1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/09/20 09:07:56 $
7 Version: $Revision: 1.280 $
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"
34 #include <ctype.h> // for isdigit
35 #include <stdlib.h> // for atoi
39 //-----------------------------------------------------------------------------
41 // Refer to Document::SetMaxSizeLoadEntry()
42 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
44 //-----------------------------------------------------------------------------
45 // Constructor / Destructor
46 // Constructors and destructors are protected to avoid user to invoke directly
49 * \brief This default constructor neither loads nor parses the file.
50 * You should then invoke \ref Document::Load.
58 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
61 Filetype = ExplicitVR;
62 // Load will set it to true if sucessfull
63 Group0002Parsed = false;
64 IsDocumentAlreadyLoaded = false;
65 IsDocumentModified = true;
66 LoadMode = LD_ALL; // default : load everything, later
71 * \brief Constructor (DEPRECATED : not to break the API)
72 * @param fileName 'Document' (File or DicomDir) to be open for parsing
74 Document::Document( std::string const &fileName )
79 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
82 Filetype = ExplicitVR;
83 Group0002Parsed = false;
84 LoadMode = LD_ALL; // Load everything, later
86 // Load will set it to true if sucessfull
87 IsDocumentAlreadyLoaded = false;
88 IsDocumentModified = true;
90 SetFileName(fileName);
94 * \brief Canonical destructor.
96 Document::~Document ()
101 //-----------------------------------------------------------------------------
105 * \brief Loader. use SetLoadMode(), SetFileName() before !
106 * @return false if file cannot be open or no swap info was found,
107 * or no tag was found.
109 bool Document::Load( )
111 if ( GetFileName() == "" )
113 gdcmWarningMacro( "Use SetFileName, before !" );
116 return DoTheLoadingDocumentJob( );
119 * \brief Loader. (DEPRECATED : not to break the API)
120 * @param fileName 'Document' (File or DicomDir) to be open for parsing
121 * @return false if file cannot be open or no swap info was found,
122 * or no tag was found.
124 bool Document::Load( std::string const &fileName )
127 return DoTheLoadingDocumentJob( );
131 * \brief Performs the Loading Job (internal use only)
132 * @return false if file cannot be open or no swap info was found,
133 * or no tag was found.
135 bool Document::DoTheLoadingDocumentJob( )
137 if ( ! IsDocumentModified ) // Nothing to do !
140 // if ( Filename == fileName )
142 // gdcmWarningMacro( "The file was already parsed inside this "
143 // << "gdcm::Document (its name is: "
144 // << Filename.c_str() );
148 //gdcmWarningMacro( "A file was already parsed inside this "
149 // << "gdcm::Document (previous name was: "
150 // << Filename.c_str() << ". New name is :"
152 // clean out the Entries, if already parsed
153 // (probabely a mistake from the user)
160 // warning already performed in OpenFile()
161 //gdcmWarningMacro( "Unable to open as an ACR/DICOM file: "
162 // << Filename.c_str() );
167 Group0002Parsed = false;
169 gdcmWarningMacro( "Starting parsing of file: " << Filename.c_str());
171 Fp->seekg(0, std::ios::end);
172 long lgt = Fp->tellg(); // total length of the file
174 Fp->seekg(0, std::ios::beg);
176 // CheckSwap returns a boolean
177 // (false if no swap info of any kind was found)
180 gdcmWarningMacro( "Neither a DICOM V3 nor an ACR-NEMA file: "
181 << Filename.c_str());
186 long beg = Fp->tellg(); // just after DICOM preamble (if any)
188 lgt -= beg; // remaining length to parse
191 // Loading is done during parsing
192 ParseDES( this, beg, lgt, false); // delim_mode is first defaulted to false
196 gdcmWarningMacro( "No tag in internal hash table for: "
197 << Filename.c_str());
201 IsDocumentAlreadyLoaded = true;
203 Fp->seekg( 0, std::ios::beg);
205 // Load 'non string' values
207 std::string PhotometricInterpretation = GetEntryValue(0x0028,0x0004);
208 if ( PhotometricInterpretation == "PALETTE COLOR " )
211 // Probabely this line should be outside the 'if'
212 // Try to find an image sample holding a 'gray LUT'
213 LoadEntryBinArea(0x0028,0x1200); // gray LUT
216 /// The tags refered by the three following lines used to be CORRECTLY
217 /// defined as having an US Value Representation in the public
218 /// dictionary. BUT the semantics implied by the three following
219 /// lines state that the corresponding tag contents are in fact
220 /// the ones of a BinEntry.
221 /// In order to fix things "Quick and Dirty" the dictionary was
222 /// altered on PURPOSE but now contains a WRONG value.
223 /// In order to fix things and restore the dictionary to its
224 /// correct value, one needs to decided of the semantics by deciding
225 /// whether the following tags are either :
226 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
227 /// also used as BinEntry, which requires the proper conversion,
228 /// - OW, and hence loaded as BinEntry, but afterwards also used
229 /// as ValEntry, which requires the proper conversion.
230 LoadEntryBinArea(0x0028,0x1201); // R LUT
231 LoadEntryBinArea(0x0028,0x1202); // G LUT
232 LoadEntryBinArea(0x0028,0x1203); // B LUT
234 // Segmented Red Palette Color LUT Data
235 LoadEntryBinArea(0x0028,0x1221);
236 // Segmented Green Palette Color LUT Data
237 LoadEntryBinArea(0x0028,0x1222);
238 // Segmented Blue Palette Color LUT Data
239 LoadEntryBinArea(0x0028,0x1223);
242 //FIXME later : how to use it?
243 SeqEntry *modLutSeq = GetSeqEntry(0x0028,0x3000);
246 SQItem *sqi= modLutSeq->GetFirstSQItem();
249 BinEntry *b = sqi->GetBinEntry(0x0028,0x3006);
252 if ( b->GetLength() != 0 )
254 LoadEntryBinArea(b); //LUT Data (CTX dependent)
260 // Force Loading some more elements if user asked to.
263 for (ListElements::iterator it = UserForceLoadList.begin();
264 it != UserForceLoadList.end();
267 gdcmWarningMacro( "Force Load " << std::hex
268 << (*it).Group << "|" <<(*it).Elem );
270 d = GetDocEntry( (*it).Group, (*it).Elem);
274 gdcmWarningMacro( "You asked toForce Load " << std::hex
275 << (*it).Group <<"|"<< (*it).Elem
276 << " that doesn't exist" );
280 if ( dynamic_cast<ValEntry *>(d) )
282 LoadDocEntry(d, true);
286 BinEntry *b = dynamic_cast<BinEntry *>(d);
290 b->SetValue(GDCM_BINLOADED);
294 if ( dynamic_cast<SeqEntry *>(d) )
296 gdcmWarningMacro( "You cannot 'ForceLoad' a SeqEntry :" << std::hex
297 << (*it).Group <<"|"<< (*it).Elem );
304 // ----------------------------
305 // Specific code to allow gdcm to read ACR-LibIDO formated images
306 // Note: ACR-LibIDO is an extension of the ACR standard that was
307 // used at CREATIS. For the time being (say a couple of years)
308 // we keep this kludge to allow CREATIS users
309 // reading their old images.
311 // if recognition code tells us we deal with a LibIDO image
312 // we switch lineNumber and columnNumber
315 RecCode = GetEntryValue(0x0008, 0x0010); // recognition code (RET)
316 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
317 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
318 // with "little-endian strings"
320 Filetype = ACR_LIBIDO;
321 std::string rows = GetEntryValue(0x0028, 0x0010);
322 std::string columns = GetEntryValue(0x0028, 0x0011);
323 SetValEntry(columns, 0x0028, 0x0010);
324 SetValEntry(rows , 0x0028, 0x0011);
326 // --- End of ACR-LibIDO kludge ---
332 * \brief Adds a new element we want to load anyway
333 * @param group Group number of the target tag.
334 * @param elem Element number of the target tag.
336 void Document::AddForceLoadElement (uint16_t group, uint16_t elem)
341 UserForceLoadList.push_back(el);
344 * \brief Get the public dictionary used
346 Dict *Document::GetPubDict()
352 * \brief Get the shadow dictionary used
354 Dict *Document::GetShaDict()
360 * \brief Set the shadow dictionary used
361 * @param dict dictionary to use in shadow
363 bool Document::SetShaDict(Dict *dict)
370 * \brief Set the shadow dictionary used
371 * @param dictName name of the dictionary to use in shadow
373 bool Document::SetShaDict(DictKey const &dictName)
375 RefShaDict = Global::GetDicts()->GetDict(dictName);
380 * \brief This predicate tells us whether or not the current Document
381 * was properly parsed and contains at least *one* Dicom Element
382 * (and nothing more, sorry).
383 * @return false when we're 150 % sure it's NOT a Dicom/Acr file,
386 bool Document::IsReadable()
388 if ( Filetype == Unknown )
390 gdcmWarningMacro( "Wrong filetype");
396 gdcmWarningMacro( "No tag in internal hash table.");
404 * \brief Predicate for dicom version 3 file.
405 * @return True when the file is a dicom version 3.
407 bool Document::IsDicomV3()
409 // Checking if Transfer Syntax exists is enough
410 // Anyway, it's too late check if the 'Preamble' was found ...
411 // And ... would it be a rich idea to check ?
412 // (some 'no Preamble' DICOM images exist !)
413 return GetDocEntry(0x0002, 0x0010) != NULL;
417 * \brief Predicate for Papyrus file
418 * Dedicated to whomsoever it may concern
419 * @return True when the file is a Papyrus file.
421 bool Document::IsPapyrus()
423 // check for Papyrus private Sequence
424 DocEntry *e = GetDocEntry(0x0041, 0x1050);
427 // check if it's actually a Sequence
428 if ( !dynamic_cast<SeqEntry*>(e) )
434 * \brief returns the File Type
435 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
436 * @return the FileType code
438 FileType Document::GetFileType()
444 * \brief Accessor to the Transfer Syntax (when present) of the
445 * current document (it internally handles reading the
446 * value from disk when only parsing occured).
447 * @return The encountered Transfer Syntax of the current document.
449 std::string Document::GetTransferSyntax()
451 DocEntry *entry = GetDocEntry(0x0002, 0x0010);
457 // The entry might be present but not loaded (parsing and loading
458 // happen at different stages): try loading and proceed with check...
459 LoadDocEntrySafe(entry);
460 if (ValEntry *valEntry = dynamic_cast< ValEntry* >(entry) )
462 std::string transfer = valEntry->GetValue();
463 // The actual transfer (as read from disk) might be padded. We
464 // first need to remove the potential padding. We can make the
465 // weak assumption that padding was not executed with digits...
466 if ( transfer.length() == 0 )
468 // for brain damaged headers
471 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
473 transfer.erase(transfer.length()-1, 1);
481 * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS
482 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
484 std::string Document::GetTransferSyntaxName()
486 // use the TS (TS : Transfer Syntax)
487 std::string transferSyntax = GetEntryValue(0x0002,0x0010);
489 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
491 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
492 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
493 return "Uncompressed ACR-NEMA";
495 if ( transferSyntax == GDCM_UNFOUND )
497 gdcmWarningMacro( "Unfound Transfer Syntax (0002,0010)");
498 return "Uncompressed ACR-NEMA";
501 // we do it only when we need it
502 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
504 // Global::GetTS() is a global static you shall never try to delete it!
508 // --------------- Swap Code ------------------
510 * \brief Swaps the bytes so they agree with the processor order
511 * @return The properly swaped 16 bits integer.
513 uint16_t Document::SwapShort(uint16_t a)
515 if ( SwapCode == 4321 || SwapCode == 2143 )
517 //a = ((( a << 8 ) & 0xff00 ) | (( a >> 8 ) & 0x00ff ) );
519 a = ( a << 8 ) | ( a >> 8 );
525 * \brief Swaps back the bytes of 4-byte long integer accordingly to
527 * @return The properly swaped 32 bits integer.
529 uint32_t Document::SwapLong(uint32_t a)
536 // a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
537 // ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
539 a=( ( a<<24) | ((a<<8) & 0x00ff0000) |
540 ((a>>8) & 0x0000ff00) | (a>>24) );
543 // a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
544 a=( (a<<16) | (a>>16) );
547 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
550 gdcmErrorMacro( "Unset swap code:" << SwapCode );
557 // -----------------File I/O ---------------
559 * \brief Tries to open the file \ref Document::Filename and
560 * checks the preamble when existing.
561 * @return The FILE pointer on success.
563 std::ifstream *Document::OpenFile()
565 HasDCMPreamble = false;
566 if (Filename.length() == 0)
573 gdcmWarningMacro( "File already open: " << Filename.c_str());
577 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
580 // Don't user gdcmErrorMacro :
581 // a spurious message will appear when you use, for instance
582 // gdcm::FileHelper *fh = new gdcm::FileHelper( outputFileName );
583 // to create outputFileName.
584 gdcmWarningMacro( "Cannot open file: " << Filename.c_str());
588 //exit(1); // No function is allowed to leave the application instead
589 // of warning the caller
593 Fp->read((char*)&zero, (size_t)2);
600 //-- ACR or DICOM with no Preamble; may start with a Shadow Group --
602 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
603 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
604 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
605 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
607 std::string msg = Util::Format(
608 "ACR/DICOM starting at the beginning of the file:(%04x)\n", zero);
609 gdcmWarningMacro( msg.c_str() );
614 Fp->seekg(126L, std::ios::cur);
615 char dicm[4]; // = {' ',' ',' ',' '};
616 Fp->read(dicm, (size_t)4);
622 if ( memcmp(dicm, "DICM", 4) == 0 )
624 HasDCMPreamble = true;
628 // -- Neither ACR/No Preamble Dicom nor DICOMV3 file
630 gdcmWarningMacro( "Neither ACR/No Preamble Dicom nor DICOMV3 file: "
631 << Filename.c_str());
636 * \brief closes the file
637 * @return TRUE if the close was successfull
639 bool Document::CloseFile()
651 * \brief Writes in a file all the Header Entries (Dicom Elements)
652 * @param fp file pointer on an already open file (actually: Output File Stream)
653 * @param filetype Type of the File to be written
654 * (ACR-NEMA, ExplicitVR, ImplicitVR)
655 * @return Always true.
657 void Document::WriteContent(std::ofstream *fp, FileType filetype)
659 // Skip if user wants to write an ACR-NEMA file
661 if ( filetype == ImplicitVR || filetype == ExplicitVR )
663 // writing Dicom File Preamble
664 char filePreamble[128];
665 memset(filePreamble, 0, 128);
666 fp->write(filePreamble, 128);
667 fp->write("DICM", 4);
671 * \todo rewrite later, if really usefull
672 * - 'Group Length' element is optional in DICOM
673 * - but un-updated odd groups lengthes can causes pb
676 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
677 * UpdateGroupLength(false,filetype);
678 * if ( filetype == ACR)
679 * UpdateGroupLength(true,ACR);
682 ElementSet::WriteContent(fp, filetype); // This one is recursive
685 // -----------------------------------------
688 * \brief Loads (from disk) the element content
689 * when a string is not suitable
690 * @param group group number of the Entry
691 * @param elem element number of the Entry
693 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
695 // Search the corresponding DocEntry
696 DocEntry *docElement = GetDocEntry(group, elem);
699 gdcmWarningMacro(std::hex << group << "|" << elem
700 << "doesn't exist" );
703 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
706 gdcmWarningMacro(std::hex << group << "|" << elem
707 << "is NOT a BinEntry");
710 LoadEntryBinArea(binElement);
714 * \brief Loads (from disk) the element content
715 * when a string is not suitable
716 * @param elem Entry whose binArea is going to be loaded
718 void Document::LoadEntryBinArea(BinEntry *elem)
720 if (elem->GetBinArea() )
727 size_t o =(size_t)elem->GetOffset();
728 Fp->seekg(o, std::ios::beg);
730 size_t l = elem->GetLength();
731 uint8_t *a = new uint8_t[l];
734 gdcmWarningMacro( "Cannot allocate BinEntry content for : "
735 << std::hex << elem->GetGroup()
736 << "|" << elem->GetElement() );
740 Fp->read((char*)a, l);
741 if ( Fp->fail() || Fp->eof() )
754 * \brief Loads the element while preserving the current
755 * underlying file position indicator as opposed to
756 * LoadDocEntry that modifies it.
757 * @param entry DocEntry whose value will be loaded.
759 void Document::LoadDocEntrySafe(DocEntry *entry)
763 long PositionOnEntry = Fp->tellg();
765 Fp->seekg(PositionOnEntry, std::ios::beg);
770 * \brief Compares two documents, according to \ref DicomDir rules
771 * \warning Does NOT work with ACR-NEMA files
772 * \todo Find a trick to solve the pb (use RET fields ?)
773 * @param document to compare with current one
774 * @return true if 'smaller'
776 bool Document::operator<(Document &document)
779 std::string s1 = GetEntryValue(0x0010,0x0010);
780 std::string s2 = document.GetEntryValue(0x0010,0x0010);
792 s1 = GetEntryValue(0x0010,0x0020);
793 s2 = document.GetEntryValue(0x0010,0x0020);
804 // Study Instance UID
805 s1 = GetEntryValue(0x0020,0x000d);
806 s2 = document.GetEntryValue(0x0020,0x000d);
817 // Serie Instance UID
818 s1 = GetEntryValue(0x0020,0x000e);
819 s2 = document.GetEntryValue(0x0020,0x000e);
834 //-----------------------------------------------------------------------------
837 * \brief Reads a supposed to be 16 Bits integer
838 * (swaps it depending on processor endianness)
841 uint16_t Document::ReadInt16()
845 Fp->read ((char*)&g, (size_t)2);
848 throw FormatError( "Document::ReadInt16()", " file error." );
852 throw FormatError( "Document::ReadInt16()", "EOF." );
859 * \brief Reads a supposed to be 32 Bits integer
860 * (swaps it depending on processor endianness)
863 uint32_t Document::ReadInt32()
867 Fp->read ((char*)&g, (size_t)4);
870 throw FormatError( "Document::ReadInt32()", " file error." );
874 throw FormatError( "Document::ReadInt32()", "EOF." );
881 * \brief skips bytes inside the source file
882 * \warning NOT end user intended method !
885 void Document::SkipBytes(uint32_t nBytes)
887 //FIXME don't dump the returned value
888 Fp->seekg((long)nBytes, std::ios::cur);
892 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
894 int Document::ComputeGroup0002Length( /*FileType filetype*/ )
900 bool found0002 = false;
902 // for each zero-level Tag in the DCM Header
903 DocEntry *entry = GetFirstEntry();
906 gr = entry->GetGroup();
912 if ( entry->GetElement() != 0x0000 )
916 // FIXME : group 0x0002 is *always* Explicit VR!
918 //if ( filetype == ExplicitVR )
920 // if ( (vr == "OB") || (vr == "OW") || (vr == "UT") || (vr == "SQ") )
921 // (no SQ, OW, UT in group 0x0002;)
924 // explicit VR AND OB, OW, SQ, UT : 4 more bytes
928 groupLength += 2 + 2 + 4 + entry->GetLength();
934 entry = GetNextEntry();
939 //-----------------------------------------------------------------------------
942 * \brief Loads all the needed Dictionaries
943 * \warning NOT end user intended method !
945 void Document::Initialize()
947 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
953 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
954 * @param set DocEntrySet we are going to parse ('zero level' or a SQItem)
955 * @param offset start of parsing
956 * @param l_max length to parse (meaningless when we are in 'delimitor mode')
957 * @param delim_mode : whether we are in 'delimitor mode' (l=0xffffff) or not
959 void Document::ParseDES(DocEntrySet *set, long offset,
960 long l_max, bool delim_mode)
962 DocEntry *newDocEntry;
963 ValEntry *newValEntry;
964 BinEntry *newBinEntry;
965 SeqEntry *newSeqEntry;
967 bool used; // will be set to false when something wrong happens to an Entry.
968 // (Entry will then be deleted)
969 bool delim_mode_intern = delim_mode;
973 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
978 newDocEntry = ReadNextDocEntry( );
980 // Uncoment this printf line to be able to 'follow' the DocEntries
981 // when something *very* strange happens
983 //printf( "%04x|%04x %s\n",newDocEntry->GetGroup(),
984 // newDocEntry->GetElement(),
985 // newDocEntry->GetVR().c_str() );
992 // an Item Starter found elsewhere but the first postition
993 // of a SeqEntry means previous entry was a Sequence
994 // but we didn't get it (private Sequence + Implicit VR)
995 // we have to backtrack.
996 if ( !first && newDocEntry->IsItemStarter() )
998 newDocEntry = Backtrack(newDocEntry);
1002 PreviousDocEntry = newDocEntry;
1006 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
1007 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
1009 if ( newValEntry || newBinEntry )
1011 //////////////////////////// ContentEntry
1014 vr = newDocEntry->GetVR();
1015 if ( Filetype == ExplicitVR &&
1016 !Global::GetVR()->IsVROfBinaryRepresentable(vr) )
1018 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1019 gdcmWarningMacro( std::hex << newDocEntry->GetGroup()
1020 << "|" << newDocEntry->GetElement()
1021 << " : Neither Valentry, nor BinEntry."
1022 "Probably unknown VR.");
1025 //////////////////// BinEntry or UNKOWN VR:
1027 // When "this" is a Document the Key is simply of the
1028 // form ( group, elem )...
1029 //if ( set == this ) // ( dynamic_cast< Document* > ( set ) )
1031 // newBinEntry->SetKey( newBinEntry->GetKey() );
1033 // but when "this" is a SQItem, we are inserting this new
1034 // valEntry in a sequence item, and the key has the
1035 // generalized form (refer to \ref BaseTagKey):
1037 // time waste hunting
1038 //if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1040 // newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1041 // + newBinEntry->GetKey() );
1044 if ( !set->AddEntry( newBinEntry ) )
1046 gdcmWarningMacro( "in ParseDES : cannot add a BinEntry "
1047 << newBinEntry->GetKey()
1049 << newBinEntry->GetOffset() << " )" );
1054 // Load only if we can add (not a duplicate key)
1055 LoadDocEntry( newBinEntry );
1060 /////////////////////// ValEntry
1062 // When "set" is a Document, then we are at the top of the
1063 // hierarchy and the Key is simply of the form ( group, elem )...
1064 //if ( set == this ) // ( dynamic_cast< Document* > ( set ) )
1066 // newValEntry->SetKey( newValEntry->GetKey() );
1068 // ...but when "set" is a SQItem, we are inserting this new
1069 // valEntry in a sequence item. Hence the key has the
1070 // generalized form (refer to \ref BaseTagKey):
1072 // time waste hunting
1073 //if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1075 // newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1076 // + newValEntry->GetKey() );
1079 if ( !set->AddEntry( newValEntry ) )
1081 gdcmWarningMacro( "in ParseDES : cannot add a ValEntry "
1082 << newValEntry->GetKey()
1084 << newValEntry->GetOffset() << " )" );
1089 // Load only if we can add (not a duplicate key)
1090 LoadDocEntry( newValEntry );
1093 if ( newValEntry->GetElement() == 0x0000 ) // if on group length
1095 if ( newValEntry->GetGroup()%2 != 0 ) // if Shadow Group
1097 if ( LoadMode & LD_NOSHADOW ) // if user asked to skip shad.gr
1099 std::string strLgrGroup = newValEntry->GetValue();
1101 if ( strLgrGroup != GDCM_UNFOUND)
1103 lgrGroup = atoi(strLgrGroup.c_str());
1104 Fp->seekg(lgrGroup, std::ios::cur);
1106 RemoveEntry( newDocEntry );
1107 // bcc 5.5 is right "assigned a value that's never used"
1115 bool delimitor = newValEntry->IsItemDelimitor();
1118 (!delim_mode && ((long)(Fp->tellg())-offset) >= l_max) )
1126 // Just to make sure we are at the beginning of next entry.
1127 SkipToNextDocEntry(newDocEntry);
1131 /////////////////////// SeqEntry : VR = "SQ"
1133 unsigned long l = newDocEntry->GetReadLength();
1134 if ( l != 0 ) // don't mess the delim_mode for 'zero-length sequence'
1136 if ( l == 0xffffffff )
1138 delim_mode_intern = true;
1142 delim_mode_intern = false;
1146 if ( (LoadMode & LD_NOSHADOWSEQ) && ! delim_mode_intern )
1148 // User asked to skip SeQuences *only* if they belong to Shadow Group
1149 if ( newDocEntry->GetGroup()%2 != 0 )
1151 Fp->seekg( l, std::ios::cur);
1156 if ( (LoadMode & LD_NOSEQ) && ! delim_mode_intern )
1158 // User asked to skip *any* SeQuence
1159 Fp->seekg( l, std::ios::cur);
1163 // delay the dynamic cast as late as possible
1164 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1166 // no other way to create the Delimitor ...
1167 newSeqEntry->SetDelimitorMode( delim_mode_intern );
1169 // At the top of the hierarchy, stands a Document. When "set"
1170 // is a Document, then we are building the first depth level.
1171 // Hence the SeqEntry we are building simply has a depth
1173 // SQItem *parentSQItem = dynamic_cast< SQItem* > ( set );
1174 if ( set == this ) // ( dynamic_cast< Document* > ( set ) )
1176 newSeqEntry->SetDepthLevel( 1 );
1177 // newSeqEntry->SetKey( newSeqEntry->GetKey() );
1179 // But when "set" is already a SQItem, we are building a nested
1180 // sequence, and hence the depth level of the new SeqEntry
1181 // we are building, is one level deeper:
1183 // time waste hunting
1184 else if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1186 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1188 // newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1189 // + newSeqEntry->GetKey() );
1193 { // Don't try to parse zero-length sequences
1194 ParseSQ( newSeqEntry,
1195 newDocEntry->GetOffset(),
1196 l, delim_mode_intern);
1198 if ( !set->AddEntry( newSeqEntry ) )
1200 gdcmWarningMacro( "in ParseDES : cannot add a SeqEntry "
1201 << newSeqEntry->GetKey()
1203 << newSeqEntry->GetOffset() << " )" );
1207 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1213 } // end SeqEntry : VR = "SQ"
1224 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1225 * @return parsed length for this level
1227 void Document::ParseSQ( SeqEntry *seqEntry,
1228 long offset, long l_max, bool delim_mode)
1230 int SQItemNumber = 0;
1232 long offsetStartCurrentSQItem = offset;
1236 // the first time, we read the fff0,e000 of the first SQItem
1237 DocEntry *newDocEntry = ReadNextDocEntry();
1241 // FIXME Should warn user
1242 gdcmWarningMacro("in ParseSQ : should never get here!");
1247 if ( newDocEntry->IsSequenceDelimitor() )
1249 seqEntry->SetDelimitationItem( newDocEntry );
1253 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1258 // create the current SQItem
1259 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1261 std::ostringstream newBase;
1262 newBase << seqEntry->GetKey()
1266 itemSQ->SetBaseTagKey( newBase.str() );
1268 unsigned int l = newDocEntry->GetReadLength();
1270 if ( l == 0xffffffff )
1279 // Let's try :------------
1280 // remove fff0,e000, created out of the SQItem
1282 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg);
1283 // fill up the current SQItem, starting at the beginning of fff0,e000
1285 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
1287 offsetStartCurrentSQItem = Fp->tellg();
1288 // end try -----------------
1290 seqEntry->AddSQItem( itemSQ, SQItemNumber );
1292 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1300 * \brief When a private Sequence + Implicit VR is encountered
1301 * we cannot guess it's a Sequence till we find the first
1302 * Item Starter. We then backtrack to do the job.
1303 * @param docEntry Item Starter that warned us
1305 DocEntry *Document::Backtrack(DocEntry *docEntry)
1307 // delete the Item Starter, built erroneously out of any Sequence
1308 // it's not yet in the HTable/chained list
1311 // Get all info we can from PreviousDocEntry
1312 uint16_t group = PreviousDocEntry->GetGroup();
1313 uint16_t elem = PreviousDocEntry->GetElement();
1314 uint32_t lgt = PreviousDocEntry->GetLength();
1315 long offset = PreviousDocEntry->GetOffset();
1317 gdcmWarningMacro( "Backtrack :" << std::hex << group
1319 << " at offset " << offset );
1320 RemoveEntry( PreviousDocEntry );
1322 // forge the Seq Entry
1323 DocEntry *newEntry = NewSeqEntry(group, elem);
1324 newEntry->SetLength(lgt);
1325 newEntry->SetOffset(offset);
1327 // Move back to the beginning of the Sequence
1328 Fp->seekg( 0, std::ios::beg);
1329 Fp->seekg(offset, std::ios::cur);
1335 * \brief Loads (or not) the element content depending if its length exceeds
1336 * or not the value specified with Document::SetMaxSizeLoadEntry()
1337 * @param entry Header Entry (Dicom Element) to be dealt with
1339 void Document::LoadDocEntry(DocEntry *entry, bool forceLoad)
1341 uint16_t group = entry->GetGroup();
1342 std::string vr = entry->GetVR();
1343 uint32_t length = entry->GetLength();
1345 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1347 // A SeQuence "contains" a set of Elements.
1348 // (fffe e000) tells us an Element is beginning
1349 // (fffe e00d) tells us an Element just ended
1350 // (fffe e0dd) tells us the current SeQuence just ended
1351 if ( group == 0xfffe )
1353 // NO more value field for SQ !
1357 // When the length is zero things are easy:
1360 ((ValEntry *)entry)->SetValue("");
1364 // The elements whose length is bigger than the specified upper bound
1365 // are not loaded. Instead we leave a short notice on the offset of
1366 // the element content and it's length.
1368 std::ostringstream s;
1372 if (length > MaxSizeLoadEntry)
1374 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1376 s << GDCM_NOTLOADED;
1377 s << " Ad.:" << (long)entry->GetOffset();
1378 s << " x(" << std::hex << entry->GetOffset() << ")";
1380 s << " Lgt:" << entry->GetLength();
1381 s << " x(" << std::hex << entry->GetLength() << ")";
1382 binEntryPtr->SetValue(s.str());
1384 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1386 s << GDCM_NOTLOADED;
1387 s << " Address:" << (long)entry->GetOffset();
1388 s << " Length:" << entry->GetLength();
1389 s << " x(" << std::hex << entry->GetLength() << ")";
1390 valEntryPtr->SetValue(s.str());
1395 gdcmErrorMacro( "MaxSizeLoadEntry exceeded, neither a BinEntry "
1396 << "nor a ValEntry ?! Should never print that !" );
1399 // to be sure we are at the end of the value ...
1400 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1406 // When we find a BinEntry not very much can be done :
1407 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1409 s << GDCM_BINLOADED;
1410 binEntryPtr->SetValue(s.str());
1411 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1415 if ( IsDocEntryAnInteger(entry) )
1419 // When short integer(s) are expected, read and convert the following
1420 // (n * 2) characters properly i.e. consider them as short integers as
1421 // opposed to strings.
1422 // Elements with Value Multiplicity > 1
1423 // contain a set of integers (not a single one)
1424 if (vr == "US" || vr == "SS")
1427 NewInt = ReadInt16();
1431 for (int i=1; i < nbInt; i++)
1434 NewInt = ReadInt16();
1439 // See above comment on multiple integers (mutatis mutandis).
1440 else if (vr == "UL" || vr == "SL")
1443 NewInt = ReadInt32();
1447 for (int i=1; i < nbInt; i++)
1450 NewInt = ReadInt32();
1455 #ifdef GDCM_NO_ANSI_STRING_STREAM
1456 s << std::ends; // to avoid oddities on Solaris
1457 #endif //GDCM_NO_ANSI_STRING_STREAM
1459 ((ValEntry *)entry)->SetValue(s.str());
1463 // FIXME: We need an additional byte for storing \0 that is not on disk
1464 char *str = new char[length+1];
1465 Fp->read(str, (size_t)length);
1466 str[length] = '\0'; //this is only useful when length is odd
1467 // Special DicomString call to properly handle \0 and even length
1468 std::string newValue;
1471 newValue = Util::DicomString(str, length+1);
1472 gdcmWarningMacro("Warning: bad length: " << length <<
1473 " For string :" << newValue.c_str());
1474 // Since we change the length of string update it length
1475 //entry->SetReadLength(length+1);
1479 newValue = Util::DicomString(str, length);
1483 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1485 if ( Fp->fail() || Fp->eof())
1488 gdcmWarningMacro("--> fail");
1490 gdcmWarningMacro("Unread element value " << valEntry->GetKey()
1491 << " lgt : " << valEntry->GetReadLength()
1492 << " at " << std::hex << valEntry->GetOffset());
1493 valEntry->SetValue(GDCM_UNREAD);
1499 // Because of correspondance with the VR dic
1500 valEntry->SetValue(newValue);
1504 valEntry->SetValue(newValue);
1509 gdcmWarningMacro("Should have a ValEntry, here ! " << valEntry->GetKey()
1510 << " lgt : " << valEntry->GetReadLength()
1511 << " at " << std::hex << valEntry->GetOffset());
1516 * \brief Find the value Length of the passed Doc Entry
1517 * @param entry Header Entry whose length of the value shall be loaded.
1519 void Document::FindDocEntryLength( DocEntry *entry )
1520 throw ( FormatError )
1522 std::string vr = entry->GetVR();
1525 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1527 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UT"
1530 // The following reserved two bytes (see PS 3.5-2003, section
1531 // "7.1.2 Data element structure with explicit vr", p 27) must be
1532 // skipped before proceeding on reading the length on 4 bytes.
1533 Fp->seekg( 2L, std::ios::cur);
1534 uint32_t length32 = ReadInt32();
1536 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1541 lengthOB = FindDocEntryLengthOBOrOW();
1543 catch ( FormatUnexpected )
1545 // Computing the length failed (this happens with broken
1546 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1547 // chance to get the pixels by deciding the element goes
1548 // until the end of the file. Hence we artificially fix the
1549 // the length and proceed.
1550 gdcmWarningMacro( " Computing the length failed for " <<
1551 entry->GetKey() <<" in " <<GetFileName());
1553 long currentPosition = Fp->tellg();
1554 Fp->seekg(0L,std::ios::end);
1556 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1557 Fp->seekg(currentPosition, std::ios::beg);
1559 entry->SetReadLength(lengthUntilEOF);
1560 entry->SetLength(lengthUntilEOF);
1563 entry->SetReadLength(lengthOB);
1564 entry->SetLength(lengthOB);
1567 FixDocEntryFoundLength(entry, length32);
1571 // Length is encoded on 2 bytes.
1572 length16 = ReadInt16();
1574 // 0xffff means that we deal with 'No Length' Sequence
1575 // or 'No Length' SQItem
1576 if ( length16 == 0xffff)
1580 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1585 // Either implicit VR or a non DICOM conformal (see note below) explicit
1586 // VR that ommited the VR of (at least) this element. Farts happen.
1587 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1588 // on Data elements "Implicit and Explicit VR Data Elements shall
1589 // not coexist in a Data Set and Data Sets nested within it".]
1590 // Length is on 4 bytes.
1592 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1593 // even if Transfer Syntax is 'Implicit VR ...'
1595 FixDocEntryFoundLength( entry, ReadInt32() );
1601 * \brief Find the Length till the next sequence delimiter
1602 * \warning NOT end user intended method !
1605 uint32_t Document::FindDocEntryLengthOBOrOW()
1606 throw( FormatUnexpected )
1608 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1609 long positionOnEntry = Fp->tellg();
1610 bool foundSequenceDelimiter = false;
1611 uint32_t totalLength = 0;
1613 while ( !foundSequenceDelimiter )
1619 group = ReadInt16();
1622 catch ( FormatError )
1624 throw FormatError("Unexpected end of file encountered during ",
1625 "Document::FindDocEntryLengthOBOrOW()");
1627 // We have to decount the group and element we just read
1629 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
1631 long filePosition = Fp->tellg();
1633 "Neither an Item tag nor a Sequence delimiter tag on :"
1634 << std::hex << group << " , " << elem
1635 << ") -before- position x(" << filePosition << ")" );
1637 Fp->seekg(positionOnEntry, std::ios::beg);
1638 throw FormatUnexpected(
1639 "Neither an Item tag nor a Sequence delimiter tag.");
1641 if ( elem == 0xe0dd )
1643 foundSequenceDelimiter = true;
1645 uint32_t itemLength = ReadInt32();
1646 // We add 4 bytes since we just read the ItemLength with ReadInt32
1647 totalLength += itemLength + 4;
1648 SkipBytes(itemLength);
1650 if ( foundSequenceDelimiter )
1655 Fp->seekg( positionOnEntry, std::ios::beg);
1660 * \brief Find the Value Representation of the current Dicom Element.
1661 * @return Value Representation of the current Entry
1663 std::string Document::FindDocEntryVR()
1665 if ( Filetype != ExplicitVR )
1666 return GDCM_UNKNOWN;
1668 long positionOnEntry = Fp->tellg();
1669 // Warning: we believe this is explicit VR (Value Representation) because
1670 // we used a heuristic that found "UL" in the first tag. Alas this
1671 // doesn't guarantee that all the tags will be in explicit VR. In some
1672 // cases (see e-film filtered files) one finds implicit VR tags mixed
1673 // within an explicit VR file. Hence we make sure the present tag
1674 // is in explicit VR and try to fix things if it happens not to be
1678 Fp->read (vr, (size_t)2);
1681 if ( !CheckDocEntryVR(vr) )
1683 Fp->seekg(positionOnEntry, std::ios::beg);
1684 return GDCM_UNKNOWN;
1690 * \brief Check the correspondance between the VR of the header entry
1691 * and the taken VR. If they are different, the header entry is
1692 * updated with the new VR.
1693 * @param vr Dicom Value Representation
1694 * @return false if the VR is incorrect of if the VR isn't referenced
1695 * otherwise, it returns true
1697 bool Document::CheckDocEntryVR(VRKey vr)
1699 if ( !Global::GetVR()->IsValidVR(vr) )
1706 * \brief Get the transformed value of the header entry. The VR value
1707 * is used to define the transformation to operate on the value
1708 * \warning NOT end user intended method !
1709 * @param entry entry to tranform
1710 * @return Transformed entry value
1712 std::string Document::GetDocEntryValue(DocEntry *entry)
1714 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1716 std::string val = ((ValEntry *)entry)->GetValue();
1717 std::string vr = entry->GetVR();
1718 uint32_t length = entry->GetLength();
1719 std::ostringstream s;
1722 // When short integer(s) are expected, read and convert the following
1723 // n * 2 bytes properly i.e. as a multivaluated strings
1724 // (each single value is separated fromthe next one by '\'
1725 // as usual for standard multivaluated filels
1726 // Elements with Value Multiplicity > 1
1727 // contain a set of short integers (not a single one)
1729 if ( vr == "US" || vr == "SS" )
1734 for (int i=0; i < nbInt; i++)
1740 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1741 newInt16 = SwapShort( newInt16 );
1746 // When integer(s) are expected, read and convert the following
1747 // n * 4 bytes properly i.e. as a multivaluated strings
1748 // (each single value is separated fromthe next one by '\'
1749 // as usual for standard multivaluated filels
1750 // Elements with Value Multiplicity > 1
1751 // contain a set of integers (not a single one)
1752 else if ( vr == "UL" || vr == "SL" )
1757 for (int i=0; i < nbInt; i++)
1763 newInt32 = ( val[4*i+0] & 0xFF )
1764 + (( val[4*i+1] & 0xFF ) << 8 )
1765 + (( val[4*i+2] & 0xFF ) << 16 )
1766 + (( val[4*i+3] & 0xFF ) << 24 );
1767 newInt32 = SwapLong( newInt32 );
1771 #ifdef GDCM_NO_ANSI_STRING_STREAM
1772 s << std::ends; // to avoid oddities on Solaris
1773 #endif //GDCM_NO_ANSI_STRING_STREAM
1776 return ((ValEntry *)entry)->GetValue();
1780 * \brief Get the reverse transformed value of the header entry. The VR
1781 * value is used to define the reverse transformation to operate on
1783 * \warning NOT end user intended method !
1784 * @param entry Entry to reverse transform
1785 * @return Reverse transformed entry value
1787 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1789 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1791 std::string vr = entry->GetVR();
1792 std::vector<std::string> tokens;
1793 std::ostringstream s;
1795 if ( vr == "US" || vr == "SS" )
1799 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1800 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1801 for (unsigned int i=0; i<tokens.size(); i++)
1803 newInt16 = atoi(tokens[i].c_str());
1804 s << ( newInt16 & 0xFF )
1805 << (( newInt16 >> 8 ) & 0xFF );
1809 if ( vr == "UL" || vr == "SL")
1813 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1814 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1815 for (unsigned int i=0; i<tokens.size();i++)
1817 newInt32 = atoi(tokens[i].c_str());
1818 s << (char)( newInt32 & 0xFF )
1819 << (char)(( newInt32 >> 8 ) & 0xFF )
1820 << (char)(( newInt32 >> 16 ) & 0xFF )
1821 << (char)(( newInt32 >> 24 ) & 0xFF );
1826 #ifdef GDCM_NO_ANSI_STRING_STREAM
1827 s << std::ends; // to avoid oddities on Solaris
1828 #endif //GDCM_NO_ANSI_STRING_STREAM
1832 return ((ValEntry *)entry)->GetValue();
1836 * \brief Skip a given Header Entry
1837 * \warning NOT end user intended method !
1838 * @param entry entry to skip
1840 void Document::SkipDocEntry(DocEntry *entry)
1842 SkipBytes(entry->GetLength());
1846 * \brief Skips to the beginning of the next Header Entry
1847 * \warning NOT end user intended method !
1848 * @param currentDocEntry entry to skip
1850 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1852 int l = currentDocEntry->GetReadLength();
1853 if ( l == -1 ) // length = 0xffff shouldn't appear here ...
1854 // ... but PMS imagers happen !
1856 Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg);
1857 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1859 Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur);
1864 * \brief When the length of an element value is obviously wrong (because
1865 * the parser went Jabberwocky) one can hope improving things by
1866 * applying some heuristics.
1867 * @param entry entry to check
1868 * @param foundLength first assumption about length
1870 void Document::FixDocEntryFoundLength(DocEntry *entry,
1871 uint32_t foundLength)
1873 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1874 if ( foundLength == 0xffffffff)
1879 uint16_t gr = entry->GetGroup();
1880 uint16_t elem = entry->GetElement();
1882 if ( foundLength % 2)
1884 gdcmWarningMacro( "Warning : Tag with uneven length " << foundLength
1885 << " in x(" << std::hex << gr << "," << elem <<")");
1888 //////// Fix for some naughty General Electric images.
1889 // Allthough not recent many such GE corrupted images are still present
1890 // on Creatis hard disks. Hence this fix shall remain when such images
1891 // are no longer in use (we are talking a few years, here)...
1892 // Note: XMedCon probably uses such a trick since it is able to read
1893 // those pesky GE images ...
1894 if ( foundLength == 13)
1896 // Only happens for this length !
1897 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1900 entry->SetReadLength(10); // a bug is to be fixed !?
1904 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1905 // Occurence of such images is quite low (unless one leaves close to a
1906 // 'Leonardo' source. Hence, one might consider commenting out the
1907 // following fix on efficiency reasons.
1908 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1911 entry->SetReadLength(4); // a bug is to be fixed !?
1914 else if ( entry->GetVR() == "SQ" )
1916 foundLength = 0; // ReadLength is unchanged
1919 //////// We encountered a 'delimiter' element i.e. a tag of the form
1920 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1921 // taken into account.
1922 else if ( gr == 0xfffe )
1924 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1925 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1926 // causes extra troubles...
1927 if ( entry->GetElement() != 0x0000 )
1932 entry->SetLength(foundLength);
1936 * \brief Apply some heuristics to predict whether the considered
1937 * element value contains/represents an integer or not.
1938 * @param entry The element value on which to apply the predicate.
1939 * @return The result of the heuristical predicate.
1941 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1943 uint16_t elem = entry->GetElement();
1944 uint16_t group = entry->GetGroup();
1945 const std::string &vr = entry->GetVR();
1946 uint32_t length = entry->GetLength();
1948 // When we have some semantics on the element we just read, and if we
1949 // a priori know we are dealing with an integer, then we shall be
1950 // able to swap it's element value properly.
1951 if ( elem == 0 ) // This is the group length of the group
1959 // Allthough this should never happen, still some images have a
1960 // corrupted group length [e.g. have a glance at offset x(8336) of
1961 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
1962 // Since for dicom compliant and well behaved headers, the present
1963 // test is useless (and might even look a bit paranoid), when we
1964 // encounter such an ill-formed image, we simply display a warning
1965 // message and proceed on parsing (while crossing fingers).
1966 long filePosition = Fp->tellg();
1967 gdcmWarningMacro( "Erroneous Group Length element length on : ("
1968 << std::hex << group << " , " << elem
1969 << ") -before- position x(" << filePosition << ")"
1970 << "lgt : " << length );
1974 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1982 * \brief Discover what the swap code is (among little endian, big endian,
1983 * bad little endian, bad big endian).
1985 * @return false when we are absolutely sure
1986 * it's neither ACR-NEMA nor DICOM
1987 * true when we hope ours assuptions are OK
1989 bool Document::CheckSwap()
1996 // First, compare HostByteOrder and NetworkByteOrder in order to
1997 // determine if we shall need to swap bytes (i.e. the Endian type).
1998 bool net2host = Util::IsCurrentProcessorBigEndian();
2000 // The easiest case is the one of a 'true' DICOM header, we just have
2001 // to look for the string "DICM" inside the file preamble.
2004 char *entCur = deb + 128;
2005 if ( memcmp(entCur, "DICM", (size_t)4) == 0 )
2007 gdcmWarningMacro( "Looks like DICOM Version3 (preamble + DCM)" );
2009 // Group 0002 should always be VR, and the first element 0000
2010 // Let's be carefull (so many wrong headers ...)
2011 // and determine the value representation (VR) :
2012 // Let's skip to the first element (0002,0000) and check there if we find
2013 // "UL" - or "OB" if the 1st one is (0002,0001) -,
2014 // in which case we (almost) know it is explicit VR.
2015 // WARNING: if it happens to be implicit VR then what we will read
2016 // is the length of the group. If this ascii representation of this
2017 // length happens to be "UL" then we shall believe it is explicit VR.
2018 // We need to skip :
2019 // * the 128 bytes of File Preamble (often padded with zeroes),
2020 // * the 4 bytes of "DICM" string,
2021 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2022 // i.e. a total of 136 bytes.
2025 // group 0x0002 *is always* Explicit VR Sometimes ,
2026 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
2027 // *Implicit* VR (see former 'gdcmData/icone.dcm')
2029 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
2030 memcmp(entCur, "OB", (size_t)2) == 0 ||
2031 memcmp(entCur, "UI", (size_t)2) == 0 ||
2032 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2033 // when Write DCM *adds*
2035 // Use Document::dicom_vr to test all the possibilities
2036 // instead of just checking for UL, OB and UI !? group 0000
2038 Filetype = ExplicitVR;
2039 gdcmWarningMacro( "Group 0002 : Explicit Value Representation");
2043 Filetype = ImplicitVR;
2044 gdcmWarningMacro( "Group 0002 :Not an explicit Value Representation;"
2045 << "Looks like a bugged Header!");
2051 gdcmWarningMacro( "HostByteOrder != NetworkByteOrder");
2056 gdcmWarningMacro( "HostByteOrder = NetworkByteOrder");
2059 // Position the file position indicator at first tag
2060 // (i.e. after the file preamble and the "DICM" string).
2062 Fp->seekg(0, std::ios::beg); // FIXME : Is it usefull?
2064 Fp->seekg ( 132L, std::ios::beg);
2066 } // ------------------------------- End of DicomV3 ----------------
2068 // Alas, this is not a DicomV3 file and whatever happens there is no file
2069 // preamble. We can reset the file position indicator to where the data
2070 // is (i.e. the beginning of the file).
2072 gdcmWarningMacro( "Not a Kosher DICOM Version3 file (no preamble)");
2074 Fp->seekg(0, std::ios::beg);
2076 // Let's check 'No Preamble Dicom File' :
2077 // Should start with group 0x0002
2078 // and be Explicit Value Representation
2080 s16 = *((uint16_t *)(deb));
2093 if ( SwapCode != 0 )
2095 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
2096 memcmp(entCur, "OB", (size_t)2) == 0 ||
2097 memcmp(entCur, "UI", (size_t)2) == 0 ||
2098 memcmp(entCur, "SH", (size_t)2) == 0 ||
2099 memcmp(entCur, "AE", (size_t)2) == 0 ||
2100 memcmp(entCur, "OB", (size_t)2) == 0 )
2102 Filetype = ExplicitVR;
2103 gdcmWarningMacro( "Group 0002 : Explicit Value Representation");
2107 // ------------------------------- End of 'No Preamble' DicomV3 -------------
2109 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2110 // By clean we mean that the length of the first group is written down.
2111 // If this is the case and since the length of the first group HAS to be
2112 // four (bytes), then determining the proper swap code is straightforward.
2115 // We assume the array of char we are considering contains the binary
2116 // representation of a 32 bits integer. Hence the following dirty
2118 s32 = *((uint32_t *)(entCur));
2138 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2139 // It is time for despaired wild guesses.
2140 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2141 // i.e. the 'group length' element is not present :
2143 // check the supposed-to-be 'group number'
2144 // in ( 0x0001 .. 0x0008 )
2145 // to determine ' SwapCode' value .
2146 // Only 0 or 4321 will be possible
2147 // (no oportunity to check for the formerly well known
2148 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2149 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2150 // the file IS NOT ACR-NEMA nor DICOM V3
2151 // Find a trick to tell it the caller...
2153 s16 = *((uint16_t *)(deb));
2180 gdcmWarningMacro("ACR/NEMA unfound swap info (Really hopeless !)");
2188 * \brief Change the Byte Swap code.
2190 void Document::SwitchByteSwapCode()
2192 gdcmWarningMacro( "Switching Byte Swap code from "<< SwapCode
2193 << " at :" <<std::hex << Fp->tellg() );
2194 if ( SwapCode == 1234 )
2198 else if ( SwapCode == 4321 )
2202 else if ( SwapCode == 3412 )
2206 else if ( SwapCode == 2143 )
2213 * \brief during parsing, Header Elements too long are not loaded in memory
2214 * @param newSize new size
2216 void Document::SetMaxSizeLoadEntry(long newSize)
2222 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2224 MaxSizeLoadEntry = 0xffffffff;
2227 MaxSizeLoadEntry = newSize;
2231 * \brief Read the next tag WITHOUT loading it's value
2232 * (read the 'Group Number', the 'Element Number',
2233 * gets the Dict Entry
2234 * gets the VR, gets the length, gets the offset value)
2235 * @return On succes : the newly created DocEntry, NULL on failure.
2237 DocEntry *Document::ReadNextDocEntry()
2244 group = ReadInt16();
2247 catch ( FormatError )
2249 // We reached the EOF (or an error occured) therefore
2250 // header parsing has to be considered as finished.
2254 // Sometimes file contains groups of tags with reversed endianess.
2255 HandleBrokenEndian(group, elem);
2257 // In 'true DICOM' files Group 0002 is always little endian
2258 if ( HasDCMPreamble )
2259 HandleOutOfGroup0002(group, elem);
2261 std::string vr = FindDocEntryVR();
2262 std::string realVR = vr;
2264 if ( vr == GDCM_UNKNOWN )
2266 if ( elem == 0x0000 ) // Group Length
2267 realVR = "UL"; // must be UL
2270 DictEntry *dictEntry = GetDictEntry(group,elem);
2273 realVR = dictEntry->GetVR();
2279 if ( Global::GetVR()->IsVROfSequence(realVR) )
2280 newEntry = NewSeqEntry(group, elem);
2281 else if ( Global::GetVR()->IsVROfStringRepresentable(realVR) )
2282 newEntry = NewValEntry(group, elem,vr);
2284 newEntry = NewBinEntry(group, elem,vr);
2286 if ( vr == GDCM_UNKNOWN )
2288 if ( Filetype == ExplicitVR )
2290 // We thought this was explicit VR, but we end up with an
2291 // implicit VR tag. Let's backtrack.
2292 if ( newEntry->GetGroup() != 0xfffe )
2295 int offset = Fp->tellg();
2296 msg = Util::Format("Entry (%04x,%04x) at 0x(%x) should be Explicit VR\n",
2297 newEntry->GetGroup(), newEntry->GetElement(), offset );
2298 gdcmWarningMacro( msg.c_str() );
2301 newEntry->SetImplicitVR();
2306 FindDocEntryLength(newEntry);
2308 catch ( FormatError )
2315 newEntry->SetOffset(Fp->tellg());
2321 * \brief Handle broken private tag from Philips NTSCAN
2322 * where the endianess is being switched to BigEndian
2323 * for no apparent reason
2326 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
2328 // Endian reversion. Some files contain groups of tags with reversed endianess.
2329 static int reversedEndian = 0;
2330 // try to fix endian switching in the middle of headers
2331 if ((group == 0xfeff) && (elem == 0x00e0))
2333 // start endian swap mark for group found
2335 SwitchByteSwapCode();
2340 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2342 // end of reversed endian group
2344 SwitchByteSwapCode();
2349 * \brief Group 0002 is always coded Little Endian
2350 * whatever Transfer Syntax is
2353 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2355 // Endian reversion. Some files contain groups of tags with reversed endianess.
2356 if ( !Group0002Parsed && group != 0x0002)
2358 Group0002Parsed = true;
2359 // we just came out of group 0002
2360 // if Transfer syntax is Big Endian we have to change CheckSwap
2362 std::string ts = GetTransferSyntax();
2363 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2365 gdcmWarningMacro("True DICOM File, with NO Tansfer Syntax: " << ts );
2369 // Group 0002 is always 'Explicit ...' enven when Transfer Syntax says 'Implicit ..."
2371 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ImplicitVRLittleEndian )
2373 Filetype = ImplicitVR;
2376 // FIXME Strangely, this works with
2377 //'Implicit VR Transfer Syntax (GE Private)
2378 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ExplicitVRBigEndian )
2380 gdcmWarningMacro("Transfer Syntax Name = ["
2381 << GetTransferSyntaxName() << "]" );
2382 SwitchByteSwapCode();
2383 group = SwapShort(group);
2384 elem = SwapShort(elem);
2389 //-----------------------------------------------------------------------------
2392 //-----------------------------------------------------------------------------
2393 } // end namespace gdcm