1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/10/18 11:10:45 $
7 Version: $Revision: 1.292 $
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 "gdcmSeqEntry.h"
21 #include "gdcmGlobal.h"
23 #include "gdcmDebug.h"
25 #include "gdcmDictSet.h"
26 #include "gdcmDocEntrySet.h"
27 #include "gdcmSQItem.h"
28 #include "gdcmDataEntry.h"
33 #include <ctype.h> // for isdigit
34 #include <stdlib.h> // for atoi
38 //-----------------------------------------------------------------------------
40 // Refer to Document::SetMaxSizeLoadEntry()
41 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
43 //-----------------------------------------------------------------------------
44 // Constructor / Destructor
45 // Constructors and destructors are protected to avoid user to invoke directly
48 * \brief This default constructor neither loads nor parses the file.
49 * You should then invoke \ref Document::Load.
57 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
60 Filetype = ExplicitVR;
61 // Load will set it to true if sucessfull
62 Group0002Parsed = false;
63 IsDocumentAlreadyLoaded = false;
64 IsDocumentModified = true;
65 LoadMode = LD_ALL; // default : load everything, later
70 * \brief Canonical destructor.
72 Document::~Document ()
77 //-----------------------------------------------------------------------------
81 * \brief Loader. use SetLoadMode(), SetFileName() before !
82 * @return false if file cannot be open or no swap info was found,
83 * or no tag was found.
85 bool Document::Load( )
87 if ( GetFileName() == "" )
89 gdcmWarningMacro( "Use SetFileName, before !" );
92 return DoTheLoadingDocumentJob( );
95 * \brief Loader. (DEPRECATED : not to break the API)
96 * @param fileName 'Document' (File or DicomDir) to be open for parsing
97 * @return false if file cannot be open or no swap info was found,
98 * or no tag was found.
100 bool Document::Load( std::string const &fileName )
103 return DoTheLoadingDocumentJob( );
107 * \brief Performs the Loading Job (internal use only)
108 * @return false if file cannot be open or no swap info was found,
109 * or no tag was found.
111 bool Document::DoTheLoadingDocumentJob( )
113 if ( ! IsDocumentModified ) // Nothing to do !
121 // warning already performed in OpenFile()
122 //gdcmWarningMacro( "Unable to open as an ACR/DICOM file: "
123 // << Filename.c_str() );
128 Group0002Parsed = false;
130 gdcmWarningMacro( "Starting parsing of file: " << Filename.c_str());
132 Fp->seekg(0, std::ios::end);
133 long lgt = Fp->tellg(); // total length of the file
135 Fp->seekg(0, std::ios::beg);
137 // CheckSwap returns a boolean
138 // (false if no swap info of any kind was found)
141 gdcmWarningMacro( "Neither a DICOM V3 nor an ACR-NEMA file: "
142 << Filename.c_str());
147 long beg = Fp->tellg(); // just after DICOM preamble (if any)
149 lgt -= beg; // remaining length to parse
152 // Loading is done during parsing
153 ParseDES( this, beg, lgt, false); // delim_mode is first defaulted to false
157 gdcmWarningMacro( "No tag in internal hash table for: "
158 << Filename.c_str());
162 IsDocumentAlreadyLoaded = true;
164 Fp->seekg( 0, std::ios::beg);
166 // Load 'non string' values
168 std::string PhotometricInterpretation = GetEntryString(0x0028,0x0004);
169 if ( PhotometricInterpretation == "PALETTE COLOR " )
172 // Probabely this line should be outside the 'if'
173 // Try to find an image sample holding a 'gray LUT'
174 LoadEntryBinArea(0x0028,0x1200); // gray LUT
177 /// The tags refered by the three following lines used to be CORRECTLY
178 /// defined as having an US Value Representation in the public
179 /// dictionary. BUT the semantics implied by the three following
180 /// lines state that the corresponding tag contents are in fact
181 /// the ones of a DataEntry.
182 /// In order to fix things "Quick and Dirty" the dictionary was
183 /// altered on PURPOSE but now contains a WRONG value.
184 /// In order to fix things and restore the dictionary to its
185 /// correct value, one needs to decided of the semantics by deciding
186 /// whether the following tags are either :
187 /// - multivaluated US, and hence loaded as DataEntry, but afterwards
188 /// also used as DataEntry, which requires the proper conversion,
189 /// - OW, and hence loaded as DataEntry, but afterwards also used
190 /// as DataEntry, which requires the proper conversion.
191 LoadEntryBinArea(0x0028,0x1201); // R LUT
192 LoadEntryBinArea(0x0028,0x1202); // G LUT
193 LoadEntryBinArea(0x0028,0x1203); // B LUT
195 // Segmented Red Palette Color LUT Data
196 LoadEntryBinArea(0x0028,0x1221);
197 // Segmented Green Palette Color LUT Data
198 LoadEntryBinArea(0x0028,0x1222);
199 // Segmented Blue Palette Color LUT Data
200 LoadEntryBinArea(0x0028,0x1223);
203 //FIXME later : how to use it?
204 SeqEntry *modLutSeq = GetSeqEntry(0x0028,0x3000);
207 SQItem *sqi= modLutSeq->GetFirstSQItem();
210 DataEntry *dataEntry = sqi->GetDataEntry(0x0028,0x3006);
211 if ( dataEntry != 0 )
213 if ( dataEntry->GetLength() != 0 )
215 LoadEntryBinArea(dataEntry); //LUT Data (CTX dependent)
221 // Force Loading some more elements if user asked to.
224 for (ListElements::iterator it = UserForceLoadList.begin();
225 it != UserForceLoadList.end();
228 gdcmWarningMacro( "Force Load " << std::hex
229 << (*it).Group << "|" <<(*it).Elem );
231 d = GetDocEntry( (*it).Group, (*it).Elem);
235 gdcmWarningMacro( "You asked toForce Load " << std::hex
236 << (*it).Group <<"|"<< (*it).Elem
237 << " that doesn't exist" );
241 LoadDocEntry(d, true);
246 // ----------------------------
247 // Specific code to allow gdcm to read ACR-LibIDO formated images
248 // Note: ACR-LibIDO is an extension of the ACR standard that was
249 // used at CREATIS. For the time being (say a couple of years)
250 // we keep this kludge to allow CREATIS users
251 // reading their old images.
253 // if recognition code tells us we deal with a LibIDO image
254 // we switch lineNumber and columnNumber
257 RecCode = GetEntryString(0x0008, 0x0010); // recognition code (RET)
258 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
259 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
260 // with "little-endian strings"
262 Filetype = ACR_LIBIDO;
263 std::string rows = GetEntryString(0x0028, 0x0010);
264 std::string columns = GetEntryString(0x0028, 0x0011);
265 SetEntryString(columns, 0x0028, 0x0010);
266 SetEntryString(rows , 0x0028, 0x0011);
268 // --- End of ACR-LibIDO kludge ---
274 * \brief Adds a new element we want to load anyway
275 * @param group Group number of the target tag.
276 * @param elem Element number of the target tag.
278 void Document::AddForceLoadElement (uint16_t group, uint16_t elem)
283 UserForceLoadList.push_back(el);
286 * \brief Get the public dictionary used
288 Dict *Document::GetPubDict()
294 * \brief Get the shadow dictionary used
296 Dict *Document::GetShaDict()
302 * \brief Set the shadow dictionary used
303 * @param dict dictionary to use in shadow
305 bool Document::SetShaDict(Dict *dict)
312 * \brief Set the shadow dictionary used
313 * @param dictName name of the dictionary to use in shadow
315 bool Document::SetShaDict(DictKey const &dictName)
317 RefShaDict = Global::GetDicts()->GetDict(dictName);
322 * \brief This predicate tells us whether or not the current Document
323 * was properly parsed and contains at least *one* Dicom Element
324 * (and nothing more, sorry).
325 * @return false when we're 150 % sure it's NOT a Dicom/Acr file,
328 bool Document::IsReadable()
330 if ( Filetype == Unknown )
332 gdcmWarningMacro( "Wrong filetype");
338 gdcmWarningMacro( "No tag in internal hash table.");
346 * \brief Predicate for dicom version 3 file.
347 * @return True when the file is a dicom version 3.
349 bool Document::IsDicomV3()
351 // Checking if Transfer Syntax exists is enough
352 // Anyway, it's too late check if the 'Preamble' was found ...
353 // And ... would it be a rich idea to check ?
354 // (some 'no Preamble' DICOM images exist !)
355 return GetDocEntry(0x0002, 0x0010) != NULL;
359 * \brief Predicate for Papyrus file
360 * Dedicated to whomsoever it may concern
361 * @return True when the file is a Papyrus file.
363 bool Document::IsPapyrus()
365 // check for Papyrus private Sequence
366 DocEntry *e = GetDocEntry(0x0041, 0x1050);
369 // check if it's actually a Sequence
370 if ( !dynamic_cast<SeqEntry*>(e) )
376 * \brief returns the File Type
377 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
378 * @return the FileType code
380 FileType Document::GetFileType()
386 * \brief Accessor to the Transfer Syntax (when present) of the
387 * current document (it internally handles reading the
388 * value from disk when only parsing occured).
389 * @return The encountered Transfer Syntax of the current document.
391 std::string Document::GetTransferSyntax()
393 DocEntry *entry = GetDocEntry(0x0002, 0x0010);
399 // The entry might be present but not loaded (parsing and loading
400 // happen at different stages): try loading and proceed with check...
401 LoadDocEntrySafe(entry);
402 if (DataEntry *dataEntry = dynamic_cast<DataEntry *>(entry) )
404 std::string transfer = dataEntry->GetString();
405 // The actual transfer (as read from disk) might be padded. We
406 // first need to remove the potential padding. We can make the
407 // weak assumption that padding was not executed with digits...
408 if ( transfer.length() == 0 )
410 // for brain damaged headers
413 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
415 transfer.erase(transfer.length()-1, 1);
416 if ( transfer.length() == 0 )
418 // for brain damaged headers
428 * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS
429 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
431 std::string Document::GetTransferSyntaxName()
433 // use the TS (TS : Transfer Syntax)
434 std::string transferSyntax = GetEntryString(0x0002,0x0010);
436 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
438 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
439 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
440 return "Uncompressed ACR-NEMA";
442 if ( transferSyntax == GDCM_UNFOUND )
444 gdcmWarningMacro( "Unfound Transfer Syntax (0002,0010)");
445 return "Uncompressed ACR-NEMA";
448 // we do it only when we need it
449 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
451 // Global::GetTS() is a global static you shall never try to delete it!
455 // --------------- Swap Code ------------------
457 * \brief Swaps the bytes so they agree with the processor order
458 * @return The properly swaped 16 bits integer.
460 uint16_t Document::SwapShort(uint16_t a)
462 if ( SwapCode == 4321 || SwapCode == 2143 )
464 //a = ((( a << 8 ) & 0xff00 ) | (( a >> 8 ) & 0x00ff ) );
466 a = ( a << 8 ) | ( a >> 8 );
472 * \brief Swaps back the bytes of 4-byte long integer accordingly to
474 * @return The properly swaped 32 bits integer.
476 uint32_t Document::SwapLong(uint32_t a)
483 // a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
484 // ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
486 a=( ( a<<24) | ((a<<8) & 0x00ff0000) |
487 ((a>>8) & 0x0000ff00) | (a>>24) );
490 // a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
491 a=( (a<<16) | (a>>16) );
494 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
497 gdcmErrorMacro( "Unset swap code:" << SwapCode );
504 // -----------------File I/O ---------------
506 * \brief Tries to open the file \ref Document::Filename and
507 * checks the preamble when existing.
508 * @return The FILE pointer on success.
510 std::ifstream *Document::OpenFile()
512 HasDCMPreamble = false;
513 if (Filename.length() == 0)
520 gdcmWarningMacro( "File already open: " << Filename.c_str());
524 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
527 // Don't user gdcmErrorMacro :
528 // a spurious message will appear when you use, for instance
529 // gdcm::FileHelper *fh = new gdcm::FileHelper( outputFileName );
530 // to create outputFileName.
531 gdcmWarningMacro( "Cannot open file: " << Filename.c_str());
535 //exit(1); // No function is allowed to leave the application instead
536 // of warning the caller
540 Fp->read((char*)&zero, (size_t)2);
547 //-- ACR or DICOM with no Preamble; may start with a Shadow Group --
549 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
550 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
551 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
552 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
554 std::string msg = Util::Format(
555 "ACR/DICOM starting at the beginning of the file:(%04x)\n", zero);
556 gdcmWarningMacro( msg.c_str() );
561 Fp->seekg(126L, std::ios::cur);
562 char dicm[4]; // = {' ',' ',' ',' '};
563 Fp->read(dicm, (size_t)4);
569 if ( memcmp(dicm, "DICM", 4) == 0 )
571 HasDCMPreamble = true;
575 // -- Neither ACR/No Preamble Dicom nor DICOMV3 file
577 gdcmWarningMacro( "Neither ACR/No Preamble Dicom nor DICOMV3 file: "
578 << Filename.c_str());
583 * \brief closes the file
584 * @return TRUE if the close was successfull
586 bool Document::CloseFile()
598 * \brief Writes in a file all the Entries (Dicom Elements)
599 * @param fp file pointer on an already open file (actually: Output File Stream)
600 * @param filetype Type of the File to be written
601 * (ACR-NEMA, ExplicitVR, ImplicitVR)
603 void Document::WriteContent(std::ofstream *fp, FileType filetype)
605 // Skip if user wants to write an ACR-NEMA file
607 if ( filetype == ImplicitVR || filetype == ExplicitVR )
609 // writing Dicom File Preamble
610 char filePreamble[128];
611 memset(filePreamble, 0, 128);
612 fp->write(filePreamble, 128);
613 fp->write("DICM", 4);
617 * \todo rewrite later, if really usefull
618 * - 'Group Length' element is optional in DICOM
619 * - but un-updated odd groups lengthes can causes pb
622 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
623 * UpdateGroupLength(false,filetype);
624 * if ( filetype == ACR)
625 * UpdateGroupLength(true,ACR);
628 ElementSet::WriteContent(fp, filetype); // This one is recursive
631 // -----------------------------------------
634 * \brief Loads (from disk) the element content
635 * when a string is not suitable
636 * @param group group number of the Entry
637 * @param elem element number of the Entry
639 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
641 // Search the corresponding DocEntry
642 DocEntry *docEntry = GetDocEntry(group, elem);
645 gdcmWarningMacro(std::hex << group << "|" << elem
646 << "doesn't exist" );
649 DataEntry *dataEntry = dynamic_cast<DataEntry *>(docEntry);
652 gdcmWarningMacro(std::hex << group << "|" << elem
653 << "is NOT a DataEntry");
656 LoadEntryBinArea(dataEntry);
660 * \brief Loads (from disk) the element content
661 * when a string is not suitable
662 * @param elem Entry whose binArea is going to be loaded
664 void Document::LoadEntryBinArea(DataEntry *entry)
666 if( entry->GetBinArea() )
673 size_t o =(size_t)entry->GetOffset();
674 Fp->seekg(o, std::ios::beg);
676 size_t l = entry->GetLength();
677 uint8_t *data = new uint8_t[l];
680 gdcmWarningMacro( "Cannot allocate DataEntry content for : "
681 << std::hex << entry->GetGroup()
682 << "|" << entry->GetElement() );
687 Fp->read((char*)data, l);
688 if ( Fp->fail() || Fp->eof() )
691 entry->SetState(DataEntry::STATE_UNREAD);
695 // Swap the data content if necessary
697 unsigned short vrLgth = Global::GetVR()->GetAtomicElementLength(entry->GetVR());
698 if( entry->GetVR() == "OW" )
705 uint16_t *data16 = (uint16_t *)data;
706 for(i=0;i<l/vrLgth;i++)
707 data16[i] = SwapShort(data16[i]);
712 uint32_t *data32 = (uint32_t *)data;
713 for(i=0;i<l/vrLgth;i++)
714 data32[i] = SwapLong(data32[i]);
719 gdcmWarningMacro("Can't swap 64 bits datas");
720 /* uint64_t *data64 = (uint64_t *)data;
721 for(i=0;i<l/vrLgth;i++)
722 data64[i] = SwapLongLong(data64[i]);*/
727 entry->SetBinArea(data);
734 * \brief Loads the element while preserving the current
735 * underlying file position indicator as opposed to
736 * LoadDocEntry that modifies it.
737 * @param entry DocEntry whose value will be loaded.
739 void Document::LoadDocEntrySafe(DocEntry *entry)
743 long PositionOnEntry = Fp->tellg();
745 Fp->seekg(PositionOnEntry, std::ios::beg);
750 * \brief Compares two documents, according to \ref DicomDir rules
751 * \warning Does NOT work with ACR-NEMA files
752 * \todo Find a trick to solve the pb (use RET fields ?)
753 * @param document to compare with current one
754 * @return true if 'smaller'
756 bool Document::operator<(Document &document)
759 std::string s1 = GetEntryString(0x0010,0x0010);
760 std::string s2 = document.GetEntryString(0x0010,0x0010);
772 s1 = GetEntryString(0x0010,0x0020);
773 s2 = document.GetEntryString(0x0010,0x0020);
784 // Study Instance UID
785 s1 = GetEntryString(0x0020,0x000d);
786 s2 = document.GetEntryString(0x0020,0x000d);
797 // Serie Instance UID
798 s1 = GetEntryString(0x0020,0x000e);
799 s2 = document.GetEntryString(0x0020,0x000e);
814 //-----------------------------------------------------------------------------
817 * \brief Reads a supposed to be 16 Bits integer
818 * (swaps it depending on processor endianness)
821 uint16_t Document::ReadInt16()
825 Fp->read ((char*)&g, (size_t)2);
828 throw FormatError( "Document::ReadInt16()", " file error." );
832 throw FormatError( "Document::ReadInt16()", "EOF." );
839 * \brief Reads a supposed to be 32 Bits integer
840 * (swaps it depending on processor endianness)
843 uint32_t Document::ReadInt32()
847 Fp->read ((char*)&g, (size_t)4);
850 throw FormatError( "Document::ReadInt32()", " file error." );
854 throw FormatError( "Document::ReadInt32()", "EOF." );
861 * \brief skips bytes inside the source file
862 * \warning NOT end user intended method !
865 void Document::SkipBytes(uint32_t nBytes)
867 //FIXME don't dump the returned value
868 Fp->seekg((long)nBytes, std::ios::cur);
872 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
874 int Document::ComputeGroup0002Length( /*FileType filetype*/ )
880 bool found0002 = false;
882 // for each zero-level Tag in the DCM Header
883 DocEntry *entry = GetFirstEntry();
886 gr = entry->GetGroup();
892 if ( entry->GetElement() != 0x0000 )
896 // FIXME : group 0x0002 is *always* Explicit VR!
898 //if ( filetype == ExplicitVR )
900 // if ( (vr == "OB") || (vr == "OW") || (vr == "UT") || (vr == "SQ") )
901 // (no SQ, OW, UT in group 0x0002;)
904 // explicit VR AND OB, OW, SQ, UT : 4 more bytes
908 groupLength += 2 + 2 + 4 + entry->GetLength();
914 entry = GetNextEntry();
919 //-----------------------------------------------------------------------------
922 * \brief Loads all the needed Dictionaries
923 * \warning NOT end user intended method !
925 void Document::Initialize()
927 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
933 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
934 * @param set DocEntrySet we are going to parse ('zero level' or a SQItem)
935 * @param offset start of parsing
936 * @param l_max length to parse (meaningless when we are in 'delimitor mode')
937 * @param delim_mode : whether we are in 'delimitor mode' (l=0xffffff) or not
939 void Document::ParseDES(DocEntrySet *set, long offset,
940 long l_max, bool delim_mode)
942 DocEntry *newDocEntry;
943 DataEntry *newDataEntry;
944 SeqEntry *newSeqEntry;
946 bool used; // will be set to false when something wrong happens to an Entry.
947 // (Entry will then be deleted)
948 bool delim_mode_intern = delim_mode;
950 gdcmWarningMacro( "Enter in ParseDES, delim-mode " << delim_mode
951 << " at offset " << std::hex << offset );
954 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
959 newDocEntry = ReadNextDocEntry( );
962 // Private tag, in IMplicit VR are defaulted as a DataEntry,
963 // Very often they are only composed of Printable characters,
964 // and could be defaulted as a DataEntry.
965 // It's too late to do the Job
966 // (we should check the value, but we know it after LoadDocEntry ...)
968 // Uncoment this printf line to be able to 'follow' the DocEntries
969 // when something *very* strange happens
971 //printf( "%04x|%04x %s\n",newDocEntry->GetGroup(),
972 // newDocEntry->GetElement(),
973 // newDocEntry->GetVR().c_str() );
980 // an Item Starter found elsewhere but the first position
981 // of a SeqEntry means previous entry was a Sequence
982 // but we didn't get it (private Sequence + Implicit VR)
983 // we have to backtrack.
984 if ( !first && newDocEntry->IsItemStarter() )
986 newDocEntry = Backtrack(newDocEntry);
990 PreviousDocEntry = newDocEntry;
994 newDataEntry = dynamic_cast<DataEntry*>(newDocEntry);
998 //////////////////////////// DataEntry
999 vr = newDocEntry->GetVR();
1000 if ( Filetype == ExplicitVR &&
1001 !Global::GetVR()->IsVROfBinaryRepresentable(vr) )
1003 ////// No DataEntry: should mean UNKOWN VR
1004 gdcmWarningMacro( std::hex << newDocEntry->GetGroup()
1005 << "|" << newDocEntry->GetElement()
1006 << " : No DataEntry."
1007 "Probably unknown VR.");
1010 if ( !set->AddEntry( newDataEntry ) )
1012 gdcmWarningMacro( "in ParseDES : cannot add a DataEntry "
1013 << newDataEntry->GetKey()
1015 << newDataEntry->GetOffset() << " )" );
1020 // Load only if we can add (not a duplicate key)
1021 LoadDocEntry( newDataEntry );
1024 if ( newDataEntry->GetElement() == 0x0000 ) // if on group length
1026 if ( newDataEntry->GetGroup()%2 != 0 ) // if Shadow Group
1028 if ( LoadMode & LD_NOSHADOW ) // if user asked to skip shad.gr
1030 std::string strLgrGroup = newDataEntry->GetString();
1032 if ( newDataEntry->IsUnfound() )
1034 lgrGroup = atoi(strLgrGroup.c_str());
1035 Fp->seekg(lgrGroup, std::ios::cur);
1036 //used = false; // never used
1037 RemoveEntry( newDocEntry ); // Remove and delete
1038 // bcc 5.5 is right "assigned a value that's never used"
1046 bool delimitor = newDataEntry->IsItemDelimitor();
1049 (!delim_mode && ((long)(Fp->tellg())-offset) >= l_max) )
1056 // Just to make sure we are at the beginning of next entry.
1057 SkipToNextDocEntry(newDocEntry);
1061 /////////////////////// SeqEntry : VR = "SQ"
1063 unsigned long l = newDocEntry->GetReadLength();
1064 if ( l != 0 ) // don't mess the delim_mode for 'zero-length sequence'
1066 if ( l == 0xffffffff )
1068 delim_mode_intern = true;
1072 delim_mode_intern = false;
1076 if ( (LoadMode & LD_NOSHADOWSEQ) && ! delim_mode_intern )
1078 // User asked to skip SeQuences *only* if they belong to Shadow Group
1079 if ( newDocEntry->GetGroup()%2 != 0 )
1081 Fp->seekg( l, std::ios::cur);
1082 RemoveEntry( newDocEntry ); // Remove and delete
1083 //used = false; // never used
1087 if ( (LoadMode & LD_NOSEQ) && ! delim_mode_intern )
1089 // User asked to skip *any* SeQuence
1090 Fp->seekg( l, std::ios::cur);
1091 //used = false; // never used
1092 RemoveEntry( newDocEntry ); // Remove and delete
1095 // delay the dynamic cast as late as possible
1096 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1098 // no other way to create the Delimitor ...
1099 newSeqEntry->SetDelimitorMode( delim_mode_intern );
1101 // At the top of the hierarchy, stands a Document. When "set"
1102 // is a Document, then we are building the first depth level.
1103 // Hence the SeqEntry we are building simply has a depth
1105 // SQItem *parentSQItem = dynamic_cast< SQItem* > ( set );
1106 if ( set == this ) // ( dynamic_cast< Document* > ( set ) )
1108 newSeqEntry->SetDepthLevel( 1 );
1109 // newSeqEntry->SetKey( newSeqEntry->GetKey() );
1111 // But when "set" is already a SQItem, we are building a nested
1112 // sequence, and hence the depth level of the new SeqEntry
1113 // we are building, is one level deeper:
1115 // time waste hunting
1116 else if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1118 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1120 // newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1121 // + newSeqEntry->GetKey() );
1125 { // Don't try to parse zero-length sequences
1127 gdcmWarningMacro( "Entry in ParseSQ, delim " << delim_mode_intern
1128 << " at offset " << std::hex
1129 << newDocEntry->GetOffset() );
1131 ParseSQ( newSeqEntry,
1132 newDocEntry->GetOffset(),
1133 l, delim_mode_intern);
1135 gdcmWarningMacro( "Exit from ParseSQ, delim " << delim_mode_intern);
1138 if ( !set->AddEntry( newSeqEntry ) )
1140 gdcmWarningMacro( "in ParseDES : cannot add a SeqEntry "
1141 << newSeqEntry->GetKey()
1143 << newSeqEntry->GetOffset() << " )" );
1147 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1153 } // end SeqEntry : VR = "SQ"
1161 gdcmWarningMacro( "Exit from ParseDES, delim-mode " << delim_mode );
1165 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1166 * @return parsed length for this level
1168 void Document::ParseSQ( SeqEntry *seqEntry,
1169 long offset, long l_max, bool delim_mode)
1171 int SQItemNumber = 0;
1173 long offsetStartCurrentSQItem = offset;
1177 // the first time, we read the fff0,e000 of the first SQItem
1178 DocEntry *newDocEntry = ReadNextDocEntry();
1182 // FIXME Should warn user
1183 gdcmWarningMacro("in ParseSQ : should never get here!");
1188 if ( newDocEntry->IsSequenceDelimitor() )
1190 seqEntry->SetDelimitationItem( newDocEntry );
1194 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1199 // create the current SQItem
1200 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1201 unsigned int l = newDocEntry->GetReadLength();
1203 if ( l == 0xffffffff )
1212 // Let's try :------------
1213 // remove fff0,e000, created out of the SQItem
1215 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg);
1216 // fill up the current SQItem, starting at the beginning of fff0,e000
1218 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
1220 offsetStartCurrentSQItem = Fp->tellg();
1221 // end try -----------------
1223 seqEntry->AddSQItem( itemSQ, SQItemNumber );
1225 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1233 * \brief When a private Sequence + Implicit VR is encountered
1234 * we cannot guess it's a Sequence till we find the first
1235 * Item Starter. We then backtrack to do the job.
1236 * @param docEntry Item Starter that warned us
1238 DocEntry *Document::Backtrack(DocEntry *docEntry)
1240 // delete the Item Starter, built erroneously out of any Sequence
1241 // it's not yet in the HTable/chained list
1244 // Get all info we can from PreviousDocEntry
1245 uint16_t group = PreviousDocEntry->GetGroup();
1246 uint16_t elem = PreviousDocEntry->GetElement();
1247 uint32_t lgt = PreviousDocEntry->GetLength();
1248 long offset = PreviousDocEntry->GetOffset();
1250 gdcmWarningMacro( "Backtrack :" << std::hex << group
1252 << " at offset " << offset );
1253 RemoveEntry( PreviousDocEntry );
1255 // forge the Seq Entry
1256 DocEntry *newEntry = NewSeqEntry(group, elem);
1257 newEntry->SetLength(lgt);
1258 newEntry->SetOffset(offset);
1260 // Move back to the beginning of the Sequence
1261 Fp->seekg( 0, std::ios::beg);
1262 Fp->seekg(offset, std::ios::cur);
1268 * \brief Loads (or not) the element content depending if its length exceeds
1269 * or not the value specified with Document::SetMaxSizeLoadEntry()
1270 * @param entry Header Entry (Dicom Element) to be dealt with
1271 * @param forceLoad whether you want to force loading of 'long' elements
1273 void Document::LoadDocEntry(DocEntry *entry, bool forceLoad)
1275 uint16_t group = entry->GetGroup();
1276 uint16_t elem = entry->GetElement();
1277 std::string vr = entry->GetVR();
1278 uint32_t length = entry->GetLength();
1280 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1282 // A SeQuence "contains" a set of Elements.
1283 // (fffe e000) tells us an Element is beginning
1284 // (fffe e00d) tells us an Element just ended
1285 // (fffe e0dd) tells us the current SeQuence just ended
1287 // (fffe 0000) is an 'impossible' tag value,
1288 // found in MR-PHILIPS-16-Multi-Seq.dcm
1290 if ( (group == 0xfffe && elem != 0x0000 ) || vr == "SQ" )
1292 // NO more value field for SQ !
1296 DataEntry *dataEntryPtr = dynamic_cast< DataEntry* >(entry);
1302 // When the length is zero things are easy:
1305 dataEntryPtr->SetBinArea(NULL,true);
1309 // The elements whose length is bigger than the specified upper bound
1310 // are not loaded. Instead we leave a short notice on the offset of
1311 // the element content and it's length.
1313 std::ostringstream s;
1317 if (length > MaxSizeLoadEntry)
1319 dataEntryPtr->SetBinArea(NULL,true);
1320 dataEntryPtr->SetState(DataEntry::STATE_NOTLOADED);
1322 // to be sure we are at the end of the value ...
1323 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1329 LoadEntryBinArea(dataEntryPtr); // last one, not to erase length !
1333 * \brief Find the value Length of the passed Doc Entry
1334 * @param entry Header Entry whose length of the value shall be loaded.
1336 void Document::FindDocEntryLength( DocEntry *entry )
1337 throw ( FormatError )
1339 std::string vr = entry->GetVR();
1342 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1344 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UT"
1347 // The following reserved two bytes (see PS 3.5-2003, section
1348 // "7.1.2 Data element structure with explicit vr", p 27) must be
1349 // skipped before proceeding on reading the length on 4 bytes.
1350 Fp->seekg( 2L, std::ios::cur);
1351 uint32_t length32 = ReadInt32();
1353 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1358 lengthOB = FindDocEntryLengthOBOrOW();
1360 catch ( FormatUnexpected )
1362 // Computing the length failed (this happens with broken
1363 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1364 // chance to get the pixels by deciding the element goes
1365 // until the end of the file. Hence we artificially fix the
1366 // the length and proceed.
1367 gdcmWarningMacro( " Computing the length failed for " <<
1368 entry->GetKey() <<" in " <<GetFileName());
1370 long currentPosition = Fp->tellg();
1371 Fp->seekg(0L,std::ios::end);
1373 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1374 Fp->seekg(currentPosition, std::ios::beg);
1376 entry->SetReadLength(lengthUntilEOF);
1377 entry->SetLength(lengthUntilEOF);
1380 entry->SetReadLength(lengthOB);
1381 entry->SetLength(lengthOB);
1384 FixDocEntryFoundLength(entry, length32);
1388 // Length is encoded on 2 bytes.
1389 length16 = ReadInt16();
1391 // 0xffff means that we deal with 'No Length' Sequence
1392 // or 'No Length' SQItem
1393 if ( length16 == 0xffff)
1397 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1402 // Either implicit VR or a non DICOM conformal (see note below) explicit
1403 // VR that ommited the VR of (at least) this element. Farts happen.
1404 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1405 // on Data elements "Implicit and Explicit VR Data Elements shall
1406 // not coexist in a Data Set and Data Sets nested within it".]
1407 // Length is on 4 bytes.
1409 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1410 // even if Transfer Syntax is 'Implicit VR ...'
1412 FixDocEntryFoundLength( entry, ReadInt32() );
1418 * \brief Find the Length till the next sequence delimiter
1419 * \warning NOT end user intended method !
1422 uint32_t Document::FindDocEntryLengthOBOrOW()
1423 throw( FormatUnexpected )
1425 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1426 long positionOnEntry = Fp->tellg();
1427 bool foundSequenceDelimiter = false;
1428 uint32_t totalLength = 0;
1430 while ( !foundSequenceDelimiter )
1436 group = ReadInt16();
1439 catch ( FormatError )
1441 throw FormatError("Unexpected end of file encountered during ",
1442 "Document::FindDocEntryLengthOBOrOW()");
1444 // We have to decount the group and element we just read
1446 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
1448 long filePosition = Fp->tellg();
1450 "Neither an Item tag nor a Sequence delimiter tag on :"
1451 << std::hex << group << " , " << elem
1452 << ") -before- position x(" << filePosition << ")" );
1454 Fp->seekg(positionOnEntry, std::ios::beg);
1455 throw FormatUnexpected(
1456 "Neither an Item tag nor a Sequence delimiter tag.");
1458 if ( elem == 0xe0dd )
1460 foundSequenceDelimiter = true;
1462 uint32_t itemLength = ReadInt32();
1463 // We add 4 bytes since we just read the ItemLength with ReadInt32
1464 totalLength += itemLength + 4;
1465 SkipBytes(itemLength);
1467 if ( foundSequenceDelimiter )
1472 Fp->seekg( positionOnEntry, std::ios::beg);
1477 * \brief Find the Value Representation of the current Dicom Element.
1478 * @return Value Representation of the current Entry
1480 std::string Document::FindDocEntryVR()
1482 if ( Filetype != ExplicitVR )
1483 return GDCM_UNKNOWN;
1485 long positionOnEntry = Fp->tellg();
1486 // Warning: we believe this is explicit VR (Value Representation) because
1487 // we used a heuristic that found "UL" in the first tag. Alas this
1488 // doesn't guarantee that all the tags will be in explicit VR. In some
1489 // cases (see e-film filtered files) one finds implicit VR tags mixed
1490 // within an explicit VR file. Hence we make sure the present tag
1491 // is in explicit VR and try to fix things if it happens not to be
1495 Fp->read (vr, (size_t)2);
1498 if ( !CheckDocEntryVR(vr) )
1500 Fp->seekg(positionOnEntry, std::ios::beg);
1501 return GDCM_UNKNOWN;
1507 * \brief Check the correspondance between the VR of the header entry
1508 * and the taken VR. If they are different, the header entry is
1509 * updated with the new VR.
1510 * @param vr Dicom Value Representation
1511 * @return false if the VR is incorrect or if the VR isn't referenced
1512 * otherwise, it returns true
1514 bool Document::CheckDocEntryVR(VRKey vr)
1516 if ( !Global::GetVR()->IsValidVR(vr) )
1523 * \brief Skip a given Header Entry
1524 * \warning NOT end user intended method !
1525 * @param entry entry to skip
1527 void Document::SkipDocEntry(DocEntry *entry)
1529 SkipBytes(entry->GetLength());
1533 * \brief Skips to the beginning of the next Header Entry
1534 * \warning NOT end user intended method !
1535 * @param currentDocEntry entry to skip
1537 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1539 int l = currentDocEntry->GetReadLength();
1540 if ( l == -1 ) // length = 0xffff shouldn't appear here ...
1541 // ... but PMS imagers happen !
1543 Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg);
1544 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1546 Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur);
1551 * \brief When the length of an element value is obviously wrong (because
1552 * the parser went Jabberwocky) one can hope improving things by
1553 * applying some heuristics.
1554 * @param entry entry to check
1555 * @param foundLength first assumption about length
1557 void Document::FixDocEntryFoundLength(DocEntry *entry,
1558 uint32_t foundLength)
1560 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1561 if ( foundLength == 0xffffffff)
1566 uint16_t gr = entry->GetGroup();
1567 uint16_t elem = entry->GetElement();
1569 if ( foundLength % 2)
1571 gdcmWarningMacro( "Warning : Tag with uneven length " << foundLength
1572 << " in x(" << std::hex << gr << "," << elem <<")");
1575 //////// Fix for some naughty General Electric images.
1576 // Allthough not recent many such GE corrupted images are still present
1577 // on Creatis hard disks. Hence this fix shall remain when such images
1578 // are no longer in use (we are talking a few years, here)...
1579 // Note: XMedCon probably uses such a trick since it is able to read
1580 // those pesky GE images ...
1581 if ( foundLength == 13)
1583 // Only happens for this length !
1584 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1587 entry->SetReadLength(10); // a bug is to be fixed !?
1591 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1592 // Occurence of such images is quite low (unless one leaves close to a
1593 // 'Leonardo' source. Hence, one might consider commenting out the
1594 // following fix on efficiency reasons.
1595 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1598 entry->SetReadLength(4); // a bug is to be fixed !
1601 else if ( entry->GetVR() == "SQ" )
1603 foundLength = 0; // ReadLength is unchanged
1606 //////// We encountered a 'delimiter' element i.e. a tag of the form
1607 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1608 // taken into account.
1609 else if ( gr == 0xfffe )
1611 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1612 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1613 // causes extra troubles...
1614 if ( entry->GetElement() != 0x0000 )
1620 foundLength=12; // to skip the mess that follows this bugged Tag !
1623 entry->SetLength(foundLength);
1627 * \brief Apply some heuristics to predict whether the considered
1628 * element value contains/represents an integer or not.
1629 * @param entry The element value on which to apply the predicate.
1630 * @return The result of the heuristical predicate.
1632 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1634 uint16_t elem = entry->GetElement();
1635 uint16_t group = entry->GetGroup();
1636 const std::string &vr = entry->GetVR();
1637 uint32_t length = entry->GetLength();
1639 // When we have some semantics on the element we just read, and if we
1640 // a priori know we are dealing with an integer, then we shall be
1641 // able to swap it's element value properly.
1642 if ( elem == 0 ) // This is the group length of the group
1650 // Allthough this should never happen, still some images have a
1651 // corrupted group length [e.g. have a glance at offset x(8336) of
1652 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
1653 // Since for dicom compliant and well behaved headers, the present
1654 // test is useless (and might even look a bit paranoid), when we
1655 // encounter such an ill-formed image, we simply display a warning
1656 // message and proceed on parsing (while crossing fingers).
1657 long filePosition = Fp->tellg();
1658 gdcmWarningMacro( "Erroneous Group Length element length on : ("
1659 << std::hex << group << " , " << elem
1660 << ") -before- position x(" << filePosition << ")"
1661 << "lgt : " << length );
1665 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1673 * \brief Discover what the swap code is (among little endian, big endian,
1674 * bad little endian, bad big endian).
1676 * @return false when we are absolutely sure
1677 * it's neither ACR-NEMA nor DICOM
1678 * true when we hope ours assuptions are OK
1680 bool Document::CheckSwap()
1687 // First, compare HostByteOrder and NetworkByteOrder in order to
1688 // determine if we shall need to swap bytes (i.e. the Endian type).
1689 bool net2host = Util::IsCurrentProcessorBigEndian();
1691 // The easiest case is the one of a 'true' DICOM header, we just have
1692 // to look for the string "DICM" inside the file preamble.
1695 char *entCur = deb + 128;
1696 if ( memcmp(entCur, "DICM", (size_t)4) == 0 )
1698 gdcmWarningMacro( "Looks like DICOM Version3 (preamble + DCM)" );
1700 // Group 0002 should always be VR, and the first element 0000
1701 // Let's be carefull (so many wrong headers ...)
1702 // and determine the value representation (VR) :
1703 // Let's skip to the first element (0002,0000) and check there if we find
1704 // "UL" - or "OB" if the 1st one is (0002,0001) -,
1705 // in which case we (almost) know it is explicit VR.
1706 // WARNING: if it happens to be implicit VR then what we will read
1707 // is the length of the group. If this ascii representation of this
1708 // length happens to be "UL" then we shall believe it is explicit VR.
1709 // We need to skip :
1710 // * the 128 bytes of File Preamble (often padded with zeroes),
1711 // * the 4 bytes of "DICM" string,
1712 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
1713 // i.e. a total of 136 bytes.
1716 // group 0x0002 *is always* Explicit VR Sometimes ,
1717 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
1718 // *Implicit* VR (see former 'gdcmData/icone.dcm')
1720 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
1721 memcmp(entCur, "OB", (size_t)2) == 0 ||
1722 memcmp(entCur, "UI", (size_t)2) == 0 ||
1723 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
1724 // when Write DCM *adds*
1726 // Use Document::dicom_vr to test all the possibilities
1727 // instead of just checking for UL, OB and UI !? group 0000
1729 Filetype = ExplicitVR;
1730 gdcmWarningMacro( "Group 0002 : Explicit Value Representation");
1734 Filetype = ImplicitVR;
1735 gdcmWarningMacro( "Group 0002 :Not an explicit Value Representation;"
1736 << "Looks like a bugged Header!");
1742 gdcmWarningMacro( "HostByteOrder != NetworkByteOrder");
1747 gdcmWarningMacro( "HostByteOrder = NetworkByteOrder");
1750 // Position the file position indicator at first tag
1751 // (i.e. after the file preamble and the "DICM" string).
1753 Fp->seekg(0, std::ios::beg); // FIXME : Is it usefull?
1755 Fp->seekg ( 132L, std::ios::beg);
1757 } // ------------------------------- End of DicomV3 ----------------
1759 // Alas, this is not a DicomV3 file and whatever happens there is no file
1760 // preamble. We can reset the file position indicator to where the data
1761 // is (i.e. the beginning of the file).
1763 gdcmWarningMacro( "Not a Kosher DICOM Version3 file (no preamble)");
1765 Fp->seekg(0, std::ios::beg);
1767 // Let's check 'No Preamble Dicom File' :
1768 // Should start with group 0x0002
1769 // and be Explicit Value Representation
1771 s16 = *((uint16_t *)(deb));
1784 if ( SwapCode != 0 )
1786 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
1787 memcmp(entCur, "OB", (size_t)2) == 0 ||
1788 memcmp(entCur, "UI", (size_t)2) == 0 ||
1789 memcmp(entCur, "SH", (size_t)2) == 0 ||
1790 memcmp(entCur, "AE", (size_t)2) == 0 ||
1791 memcmp(entCur, "OB", (size_t)2) == 0 )
1793 Filetype = ExplicitVR;
1794 gdcmWarningMacro( "Group 0002 : Explicit Value Representation");
1798 // ------------------------------- End of 'No Preamble' DicomV3 -------------
1800 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
1801 // By clean we mean that the length of the first group is written down.
1802 // If this is the case and since the length of the first group HAS to be
1803 // four (bytes), then determining the proper swap code is straightforward.
1806 // We assume the array of char we are considering contains the binary
1807 // representation of a 32 bits integer. Hence the following dirty
1809 s32 = *((uint32_t *)(entCur));
1829 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
1830 // It is time for despaired wild guesses.
1831 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
1832 // i.e. the 'group length' element is not present :
1834 // check the supposed-to-be 'group number'
1835 // in ( 0x0001 .. 0x0008 )
1836 // to determine ' SwapCode' value .
1837 // Only 0 or 4321 will be possible
1838 // (no oportunity to check for the formerly well known
1839 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
1840 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
1841 // the file IS NOT ACR-NEMA nor DICOM V3
1842 // Find a trick to tell it the caller...
1844 s16 = *((uint16_t *)(deb));
1871 gdcmWarningMacro("ACR/NEMA unfound swap info (Really hopeless !)");
1879 * \brief Change the Byte Swap code.
1881 void Document::SwitchByteSwapCode()
1883 gdcmWarningMacro( "Switching Byte Swap code from "<< SwapCode
1884 << " at :" <<std::hex << Fp->tellg() );
1885 if ( SwapCode == 1234 )
1889 else if ( SwapCode == 4321 )
1893 else if ( SwapCode == 3412 )
1897 else if ( SwapCode == 2143 )
1904 * \brief during parsing, Header Elements too long are not loaded in memory
1905 * @param newSize new size
1907 void Document::SetMaxSizeLoadEntry(long newSize)
1913 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
1915 MaxSizeLoadEntry = 0xffffffff;
1918 MaxSizeLoadEntry = newSize;
1922 * \brief Read the next tag WITHOUT loading it's value
1923 * (read the 'Group Number', the 'Element Number',
1924 * gets the Dict Entry
1925 * gets the VR, gets the length, gets the offset value)
1926 * @return On succes : the newly created DocEntry, NULL on failure.
1928 DocEntry *Document::ReadNextDocEntry()
1935 group = ReadInt16();
1938 catch ( FormatError )
1940 // We reached the EOF (or an error occured) therefore
1941 // header parsing has to be considered as finished.
1945 // Sometimes file contains groups of tags with reversed endianess.
1946 HandleBrokenEndian(group, elem);
1948 // In 'true DICOM' files Group 0002 is always little endian
1949 if ( HasDCMPreamble )
1950 HandleOutOfGroup0002(group, elem);
1952 std::string vr = FindDocEntryVR();
1953 std::string realVR = vr;
1955 if ( vr == GDCM_UNKNOWN )
1957 if ( elem == 0x0000 ) // Group Length
1959 realVR = "UL"; // must be UL
1961 else if (group%2 == 1 && (elem >= 0x0010 && elem <=0x00ff ))
1963 // DICOM PS 3-5 7.8.1 a) states that those
1964 // (gggg-0010->00FF where gggg is odd) attributes have to be LO
1969 DictEntry *dictEntry = GetDictEntry(group,elem);
1972 realVR = dictEntry->GetVR();
1978 if ( Global::GetVR()->IsVROfSequence(realVR) )
1979 newEntry = NewSeqEntry(group, elem);
1982 newEntry = NewDataEntry(group, elem, realVR);
1983 static_cast<DataEntry *>(newEntry)->SetState(DataEntry::STATE_NOTLOADED);
1986 if ( vr == GDCM_UNKNOWN )
1988 if ( Filetype == ExplicitVR )
1990 // We thought this was explicit VR, but we end up with an
1991 // implicit VR tag. Let's backtrack.
1992 if ( newEntry->GetGroup() != 0xfffe )
1995 int offset = Fp->tellg();
1996 msg = Util::Format("Entry (%04x,%04x) at 0x(%x) should be Explicit VR\n",
1997 newEntry->GetGroup(), newEntry->GetElement(), offset );
1998 gdcmWarningMacro( msg.c_str() );
2001 newEntry->SetImplicitVR();
2006 FindDocEntryLength(newEntry);
2008 catch ( FormatError )
2015 newEntry->SetOffset(Fp->tellg());
2021 * \brief Handle broken private tag from Philips NTSCAN
2022 * where the endianess is being switched to BigEndian
2023 * for no apparent reason
2026 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
2028 // Endian reversion. Some files contain groups of tags with reversed endianess.
2029 static int reversedEndian = 0;
2030 // try to fix endian switching in the middle of headers
2031 if ((group == 0xfeff) && (elem == 0x00e0))
2033 // start endian swap mark for group found
2035 SwitchByteSwapCode();
2040 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2042 // end of reversed endian group
2044 SwitchByteSwapCode();
2046 else if (group == 0xfeff && elem == 0xdde0)
2048 // reversed Sequence Terminator found
2049 // probabely a bug in the header !
2050 // Do what you want, it breaks !
2052 //SwitchByteSwapCode();
2053 gdcmWarningMacro( "Should never get here! reversed Sequence Terminator!" );
2058 else if (group == 0xfffe && elem == 0xe0dd)
2060 gdcmWarningMacro( "Straight Sequence Terminator." );
2065 * \brief Group 0002 is always coded Little Endian
2066 * whatever Transfer Syntax is
2069 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2071 // Endian reversion. Some files contain groups of tags with reversed endianess.
2072 if ( !Group0002Parsed && group != 0x0002)
2074 Group0002Parsed = true;
2075 // we just came out of group 0002
2076 // if Transfer syntax is Big Endian we have to change CheckSwap
2078 std::string ts = GetTransferSyntax();
2079 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2081 gdcmWarningMacro("True DICOM File, with NO Tansfer Syntax: " << ts );
2085 // Group 0002 is always 'Explicit ...' even when Transfer Syntax says 'Implicit ..."
2087 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ImplicitVRLittleEndian )
2089 Filetype = ImplicitVR;
2092 // FIXME Strangely, this works with
2093 //'Implicit VR BigEndian Transfer Syntax (GE Private)
2095 // --> Probabely normal, since we considered we never have to trust manufacturers.
2096 // (we find very often 'Implicit VR' tag, even when Transfer Syntax tells us it's Explicit ...
2097 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ExplicitVRBigEndian )
2099 gdcmWarningMacro("Transfer Syntax Name = ["
2100 << GetTransferSyntaxName() << "]" );
2101 SwitchByteSwapCode();
2102 group = SwapShort(group);
2103 elem = SwapShort(elem);
2108 //-----------------------------------------------------------------------------
2111 //-----------------------------------------------------------------------------
2112 } // end namespace gdcm