1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/09/22 14:41:24 $
7 Version: $Revision: 1.284 $
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 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)
656 void Document::WriteContent(std::ofstream *fp, FileType filetype)
658 // Skip if user wants to write an ACR-NEMA file
660 if ( filetype == ImplicitVR || filetype == ExplicitVR )
662 // writing Dicom File Preamble
663 char filePreamble[128];
664 memset(filePreamble, 0, 128);
665 fp->write(filePreamble, 128);
666 fp->write("DICM", 4);
670 * \todo rewrite later, if really usefull
671 * - 'Group Length' element is optional in DICOM
672 * - but un-updated odd groups lengthes can causes pb
675 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
676 * UpdateGroupLength(false,filetype);
677 * if ( filetype == ACR)
678 * UpdateGroupLength(true,ACR);
681 ElementSet::WriteContent(fp, filetype); // This one is recursive
684 // -----------------------------------------
687 * \brief Loads (from disk) the element content
688 * when a string is not suitable
689 * @param group group number of the Entry
690 * @param elem element number of the Entry
692 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
694 // Search the corresponding DocEntry
695 DocEntry *docElement = GetDocEntry(group, elem);
698 gdcmWarningMacro(std::hex << group << "|" << elem
699 << "doesn't exist" );
702 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
705 gdcmWarningMacro(std::hex << group << "|" << elem
706 << "is NOT a BinEntry");
709 LoadEntryBinArea(binElement);
713 * \brief Loads (from disk) the element content
714 * when a string is not suitable
715 * @param elem Entry whose binArea is going to be loaded
717 void Document::LoadEntryBinArea(BinEntry *elem)
719 if (elem->GetBinArea() )
726 size_t o =(size_t)elem->GetOffset();
727 Fp->seekg(o, std::ios::beg);
729 size_t l = elem->GetLength();
730 uint8_t *a = new uint8_t[l];
733 gdcmWarningMacro( "Cannot allocate BinEntry content for : "
734 << std::hex << elem->GetGroup()
735 << "|" << elem->GetElement() );
739 Fp->read((char*)a, l);
740 if ( Fp->fail() || Fp->eof() )
753 * \brief Loads the element while preserving the current
754 * underlying file position indicator as opposed to
755 * LoadDocEntry that modifies it.
756 * @param entry DocEntry whose value will be loaded.
758 void Document::LoadDocEntrySafe(DocEntry *entry)
762 long PositionOnEntry = Fp->tellg();
764 Fp->seekg(PositionOnEntry, std::ios::beg);
769 * \brief Compares two documents, according to \ref DicomDir rules
770 * \warning Does NOT work with ACR-NEMA files
771 * \todo Find a trick to solve the pb (use RET fields ?)
772 * @param document to compare with current one
773 * @return true if 'smaller'
775 bool Document::operator<(Document &document)
778 std::string s1 = GetEntryValue(0x0010,0x0010);
779 std::string s2 = document.GetEntryValue(0x0010,0x0010);
791 s1 = GetEntryValue(0x0010,0x0020);
792 s2 = document.GetEntryValue(0x0010,0x0020);
803 // Study Instance UID
804 s1 = GetEntryValue(0x0020,0x000d);
805 s2 = document.GetEntryValue(0x0020,0x000d);
816 // Serie Instance UID
817 s1 = GetEntryValue(0x0020,0x000e);
818 s2 = document.GetEntryValue(0x0020,0x000e);
833 //-----------------------------------------------------------------------------
836 * \brief Reads a supposed to be 16 Bits integer
837 * (swaps it depending on processor endianness)
840 uint16_t Document::ReadInt16()
844 Fp->read ((char*)&g, (size_t)2);
847 throw FormatError( "Document::ReadInt16()", " file error." );
851 throw FormatError( "Document::ReadInt16()", "EOF." );
858 * \brief Reads a supposed to be 32 Bits integer
859 * (swaps it depending on processor endianness)
862 uint32_t Document::ReadInt32()
866 Fp->read ((char*)&g, (size_t)4);
869 throw FormatError( "Document::ReadInt32()", " file error." );
873 throw FormatError( "Document::ReadInt32()", "EOF." );
880 * \brief skips bytes inside the source file
881 * \warning NOT end user intended method !
884 void Document::SkipBytes(uint32_t nBytes)
886 //FIXME don't dump the returned value
887 Fp->seekg((long)nBytes, std::ios::cur);
891 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
893 int Document::ComputeGroup0002Length( /*FileType filetype*/ )
899 bool found0002 = false;
901 // for each zero-level Tag in the DCM Header
902 DocEntry *entry = GetFirstEntry();
905 gr = entry->GetGroup();
911 if ( entry->GetElement() != 0x0000 )
915 // FIXME : group 0x0002 is *always* Explicit VR!
917 //if ( filetype == ExplicitVR )
919 // if ( (vr == "OB") || (vr == "OW") || (vr == "UT") || (vr == "SQ") )
920 // (no SQ, OW, UT in group 0x0002;)
923 // explicit VR AND OB, OW, SQ, UT : 4 more bytes
927 groupLength += 2 + 2 + 4 + entry->GetLength();
933 entry = GetNextEntry();
938 //-----------------------------------------------------------------------------
941 * \brief Loads all the needed Dictionaries
942 * \warning NOT end user intended method !
944 void Document::Initialize()
946 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
952 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
953 * @param set DocEntrySet we are going to parse ('zero level' or a SQItem)
954 * @param offset start of parsing
955 * @param l_max length to parse (meaningless when we are in 'delimitor mode')
956 * @param delim_mode : whether we are in 'delimitor mode' (l=0xffffff) or not
958 void Document::ParseDES(DocEntrySet *set, long offset,
959 long l_max, bool delim_mode)
961 DocEntry *newDocEntry;
962 ValEntry *newValEntry;
963 BinEntry *newBinEntry;
964 SeqEntry *newSeqEntry;
966 bool used; // will be set to false when something wrong happens to an Entry.
967 // (Entry will then be deleted)
968 bool delim_mode_intern = delim_mode;
972 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
977 newDocEntry = ReadNextDocEntry( );
980 // Private tag, in IMplicit VR are defaulted as a BinEntry,
981 // Very often they are only composed of Printable characters,
982 // and could be defaulted as a ValEntry.
983 // It's too late to do the Job
984 // (we should check the value, but we know it after LoadDocEntry ...)
985 // --> in next gdcm major release let's unify ValEntry and BinEntry !
987 // Uncoment this printf line to be able to 'follow' the DocEntries
988 // when something *very* strange happens
990 //printf( "%04x|%04x %s\n",newDocEntry->GetGroup(),
991 // newDocEntry->GetElement(),
992 // newDocEntry->GetVR().c_str() );
999 // an Item Starter found elsewhere but the first position
1000 // of a SeqEntry means previous entry was a Sequence
1001 // but we didn't get it (private Sequence + Implicit VR)
1002 // we have to backtrack.
1003 if ( !first && newDocEntry->IsItemStarter() )
1005 newDocEntry = Backtrack(newDocEntry);
1009 PreviousDocEntry = newDocEntry;
1013 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
1014 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
1016 if ( newValEntry || newBinEntry )
1018 //////////////////////////// ContentEntry
1021 vr = newDocEntry->GetVR();
1022 if ( Filetype == ExplicitVR &&
1023 !Global::GetVR()->IsVROfBinaryRepresentable(vr) )
1025 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
1026 gdcmWarningMacro( std::hex << newDocEntry->GetGroup()
1027 << "|" << newDocEntry->GetElement()
1028 << " : Neither Valentry, nor BinEntry."
1029 "Probably unknown VR.");
1032 //////////////////// BinEntry or UNKOWN VR:
1034 // When "this" is a Document the Key is simply of the
1035 // form ( group, elem )...
1036 //if ( set == this ) // ( dynamic_cast< Document* > ( set ) )
1038 // newBinEntry->SetKey( newBinEntry->GetKey() );
1040 // but when "this" is a SQItem, we are inserting this new
1041 // valEntry in a sequence item, and the key has the
1042 // generalized form (refer to \ref BaseTagKey):
1044 // time waste hunting
1045 //if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1047 // newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
1048 // + newBinEntry->GetKey() );
1051 if ( !set->AddEntry( newBinEntry ) )
1053 gdcmWarningMacro( "in ParseDES : cannot add a BinEntry "
1054 << newBinEntry->GetKey()
1056 << newBinEntry->GetOffset() << " )" );
1061 // Load only if we can add (not a duplicate key)
1062 LoadDocEntry( newBinEntry );
1067 /////////////////////// ValEntry
1069 // When "set" is a Document, then we are at the top of the
1070 // hierarchy and the Key is simply of the form ( group, elem )...
1071 //if ( set == this ) // ( dynamic_cast< Document* > ( set ) )
1073 // newValEntry->SetKey( newValEntry->GetKey() );
1075 // ...but when "set" is a SQItem, we are inserting this new
1076 // valEntry in a sequence item. Hence the key has the
1077 // generalized form (refer to \ref BaseTagKey):
1079 // time waste hunting
1080 //if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1082 // newValEntry->SetKey( parentSQItem->GetBaseTagKey()
1083 // + newValEntry->GetKey() );
1086 if ( !set->AddEntry( newValEntry ) )
1088 gdcmWarningMacro( "in ParseDES : cannot add a ValEntry "
1089 << newValEntry->GetKey()
1091 << newValEntry->GetOffset() << " )" );
1096 // Load only if we can add (not a duplicate key)
1097 LoadDocEntry( newValEntry );
1100 if ( newValEntry->GetElement() == 0x0000 ) // if on group length
1102 if ( newValEntry->GetGroup()%2 != 0 ) // if Shadow Group
1104 if ( LoadMode & LD_NOSHADOW ) // if user asked to skip shad.gr
1106 std::string strLgrGroup = newValEntry->GetValue();
1108 if ( strLgrGroup != GDCM_UNFOUND)
1110 lgrGroup = atoi(strLgrGroup.c_str());
1111 Fp->seekg(lgrGroup, std::ios::cur);
1112 //used = false; // never used
1113 RemoveEntry( newDocEntry ); // Remove and delete
1114 // bcc 5.5 is right "assigned a value that's never used"
1122 bool delimitor = newValEntry->IsItemDelimitor();
1125 (!delim_mode && ((long)(Fp->tellg())-offset) >= l_max) )
1133 // Just to make sure we are at the beginning of next entry.
1134 SkipToNextDocEntry(newDocEntry);
1138 /////////////////////// SeqEntry : VR = "SQ"
1140 unsigned long l = newDocEntry->GetReadLength();
1141 if ( l != 0 ) // don't mess the delim_mode for 'zero-length sequence'
1143 if ( l == 0xffffffff )
1145 delim_mode_intern = true;
1149 delim_mode_intern = false;
1153 if ( (LoadMode & LD_NOSHADOWSEQ) && ! delim_mode_intern )
1155 // User asked to skip SeQuences *only* if they belong to Shadow Group
1156 if ( newDocEntry->GetGroup()%2 != 0 )
1158 Fp->seekg( l, std::ios::cur);
1159 RemoveEntry( newDocEntry ); // Remove and delete
1160 //used = false; // never used
1164 if ( (LoadMode & LD_NOSEQ) && ! delim_mode_intern )
1166 // User asked to skip *any* SeQuence
1167 Fp->seekg( l, std::ios::cur);
1168 //used = false; // never used
1169 RemoveEntry( newDocEntry ); // Remove and delete
1172 // delay the dynamic cast as late as possible
1173 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1175 // no other way to create the Delimitor ...
1176 newSeqEntry->SetDelimitorMode( delim_mode_intern );
1178 // At the top of the hierarchy, stands a Document. When "set"
1179 // is a Document, then we are building the first depth level.
1180 // Hence the SeqEntry we are building simply has a depth
1182 // SQItem *parentSQItem = dynamic_cast< SQItem* > ( set );
1183 if ( set == this ) // ( dynamic_cast< Document* > ( set ) )
1185 newSeqEntry->SetDepthLevel( 1 );
1186 // newSeqEntry->SetKey( newSeqEntry->GetKey() );
1188 // But when "set" is already a SQItem, we are building a nested
1189 // sequence, and hence the depth level of the new SeqEntry
1190 // we are building, is one level deeper:
1192 // time waste hunting
1193 else if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1195 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1197 // newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
1198 // + newSeqEntry->GetKey() );
1202 { // Don't try to parse zero-length sequences
1203 ParseSQ( newSeqEntry,
1204 newDocEntry->GetOffset(),
1205 l, delim_mode_intern);
1207 if ( !set->AddEntry( newSeqEntry ) )
1209 gdcmWarningMacro( "in ParseDES : cannot add a SeqEntry "
1210 << newSeqEntry->GetKey()
1212 << newSeqEntry->GetOffset() << " )" );
1216 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1222 } // end SeqEntry : VR = "SQ"
1233 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1234 * @return parsed length for this level
1236 void Document::ParseSQ( SeqEntry *seqEntry,
1237 long offset, long l_max, bool delim_mode)
1239 int SQItemNumber = 0;
1241 long offsetStartCurrentSQItem = offset;
1245 // the first time, we read the fff0,e000 of the first SQItem
1246 DocEntry *newDocEntry = ReadNextDocEntry();
1250 // FIXME Should warn user
1251 gdcmWarningMacro("in ParseSQ : should never get here!");
1256 if ( newDocEntry->IsSequenceDelimitor() )
1258 seqEntry->SetDelimitationItem( newDocEntry );
1262 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
1267 // create the current SQItem
1268 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
1270 std::ostringstream newBase;
1271 newBase << seqEntry->GetKey()
1275 itemSQ->SetBaseTagKey( newBase.str() );
1277 unsigned int l = newDocEntry->GetReadLength();
1279 if ( l == 0xffffffff )
1288 // Let's try :------------
1289 // remove fff0,e000, created out of the SQItem
1291 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg);
1292 // fill up the current SQItem, starting at the beginning of fff0,e000
1294 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
1296 offsetStartCurrentSQItem = Fp->tellg();
1297 // end try -----------------
1299 seqEntry->AddSQItem( itemSQ, SQItemNumber );
1301 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
1309 * \brief When a private Sequence + Implicit VR is encountered
1310 * we cannot guess it's a Sequence till we find the first
1311 * Item Starter. We then backtrack to do the job.
1312 * @param docEntry Item Starter that warned us
1314 DocEntry *Document::Backtrack(DocEntry *docEntry)
1316 // delete the Item Starter, built erroneously out of any Sequence
1317 // it's not yet in the HTable/chained list
1320 // Get all info we can from PreviousDocEntry
1321 uint16_t group = PreviousDocEntry->GetGroup();
1322 uint16_t elem = PreviousDocEntry->GetElement();
1323 uint32_t lgt = PreviousDocEntry->GetLength();
1324 long offset = PreviousDocEntry->GetOffset();
1326 gdcmWarningMacro( "Backtrack :" << std::hex << group
1328 << " at offset " << offset );
1329 RemoveEntry( PreviousDocEntry );
1331 // forge the Seq Entry
1332 DocEntry *newEntry = NewSeqEntry(group, elem);
1333 newEntry->SetLength(lgt);
1334 newEntry->SetOffset(offset);
1336 // Move back to the beginning of the Sequence
1337 Fp->seekg( 0, std::ios::beg);
1338 Fp->seekg(offset, std::ios::cur);
1344 * \brief Loads (or not) the element content depending if its length exceeds
1345 * or not the value specified with Document::SetMaxSizeLoadEntry()
1346 * @param entry Header Entry (Dicom Element) to be dealt with
1348 void Document::LoadDocEntry(DocEntry *entry, bool forceLoad)
1350 uint16_t group = entry->GetGroup();
1351 std::string vr = entry->GetVR();
1352 uint32_t length = entry->GetLength();
1354 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
1356 // A SeQuence "contains" a set of Elements.
1357 // (fffe e000) tells us an Element is beginning
1358 // (fffe e00d) tells us an Element just ended
1359 // (fffe e0dd) tells us the current SeQuence just ended
1360 if ( group == 0xfffe )
1362 // NO more value field for SQ !
1366 // When the length is zero things are easy:
1369 ((ValEntry *)entry)->SetValue("");
1373 // The elements whose length is bigger than the specified upper bound
1374 // are not loaded. Instead we leave a short notice on the offset of
1375 // the element content and it's length.
1377 std::ostringstream s;
1381 if (length > MaxSizeLoadEntry)
1383 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1385 s << GDCM_NOTLOADED;
1386 s << " Ad.:" << (long)entry->GetOffset();
1387 s << " x(" << std::hex << entry->GetOffset() << ")";
1389 s << " Lgt:" << entry->GetLength();
1390 s << " x(" << std::hex << entry->GetLength() << ")";
1391 binEntryPtr->SetValue(s.str());
1393 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
1395 s << GDCM_NOTLOADED;
1396 s << " Address:" << (long)entry->GetOffset();
1397 s << " Length:" << entry->GetLength();
1398 s << " x(" << std::hex << entry->GetLength() << ")";
1399 valEntryPtr->SetValue(s.str());
1404 gdcmErrorMacro( "MaxSizeLoadEntry exceeded, neither a BinEntry "
1405 << "nor a ValEntry ?! Should never print that !" );
1408 // to be sure we are at the end of the value ...
1409 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1415 // When we find a BinEntry not very much can be done :
1416 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
1418 s << GDCM_BINLOADED;
1419 binEntryPtr->SetValue(s.str());
1420 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
1424 if ( IsDocEntryAnInteger(entry) )
1428 // When short integer(s) are expected, read and convert the following
1429 // (n * 2) characters properly i.e. consider them as short integers as
1430 // opposed to strings.
1431 // Elements with Value Multiplicity > 1
1432 // contain a set of integers (not a single one)
1433 if (vr == "US" || vr == "SS")
1436 NewInt = ReadInt16();
1440 for (int i=1; i < nbInt; i++)
1443 NewInt = ReadInt16();
1448 // See above comment on multiple integers (mutatis mutandis).
1449 else if (vr == "UL" || vr == "SL")
1452 NewInt = ReadInt32();
1456 for (int i=1; i < nbInt; i++)
1459 NewInt = ReadInt32();
1464 #ifdef GDCM_NO_ANSI_STRING_STREAM
1465 s << std::ends; // to avoid oddities on Solaris
1466 #endif //GDCM_NO_ANSI_STRING_STREAM
1468 ((ValEntry *)entry)->SetValue(s.str());
1472 // FIXME: We need an additional byte for storing \0 that is not on disk
1473 char *str = new char[length+1];
1474 Fp->read(str, (size_t)length);
1475 str[length] = '\0'; //this is only useful when length is odd
1476 // Special DicomString call to properly handle \0 and even length
1477 std::string newValue;
1480 newValue = Util::DicomString(str, length+1);
1481 gdcmWarningMacro("Warning: bad length: " << length <<
1482 " For string :" << newValue.c_str());
1483 // Since we change the length of string update it length
1484 //entry->SetReadLength(length+1);
1488 newValue = Util::DicomString(str, length);
1492 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1494 if ( Fp->fail() || Fp->eof())
1497 gdcmWarningMacro("--> fail");
1499 gdcmWarningMacro("Unread element value " << valEntry->GetKey()
1500 << " lgt : " << valEntry->GetReadLength()
1501 << " at " << std::hex << valEntry->GetOffset());
1502 valEntry->SetValue(GDCM_UNREAD);
1508 // Because of correspondance with the VR dic
1509 valEntry->SetValue(newValue);
1513 valEntry->SetValue(newValue);
1518 gdcmWarningMacro("Should have a ValEntry, here ! " << valEntry->GetKey()
1519 << " lgt : " << valEntry->GetReadLength()
1520 << " at " << std::hex << valEntry->GetOffset());
1525 * \brief Find the value Length of the passed Doc Entry
1526 * @param entry Header Entry whose length of the value shall be loaded.
1528 void Document::FindDocEntryLength( DocEntry *entry )
1529 throw ( FormatError )
1531 std::string vr = entry->GetVR();
1534 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1536 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UT"
1539 // The following reserved two bytes (see PS 3.5-2003, section
1540 // "7.1.2 Data element structure with explicit vr", p 27) must be
1541 // skipped before proceeding on reading the length on 4 bytes.
1542 Fp->seekg( 2L, std::ios::cur);
1543 uint32_t length32 = ReadInt32();
1545 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1550 lengthOB = FindDocEntryLengthOBOrOW();
1552 catch ( FormatUnexpected )
1554 // Computing the length failed (this happens with broken
1555 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1556 // chance to get the pixels by deciding the element goes
1557 // until the end of the file. Hence we artificially fix the
1558 // the length and proceed.
1559 gdcmWarningMacro( " Computing the length failed for " <<
1560 entry->GetKey() <<" in " <<GetFileName());
1562 long currentPosition = Fp->tellg();
1563 Fp->seekg(0L,std::ios::end);
1565 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1566 Fp->seekg(currentPosition, std::ios::beg);
1568 entry->SetReadLength(lengthUntilEOF);
1569 entry->SetLength(lengthUntilEOF);
1572 entry->SetReadLength(lengthOB);
1573 entry->SetLength(lengthOB);
1576 FixDocEntryFoundLength(entry, length32);
1580 // Length is encoded on 2 bytes.
1581 length16 = ReadInt16();
1583 // 0xffff means that we deal with 'No Length' Sequence
1584 // or 'No Length' SQItem
1585 if ( length16 == 0xffff)
1589 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1594 // Either implicit VR or a non DICOM conformal (see note below) explicit
1595 // VR that ommited the VR of (at least) this element. Farts happen.
1596 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1597 // on Data elements "Implicit and Explicit VR Data Elements shall
1598 // not coexist in a Data Set and Data Sets nested within it".]
1599 // Length is on 4 bytes.
1601 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1602 // even if Transfer Syntax is 'Implicit VR ...'
1604 FixDocEntryFoundLength( entry, ReadInt32() );
1610 * \brief Find the Length till the next sequence delimiter
1611 * \warning NOT end user intended method !
1614 uint32_t Document::FindDocEntryLengthOBOrOW()
1615 throw( FormatUnexpected )
1617 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1618 long positionOnEntry = Fp->tellg();
1619 bool foundSequenceDelimiter = false;
1620 uint32_t totalLength = 0;
1622 while ( !foundSequenceDelimiter )
1628 group = ReadInt16();
1631 catch ( FormatError )
1633 throw FormatError("Unexpected end of file encountered during ",
1634 "Document::FindDocEntryLengthOBOrOW()");
1636 // We have to decount the group and element we just read
1638 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
1640 long filePosition = Fp->tellg();
1642 "Neither an Item tag nor a Sequence delimiter tag on :"
1643 << std::hex << group << " , " << elem
1644 << ") -before- position x(" << filePosition << ")" );
1646 Fp->seekg(positionOnEntry, std::ios::beg);
1647 throw FormatUnexpected(
1648 "Neither an Item tag nor a Sequence delimiter tag.");
1650 if ( elem == 0xe0dd )
1652 foundSequenceDelimiter = true;
1654 uint32_t itemLength = ReadInt32();
1655 // We add 4 bytes since we just read the ItemLength with ReadInt32
1656 totalLength += itemLength + 4;
1657 SkipBytes(itemLength);
1659 if ( foundSequenceDelimiter )
1664 Fp->seekg( positionOnEntry, std::ios::beg);
1669 * \brief Find the Value Representation of the current Dicom Element.
1670 * @return Value Representation of the current Entry
1672 std::string Document::FindDocEntryVR()
1674 if ( Filetype != ExplicitVR )
1675 return GDCM_UNKNOWN;
1677 long positionOnEntry = Fp->tellg();
1678 // Warning: we believe this is explicit VR (Value Representation) because
1679 // we used a heuristic that found "UL" in the first tag. Alas this
1680 // doesn't guarantee that all the tags will be in explicit VR. In some
1681 // cases (see e-film filtered files) one finds implicit VR tags mixed
1682 // within an explicit VR file. Hence we make sure the present tag
1683 // is in explicit VR and try to fix things if it happens not to be
1687 Fp->read (vr, (size_t)2);
1690 if ( !CheckDocEntryVR(vr) )
1692 Fp->seekg(positionOnEntry, std::ios::beg);
1693 return GDCM_UNKNOWN;
1699 * \brief Check the correspondance between the VR of the header entry
1700 * and the taken VR. If they are different, the header entry is
1701 * updated with the new VR.
1702 * @param vr Dicom Value Representation
1703 * @return false if the VR is incorrect or if the VR isn't referenced
1704 * otherwise, it returns true
1706 bool Document::CheckDocEntryVR(VRKey vr)
1708 if ( !Global::GetVR()->IsValidVR(vr) )
1715 * \brief Get the transformed value of the header entry. The VR value
1716 * is used to define the transformation to operate on the value
1717 * \warning NOT end user intended method !
1718 * @param entry entry to tranform
1719 * @return Transformed entry value
1721 std::string Document::GetDocEntryValue(DocEntry *entry)
1723 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1725 std::string val = ((ValEntry *)entry)->GetValue();
1726 std::string vr = entry->GetVR();
1727 uint32_t length = entry->GetLength();
1728 std::ostringstream s;
1731 // When short integer(s) are expected, read and convert the following
1732 // n * 2 bytes properly i.e. as a multivaluated strings
1733 // (each single value is separated fromthe next one by '\'
1734 // as usual for standard multivaluated filels
1735 // Elements with Value Multiplicity > 1
1736 // contain a set of short integers (not a single one)
1738 if ( vr == "US" || vr == "SS" )
1743 for (int i=0; i < nbInt; i++)
1749 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1750 newInt16 = SwapShort( newInt16 );
1755 // When integer(s) are expected, read and convert the following
1756 // n * 4 bytes properly i.e. as a multivaluated strings
1757 // (each single value is separated fromthe next one by '\'
1758 // as usual for standard multivaluated filels
1759 // Elements with Value Multiplicity > 1
1760 // contain a set of integers (not a single one)
1761 else if ( vr == "UL" || vr == "SL" )
1766 for (int i=0; i < nbInt; i++)
1772 newInt32 = ( val[4*i+0] & 0xFF )
1773 + (( val[4*i+1] & 0xFF ) << 8 )
1774 + (( val[4*i+2] & 0xFF ) << 16 )
1775 + (( val[4*i+3] & 0xFF ) << 24 );
1776 newInt32 = SwapLong( newInt32 );
1780 #ifdef GDCM_NO_ANSI_STRING_STREAM
1781 s << std::ends; // to avoid oddities on Solaris
1782 #endif //GDCM_NO_ANSI_STRING_STREAM
1785 return ((ValEntry *)entry)->GetValue();
1789 * \brief Get the reverse transformed value of the header entry. The VR
1790 * value is used to define the reverse transformation to operate on
1792 * \warning NOT end user intended method !
1793 * @param entry Entry to reverse transform
1794 * @return Reverse transformed entry value
1796 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1798 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1800 std::string vr = entry->GetVR();
1801 std::vector<std::string> tokens;
1802 std::ostringstream s;
1804 if ( vr == "US" || vr == "SS" )
1808 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1809 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1810 for (unsigned int i=0; i<tokens.size(); i++)
1812 newInt16 = atoi(tokens[i].c_str());
1813 s << ( newInt16 & 0xFF )
1814 << (( newInt16 >> 8 ) & 0xFF );
1818 if ( vr == "UL" || vr == "SL")
1822 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1823 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1824 for (unsigned int i=0; i<tokens.size();i++)
1826 newInt32 = atoi(tokens[i].c_str());
1827 s << (char)( newInt32 & 0xFF )
1828 << (char)(( newInt32 >> 8 ) & 0xFF )
1829 << (char)(( newInt32 >> 16 ) & 0xFF )
1830 << (char)(( newInt32 >> 24 ) & 0xFF );
1835 #ifdef GDCM_NO_ANSI_STRING_STREAM
1836 s << std::ends; // to avoid oddities on Solaris
1837 #endif //GDCM_NO_ANSI_STRING_STREAM
1841 return ((ValEntry *)entry)->GetValue();
1845 * \brief Skip a given Header Entry
1846 * \warning NOT end user intended method !
1847 * @param entry entry to skip
1849 void Document::SkipDocEntry(DocEntry *entry)
1851 SkipBytes(entry->GetLength());
1855 * \brief Skips to the beginning of the next Header Entry
1856 * \warning NOT end user intended method !
1857 * @param currentDocEntry entry to skip
1859 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1861 int l = currentDocEntry->GetReadLength();
1862 if ( l == -1 ) // length = 0xffff shouldn't appear here ...
1863 // ... but PMS imagers happen !
1865 Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg);
1866 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1868 Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur);
1873 * \brief When the length of an element value is obviously wrong (because
1874 * the parser went Jabberwocky) one can hope improving things by
1875 * applying some heuristics.
1876 * @param entry entry to check
1877 * @param foundLength first assumption about length
1879 void Document::FixDocEntryFoundLength(DocEntry *entry,
1880 uint32_t foundLength)
1882 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1883 if ( foundLength == 0xffffffff)
1888 uint16_t gr = entry->GetGroup();
1889 uint16_t elem = entry->GetElement();
1891 if ( foundLength % 2)
1893 gdcmWarningMacro( "Warning : Tag with uneven length " << foundLength
1894 << " in x(" << std::hex << gr << "," << elem <<")");
1897 //////// Fix for some naughty General Electric images.
1898 // Allthough not recent many such GE corrupted images are still present
1899 // on Creatis hard disks. Hence this fix shall remain when such images
1900 // are no longer in use (we are talking a few years, here)...
1901 // Note: XMedCon probably uses such a trick since it is able to read
1902 // those pesky GE images ...
1903 if ( foundLength == 13)
1905 // Only happens for this length !
1906 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1909 entry->SetReadLength(10); // a bug is to be fixed !?
1913 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1914 // Occurence of such images is quite low (unless one leaves close to a
1915 // 'Leonardo' source. Hence, one might consider commenting out the
1916 // following fix on efficiency reasons.
1917 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1920 entry->SetReadLength(4); // a bug is to be fixed !?
1923 else if ( entry->GetVR() == "SQ" )
1925 foundLength = 0; // ReadLength is unchanged
1928 //////// We encountered a 'delimiter' element i.e. a tag of the form
1929 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1930 // taken into account.
1931 else if ( gr == 0xfffe )
1933 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1934 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1935 // causes extra troubles...
1936 if ( entry->GetElement() != 0x0000 )
1941 entry->SetLength(foundLength);
1945 * \brief Apply some heuristics to predict whether the considered
1946 * element value contains/represents an integer or not.
1947 * @param entry The element value on which to apply the predicate.
1948 * @return The result of the heuristical predicate.
1950 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1952 uint16_t elem = entry->GetElement();
1953 uint16_t group = entry->GetGroup();
1954 const std::string &vr = entry->GetVR();
1955 uint32_t length = entry->GetLength();
1957 // When we have some semantics on the element we just read, and if we
1958 // a priori know we are dealing with an integer, then we shall be
1959 // able to swap it's element value properly.
1960 if ( elem == 0 ) // This is the group length of the group
1968 // Allthough this should never happen, still some images have a
1969 // corrupted group length [e.g. have a glance at offset x(8336) of
1970 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
1971 // Since for dicom compliant and well behaved headers, the present
1972 // test is useless (and might even look a bit paranoid), when we
1973 // encounter such an ill-formed image, we simply display a warning
1974 // message and proceed on parsing (while crossing fingers).
1975 long filePosition = Fp->tellg();
1976 gdcmWarningMacro( "Erroneous Group Length element length on : ("
1977 << std::hex << group << " , " << elem
1978 << ") -before- position x(" << filePosition << ")"
1979 << "lgt : " << length );
1983 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1991 * \brief Discover what the swap code is (among little endian, big endian,
1992 * bad little endian, bad big endian).
1994 * @return false when we are absolutely sure
1995 * it's neither ACR-NEMA nor DICOM
1996 * true when we hope ours assuptions are OK
1998 bool Document::CheckSwap()
2005 // First, compare HostByteOrder and NetworkByteOrder in order to
2006 // determine if we shall need to swap bytes (i.e. the Endian type).
2007 bool net2host = Util::IsCurrentProcessorBigEndian();
2009 // The easiest case is the one of a 'true' DICOM header, we just have
2010 // to look for the string "DICM" inside the file preamble.
2013 char *entCur = deb + 128;
2014 if ( memcmp(entCur, "DICM", (size_t)4) == 0 )
2016 gdcmWarningMacro( "Looks like DICOM Version3 (preamble + DCM)" );
2018 // Group 0002 should always be VR, and the first element 0000
2019 // Let's be carefull (so many wrong headers ...)
2020 // and determine the value representation (VR) :
2021 // Let's skip to the first element (0002,0000) and check there if we find
2022 // "UL" - or "OB" if the 1st one is (0002,0001) -,
2023 // in which case we (almost) know it is explicit VR.
2024 // WARNING: if it happens to be implicit VR then what we will read
2025 // is the length of the group. If this ascii representation of this
2026 // length happens to be "UL" then we shall believe it is explicit VR.
2027 // We need to skip :
2028 // * the 128 bytes of File Preamble (often padded with zeroes),
2029 // * the 4 bytes of "DICM" string,
2030 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2031 // i.e. a total of 136 bytes.
2034 // group 0x0002 *is always* Explicit VR Sometimes ,
2035 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
2036 // *Implicit* VR (see former 'gdcmData/icone.dcm')
2038 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
2039 memcmp(entCur, "OB", (size_t)2) == 0 ||
2040 memcmp(entCur, "UI", (size_t)2) == 0 ||
2041 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2042 // when Write DCM *adds*
2044 // Use Document::dicom_vr to test all the possibilities
2045 // instead of just checking for UL, OB and UI !? group 0000
2047 Filetype = ExplicitVR;
2048 gdcmWarningMacro( "Group 0002 : Explicit Value Representation");
2052 Filetype = ImplicitVR;
2053 gdcmWarningMacro( "Group 0002 :Not an explicit Value Representation;"
2054 << "Looks like a bugged Header!");
2060 gdcmWarningMacro( "HostByteOrder != NetworkByteOrder");
2065 gdcmWarningMacro( "HostByteOrder = NetworkByteOrder");
2068 // Position the file position indicator at first tag
2069 // (i.e. after the file preamble and the "DICM" string).
2071 Fp->seekg(0, std::ios::beg); // FIXME : Is it usefull?
2073 Fp->seekg ( 132L, std::ios::beg);
2075 } // ------------------------------- End of DicomV3 ----------------
2077 // Alas, this is not a DicomV3 file and whatever happens there is no file
2078 // preamble. We can reset the file position indicator to where the data
2079 // is (i.e. the beginning of the file).
2081 gdcmWarningMacro( "Not a Kosher DICOM Version3 file (no preamble)");
2083 Fp->seekg(0, std::ios::beg);
2085 // Let's check 'No Preamble Dicom File' :
2086 // Should start with group 0x0002
2087 // and be Explicit Value Representation
2089 s16 = *((uint16_t *)(deb));
2102 if ( SwapCode != 0 )
2104 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
2105 memcmp(entCur, "OB", (size_t)2) == 0 ||
2106 memcmp(entCur, "UI", (size_t)2) == 0 ||
2107 memcmp(entCur, "SH", (size_t)2) == 0 ||
2108 memcmp(entCur, "AE", (size_t)2) == 0 ||
2109 memcmp(entCur, "OB", (size_t)2) == 0 )
2111 Filetype = ExplicitVR;
2112 gdcmWarningMacro( "Group 0002 : Explicit Value Representation");
2116 // ------------------------------- End of 'No Preamble' DicomV3 -------------
2118 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2119 // By clean we mean that the length of the first group is written down.
2120 // If this is the case and since the length of the first group HAS to be
2121 // four (bytes), then determining the proper swap code is straightforward.
2124 // We assume the array of char we are considering contains the binary
2125 // representation of a 32 bits integer. Hence the following dirty
2127 s32 = *((uint32_t *)(entCur));
2147 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2148 // It is time for despaired wild guesses.
2149 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2150 // i.e. the 'group length' element is not present :
2152 // check the supposed-to-be 'group number'
2153 // in ( 0x0001 .. 0x0008 )
2154 // to determine ' SwapCode' value .
2155 // Only 0 or 4321 will be possible
2156 // (no oportunity to check for the formerly well known
2157 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2158 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
2159 // the file IS NOT ACR-NEMA nor DICOM V3
2160 // Find a trick to tell it the caller...
2162 s16 = *((uint16_t *)(deb));
2189 gdcmWarningMacro("ACR/NEMA unfound swap info (Really hopeless !)");
2197 * \brief Change the Byte Swap code.
2199 void Document::SwitchByteSwapCode()
2201 gdcmWarningMacro( "Switching Byte Swap code from "<< SwapCode
2202 << " at :" <<std::hex << Fp->tellg() );
2203 if ( SwapCode == 1234 )
2207 else if ( SwapCode == 4321 )
2211 else if ( SwapCode == 3412 )
2215 else if ( SwapCode == 2143 )
2222 * \brief during parsing, Header Elements too long are not loaded in memory
2223 * @param newSize new size
2225 void Document::SetMaxSizeLoadEntry(long newSize)
2231 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2233 MaxSizeLoadEntry = 0xffffffff;
2236 MaxSizeLoadEntry = newSize;
2240 * \brief Read the next tag WITHOUT loading it's value
2241 * (read the 'Group Number', the 'Element Number',
2242 * gets the Dict Entry
2243 * gets the VR, gets the length, gets the offset value)
2244 * @return On succes : the newly created DocEntry, NULL on failure.
2246 DocEntry *Document::ReadNextDocEntry()
2253 group = ReadInt16();
2256 catch ( FormatError )
2258 // We reached the EOF (or an error occured) therefore
2259 // header parsing has to be considered as finished.
2263 // Sometimes file contains groups of tags with reversed endianess.
2264 HandleBrokenEndian(group, elem);
2266 // In 'true DICOM' files Group 0002 is always little endian
2267 if ( HasDCMPreamble )
2268 HandleOutOfGroup0002(group, elem);
2270 std::string vr = FindDocEntryVR();
2271 std::string realVR = vr;
2273 if ( vr == GDCM_UNKNOWN )
2275 if ( elem == 0x0000 ) // Group Length
2277 realVR = "UL"; // must be UL
2279 else if (group%2 == 1 && (elem >= 0x0010 && elem <=0x00ff ))
2281 // DICOM PS 3-5 7.8.1 a) states that those
2282 // (gggg-0010->00FF where gggg is odd) attributes have to be LO
2287 DictEntry *dictEntry = GetDictEntry(group,elem);
2290 realVR = dictEntry->GetVR();
2296 if ( Global::GetVR()->IsVROfSequence(realVR) )
2297 newEntry = NewSeqEntry(group, elem);
2298 else if ( Global::GetVR()->IsVROfStringRepresentable(realVR) )
2299 newEntry = NewValEntry(group, elem, realVR);
2301 newEntry = NewBinEntry(group, elem, realVR);
2303 if ( vr == GDCM_UNKNOWN )
2305 if ( Filetype == ExplicitVR )
2307 // We thought this was explicit VR, but we end up with an
2308 // implicit VR tag. Let's backtrack.
2309 if ( newEntry->GetGroup() != 0xfffe )
2312 int offset = Fp->tellg();
2313 msg = Util::Format("Entry (%04x,%04x) at 0x(%x) should be Explicit VR\n",
2314 newEntry->GetGroup(), newEntry->GetElement(), offset );
2315 gdcmWarningMacro( msg.c_str() );
2318 newEntry->SetImplicitVR();
2323 FindDocEntryLength(newEntry);
2325 catch ( FormatError )
2332 newEntry->SetOffset(Fp->tellg());
2338 * \brief Handle broken private tag from Philips NTSCAN
2339 * where the endianess is being switched to BigEndian
2340 * for no apparent reason
2343 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
2345 // Endian reversion. Some files contain groups of tags with reversed endianess.
2346 static int reversedEndian = 0;
2347 // try to fix endian switching in the middle of headers
2348 if ((group == 0xfeff) && (elem == 0x00e0))
2350 // start endian swap mark for group found
2352 SwitchByteSwapCode();
2357 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2359 // end of reversed endian group
2361 SwitchByteSwapCode();
2366 * \brief Group 0002 is always coded Little Endian
2367 * whatever Transfer Syntax is
2370 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2372 // Endian reversion. Some files contain groups of tags with reversed endianess.
2373 if ( !Group0002Parsed && group != 0x0002)
2375 Group0002Parsed = true;
2376 // we just came out of group 0002
2377 // if Transfer syntax is Big Endian we have to change CheckSwap
2379 std::string ts = GetTransferSyntax();
2380 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2382 gdcmWarningMacro("True DICOM File, with NO Tansfer Syntax: " << ts );
2386 // Group 0002 is always 'Explicit ...' enven when Transfer Syntax says 'Implicit ..."
2388 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ImplicitVRLittleEndian )
2390 Filetype = ImplicitVR;
2393 // FIXME Strangely, this works with
2394 //'Implicit VR Transfer Syntax (GE Private)
2395 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ExplicitVRBigEndian )
2397 gdcmWarningMacro("Transfer Syntax Name = ["
2398 << GetTransferSyntaxName() << "]" );
2399 SwitchByteSwapCode();
2400 group = SwapShort(group);
2401 elem = SwapShort(elem);
2406 //-----------------------------------------------------------------------------
2409 //-----------------------------------------------------------------------------
2410 } // end namespace gdcm