0010)");
360 return "Uncompressed ACR-NEMA";
363 // we do it only when we need it
364 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
366 // Global::GetTS() is a global static you shall never try to delete it!
370 // --------------- Swap Code ------------------
372 * \brief Swaps the bytes so they agree with the processor order
373 * @return The properly swaped 16 bits integer.
375 uint16_t Document::SwapShort(uint16_t a)
377 if ( SwapCode == 4321 || SwapCode == 2143 )
379 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
385 * \brief Swaps back the bytes of 4-byte long integer accordingly to
387 * @return The properly swaped 32 bits integer.
389 uint32_t Document::SwapLong(uint32_t a)
396 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
397 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
400 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
403 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
406 gdcmErrorMacro( "Unset swap code:" << SwapCode );
413 // -----------------File I/O ---------------
415 * \brief Tries to open the file \ref Document::Filename and
416 * checks the preamble when existing.
417 * @return The FILE pointer on success.
419 std::ifstream *Document::OpenFile()
421 HasDCMPreamble = false;
422 if (Filename.length() == 0)
429 gdcmVerboseMacro( "File already open: " << Filename.c_str());
433 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
436 gdcmDebugMacro( "Cannot open file: " << Filename.c_str());
443 Fp->read((char*)&zero, (size_t)2);
450 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
452 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
453 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
454 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
455 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
458 = Util::Format("ACR/DICOM with no preamble: (%04x)\n", zero);
459 gdcmVerboseMacro( msg.c_str() );
464 Fp->seekg(126L, std::ios::cur);
466 Fp->read(dicm, (size_t)4);
472 if( memcmp(dicm, "DICM", 4) == 0 )
474 HasDCMPreamble = true;
479 gdcmVerboseMacro( "Not DICOM/ACR (missing preamble)" << Filename.c_str());
485 * \brief closes the file
486 * @return TRUE if the close was successfull
488 bool Document::CloseFile()
496 return true; //FIXME how do we detect a non-closed ifstream ?
500 * \brief Writes in a file all the Header Entries (Dicom Elements)
501 * @param fp file pointer on an already open file (actually: Output File Stream)
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
511 // (necessary if user wants to write a DICOM V3 file
512 // starting from an ACR-NEMA (V2) Header
514 if ( filetype == ImplicitVR || filetype == ExplicitVR )
516 // writing Dicom File Preamble
517 char filePreamble[128];
518 memset(filePreamble, 0, 128);
519 fp->write(filePreamble, 128);
520 fp->write("DICM", 4);
524 * \todo rewrite later, if really usefull
525 * - 'Group Length' element is optional in DICOM
526 * - but un-updated odd groups lengthes can causes pb
529 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
530 * UpdateGroupLength(false,filetype);
531 * if ( filetype == ACR)
532 * UpdateGroupLength(true,ACR);
535 ElementSet::WriteContent(fp, filetype); // This one is recursive
538 // -----------------------------------------
542 * \brief Loads (from disk) the element content
543 * when a string is not suitable
544 * @param group group number of the Entry
545 * @param elem element number of the Entry
547 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
549 // Search the corresponding DocEntry
550 DocEntry *docElement = GetDocEntry(group, elem);
554 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
558 LoadEntryBinArea(binElement);
562 * \brief Loads (from disk) the element content
563 * when a string is not suitable
564 * @param elem Entry whose binArea is going to be loaded
566 void Document::LoadEntryBinArea(BinEntry *elem)
568 if(elem->GetBinArea())
575 size_t o =(size_t)elem->GetOffset();
576 Fp->seekg(o, std::ios::beg);
578 size_t l = elem->GetLength();
579 uint8_t *a = new uint8_t[l];
582 gdcmVerboseMacro( "Cannot allocate BinEntry content");
586 /// \todo check the result
587 Fp->read((char*)a, l);
588 if( Fp->fail() || Fp->eof())
601 * \brief Loads the element while preserving the current
602 * underlying file position indicator as opposed to
603 * LoadDocEntry that modifies it.
604 * @param entry Header Entry whose value will be loaded.
607 void Document::LoadDocEntrySafe(DocEntry *entry)
611 long PositionOnEntry = Fp->tellg();
613 Fp->seekg(PositionOnEntry, std::ios::beg);
617 //-----------------------------------------------------------------------------
620 // Constructors and destructors are protected to avoid user to invoke directly
623 * \brief Reads a supposed to be 16 Bits integer
624 * (swaps it depending on processor endianity)
627 uint16_t Document::ReadInt16()
631 Fp->read ((char*)&g, (size_t)2);
634 throw FormatError( "Document::ReadInt16()", " file error." );
638 throw FormatError( "Document::ReadInt16()", "EOF." );
645 * \brief Reads a supposed to be 32 Bits integer
646 * (swaps it depending on processor endianity)
649 uint32_t Document::ReadInt32()
653 Fp->read ((char*)&g, (size_t)4);
656 throw FormatError( "Document::ReadInt32()", " file error." );
660 throw FormatError( "Document::ReadInt32()", "EOF." );
667 * \brief skips bytes inside the source file
668 * \warning NOT end user intended method !
671 void Document::SkipBytes(uint32_t nBytes)
673 //FIXME don't dump the returned value
674 Fp->seekg((long)nBytes, std::ios::cur);
678 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
679 * @param filetype Type of the File to be written
681 int Document::ComputeGroup0002Length( FileType filetype )
687 bool found0002 = false;
689 // for each zero-level Tag in the DCM Header
690 DocEntry *entry = GetFirstEntry();
693 gr = entry->GetGroup();
699 if( entry->GetElement() != 0x0000 )
703 if( filetype == ExplicitVR )
705 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
707 // explicit VR AND OB, OW, SQ : 4 more bytes
711 groupLength += 2 + 2 + 4 + entry->GetLength();
717 entry = GetNextEntry();
722 //-----------------------------------------------------------------------------
726 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
727 * @return length of the parsed set.
729 void Document::ParseDES(DocEntrySet *set, long offset,
730 long l_max, bool delim_mode)
732 DocEntry *newDocEntry = 0;
733 ValEntry *newValEntry;
734 BinEntry *newBinEntry;
735 SeqEntry *newSeqEntry;
741 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
747 newDocEntry = ReadNextDocEntry( );
754 vr = newDocEntry->GetVR();
755 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
756 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
757 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
759 if ( newValEntry || newBinEntry )
763 if ( Filetype == ExplicitVR &&
764 !Global::GetVR()->IsVROfBinaryRepresentable(vr) )
766 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
767 gdcmVerboseMacro( std::hex << newDocEntry->GetGroup()
768 << "|" << newDocEntry->GetElement()
769 << " : Neither Valentry, nor BinEntry."
770 "Probably unknown VR.");
773 //////////////////// BinEntry or UNKOWN VR:
774 // When "this" is a Document the Key is simply of the
775 // form ( group, elem )...
776 if ( dynamic_cast< Document* > ( set ) )
778 newBinEntry->SetKey( newBinEntry->GetKey() );
780 // but when "this" is a SQItem, we are inserting this new
781 // valEntry in a sequence item, and the key has the
782 // generalized form (refer to \ref BaseTagKey):
783 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
785 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
786 + newBinEntry->GetKey() );
789 LoadDocEntry( newBinEntry );
790 if( !set->AddEntry( newBinEntry ) )
792 //Expect big troubles if here
793 //delete newBinEntry;
799 /////////////////////// ValEntry
800 // When "set" is a Document, then we are at the top of the
801 // hierarchy and the Key is simply of the form ( group, elem )...
802 if ( dynamic_cast< Document* > ( set ) )
804 newValEntry->SetKey( newValEntry->GetKey() );
806 // ...but when "set" is a SQItem, we are inserting this new
807 // valEntry in a sequence item. Hence the key has the
808 // generalized form (refer to \ref BaseTagKey):
809 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
811 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
812 + newValEntry->GetKey() );
815 LoadDocEntry( newValEntry );
816 bool delimitor=newValEntry->IsItemDelimitor();
817 if( !set->AddEntry( newValEntry ) )
819 // If here expect big troubles
820 //delete newValEntry; //otherwise mem leak
830 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
838 // Just to make sure we are at the beginning of next entry.
839 SkipToNextDocEntry(newDocEntry);
844 unsigned long l = newDocEntry->GetReadLength();
845 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
847 if ( l == 0xffffffff )
856 // no other way to create it ...
857 newSeqEntry->SetDelimitorMode( delim_mode );
859 // At the top of the hierarchy, stands a Document. When "set"
860 // is a Document, then we are building the first depth level.
861 // Hence the SeqEntry we are building simply has a depth
863 if (/*Document *dummy =*/ dynamic_cast< Document* > ( set ) )
866 newSeqEntry->SetDepthLevel( 1 );
867 newSeqEntry->SetKey( newSeqEntry->GetKey() );
869 // But when "set" is already a SQItem, we are building a nested
870 // sequence, and hence the depth level of the new SeqEntry
871 // we are building, is one level deeper:
872 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
874 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
875 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
876 + newSeqEntry->GetKey() );
880 { // Don't try to parse zero-length sequences
881 ParseSQ( newSeqEntry,
882 newDocEntry->GetOffset(),
885 if( !set->AddEntry( newSeqEntry ) )
889 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
903 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
904 * @return parsed length for this level
906 void Document::ParseSQ( SeqEntry *seqEntry,
907 long offset, long l_max, bool delim_mode)
909 int SQItemNumber = 0;
911 long offsetStartCurrentSQItem = offset;
915 // the first time, we read the fff0,e000 of the first SQItem
916 DocEntry *newDocEntry = ReadNextDocEntry();
920 // FIXME Should warn user
925 if ( newDocEntry->IsSequenceDelimitor() )
927 seqEntry->SetDelimitationItem( newDocEntry );
931 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
936 // create the current SQItem
937 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
938 std::ostringstream newBase;
939 newBase << seqEntry->GetKey()
943 itemSQ->SetBaseTagKey( newBase.str() );
944 unsigned int l = newDocEntry->GetReadLength();
946 if ( l == 0xffffffff )
955 // when we're here, element fffe,e000 is already passed.
956 // it's lost for the SQItem we're going to process !!
958 //ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
959 //delete newDocEntry; // FIXME well ... it's too late to use it !
961 // Let's try :------------
962 // remove fff0,e000, created out of the SQItem
964 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg);
965 // fill up the current SQItem, starting at the beginning of fff0,e000
966 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
967 offsetStartCurrentSQItem = Fp->tellg();
968 // end try -----------------
970 seqEntry->AddSQItem( itemSQ, SQItemNumber );
972 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
980 * \brief Loads the element content if its length doesn't exceed
981 * the value specified with Document::SetMaxSizeLoadEntry()
982 * @param entry Header Entry (Dicom Element) to be dealt with
984 void Document::LoadDocEntry(DocEntry *entry)
986 uint16_t group = entry->GetGroup();
987 std::string vr = entry->GetVR();
988 uint32_t length = entry->GetLength();
990 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
992 // A SeQuence "contains" a set of Elements.
993 // (fffe e000) tells us an Element is beginning
994 // (fffe e00d) tells us an Element just ended
995 // (fffe e0dd) tells us the current SeQuence just ended
996 if( group == 0xfffe )
998 // NO more value field for SQ !
1002 // When the length is zero things are easy:
1005 ((ValEntry *)entry)->SetValue("");
1009 // The elements whose length is bigger than the specified upper bound
1010 // are not loaded. Instead we leave a short notice of the offset of
1011 // the element content and it's length.
1013 std::ostringstream s;
1014 if (length > MaxSizeLoadEntry)
1016 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1018 //s << "gdcm::NotLoaded (BinEntry)";
1019 s << GDCM_NOTLOADED;
1020 s << " Address:" << (long)entry->GetOffset();
1021 s << " Length:" << entry->GetLength();
1022 s << " x(" << std::hex << entry->GetLength() << ")";
1023 binEntryPtr->SetValue(s.str());
1025 // Be carefull : a BinEntry IS_A ValEntry ...
1026 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1028 // s << "gdcm::NotLoaded. (ValEntry)";
1029 s << GDCM_NOTLOADED;
1030 s << " Address:" << (long)entry->GetOffset();
1031 s << " Length:" << entry->GetLength();
1032 s << " x(" << std::hex << entry->GetLength() << ")";
1033 valEntryPtr->SetValue(s.str());
1038 gdcmErrorMacro( "MaxSizeLoadEntry exceeded, neither a BinEntry "
1039 << "nor a ValEntry ?! Should never print that !" );
1042 // to be sure we are at the end of the value ...
1043 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1048 // When we find a BinEntry not very much can be done :
1049 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1051 s << GDCM_BINLOADED;
1052 binEntryPtr->SetValue(s.str());
1053 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1057 /// \todo Any compacter code suggested (?)
1058 if ( IsDocEntryAnInteger(entry) )
1062 // When short integer(s) are expected, read and convert the following
1063 // n *two characters properly i.e. consider them as short integers as
1064 // opposed to strings.
1065 // Elements with Value Multiplicity > 1
1066 // contain a set of integers (not a single one)
1067 if (vr == "US" || vr == "SS")
1070 NewInt = ReadInt16();
1074 for (int i=1; i < nbInt; i++)
1077 NewInt = ReadInt16();
1082 // See above comment on multiple integers (mutatis mutandis).
1083 else if (vr == "UL" || vr == "SL")
1086 NewInt = ReadInt32();
1090 for (int i=1; i < nbInt; i++)
1093 NewInt = ReadInt32();
1098 #ifdef GDCM_NO_ANSI_STRING_STREAM
1099 s << std::ends; // to avoid oddities on Solaris
1100 #endif //GDCM_NO_ANSI_STRING_STREAM
1102 ((ValEntry *)entry)->SetValue(s.str());
1106 // FIXME: We need an additional byte for storing \0 that is not on disk
1107 char *str = new char[length+1];
1108 Fp->read(str, (size_t)length);
1109 str[length] = '\0'; //this is only useful when length is odd
1110 // Special DicomString call to properly handle \0 and even length
1111 std::string newValue;
1114 newValue = Util::DicomString(str, length+1);
1115 gdcmVerboseMacro("Warning: bad length: " << length <<
1116 ",For string :" << newValue.c_str());
1117 // Since we change the length of string update it length
1118 //entry->SetReadLength(length+1);
1122 newValue = Util::DicomString(str, length);
1126 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1128 if ( Fp->fail() || Fp->eof())
1130 gdcmVerboseMacro("Unread element value");
1131 valEntry->SetValue(GDCM_UNREAD);
1137 // Because of correspondance with the VR dic
1138 valEntry->SetValue(newValue);
1142 valEntry->SetValue(newValue);
1147 gdcmErrorMacro( "Should have a ValEntry, here !");
1152 * \brief Find the value Length of the passed Header Entry
1153 * @param entry Header Entry whose length of the value shall be loaded.
1155 void Document::FindDocEntryLength( DocEntry *entry )
1156 throw ( FormatError )
1158 std::string vr = entry->GetVR();
1161 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1163 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1165 // The following reserved two bytes (see PS 3.5-2003, section
1166 // "7.1.2 Data element structure with explicit vr", p 27) must be
1167 // skipped before proceeding on reading the length on 4 bytes.
1168 Fp->seekg( 2L, std::ios::cur);
1169 uint32_t length32 = ReadInt32();
1171 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1176 lengthOB = FindDocEntryLengthOBOrOW();
1178 catch ( FormatUnexpected )
1180 // Computing the length failed (this happens with broken
1181 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1182 // chance to get the pixels by deciding the element goes
1183 // until the end of the file. Hence we artificially fix the
1184 // the length and proceed.
1185 long currentPosition = Fp->tellg();
1186 Fp->seekg(0L,std::ios::end);
1188 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1189 Fp->seekg(currentPosition, std::ios::beg);
1191 entry->SetReadLength(lengthUntilEOF);
1192 entry->SetLength(lengthUntilEOF);
1195 entry->SetReadLength(lengthOB);
1196 entry->SetLength(lengthOB);
1199 FixDocEntryFoundLength(entry, length32);
1203 // Length is encoded on 2 bytes.
1204 length16 = ReadInt16();
1206 // FIXME : This heuristic supposes that the first group following
1207 // group 0002 *has* and element 0000.
1208 // BUT ... Element 0000 is optionnal :-(
1211 // Fixed using : HandleOutOfGroup0002()
1212 // (first hereafter strategy ...)
1214 // We can tell the current file is encoded in big endian (like
1215 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1216 // and it's value is the one of the encoding of a big endian file.
1217 // In order to deal with such big endian encoded files, we have
1218 // (at least) two strategies:
1219 // * when we load the "Transfer Syntax" tag with value of big endian
1220 // encoding, we raise the proper flags. Then we wait for the end
1221 // of the META group (0x0002) among which is "Transfer Syntax",
1222 // before switching the swap code to big endian. We have to postpone
1223 // the switching of the swap code since the META group is fully encoded
1224 // in little endian, and big endian coding only starts at the next
1225 // group. The corresponding code can be hard to analyse and adds
1226 // many additional unnecessary tests for regular tags.
1227 // * the second strategy consists in waiting for trouble, that shall
1228 // appear when we find the first group with big endian encoding. This
1229 // is easy to detect since the length of a "Group Length" tag (the
1230 // ones with zero as element number) has to be of 4 (0x0004). When we
1231 // encounter 1024 (0x0400) chances are the encoding changed and we
1232 // found a group with big endian encoding.
1233 //---> Unfortunately, element 0000 is optional.
1234 //---> This will not work when missing!
1235 // We shall use this second strategy. In order to make sure that we
1236 // can interpret the presence of an apparently big endian encoded
1237 // length of a "Group Length" without committing a big mistake, we
1238 // add an additional check: we look in the already parsed elements
1239 // for the presence of a "Transfer Syntax" whose value has to be "big
1240 // endian encoding". When this is the case, chances are we have got our
1241 // hands on a big endian encoded file: we switch the swap code to
1242 // big endian and proceed...
1244 // if ( element == 0x0000 && length16 == 0x0400 )
1246 // std::string ts = GetTransferSyntax();
1247 // if ( Global::GetTS()->GetSpecialTransferSyntax(ts)
1248 // != TS::ExplicitVRBigEndian )
1250 // throw FormatError( "Document::FindDocEntryLength()",
1251 // " not explicit VR." );
1255 // SwitchByteSwapCode();
1257 // // Restore the unproperly loaded values i.e. the group, the element
1258 // // and the dictionary entry depending on them.
1259 // uint16_t correctGroup = SwapShort( entry->GetGroup() );
1260 // uint16_t correctElem = SwapShort( entry->GetElement() );
1261 // DictEntry *newTag = GetDictEntry( correctGroup, correctElem );
1264 // // This correct tag is not in the dictionary. Create a new one.
1265 // newTag = NewVirtualDictEntry(correctGroup, correctElem);
1267 // // FIXME this can create a memory leaks on the old entry that be
1268 // // left unreferenced.
1269 // entry->SetDictEntry( newTag );
1272 // 0xffff means that we deal with 'No Length' Sequence
1273 // or 'No Length' SQItem
1274 if ( length16 == 0xffff)
1278 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1283 // Either implicit VR or a non DICOM conformal (see note below) explicit
1284 // VR that ommited the VR of (at least) this element. Farts happen.
1285 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1286 // on Data elements "Implicit and Explicit VR Data Elements shall
1287 // not coexist in a Data Set and Data Sets nested within it".]
1288 // Length is on 4 bytes.
1290 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1291 // even if Transfer Syntax is 'Implicit VR ...'
1293 FixDocEntryFoundLength( entry, ReadInt32() );
1299 * \brief Find the Length till the next sequence delimiter
1300 * \warning NOT end user intended method !
1303 uint32_t Document::FindDocEntryLengthOBOrOW()
1304 throw( FormatUnexpected )
1306 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1307 long positionOnEntry = Fp->tellg();
1308 bool foundSequenceDelimiter = false;
1309 uint32_t totalLength = 0;
1311 while ( !foundSequenceDelimiter )
1317 group = ReadInt16();
1320 catch ( FormatError )
1322 throw FormatError("Unexpected end of file encountered during ",
1323 "Document::FindDocEntryLengthOBOrOW()");
1325 // We have to decount the group and element we just read
1327 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
1329 long filePosition = Fp->tellg();
1330 gdcmVerboseMacro( "Neither an Item tag nor a Sequence delimiter tag on :"
1331 << std::hex << group << " , " << elem
1332 << ") -before- position x(" << filePosition << ")" );
1334 Fp->seekg(positionOnEntry, std::ios::beg);
1335 throw FormatUnexpected( "Neither an Item tag nor a Sequence delimiter tag.");
1337 if ( elem == 0xe0dd )
1339 foundSequenceDelimiter = true;
1341 uint32_t itemLength = ReadInt32();
1342 // We add 4 bytes since we just read the ItemLength with ReadInt32
1343 totalLength += itemLength + 4;
1344 SkipBytes(itemLength);
1346 if ( foundSequenceDelimiter )
1351 Fp->seekg( positionOnEntry, std::ios::beg);
1356 * \brief Find the Value Representation of the current Dicom Element.
1357 * @return Value Representation of the current Entry
1359 std::string Document::FindDocEntryVR()
1361 if ( Filetype != ExplicitVR )
1362 return GDCM_UNKNOWN;
1364 long positionOnEntry = Fp->tellg();
1365 // Warning: we believe this is explicit VR (Value Representation) because
1366 // we used a heuristic that found "UL" in the first tag. Alas this
1367 // doesn't guarantee that all the tags will be in explicit VR. In some
1368 // cases (see e-film filtered files) one finds implicit VR tags mixed
1369 // within an explicit VR file. Hence we make sure the present tag
1370 // is in explicit VR and try to fix things if it happens not to be
1374 Fp->read (vr, (size_t)2);
1377 if( !CheckDocEntryVR(vr) )
1379 Fp->seekg(positionOnEntry, std::ios::beg);
1380 return GDCM_UNKNOWN;
1386 * \brief Check the correspondance between the VR of the header entry
1387 * and the taken VR. If they are different, the header entry is
1388 * updated with the new VR.
1389 * @param vr Dicom Value Representation
1390 * @return false if the VR is incorrect of if the VR isn't referenced
1391 * otherwise, it returns true
1393 bool Document::CheckDocEntryVR(VRKey vr)
1395 // CLEANME searching the dicom_vr at each occurence is expensive.
1396 // PostPone this test in an optional integrity check at the end
1397 // of parsing or only in debug mode.
1398 if ( !Global::GetVR()->IsValidVR(vr) )
1405 * \brief Get the transformed value of the header entry. The VR value
1406 * is used to define the transformation to operate on the value
1407 * \warning NOT end user intended method !
1408 * @param entry entry to tranform
1409 * @return Transformed entry value
1411 std::string Document::GetDocEntryValue(DocEntry *entry)
1413 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1415 std::string val = ((ValEntry *)entry)->GetValue();
1416 std::string vr = entry->GetVR();
1417 uint32_t length = entry->GetLength();
1418 std::ostringstream s;
1421 // When short integer(s) are expected, read and convert the following
1422 // n * 2 bytes properly i.e. as a multivaluated strings
1423 // (each single value is separated fromthe next one by '\'
1424 // as usual for standard multivaluated filels
1425 // Elements with Value Multiplicity > 1
1426 // contain a set of short integers (not a single one)
1428 if( vr == "US" || vr == "SS" )
1433 for (int i=0; i < nbInt; i++)
1439 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1440 newInt16 = SwapShort( newInt16 );
1445 // When integer(s) are expected, read and convert the following
1446 // n * 4 bytes properly i.e. as a multivaluated strings
1447 // (each single value is separated fromthe next one by '\'
1448 // as usual for standard multivaluated filels
1449 // Elements with Value Multiplicity > 1
1450 // contain a set of integers (not a single one)
1451 else if( vr == "UL" || vr == "SL" )
1456 for (int i=0; i < nbInt; i++)
1462 newInt32 = ( val[4*i+0] & 0xFF )
1463 + (( val[4*i+1] & 0xFF ) << 8 )
1464 + (( val[4*i+2] & 0xFF ) << 16 )
1465 + (( val[4*i+3] & 0xFF ) << 24 );
1466 newInt32 = SwapLong( newInt32 );
1470 #ifdef GDCM_NO_ANSI_STRING_STREAM
1471 s << std::ends; // to avoid oddities on Solaris
1472 #endif //GDCM_NO_ANSI_STRING_STREAM
1475 return ((ValEntry *)entry)->GetValue();
1479 * \brief Get the reverse transformed value of the header entry. The VR
1480 * value is used to define the reverse transformation to operate on
1482 * \warning NOT end user intended method !
1483 * @param entry Entry to reverse transform
1484 * @return Reverse transformed entry value
1486 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1488 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1490 std::string vr = entry->GetVR();
1491 std::vector<std::string> tokens;
1492 std::ostringstream s;
1494 if ( vr == "US" || vr == "SS" )
1498 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1499 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1500 for (unsigned int i=0; i<tokens.size(); i++)
1502 newInt16 = atoi(tokens[i].c_str());
1503 s << ( newInt16 & 0xFF )
1504 << (( newInt16 >> 8 ) & 0xFF );
1508 if ( vr == "UL" || vr == "SL")
1512 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1513 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1514 for (unsigned int i=0; i<tokens.size();i++)
1516 newInt32 = atoi(tokens[i].c_str());
1517 s << (char)( newInt32 & 0xFF )
1518 << (char)(( newInt32 >> 8 ) & 0xFF )
1519 << (char)(( newInt32 >> 16 ) & 0xFF )
1520 << (char)(( newInt32 >> 24 ) & 0xFF );
1525 #ifdef GDCM_NO_ANSI_STRING_STREAM
1526 s << std::ends; // to avoid oddities on Solaris
1527 #endif //GDCM_NO_ANSI_STRING_STREAM
1531 return ((ValEntry *)entry)->GetValue();
1535 * \brief Skip a given Header Entry
1536 * \warning NOT end user intended method !
1537 * @param entry entry to skip
1539 void Document::SkipDocEntry(DocEntry *entry)
1541 SkipBytes(entry->GetLength());
1545 * \brief Skips to the beginning of the next Header Entry
1546 * \warning NOT end user intended method !
1547 * @param currentDocEntry entry to skip
1549 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1551 Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg);
1552 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1553 Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur);
1557 * \brief When the length of an element value is obviously wrong (because
1558 * the parser went Jabberwocky) one can hope improving things by
1559 * applying some heuristics.
1560 * @param entry entry to check
1561 * @param foundLength first assumption about length
1563 void Document::FixDocEntryFoundLength(DocEntry *entry,
1564 uint32_t foundLength)
1566 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1567 if ( foundLength == 0xffffffff)
1572 uint16_t gr = entry->GetGroup();
1573 uint16_t elem = entry->GetElement();
1575 if ( foundLength % 2)
1577 gdcmVerboseMacro( "Warning : Tag with uneven length " << foundLength
1578 << " in x(" << std::hex << gr << "," << elem <<")");
1581 //////// Fix for some naughty General Electric images.
1582 // Allthough not recent many such GE corrupted images are still present
1583 // on Creatis hard disks. Hence this fix shall remain when such images
1584 // are no longer in use (we are talking a few years, here)...
1585 // Note: XMedCom probably uses such a trick since it is able to read
1586 // those pesky GE images ...
1587 if ( foundLength == 13)
1589 // Only happens for this length !
1590 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1593 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
1597 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1598 // Occurence of such images is quite low (unless one leaves close to a
1599 // 'Leonardo' source. Hence, one might consider commenting out the
1600 // following fix on efficiency reasons.
1601 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1604 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
1607 else if ( entry->GetVR() == "SQ" )
1609 foundLength = 0; // ReadLength is unchanged
1612 //////// We encountered a 'delimiter' element i.e. a tag of the form
1613 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1614 // taken into account.
1615 else if( gr == 0xfffe )
1617 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1618 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1619 // causes extra troubles...
1620 if( entry->GetElement() != 0x0000 )
1625 entry->SetLength(foundLength);
1629 * \brief Apply some heuristics to predict whether the considered
1630 * element value contains/represents an integer or not.
1631 * @param entry The element value on which to apply the predicate.
1632 * @return The result of the heuristical predicate.
1634 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1636 uint16_t elem = entry->GetElement();
1637 uint16_t group = entry->GetGroup();
1638 const std::string &vr = entry->GetVR();
1639 uint32_t length = entry->GetLength();
1641 // When we have some semantics on the element we just read, and if we
1642 // a priori know we are dealing with an integer, then we shall be
1643 // able to swap it's element value properly.
1644 if ( elem == 0 ) // This is the group length of the group
1652 // Allthough this should never happen, still some images have a
1653 // corrupted group length [e.g. have a glance at offset x(8336) of
1654 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
1655 // Since for dicom compliant and well behaved headers, the present
1656 // test is useless (and might even look a bit paranoid), when we
1657 // encounter such an ill-formed image, we simply display a warning
1658 // message and proceed on parsing (while crossing fingers).
1659 long filePosition = Fp->tellg();
1660 gdcmVerboseMacro( "Erroneous Group Length element length on : ("
1661 << std::hex << group << " , " << elem
1662 << ") -before- position x(" << filePosition << ")"
1663 << "lgt : " << length );
1667 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1675 * \brief Loads all the needed Dictionaries
1676 * \warning NOT end user intended method !
1678 void Document::Initialize()
1680 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
1686 * \brief Discover what the swap code is (among little endian, big endian,
1687 * bad little endian, bad big endian).
1689 * @return false when we are absolutely sure
1690 * it's neither ACR-NEMA nor DICOM
1691 * true when we hope ours assuptions are OK
1693 bool Document::CheckSwap()
1695 // The only guaranted way of finding the swap code is to find a
1696 // group tag since we know it's length has to be of four bytes i.e.
1697 // 0x00000004. Finding the swap code in then straigthforward. Trouble
1698 // occurs when we can't find such group...
1700 uint32_t x = 4; // x : for ntohs
1701 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
1707 // First, compare HostByteOrder and NetworkByteOrder in order to
1708 // determine if we shall need to swap bytes (i.e. the Endian type).
1709 if ( x == ntohs(x) )
1718 // The easiest case is the one of a 'true' DICOM header, we just have
1719 // to look for the string "DICM" inside the file preamble.
1722 char *entCur = deb + 128;
1723 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
1725 gdcmVerboseMacro( "Looks like DICOM Version3 (preamble + DCM)" );
1727 // Group 0002 should always be VR, and the first element 0000
1728 // Let's be carefull (so many wrong headers ...)
1729 // and determine the value representation (VR) :
1730 // Let's skip to the first element (0002,0000) and check there if we find
1731 // "UL" - or "OB" if the 1st one is (0002,0001) -,
1732 // in which case we (almost) know it is explicit VR.
1733 // WARNING: if it happens to be implicit VR then what we will read
1734 // is the length of the group. If this ascii representation of this
1735 // length happens to be "UL" then we shall believe it is explicit VR.
1736 // We need to skip :
1737 // * the 128 bytes of File Preamble (often padded with zeroes),
1738 // * the 4 bytes of "DICM" string,
1739 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
1740 // i.e. a total of 136 bytes.
1743 // group 0x0002 *is always* Explicit VR Sometimes ,
1744 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
1745 // *Implicit* VR (see former 'gdcmData/icone.dcm')
1747 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
1748 memcmp(entCur, "OB", (size_t)2) == 0 ||
1749 memcmp(entCur, "UI", (size_t)2) == 0 ||
1750 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
1751 // when Write DCM *adds*
1753 // Use Document::dicom_vr to test all the possibilities
1754 // instead of just checking for UL, OB and UI !? group 0000
1756 Filetype = ExplicitVR;
1757 gdcmVerboseMacro( "Group 0002 : Explicit Value Representation");
1761 Filetype = ImplicitVR;
1762 gdcmVerboseMacro( "Group 0002 :Not an explicit Value Representation;"
1763 << "Looks like a bugged Header!");
1769 gdcmVerboseMacro( "HostByteOrder != NetworkByteOrder");
1774 gdcmVerboseMacro( "HostByteOrder = NetworkByteOrder");
1777 // Position the file position indicator at first tag
1778 // (i.e. after the file preamble and the "DICM" string).
1779 Fp->seekg(0, std::ios::beg);
1780 Fp->seekg ( 132L, std::ios::beg);
1784 // Alas, this is not a DicomV3 file and whatever happens there is no file
1785 // preamble. We can reset the file position indicator to where the data
1786 // is (i.e. the beginning of the file).
1787 gdcmVerboseMacro( "Not a DICOM Version3 file");
1788 Fp->seekg(0, std::ios::beg);
1790 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
1791 // By clean we mean that the length of the first tag is written down.
1792 // If this is the case and since the length of the first group HAS to be
1793 // four (bytes), then determining the proper swap code is straightforward.
1796 // We assume the array of char we are considering contains the binary
1797 // representation of a 32 bits integer. Hence the following dirty
1799 s32 = *((uint32_t *)(entCur));
1820 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
1821 // It is time for despaired wild guesses.
1822 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
1823 // i.e. the 'group length' element is not present :
1825 // check the supposed-to-be 'group number'
1826 // in ( 0x0001 .. 0x0008 )
1827 // to determine ' SwapCode' value .
1828 // Only 0 or 4321 will be possible
1829 // (no oportunity to check for the formerly well known
1830 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
1831 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
1832 // the file IS NOT ACR-NEMA nor DICOM V3
1833 // Find a trick to tell it the caller...
1835 s16 = *((uint16_t *)(deb));
1862 gdcmVerboseMacro( "ACR/NEMA unfound swap info (Really hopeless !)");
1870 * \brief Change the Byte Swap code.
1872 void Document::SwitchByteSwapCode()
1874 gdcmVerboseMacro( "Switching Byte Swap code from "<< SwapCode);
1875 if ( SwapCode == 1234 )
1879 else if ( SwapCode == 4321 )
1883 else if ( SwapCode == 3412 )
1887 else if ( SwapCode == 2143 )
1894 * \brief during parsing, Header Elements too long are not loaded in memory
1897 void Document::SetMaxSizeLoadEntry(long newSize)
1903 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
1905 MaxSizeLoadEntry = 0xffffffff;
1908 MaxSizeLoadEntry = newSize;
1912 * \brief Header Elements too long will not be printed
1913 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
1916 void Document::SetMaxSizePrintEntry(long newSize)
1922 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
1924 MaxSizePrintEntry = 0xffffffff;
1927 MaxSizePrintEntry = newSize;
1932 * \brief Read the next tag but WITHOUT loading it's value
1933 * (read the 'Group Number', the 'Element Number',
1934 * gets the Dict Entry
1935 * gets the VR, gets the length, gets the offset value)
1936 * @return On succes the newly created DocEntry, NULL on failure.
1938 DocEntry *Document::ReadNextDocEntry()
1945 group = ReadInt16();
1948 catch ( FormatError e )
1950 // We reached the EOF (or an error occured) therefore
1951 // header parsing has to be considered as finished.
1956 // Sometimes file contains groups of tags with reversed endianess.
1957 HandleBrokenEndian(group, elem);
1959 // In 'true DICOM' files Group 0002 is always little endian
1960 if ( HasDCMPreamble )
1961 HandleOutOfGroup0002(group, elem);
1963 std::string vr = FindDocEntryVR();
1964 std::string realVR = vr;
1966 if( vr == GDCM_UNKNOWN)
1968 DictEntry *dictEntry = GetDictEntry(group,elem);
1970 realVR = dictEntry->GetVR();
1974 if( Global::GetVR()->IsVROfSequence(realVR) )
1975 newEntry = NewSeqEntry(group, elem);
1976 else if( Global::GetVR()->IsVROfStringRepresentable(realVR) )
1977 newEntry = NewValEntry(group, elem,vr);
1979 newEntry = NewBinEntry(group, elem,vr);
1981 if( vr == GDCM_UNKNOWN )
1983 if( Filetype == ExplicitVR )
1985 // We thought this was explicit VR, but we end up with an
1986 // implicit VR tag. Let's backtrack.
1987 if ( newEntry->GetGroup() != 0xfffe )
1990 msg = Util::Format("Entry (%04x,%04x) should be Explicit VR\n",
1991 newEntry->GetGroup(), newEntry->GetElement());
1992 gdcmVerboseMacro( msg.c_str() );
1995 newEntry->SetImplicitVR();
2000 FindDocEntryLength(newEntry);
2002 catch ( FormatError e )
2010 newEntry->SetOffset(Fp->tellg());
2016 * \brief Handle broken private tag from Philips NTSCAN
2017 * where the endianess is being switch to BigEndian for no
2021 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
2023 // Endian reversion. Some files contain groups of tags with reversed endianess.
2024 static int reversedEndian = 0;
2025 // try to fix endian switching in the middle of headers
2026 if ((group == 0xfeff) && (elem == 0x00e0))
2028 // start endian swap mark for group found
2030 SwitchByteSwapCode();
2035 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2037 // end of reversed endian group
2039 SwitchByteSwapCode();
2044 * \brief Group 0002 is always coded Little Endian
2045 * whatever Transfer Syntax is
2048 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2050 // Endian reversion. Some files contain groups of tags with reversed endianess.
2051 if ( !Group0002Parsed && group != 0x0002)
2053 Group0002Parsed = true;
2054 // we just came out of group 0002
2055 // if Transfer syntax is Big Endian we have to change CheckSwap
2057 std::string ts = GetTransferSyntax();
2058 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2060 gdcmVerboseMacro("True DICOM File, with NO Tansfer Syntax: " << ts );
2064 // Group 0002 is always 'Explicit ...' enven when Transfer Syntax says 'Implicit ..."
2066 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ImplicitVRLittleEndian )
2068 Filetype = ImplicitVR;
2071 // FIXME Strangely, this works with
2072 //'Implicit VR Transfer Syntax (GE Private)
2073 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ExplicitVRBigEndian )
2075 gdcmVerboseMacro("Transfer Syntax Name = ["
2076 << GetTransferSyntaxName() << "]" );
2077 SwitchByteSwapCode();
2078 group = SwapShort(group);
2079 elem = SwapShort(elem);
2084 // GenerateFreeTagKeyInGroup?
2085 // --> What was it designed for ?!?
2087 * \brief Generate a free TagKey i.e. a TagKey that is not present
2088 * in the TagHt dictionary.
2089 * @param group The generated tag must belong to this group.
2090 * @return The element of tag with given group which is fee.
2092 //uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2094 // for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2096 // TagKey key = DictEntry::TranslateToKey(group, elem);
2097 // if (TagHT.count(key) == 0)
2102 // return UINT32_MAX;
2106 * \brief Compares two documents, according to \ref DicomDir rules
2107 * \warning Does NOT work with ACR-NEMA files
2108 * \todo Find a trick to solve the pb (use RET fields ?)
2110 * @return true if 'smaller'
2112 bool Document::operator<(Document &document)
2115 std::string s1 = GetEntryValue(0x0010,0x0010);
2116 std::string s2 = document.GetEntryValue(0x0010,0x0010);
2128 s1 = GetEntryValue(0x0010,0x0020);
2129 s2 = document.GetEntryValue(0x0010,0x0020);
2140 // Study Instance UID
2141 s1 = GetEntryValue(0x0020,0x000d);
2142 s2 = document.GetEntryValue(0x0020,0x000d);
2153 // Serie Instance UID
2154 s1 = GetEntryValue(0x0020,0x000e);
2155 s2 = document.GetEntryValue(0x0020,0x000e);
2171 * \brief Walk recursively the given \ref DocEntrySet, and feed
2172 * the given hash table (\ref TagDocEntryHT) with all the
2173 * \ref DocEntry (Dicom entries) encountered.
2174 * This method does the job for \ref BuildFlatHashTable.
2175 * @param builtHT Where to collect all the \ref DocEntry encountered
2176 * when recursively walking the given set.
2177 * @param set The structure to be traversed (recursively).
2179 /*void Document::BuildFlatHashTableRecurse( TagDocEntryHT &builtHT,
2182 if (ElementSet *elementSet = dynamic_cast< ElementSet* > ( set ) )
2184 TagDocEntryHT const ¤tHT = elementSet->GetTagHT();
2185 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2186 i != currentHT.end();
2189 DocEntry *entry = i->second;
2190 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2192 const ListSQItem& items = seqEntry->GetSQItems();
2193 for( ListSQItem::const_iterator item = items.begin();
2194 item != items.end();
2197 BuildFlatHashTableRecurse( builtHT, *item );
2201 builtHT[entry->GetKey()] = entry;
2206 if (SQItem *SQItemSet = dynamic_cast< SQItem* > ( set ) )
2208 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
2209 for (ListDocEntry::const_iterator i = currentList.begin();
2210 i != currentList.end();
2213 DocEntry *entry = *i;
2214 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2216 const ListSQItem& items = seqEntry->GetSQItems();
2217 for( ListSQItem::const_iterator item = items.begin();
2218 item != items.end();
2221 BuildFlatHashTableRecurse( builtHT, *item );
2225 builtHT[entry->GetKey()] = entry;
2232 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2235 * The structure used by a Document (through \ref ElementSet),
2236 * in order to hold the parsed entries of a Dicom header, is a recursive
2237 * one. This is due to the fact that the sequences (when present)
2238 * can be nested. Additionaly, the sequence items (represented in
2239 * gdcm as \ref SQItem) add an extra complexity to the data
2240 * structure. Hence, a gdcm user whishing to visit all the entries of
2241 * a Dicom header will need to dig in the gdcm internals (which
2242 * implies exposing all the internal data structures to the API).
2243 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2244 * recursively builds a temporary hash table, which holds all the
2245 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
2247 * \warning Of course there is NO integrity constrain between the
2248 * returned \ref TagDocEntryHT and the \ref ElementSet used
2249 * to build it. Hence if the underlying \ref ElementSet is
2250 * altered, then it is the caller responsability to invoke
2251 * \ref BuildFlatHashTable again...
2252 * @return The flat std::map<> we juste build.
2254 /*TagDocEntryHT *Document::BuildFlatHashTable()
2256 TagDocEntryHT *FlatHT = new TagDocEntryHT;
2257 BuildFlatHashTableRecurse( *FlatHT, this );
2261 } // end namespace gdcm
2263 //-----------------------------------------------------------------------------