1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2008/09/15 15:49:22 $
7 Version: $Revision: 1.381 $
9 Copyright (c) CREATIS (Centre de Recherche et d'Applications en Traitement de
10 l'Image). All rights reserved. See Doc/License.txt or
11 http://www.creatis.insa-lyon.fr/Public/Gdcm/License.html for details.
13 This software is distributed WITHOUT ANY WARRANTY; without even
14 the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
15 PURPOSE. See the above copyright notices for more information.
17 =========================================================================*/
19 #include "gdcmDocument.h"
20 #include "gdcmSeqEntry.h"
21 #include "gdcmGlobal.h"
23 #include "gdcmDebug.h"
25 #include "gdcmDictSet.h"
26 #include "gdcmDocEntrySet.h"
27 #include "gdcmSQItem.h"
28 #include "gdcmDataEntry.h"
33 #include <ctype.h> // for isdigit
34 #include <stdlib.h> // for atoi
36 #if defined(__BORLANDC__)
37 #include <mem.h> // for memset
40 namespace GDCM_NAME_SPACE
42 //-----------------------------------------------------------------------------
44 // Refer to Document::SetMaxSizeLoadEntry()
45 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
47 //-----------------------------------------------------------------------------
48 // Constructor / Destructor
49 // Constructors and destructors are protected to avoid user to invoke directly
52 * \brief This default constructor neither loads nor parses the file.
53 * You should then invoke Document::Load.
61 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
64 Filetype = ExplicitVR;
65 CurrentOffsetPosition = 0;
66 OffsetOfPreviousParseDES =0;
67 // Load will set it to true if sucessfull
68 Group0002Parsed = false;
69 IsDocumentAlreadyLoaded = false;
70 IsDocumentModified = true;
71 LoadMode = LD_ALL; // default : load everything, later
79 * \brief Canonical destructor.
81 Document::~Document ()
86 //-----------------------------------------------------------------------------
89 * \brief Loader. use SetLoadMode(), SetFileName() before !
90 * @return false if file cannot be open or no swap info was found,
91 * or no tag was found.
93 bool Document::Load( )
95 if ( GetFileName() == "" )
97 gdcmWarningMacro( "Use SetFileName, before !" );
100 return DoTheLoadingDocumentJob( );
104 //#ifndef GDCM_LEGACY_REMOVE
106 * \brief Loader. (DEPRECATED : not to break the API)
107 * @param fileName 'Document' (File or DicomDir) to be open for parsing
108 * @return false if file cannot be open or no swap info was found,
109 * or no tag was found.
112 bool Document::Load( std::string const &fileName )
115 return DoTheLoadingDocumentJob( );
121 * \brief Performs the Loading Job (internal use only)
122 * @return false if file cannot be open or no swap info was found,
123 * or no tag was found.
125 bool Document::DoTheLoadingDocumentJob( )
127 if ( ! IsDocumentModified ) // Nothing to do !
135 // warning already performed in OpenFile()
140 Group0002Parsed = false;
142 gdcmDebugMacro( "Starting parsing of file: " << Filename.c_str());
144 // Computes the total length of the file
145 Fp->seekg(0, std::ios::end); // Once per Document !
146 long lgt = Fp->tellg(); // Once per Document !
147 Fp->seekg(0, std::ios::beg); // Once per Document !
149 // CheckSwap returns a boolean
150 // (false if no swap info of any kind was found)
153 gdcmWarningMacro( "Neither a DICOM V3 nor an ACR-NEMA file: "
154 << Filename.c_str());
159 long beg = Fp->tellg(); // just after DICOM preamble (if any)
161 lgt -= beg; // remaining length to parse
164 // Loading is done during parsing
165 OffsetOfPreviousParseDES = beg;
166 ParseDES( this, beg, lgt, false); // delim_mode is first defaulted to false
170 gdcmErrorMacro( "No tag in internal hash table for: "
171 << Filename.c_str());
175 IsDocumentAlreadyLoaded = true;
177 //Fp->seekg(0, std::ios::beg); // Once per Document!
179 // Load 'non string' values
181 std::string PhotometricInterpretation = GetEntryString(0x0028,0x0004);
182 if ( PhotometricInterpretation == "PALETTE COLOR " )
185 // Probabely this line should be outside the 'if'
186 // Try to find an image sample holding a 'gray LUT'
187 LoadEntryBinArea(0x0028,0x1200); // gray LUT
190 /// --> FIXME : The difference between BinEntry and DataEntry
191 /// --> no longer exists, but the alteration of Dicom Dictionary remains.
192 /// --> Old comment restored on purpose.
193 /// --> New one (replacing both BinEntry and ValEntry by DataEntry)
194 /// --> had absolutely no meaning.
195 /// --> The whole comment will be removed when the stuff is cleaned !
197 /// The tags refered by the three following lines used to be CORRECTLY
198 /// defined as having an US Value Representation in the public
199 /// dictionary. BUT the semantics implied by the three following
200 /// lines state that the corresponding tag contents are in fact
201 /// the ones of a BinEntry.
202 /// In order to fix things "Quick and Dirty" the dictionary was
203 /// altered on PURPOSE but now contains a WRONG value.
204 /// In order to fix things and restore the dictionary to its
205 /// correct value, one needs to decide of the semantics by deciding
206 /// whether the following tags are either :
207 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
208 /// also used as BinEntry, which requires the proper conversion,
209 /// - OW, and hence loaded as BinEntry, but afterwards also used
210 /// as ValEntry, which requires the proper conversion.
212 // --> OB (byte aray) or OW (short int aray)
213 // The actual VR has to be deduced from other entries.
214 // Our way of loading them may fail in some cases :
215 // We must or not SwapByte depending on other field values.
217 LoadEntryBinArea(0x0028,0x1201); // R LUT
218 LoadEntryBinArea(0x0028,0x1202); // G LUT
219 LoadEntryBinArea(0x0028,0x1203); // B LUT
221 // Segmented Red Palette Color LUT Data
222 LoadEntryBinArea(0x0028,0x1221);
223 // Segmented Green Palette Color LUT Data
224 LoadEntryBinArea(0x0028,0x1222);
225 // Segmented Blue Palette Color LUT Data
226 LoadEntryBinArea(0x0028,0x1223);
229 //FIXME later : how to use it?
230 SeqEntry *modLutSeq = GetSeqEntry(0x0028,0x3000); // Modality LUT Sequence
233 SQItem *sqi= modLutSeq->GetFirstSQItem();
236 DataEntry *dataEntry = sqi->GetDataEntry(0x0028,0x3006); // LUT Data
237 if ( dataEntry != 0 )
239 if ( dataEntry->GetLength() != 0 )
241 // FIXME : CTX dependent means : contexted dependant.
242 // see upper comment.
243 LoadEntryBinArea(dataEntry); //LUT Data (CTX dependent)
249 // Force Loading some more elements if user asked to.
251 GDCM_NAME_SPACE::DocEntry *d;
252 for (ListElements::iterator it = UserForceLoadList.begin();
253 it != UserForceLoadList.end();
256 gdcmDebugMacro( "Force Load " << std::hex
257 << (*it).Group << "|" <<(*it).Elem );
259 d = GetDocEntry( (*it).Group, (*it).Elem);
263 gdcmWarningMacro( "You asked to ForceLoad " << std::hex
264 << (*it).Group <<"|"<< (*it).Elem
265 << " that doesn't exist" );
269 LoadDocEntry(d, true);
274 // ----------------------------
275 // Specific code to allow gdcm to read ACR-LibIDO formated images
276 // Note: ACR-LibIDO is an extension of the ACR standard that was
277 // used at CREATIS. For the time being (say a couple of years)
278 // we keep this kludge to allow CREATIS users
279 // reading their old images.
281 // if recognition code tells us we deal with a LibIDO image
282 // we switch lineNumber and columnNumber
285 RecCode = GetEntryString(0x0008, 0x0010); // recognition code (RET)
288 if(RecCode.find("ACRNEMA_LIBIDO") == 0 || // any version
289 RecCode.find("CANRME_AILIBOD") == 0) // for brain-damaged softwares
290 // with "little-endian strings"
293 Filetype = ACR_LIBIDO;
294 std::string rows = GetEntryString(0x0028, 0x0010);
295 std::string columns = GetEntryString(0x0028, 0x0011);
296 SetEntryString(columns, 0x0028, 0x0010);
297 SetEntryString(rows , 0x0028, 0x0011);
299 // --- End of ACR-LibIDO kludge ---
305 * \brief Adds a new element we want to load anyway
306 * @param group Group number of the target tag.
307 * @param elem Element number of the target tag.
309 void Document::AddForceLoadElement (uint16_t group, uint16_t elem)
314 UserForceLoadList.push_back(el);
317 * \brief Get the public dictionary used
319 Dict *Document::GetPubDict()
325 * \brief Get the shadow dictionary used
327 Dict *Document::GetShaDict()
333 * \brief Set the shadow dictionary used
334 * @param dict dictionary to use in shadow
336 bool Document::SetShaDict(Dict *dict)
343 * \brief Set the shadow dictionary used
344 * @param dictName name of the dictionary to use in shadow
346 bool Document::SetShaDict(DictKey const &dictName)
348 RefShaDict = Global::GetDicts()->GetDict(dictName);
353 * \brief This predicate tells us whether or not the current Document
354 * was properly parsed and contains at least *one* Dicom Element
355 * (and nothing more, sorry).
356 * @return false when we're 150 % sure it's NOT a Dicom/Acr file,
359 bool Document::IsParsable()
361 if ( Filetype == Unknown )
363 gdcmWarningMacro( "Wrong filetype for " << GetFileName());
369 gdcmWarningMacro( "No tag in internal hash table.");
376 * \brief This predicate tells us whether or not the current Document
377 * was properly parsed and contains at least *one* Dicom Element
378 * (and nothing more, sorry).
379 * @return false when we're 150 % sure it's NOT a Dicom/Acr file,
382 bool Document::IsReadable()
388 * \brief Predicate for dicom version 3 file.
389 * @return True when the file is a dicom version 3.
391 bool Document::IsDicomV3()
393 // Checking if Transfer Syntax exists is enough
394 // Anyway, it's too late check if the 'Preamble' was found ...
395 // And ... would it be a rich idea to check ?
396 // (some 'no Preamble' DICOM images exist !)
397 return GetDocEntry(0x0002, 0x0010) != NULL;
401 * \brief Predicate for Papyrus file
402 * Dedicated to whomsoever it may concern
403 * @return True when the file is a Papyrus file.
405 bool Document::IsPapyrus()
407 // check for Papyrus private Sequence
408 DocEntry *e = GetDocEntry(0x0041, 0x1050);
411 // check if it's actually a Sequence
412 if ( !dynamic_cast<SeqEntry*>(e) )
418 * \brief returns the File Type
419 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
420 * @return the FileType code
422 FileType Document::GetFileType()
428 * \brief Accessor to the Transfer Syntax (when present) of the
429 * current document (it internally handles reading the
430 * value from disk when only parsing occured).
431 * @return The encountered Transfer Syntax of the current document, if DICOM.
432 * GDCM_UNKNOWN for ACR-NEMA files (or broken headers ...)
434 std::string Document::GetTransferSyntax()
436 DocEntry *entry = GetDocEntry(0x0002, 0x0010);
442 // The entry might be present but not loaded (parsing and loading
443 // happen at different stages): try loading and proceed with check...
446 // (parsing and loading happen at the very same stage!)
447 //LoadDocEntrySafe(entry); //JPRx
448 if (DataEntry *dataEntry = dynamic_cast<DataEntry *>(entry) )
450 std::string transfer = dataEntry->GetString();
451 // The actual transfer (as read from disk) might be padded. We
452 // first need to remove the potential padding. We can make the
453 // weak assumption that padding was not executed with digits...
454 if ( transfer.length() == 0 )
456 // for brain damaged headers
457 gdcmWarningMacro( "Transfer Syntax has length = 0.");
460 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
462 transfer.erase(transfer.length()-1, 1);
463 if ( transfer.length() == 0 )
465 // for brain damaged headers
466 gdcmWarningMacro( "Transfer Syntax contains no valid character.");
476 * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS
477 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
479 std::string Document::GetTransferSyntaxName()
481 // use the TS (TS : Transfer Syntax)
482 std::string transferSyntax = GetEntryString(0x0002,0x0010);
484 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
486 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
487 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
488 return "Uncompressed ACR-NEMA";
490 if ( transferSyntax == GDCM_UNFOUND )
492 gdcmDebugMacro( "Unfound Transfer Syntax (0002,0010)");
493 return "Uncompressed ACR-NEMA";
496 // we do it only when we need it
497 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
499 // Global::GetTS() is a global static you shall never try to delete it!
503 // --------------- Swap Code ------------------
505 * \brief Swaps the bytes so they agree with the processor order
506 * @return The properly swaped 16 bits integer.
508 uint16_t Document::SwapShort(uint16_t a)
510 if ( SwapCode == 4321 || SwapCode == 2143 )
512 //a = ((( a << 8 ) & 0xff00 ) | (( a >> 8 ) & 0x00ff ) );
514 a = ( a << 8 ) | ( a >> 8 );
520 * \brief Swaps back the bytes of 4-byte long integer accordingly to
522 * @return The properly swaped 32 bits integer.
524 uint32_t Document::SwapLong(uint32_t a)
531 // a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
532 // ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
534 a=( ( a<<24) | ((a<<8) & 0x00ff0000) |
535 ((a>>8) & 0x0000ff00) | (a>>24) );
538 // a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
539 a=( (a<<16) | (a>>16) );
542 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
545 gdcmErrorMacro( "Unexpected swap code:" << SwapCode );
552 * \brief Swaps back the bytes of 8-byte long 'double' accordingly to
554 * @return The properly swaped 64 bits double.
556 double Document::SwapDouble(double a)
560 // There were no 'double' at ACR-NEMA time.
561 // We just have to deal with 'straight Little Endian' and
562 // 'straight Big Endian'
567 char *beg = (char *)&a;
570 for (unsigned int i = 0; i<7; i++)
581 gdcmErrorMacro( "Unexpected swap code:" << SwapCode );
588 // -----------------File I/O ---------------
590 * \brief Tries to open the file Document::Filename and
591 * checks the preamble when existing,
592 * or if the file starts with an ACR-NEMA look-like element.
593 * @return The FILE pointer on success, 0 on failure.
595 std::ifstream *Document::OpenFile()
597 HasDCMPreamble = false;
598 if (Filename.length() == 0)
605 gdcmDebugMacro( "File already open: " << Filename.c_str());
609 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
612 // Don't user gdcmErrorMacro :
613 // a spurious message will appear when you use, for instance
614 // gdcm::FileHelper *fh = new gdcm::FileHelper( outputFileName );
615 // to create outputFileName.
617 // FIXME : if the upper comment is still usefull
618 // --> the constructor is not so good ...
620 gdcmWarningMacro( "Cannot open file: " << Filename.c_str());
624 //exit(1); // No function is allowed to leave the application instead
625 // of warning the caller
629 Fp->read((char*)&zero, (size_t)2);
637 Fp->seekg(126L, std::ios::cur); // Once per Document
638 char dicm[4]; // = {' ',' ',' ',' '};
639 Fp->read(dicm, (size_t)4);
646 if ( memcmp(dicm, "DICM", 4) == 0 )
648 HasDCMPreamble = true;
652 //-- Broken ACR or DICOM (?) with no Preamble; may start with a Shadow Group --
653 // FIXME : We cannot be sure the preable is only zeroes..
654 // (see ACUSON-24-YBR_FULL-RLE.dcm )
656 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
657 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
658 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
659 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 ||
660 zero == 0x0028 || 0x2800 // worse : some ACR-NEMA like files
661 // start 00028 group ?!?
664 std::string msg = Util::Format(
665 "ACR/DICOM starting by 0x(%04x) at the beginning of the file\n", zero);
666 // FIXME : is it a Warning message, or a Debug message?
667 gdcmWarningMacro( msg.c_str() );
671 // -- Neither ACR/No Preamble Dicom nor DICOMV3 file
673 // Don't user Warning nor Error, not to pollute the output
674 // while directory recursive parsing ...
675 gdcmDebugMacro( "Neither ACR/No Preamble Dicom nor DICOMV3 file: "
676 << Filename.c_str());
681 * \brief closes the file
682 * @return TRUE if the close was successfull
684 bool Document::CloseFile()
696 * \brief Writes in a file all the Entries (Dicom Elements)
697 * @param fp file pointer on an already open file (actually: Output File Stream)
698 * @param filetype Type of the File to be written
699 * (ACR-NEMA, ExplicitVR, ImplicitVR)
701 void Document::WriteContent(std::ofstream *fp, FileType filetype, bool, bool)
703 // Skip if user wants to write an ACR-NEMA file
705 if ( filetype == ImplicitVR || filetype == ExplicitVR ||
706 filetype == JPEG || filetype == JPEG2000 )
708 // writing Dicom File Preamble
709 char filePreamble[128];
710 memset(filePreamble, 0, 128);
711 fp->write(filePreamble, 128);
712 fp->write("DICM", 4);
715 * \todo rewrite later, if really usefull
716 * - 'Group Length' element is optional in DICOM
717 * - but un-updated odd groups lengthes can causes pb
720 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
721 * UpdateGroupLength(false,filetype);
722 * if ( filetype == ACR)
723 * UpdateGroupLength(true,ACR);
725 * --> Computing group length for groups with embeded Sequences
726 * --> was too much tricky / we were [in a hurry / too lazy]
727 * --> We don't write the element 0x0000 (group length)
729 // This one is recursive
730 // false : outside MetaElements
731 // false : outside Sequence
732 ElementSet::WriteContent(fp, filetype, false, false);
735 // -----------------------------------------
738 * \brief Loads (from disk) the element content
739 * when a string is not suitable
740 * @param group group number of the Entry
741 * @param elem element number of the Entry
743 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
745 // Search the corresponding DocEntry
746 DocEntry *docEntry = GetDocEntry(group, elem);
749 gdcmDebugMacro(std::hex << group << "|" << elem
750 << " doesn't exist" );
753 DataEntry *dataEntry = dynamic_cast<DataEntry *>(docEntry);
756 gdcmWarningMacro(std::hex << group << "|" << elem
757 << " is NOT a DataEntry");
760 LoadEntryBinArea(dataEntry);
764 * \brief Loads (from disk) the element content
765 * when a string is not suitable
766 * @param entry Entry whose binArea is going to be loaded
768 void Document::LoadEntryBinArea(DataEntry *entry)
770 if( entry->GetBinArea() )
777 //size_t o =(size_t)entry->GetOffset();
778 Fp->seekg((size_t)entry->GetOffset(), std::ios::beg); // FIXME : for each DataEntry !
780 size_t l = entry->GetLength();
781 uint8_t *data = new uint8_t[l];
784 gdcmWarningMacro( "Cannot allocate DataEntry content for : "
785 << std::hex << entry->GetGroup()
786 << "|" << entry->GetElement() );
791 Fp->read((char*)data, l);
792 if ( Fp->fail() || Fp->eof() )
795 entry->SetState(DataEntry::STATE_UNREAD);
799 // Swap the data content if necessary
801 unsigned short vrLgth =
802 Global::GetVR()->GetAtomicElementLength(entry->GetVR());
804 // FIXME : trouble expected if we read an ... OW Entry (LUT, etc ..)
805 // if( entry->GetVR() == "OW" )
816 uint16_t *data16 = (uint16_t *)data;
817 for(i=0;i<l/vrLgth;i++)
818 data16[i] = SwapShort(data16[i]);
823 uint32_t *data32 = (uint32_t *)data;
824 for(i=0;i<l/vrLgth;i++)
825 data32[i] = SwapLong(data32[i]);
830 double *data64 = (double *)data;
831 for(i=0;i<l/vrLgth;i++)
832 data64[i] = SwapDouble(data64[i]);
837 entry->SetBinArea(data);
839 if ( openFile ) // The file is left in the state (open/close) it was at entrance
844 * \brief Loads the element while preserving the current
845 * underlying file position indicator as opposed to
846 * LoadDocEntry that modifies it
847 * \note seems to be unused!.
848 * @param entry DocEntry whose value will be loaded.
850 //void Document::LoadDocEntrySafe(DocEntry *entry)
854 // long PositionOnEntry = Fp->tellg(); // LoadDocEntrySafe is not used
855 // LoadDocEntry(entry);
856 // Fp->seekg(PositionOnEntry, std::ios::beg); // LoadDocEntrySafe is not used
861 * \brief Compares two documents, according to DicomDir rules
862 * \warning Does NOT work with ACR-NEMA files
863 * \todo Find a trick to solve the pb (use RET fields ?)
864 * @param document to compare with current one
865 * @return true if 'smaller'
867 bool Document::operator<(Document &document)
870 std::string s1 = GetEntryString(0x0010,0x0010);
871 std::string s2 = document.GetEntryString(0x0010,0x0010);
883 s1 = GetEntryString(0x0010,0x0020);
884 s2 = document.GetEntryString(0x0010,0x0020);
895 // Study Instance UID
896 s1 = GetEntryString(0x0020,0x000d);
897 s2 = document.GetEntryString(0x0020,0x000d);
908 // Serie Instance UID
909 s1 = GetEntryString(0x0020,0x000e);
910 s2 = document.GetEntryString(0x0020,0x000e);
925 //-----------------------------------------------------------------------------
929 * \brief Reads a given length of bytes
930 * (in order to avoid to many CPU time-consuming fread-s)
931 * @param l length to read
933 void Document::ReadBegBuffer(size_t l)
936 Fp->read (BegBuffer, (size_t)l);
939 throw FormatError( "Document::ReadBegBuffer()", " file error." );
943 throw FormatError( "Document::ReadBegBuffer()", "EOF." );
945 PtrBegBuffer = BegBuffer;
946 CurrentOffsetPosition+=l;
949 * \brief Reads a supposed to be 16 Bits integer
950 * (swaps it depending on processor endianness)
953 uint16_t Document::ReadInt16()
957 Fp->read ((char*)&g, (size_t)2);
960 throw FormatError( "Document::ReadInt16()", " file error." );
964 throw FormatError( "Document::ReadInt16()", "EOF." );
971 * \brief Gets from BegBuffer a supposed to be 16 Bits integer
972 * (swaps it depending on processor endianness)
975 uint16_t Document::GetInt16()
977 uint16_t g = *((uint16_t*)PtrBegBuffer);
983 * \brief Reads a supposed to be 32 Bits integer
984 * (swaps it depending on processor endianness)
987 uint32_t Document::ReadInt32()
991 Fp->read ((char*)&g, (size_t)4);
994 throw FormatError( "Document::ReadInt32()", " file error." );
998 throw FormatError( "Document::ReadInt32()", "EOF." );
1005 * \brief Gets from BegBuffer a supposed to be 32 Bits integer
1006 * (swaps it depending on processor endianness)
1007 * @return read value
1009 uint32_t Document::GetInt32()
1011 uint32_t g = *((uint32_t*)PtrBegBuffer);
1018 * \brief Re-computes the length of the Dicom group 0002.
1020 int Document::ComputeGroup0002Length( )
1025 int groupLength = 0;
1026 bool found0002 = false;
1028 // for each zero-level Tag in the DCM Header
1029 DocEntry *entry = GetFirstEntry();
1032 gr = entry->GetGroup();
1038 if ( entry->GetElement() != 0x0000 )
1040 vr = entry->GetVR();
1042 //if ( (vr == "OB")||(vr == "OW")||(vr == "UT")||(vr == "SQ"))
1043 // (no SQ, OW, OL, UT in group 0x0002;)
1046 // explicit VR AND (OB, OW, OL, SQ, UT, UN) : 4 more bytes
1049 groupLength += 2 + 2 + 4 + entry->GetLength();
1052 else if (found0002 )
1055 entry = GetNextEntry();
1061 * \brief CallStartMethod
1063 void Document::CallStartMethod()
1067 CommandManager::ExecuteCommand(this,CMD_STARTPROGRESS);
1071 * \brief CallProgressMethod
1073 void Document::CallProgressMethod()
1075 CommandManager::ExecuteCommand(this,CMD_PROGRESS);
1079 * \brief CallEndMethod
1081 void Document::CallEndMethod()
1084 CommandManager::ExecuteCommand(this,CMD_ENDPROGRESS);
1087 //-----------------------------------------------------------------------------
1090 * \brief Loads all the needed Dictionaries
1091 * \warning NOT end user intended method !
1093 void Document::Initialize()
1095 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
1101 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1102 * @param set DocEntrySet we are going to parse ('zero level' or a SQItem)
1103 * @param offset start of parsing
1104 * @param l_max length to parse (meaningless when we are in 'delimitor mode')
1105 * @param delim_mode : whether we are in 'delimitor mode' (l=0xffffff) or not
1107 void Document::ParseDES(DocEntrySet *set, long offset,
1108 long l_max, bool delim_mode)
1110 DocEntry *newDocEntry;
1111 DataEntry *newDataEntry;
1112 SeqEntry *newSeqEntry;
1114 bool used; // will be set to false when something wrong happens to an Entry.
1115 // (Entry will then be deleted)
1116 bool delim_mode_intern = delim_mode;
1118 gdcmDebugMacro( "Enter in ParseDES, delim-mode " << delim_mode
1119 << " at offset " << std::hex << "0x(" << offset << ")" );
1123 ///\todo FIXME : On 64 bits processors, tellg gives unexpected results after a while ?
1124 /// Probabely a bug in gdcm code somewhere (some memory erased ?)
1126 // Uncomment to track the bug
1128 if( Debug::GetDebugFlag() )
1129 std::cout << std::dec <<"(long)(Fp->tellg()) " << (long)(Fp->tellg()) // in Debug mode
1130 << std::hex << " 0x(" <<(long)(Fp->tellg()) << ")" << std::endl;
1133 // if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) // Once per DocEntry
1134 if ( !delim_mode ) // 'and then' doesn't exist in C++ :-(
1135 if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per DocEntry, when no delim mode
1140 newDocEntry = ReadNextDocEntry( );
1147 // Uncoment this cerr line to be able to 'follow' the DocEntries
1148 // when something *very* strange happens
1149 if( Debug::GetDebugFlag() )
1150 std::cerr<<newDocEntry->GetKey()<<" "<<newDocEntry->GetVR()<<std::endl;
1152 // an Item Starter found elsewhere but in the first position
1153 // of a SeqEntry means previous entry was a Sequence
1154 // but we didn't get it (private Sequence + Implicit VR)
1155 // we have to backtrack.
1156 if ( !first && newDocEntry->IsItemStarter() )
1158 // Debug message within the method !
1159 newDocEntry = Backtrack(newDocEntry, set);
1163 PreviousDocEntry = newDocEntry;
1167 newDataEntry = dynamic_cast<DataEntry*>(newDocEntry);
1171 //////////////////////////// DataEntry
1173 //vr = newDocEntry->GetVR(); // useless ?
1175 if ( !set->AddEntry( newDataEntry ) )
1177 gdcmDebugMacro( "in ParseDES : cannot add a DataEntry "
1178 << newDataEntry->GetKey()
1179 << " (at offset : 0x("
1180 << newDataEntry->GetOffset() << ") )" );
1185 newDataEntry->Delete();
1186 // Load only if we can add (not a duplicate key)
1187 LoadDocEntry( newDataEntry );
1189 if ( newDataEntry->GetElement() == 0x0000 ) // if on group length
1191 if ( newDataEntry->GetGroup()%2 != 0 ) // if Shadow Group
1193 if ( LoadMode & LD_NOSHADOW ) // if user asked to skip shad.gr
1195 std::string strLgrGroup = newDataEntry->GetString();
1198 //if ( newDataEntry->IsUnfound() ) /?!? JPR
1200 lgrGroup = atoi(strLgrGroup.c_str());
1201 Fp->seekg(lgrGroup, std::ios::cur); // Once per Shadow group, when NOSHADOW
1202 RemoveEntry( newDocEntry ); // Remove and delete
1209 bool delimitor = newDataEntry->IsItemDelimitor();
1210 bool outOfBounds = false;
1212 if ( ((long)(Fp->tellg())-offset) >= l_max ) //Once per DataEntry when no delim mode
1215 // 'and then', 'or else' don't exist in C++ :-(
1216 // if ( (delimitor) ||
1217 // (!delim_mode && ((long)(Fp->tellg())-offset) >= l_max) ) // Once per DataEntry
1219 if ( delimitor || outOfBounds )
1222 newDocEntry->Delete();
1226 // Just to make sure we are at the beginning of next entry.
1227 SkipToNextDocEntry(newDocEntry); // FIXME : once per DocEntry, segfault if commented out
1231 /////////////////////// SeqEntry : VR = "SQ"
1233 unsigned long l = newDocEntry->GetReadLength();
1234 if ( l != 0 ) // don't mess the delim_mode for 'zero-length sequence'
1236 if ( l == 0xffffffff )
1238 delim_mode_intern = true;
1242 delim_mode_intern = false;
1246 if ( (LoadMode & LD_NOSHADOWSEQ) && ! delim_mode_intern )
1248 // User asked to skip SeQuences *only* if they belong to Shadow Group
1249 if ( newDocEntry->GetGroup()%2 != 0 )
1251 Fp->seekg( l, std::ios::cur); // once per SQITEM, when NOSHADOWSEQ
1252 newDocEntry->Delete(); // Delete, not in the set
1256 if ( (LoadMode & LD_NOSEQ) && ! delim_mode_intern )
1258 // User asked to skip *any* SeQuence
1259 Fp->seekg( l, std::ios::cur); // Once per SQ, when NOSEQ
1260 newDocEntry->Delete(); // Delete, not in the set
1263 // delay the dynamic cast as late as possible
1264 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1266 // no other way to create the Delimitor ...
1267 newSeqEntry->SetDelimitorMode( delim_mode_intern );
1269 // At the top of the hierarchy, stands a Document. When "set"
1270 // is a Document, then we are building the first depth level.
1271 // Hence the SeqEntry we are building simply has a depth
1273 if ( set == this ) // ( dynamic_cast< Document* > ( set ) )
1275 newSeqEntry->SetDepthLevel( 1 );
1277 // But when "set" is already a SQItem, we are building a nested
1278 // sequence, and hence the depth level of the new SeqEntry
1279 // we are building, is one level deeper:
1281 // time waste hunting
1282 else if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1284 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1288 { // Don't try to parse zero-length sequences
1290 gdcmDebugMacro( "Entry in ParseSQ, delim " << delim_mode_intern
1291 << " at offset 0x(" << std::hex
1292 << newDocEntry->GetOffset() << ")");
1294 bool res = ParseSQ( newSeqEntry,
1295 newDocEntry->GetOffset(),
1296 l, delim_mode_intern);
1298 gdcmDebugMacro( "Exit from ParseSQ, delim " << delim_mode_intern << " -->return : " << res);
1300 if ( !set->AddEntry( newSeqEntry ) )
1302 gdcmWarningMacro( "in ParseDES : cannot add a SeqEntry "
1303 << newSeqEntry->GetKey()
1304 << " (at offset : 0x("
1305 << newSeqEntry->GetOffset() << ") )" );
1310 newDocEntry->Delete();
1313 // if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) // Once per SeqEntry
1315 if ( !delim_mode ) // 'and then' doesn't exist in C++ :-(
1316 if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per SeqEntry when no delim mode
1320 newDocEntry->Delete();
1323 } // end SeqEntry : VR = "SQ"
1327 newDocEntry->Delete();
1331 if (UnexpectedEOF) // some terminator was missing
1336 gdcmDebugMacro( "Exit from ParseDES, delim-mode " << delim_mode );
1340 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1341 * @return false if expected fff0,e000 not found
1343 bool Document::ParseSQ( SeqEntry *seqEntry,
1344 long offset, long l_max, bool delim_mode)
1346 int SQItemNumber = 0;
1348 long offsetStartCurrentSQItem = offset;
1352 // the first time, we read the fff0,e000 of the first SQItem
1353 DocEntry *newDocEntry = ReadNextDocEntry();
1357 // The most frequent is when a SQ terminator is missing (?!?)
1358 gdcmWarningMacro("in ParseSQ : should never get here!");
1359 UnexpectedEOF = true;
1364 if ( newDocEntry->IsSequenceDelimitor() )
1366 seqEntry->SetDelimitationItem( newDocEntry );
1367 newDocEntry->Delete();
1371 else // ! delim_mode
1373 if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per SQItem when no delim mode
1375 newDocEntry->Delete();
1379 // create the current SQItem
1380 SQItem *itemSQ = SQItem::New( seqEntry->GetDepthLevel() );
1381 unsigned int l = newDocEntry->GetReadLength();
1383 if ( l == 0xffffffff )
1392 // avoid infinite loop when Bad assumption was made on illegal 'unknown length' UN //JPRx
1394 if (offsetStartCurrentSQItem <= OffsetOfPreviousParseDES)
1396 gdcmWarningMacro("Bad assumption was made on illegal 'unknown length' UN!" << std::endl <<
1397 "OffsetOfPreviousParseDES " << std::hex << OffsetOfPreviousParseDES
1398 << " offsetStartCurrentSQItem " << offsetStartCurrentSQItem);
1399 /// \todo when "Bad assumption (SQ) on illegal 'unknown length' UN", Backtrack again + try OB
1404 OffsetOfPreviousParseDES = offsetStartCurrentSQItem;
1407 // fill up the current SQItem, starting at the beginning of fff0,e000
1408 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg); // Once per SQItem
1409 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
1410 offsetStartCurrentSQItem = Fp->tellg(); // Once per SQItem
1412 seqEntry->AddSQItem( itemSQ, SQItemNumber );
1414 newDocEntry->Delete();
1416 //if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max ) //JPRx
1417 if ( !delim_mode && (offsetStartCurrentSQItem-offset ) >= l_max )
1426 * \brief When a private Sequence + Implicit VR is encountered
1427 * we cannot guess it's a Sequence till we find the first
1428 * Item Starter. We then backtrack to do the job.
1429 * @param docEntry Item Starter that warned us
1430 * @param set DocEntrySet (ElementSet/SQItem) the DocEntry will belong
1432 DocEntry *Document::Backtrack(DocEntry *docEntry, DocEntrySet *set)
1434 // delete the Item Starter, built erroneously out of any Sequence
1435 // it's not yet in the HTable/chained list
1438 // Get all info we can from PreviousDocEntry
1439 uint16_t group = PreviousDocEntry->GetGroup();
1440 uint16_t elem = PreviousDocEntry->GetElement();
1441 uint32_t lgt = PreviousDocEntry->GetLength();
1442 long offset = PreviousDocEntry->GetOffset();
1444 gdcmDebugMacro( "Backtrack :" << std::hex << group
1446 << " at offset 0x(" <<offset << ")" );
1448 set->RemoveEntry( PreviousDocEntry );
1450 // forge the Seq Entry
1451 DocEntry *newEntry = NewSeqEntry(group, elem);
1452 newEntry->SetLength(lgt);
1453 newEntry->SetOffset(offset);
1455 // Move back to the beginning of the Sequence
1457 Fp->seekg(offset, std::ios::beg); // Only for Shadow Implicit VR SQ
1458 return newEntry; // It will added where it has to be!
1462 * \brief Loads (or not) the element content depending if its length exceeds
1463 * or not the value specified with Document::SetMaxSizeLoadEntry()
1464 * @param entry Header Entry (Dicom Element) to be dealt with
1465 * @param forceLoad whether you want to force loading of 'long' elements
1467 void Document::LoadDocEntry(DocEntry *entry, bool forceLoad)
1469 uint16_t group = entry->GetGroup();
1470 uint16_t elem = entry->GetElement();
1471 const VRKey &vr = entry->GetVR();
1472 uint32_t length = entry->GetLength();
1474 // Fp->seekg((long)entry->GetOffset(), std::ios::beg); // JPRx
1476 // A SeQuence "contains" a set of Elements.
1477 // (fffe e000) tells us an Element is beginning
1478 // (fffe e00d) tells us an Element just ended
1479 // (fffe e0dd) tells us the current SeQuence just ended
1480 // (fffe 0000) is an 'impossible' tag value,
1481 // found in MR-PHILIPS-16-Multi-Seq.dcm
1483 if ( (group == 0xfffe && elem != 0x0000 ) || vr == "SQ" )
1485 // NO more value field for SQ !
1489 DataEntry *dataEntryPtr = dynamic_cast< DataEntry* >(entry);
1495 // When the length is zero things are easy:
1498 dataEntryPtr->SetBinArea(NULL,true);
1502 // The elements whose length is bigger than the specified upper bound
1507 if (length > MaxSizeLoadEntry)
1509 dataEntryPtr->SetBinArea(NULL,true);
1510 dataEntryPtr->SetState(DataEntry::STATE_NOTLOADED);
1512 // to be sure we are at the end of the value ...
1513 // Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1514 // std::ios::beg); //JPRx
1519 /// \todo: a method that *doesn't* load anything (maybe with MaxSizeLoadEntry=0 ?)
1520 /// + a ForceLoad call on the +/- 20 'usefull' fields
1521 /// Allow user to tell the fields he wants to ForceLoad
1522 /// during initial stage.
1523 /// Later, a GetString or GetBinArea will load the value from disk, if not loaded
1524 /// + a method that load *everything* that's not yet loaded
1526 LoadEntryBinArea(dataEntryPtr); // last one, not to erase length !
1530 * \brief Find the value Length of the passed Doc Entry
1531 * @param entry Header Entry whose length of the value shall be loaded.
1533 void Document::FindDocEntryLength( DocEntry *entry )
1534 throw ( FormatError )
1536 const VRKey &vr = entry->GetVR();
1538 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1543 // For some images, length of UN elements is coded on 2 bytes (instead of 4)
1544 // There are *not* readable !
1545 // You can make a quick and dirty patch, commenting out
1547 // in the following line.
1548 // (the 'straight' images will no longer be readable ...)
1550 if ( vr == "OB" || vr == "OW" || vr == "OL" || vr == "SQ" || vr == "UT"
1551 || vr == "UN" || changeFromUN == true)
1553 changeFromUN = false;
1554 // The following reserved two bytes (see PS 3.5-2003, section
1555 // "7.1.2 Data element structure with explicit vr", p 27) must be
1556 // skipped before proceeding on reading the length on 4 bytes.
1558 //Fp->seekg( 2L, std::ios::cur); // Once per OB,OW,OL,UT,UN,SQ DocEntry
1559 uint32_t length32 = ReadInt32(); // Once per OB,OW,OL,UT,UN,SQ DocEntry
1560 CurrentOffsetPosition+=4;
1561 if ( (vr == "OB" || vr == "OW" || vr == "OL") && length32 == 0xffffffff )
1566 lengthOB = FindDocEntryLengthOBOrOW();// for encapsulation of encoded pixel
1568 catch ( FormatUnexpected )
1570 // Computing the length failed (this happens with broken
1571 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1572 // chance to get the pixels by deciding the element goes
1573 // until the end of the file. Hence we artificially fix the
1574 // the length and proceed.
1575 gdcmWarningMacro( " Computing the length failed for " <<
1576 entry->GetKey() <<" in " <<GetFileName());
1578 long currentPosition = Fp->tellg(); // Only for gdcm-JPEG-LossLess3a.dcm-like
1579 Fp->seekg(0L,std::ios::end); // Only for gdcm-JPEG-LossLess3a.dcm-like
1581 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition; // Only for gdcm-JPEG-LossLess3a.dcm-like
1582 Fp->seekg(currentPosition, std::ios::beg); // Only for gdcm-JPEG-LossLess3a.dcm-like
1584 entry->SetReadLength(lengthUntilEOF);
1585 entry->SetLength(lengthUntilEOF);
1588 entry->SetReadLength(lengthOB);
1589 entry->SetLength(lengthOB);
1592 FixDocEntryFoundLength(entry, length32);
1595 // Length is encoded on 2 bytes.
1596 //length16 = ReadInt16();
1597 length16 = GetInt16();
1598 // 0xffff means that we deal with 'No Length' Sequence
1599 // or 'No Length' SQItem
1600 if ( length16 == 0xffff)
1604 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1609 // Either implicit VR or a non DICOM conformal (see note below) explicit
1610 // VR that ommited the VR of (at least) this element. Farts happen.
1611 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1612 // on Data elements "Implicit and Explicit VR Data Elements shall
1613 // not coexist in a Data Set and Data Sets nested within it".]
1614 // Length is on 4 bytes.
1616 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1617 // even if Transfer Syntax is 'Implicit VR ...'
1618 // --> Except for 'Implicit VR Big Endian Transfer Syntax GE Private'
1619 // where Group 0x0002 is *also* encoded in Implicit VR !
1621 FixDocEntryFoundLength( entry, GetInt32() /*ReadInt32()*/ );
1627 * \brief Find the Length till the next sequence delimiter
1628 * \warning NOT end user intended method !
1631 uint32_t Document::FindDocEntryLengthOBOrOW()
1632 throw( FormatUnexpected )
1634 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1636 long positionOnEntry = Fp->tellg(); // Only for OB,OW DataElements
1638 bool foundSequenceDelimiter = false;
1639 uint32_t totalLength = 0;
1641 while ( !foundSequenceDelimiter )
1647 { ///\todo make sure there is never OL encoded pixel data!
1649 //group = ReadInt16(); // Once per fragment (if any) of OB,OW DataElements
1650 //elem = ReadInt16(); // Once per fragment (if any) of OB,OW DataElements
1651 ReadBegBuffer(4); // Once per fragment (if any) of OB,OW DataElements
1653 catch ( FormatError )
1655 throw FormatError("Unexpected end of file encountered during ",
1656 "Document::FindDocEntryLengthOBOrOW()");
1661 // We have to decount the group and element we just read
1663 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
1666 "Neither an Item tag nor a Sequence delimiter tag on :"
1667 << std::hex << group << "|" << elem << ") Pos. on entry was 0x(" <<positionOnEntry<< ") "
1670 Fp->seekg(positionOnEntry, std::ios::beg); // Once per fragment (if any) of OB,OW DataElements
1671 throw FormatUnexpected(
1672 "Neither an Item tag nor a Sequence delimiter tag.");
1674 if ( elem == 0xe0dd )
1676 foundSequenceDelimiter = true;
1678 uint32_t itemLength = ReadInt32(); // Once per fragment (if any) of OB,OW DataElements
1679 // We add 4 bytes since we just read the ItemLength with ReadInt32
1680 totalLength += itemLength + 4;
1681 SkipBytes(itemLength);
1683 if ( foundSequenceDelimiter )
1688 Fp->seekg( positionOnEntry, std::ios::beg); // Only once for OB,OW DataElements
1693 * \brief Find the Value Representation of the current Dicom Element.
1694 * @return Value Representation of the current Entry
1696 VRKey Document::FindDocEntryVR()
1698 if ( Filetype != ExplicitVR )
1700 return GDCM_VRUNKNOWN;
1703 // Delimiters (0xfffe), are not explicit VR ...
1704 if ( CurrentGroup == 0xfffe )
1705 return GDCM_VRUNKNOWN;
1707 //long positionOnEntry;
1708 //if( Debug::GetWarningFlag() )
1709 // positionOnEntry = Fp->tellg(); // Only in Warning Mode
1711 // Warning: we believe this is explicit VR (Value Representation) because
1712 // we used a heuristic that found "UL" in the first tag and/or
1713 // 'Transfer Syntax' told us it is.
1714 // Alas this doesn't guarantee that all the tags will be in explicit VR.
1715 // In some cases one finds implicit VR tags mixed within an explicit VR file
1717 // 'Normaly' the only case is : group 0002 Explicit, and other groups Implicit
1719 // Hence we make sure the present tag is in explicit VR and try to fix things
1720 // if it happens not to be the case.
1723 //Fp->read(&(vr[0]),(size_t)2);
1724 vr[0] = *PtrBegBuffer++;
1725 vr[1] = *PtrBegBuffer++;
1727 //if ( !CheckDocEntryVR(vr) ) // avoid useless function call
1728 if ( !Global::GetVR()->IsValidVR(vr) )
1731 gdcmWarningMacro( "Unknown VR " << vr.GetHexaRepresentation() << std::hex
1732 << " at offset : 0x(" << CurrentOffsetPosition-4
1733 << ") for group " << std::hex << CurrentGroup );
1735 //Fp->seekg(positionOnEntry, std::ios::beg); //JPRx
1736 //Fp->seekg((long)-2, std::ios::cur);// only for unrecognized VR (?!?)
1737 //see :MR_Philips_Intera_PrivateSequenceExplicitVR.dcm
1739 return GDCM_VRUNKNOWN;
1745 * \brief Check the correspondance between the VR of the header entry
1746 * and the taken VR. If they are different, the header entry is
1747 * updated with the new VR.
1748 * @param vr Dicom Value Representation
1749 * @return false if the VR is incorrect or if the VR isn't referenced
1750 * otherwise, it returns true
1752 bool Document::CheckDocEntryVR(const VRKey &vr)
1754 return Global::GetVR()->IsValidVR(vr);
1758 * \brief Skip a given Header Entry
1759 * @param entry entry to skip
1761 void Document::SkipDocEntry(DocEntry *entry)
1763 SkipBytes(entry->GetLength());
1767 * \brief Skips to the beginning of the next Header Entry
1768 * @param currentDocEntry entry to skip
1770 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1772 long l = currentDocEntry->GetReadLength();
1773 if ( (uint32_t) l == (uint32_t)-1 ) // length = 0xffff shouldn't appear here ...
1774 // ... but PMS imagers happen !
1776 Fp->seekg((size_t)(currentDocEntry->GetOffset()), std::ios::beg); //FIXME :each DocEntry
1777 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1779 Fp->seekg( l,std::ios::cur); //FIXME :each DocEntry
1784 * \brief When the length of an element value is obviously wrong (because
1785 * the parser went Jabberwocky) one can hope improving things by
1786 * applying some heuristics.
1787 * @param entry entry to check
1788 * @param foundLength first assumption about length (before bug fix, or set to zero if =0xffffffff)
1790 void Document::FixDocEntryFoundLength(DocEntry *entry,
1791 uint32_t foundLength)
1793 entry->SetReadLength( foundLength );// will be updated only if a bug is found
1795 if ( foundLength == 0xffffffff)
1798 //entry->SetLength(foundLength);
1799 entry->SetLength(0);
1800 return; // return ASAP; don't waist time on useless tests
1803 uint16_t gr = entry->GetGroup();
1804 uint16_t elem = entry->GetElement();
1806 if ( foundLength % 2)
1808 gdcmWarningMacro( "Warning : Tag (" << std::hex << gr << "|" << elem << ") with uneven length "
1809 << std::dec << foundLength << " 0x(" << std::hex << foundLength << ") "
1810 //<< " at offset x(" << offset << ")"
1814 //////// Fix for some naughty General Electric images.
1815 // Allthough not recent many such GE corrupted images are still present
1816 // on Creatis hard disks. Hence this fix shall remain when such images
1817 // are no longer in use (we are talking a few years, here)...
1818 // Note: XMedCon probably uses such a trick since it is able to read
1819 // those pesky GE images ...
1820 if ( foundLength == 13)
1822 // Only happens for this length !
1823 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1826 entry->SetReadLength(10); // a bug is to be fixed !?
1830 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1831 // Occurence of such images is quite low (unless one leaves close to a
1832 // 'Leonardo' source. Hence, one might consider commenting out the
1833 // following fix on efficiency reasons.
1834 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1836 // Ideally we should check we are in Explicit and double check
1837 // that VR=UL... this is done properly in gdcm2
1838 if( foundLength == 6 )
1840 gdcmWarningMacro( "Replacing Length from 6 into 4" );
1842 entry->SetReadLength(4); // a bug is to be fixed !
1844 else if ( foundLength%4 )
1846 gdcmErrorMacro( "This looks like to a buggy Siemens DICOM file."
1847 "The length of this tag seems to be wrong" );
1851 else if ( entry->GetVR() == "SQ" )
1853 foundLength = 0; // ReadLength is unchanged
1856 //////// We encountered a 'delimiter' element i.e. a tag of the form
1857 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1858 // taken into account.
1859 else if ( gr == 0xfffe )
1861 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1862 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1863 // causes extra troubles...
1864 if ( elem != 0x0000 )
1870 foundLength=12; // to skip the mess that follows this bugged Tag !
1873 entry->SetLength(foundLength);
1877 * \brief Apply some heuristics to predict whether the considered
1878 * element value contains/represents an integer or not.
1879 * @param entry The element value on which to apply the predicate.
1880 * @return The result of the heuristical predicate.
1882 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1884 uint16_t elem = entry->GetElement();
1885 uint16_t group = entry->GetGroup();
1886 const VRKey &vr = entry->GetVR();
1887 uint32_t length = entry->GetLength();
1889 // When we have some semantics on the element we just read, and if we
1890 // a priori know we are dealing with an integer, then we shall be
1891 // able to swap its element value properly.
1892 if ( elem == 0 ) // This is the group length of the group
1900 // Although this should never happen, still some images have a
1901 // corrupted group length [e.g. have a glance at offset x(8336) of
1902 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm.
1903 // Since for dicom compliant and well behaved headers, the present
1904 // test is useless (and might even look a bit paranoid), when we
1905 // encounter such an ill-formed image, we simply display a warning
1906 // message and proceed on parsing (while crossing fingers).
1907 long filePosition = Fp->tellg(); // Only when elem 0x0000 length is not 4 (?!?)
1909 gdcmWarningMacro( "Erroneous Group Length element length on : ("
1910 << std::hex << group << " , " << elem
1911 << ") -before- position x(" << filePosition << ")"
1912 << "lgt : " << length );
1916 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1924 * \brief Discover what the swap code is (among little endian, big endian,
1925 * bad little endian, bad big endian).
1927 * @return false when we are absolutely sure
1928 * it's neither ACR-NEMA nor DICOM
1929 * true when we hope ours assuptions are OK
1931 bool Document::CheckSwap()
1938 // First, compare HostByteOrder and NetworkByteOrder in order to
1939 // determine if we shall need to swap bytes (i.e. the Endian type).
1940 bool net2host = Util::IsCurrentProcessorBigEndian();
1942 // The easiest case is the one of a 'true' DICOM header, we just have
1943 // to look for the string "DICM" inside the file preamble.
1946 char *entCur = deb + 128;
1947 if ( memcmp(entCur, "DICM", (size_t)4) == 0 )
1949 gdcmDebugMacro( "Looks like DICOM Version3 (preamble + DCM)" );
1951 // Group 0002 should always be VR, and the first element 0000
1952 // Let's be carefull (so many wrong headers ...)
1953 // and determine the value representation (VR) :
1954 // Let's skip to the first element (0002,0000) and check there if we find
1955 // "UL" - or "OB" if the 1st one is (0002,0001) -,
1956 // in which case we (almost) know it is explicit VR.
1957 // WARNING: if it happens to be implicit VR then what we will read
1958 // is the length of the group. If this ascii representation of this
1959 // length happens to be "UL" then we shall believe it is explicit VR.
1960 // We need to skip :
1961 // * the 128 bytes of File Preamble (often padded with zeroes),
1962 // * the 4 bytes of "DICM" string,
1963 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
1964 // i.e. a total of 136 bytes.
1967 // group 0x0002 *is always* Explicit VR Sometimes,
1968 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
1969 // *Implicit* VR (see former 'gdcmData/icone.dcm')
1971 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
1972 memcmp(entCur, "OB", (size_t)2) == 0 ||
1973 memcmp(entCur, "UI", (size_t)2) == 0 ||
1974 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
1975 // when Write DCM *adds*
1977 // Use Document::dicom_vr to test all the possibilities
1978 // instead of just checking for UL, OB and UI !? group 0000
1980 Filetype = ExplicitVR;
1981 gdcmDebugMacro( "Group 0002 : Explicit Value Representation");
1985 Filetype = ImplicitVR;
1986 gdcmWarningMacro( "Group 0002 :Not an explicit Value Representation;"
1987 << "Looks like a bugged Header!");
1990 // Here, we assume that the file IS kosher Dicom !
1991 // (The meta elements - group 0x0002 - ARE little endian !)
1995 gdcmDebugMacro( "HostByteOrder != NetworkByteOrder, SwapCode = 4321");
2000 gdcmDebugMacro( "HostByteOrder = NetworkByteOrder, SwapCode = 1234");
2003 // Position the file position indicator at first tag
2004 // (i.e. after the file preamble and the "DICM" string).
2006 Fp->seekg ( 132L, std::ios::beg); // Once per Document
2007 CurrentOffsetPosition = 132;
2009 } // ------------------------------- End of DicomV3 ----------------
2011 // Alas, this is not a DicomV3 file and whatever happens there is no file
2012 // preamble. We can reset the file position indicator to where the data
2013 // is (i.e. the beginning of the file).
2015 gdcmWarningMacro( "Not a Kosher DICOM Version3 file (no preamble)");
2017 Fp->seekg(0, std::ios::beg); // Once per ACR-NEMA Document
2018 CurrentOffsetPosition = 0;
2019 // Let's check 'No Preamble Dicom File' :
2020 // Should start with group 0x0002
2021 // and be Explicit Value Representation
2023 s16 = *((uint16_t *)(deb));
2036 if ( SwapCode != 0 )
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, "SH", (size_t)2) == 0 ||
2042 memcmp(entCur, "AE", (size_t)2) == 0 ||
2043 memcmp(entCur, "OB", (size_t)2) == 0 )
2045 Filetype = ExplicitVR; // FIXME : not enough to say it's Explicit
2046 // Wait untill reading Transfer Syntax
2047 gdcmDebugMacro( "Group 0002 : Explicit Value Representation");
2051 // ------------------------------- End of 'No Preamble' DicomV3 -------------
2053 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2054 // By clean we mean that the length of the first group is written down.
2055 // If this is the case and since the length of the first group HAS to be
2056 // four (bytes), then determining the proper swap code is straightforward.
2059 // We assume the array of char we are considering contains the binary
2060 // representation of a 32 bits integer. Hence the following dirty
2062 s32 = *((uint32_t *)(entCur));
2082 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2083 // It is time for despaired wild guesses.
2084 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2085 // i.e. the 'group length' element is not present :
2087 // check the supposed-to-be 'group number'
2088 // in ( 0x0001 .. 0x0008 )
2089 // to determine ' SwapCode' value .
2090 // Only 0 or 4321 will be possible
2091 // (no oportunity to check for the formerly well known
2092 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2093 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc-3, 4, ..., 8-)
2094 // the file IS NOT ACR-NEMA nor DICOM V3
2095 // Find a trick to tell it the caller...
2097 s16 = *((uint16_t *)(deb));
2099 gdcmDebugMacro("not a DicomV3 nor a 'clean' ACR/NEMA;"
2100 << " (->despaired wild guesses !)");
2129 s16 = *((uint16_t *)(deb));
2132 s16 = *((uint16_t *)(deb+2));
2134 Fp->seekg ( 0L, std::ios::beg); // Once per Document
2135 CurrentOffsetPosition = 0;
2136 switch(s16) // try an other trick!
2137 // -> to be able to decode 0029|1010 DataElement
2138 // -> and be not less cleaver than dcmdump ;-)
2147 gdcmWarningMacro("ACR/NEMA unfound swap info (Hopeless !)");
2151 // Check if next 2 bytes are a VR
2152 // Probabely something more time-consuming exists with std::string
2153 const char VRvalues[] = "AEASATCSDADTFLFDISLOLTPNSHSLSSSTTMUIULUSUTOBOWOLOFATUNSQRT";
2155 const char *pt = VRvalues;
2156 for (int i=0;i<nbVal;i++)
2158 if(*(deb+4) == *pt++) {
2159 if(*(deb+5) == *pt++) {
2160 Filetype = ExplicitVR;
2168 Filetype = ImplicitVR;
2175 * \brief Change the Byte Swap code.
2177 void Document::SwitchByteSwapCode()
2179 gdcmDebugMacro( "Switching Byte Swap code from "<< SwapCode
2180 << " at: 0x" << std::hex << Fp->tellg() ); // Only when DEBUG
2181 if ( SwapCode == 1234 )
2185 else if ( SwapCode == 4321 )
2189 else if ( SwapCode == 3412 )
2193 else if ( SwapCode == 2143 )
2197 gdcmDebugMacro( " Into: "<< SwapCode );
2201 * \brief during parsing, Header Elements too long are not loaded in memory
2202 * @param newSize new size
2204 void Document::SetMaxSizeLoadEntry(long newSize)
2210 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2212 MaxSizeLoadEntry = 0xffffffff;
2215 MaxSizeLoadEntry = newSize;
2219 * \brief Read the next tag WITHOUT loading it's value
2220 * (read the 'Group Number', the 'Element Number',
2221 * gets the Dict Entry
2222 * gets the VR, gets the length, gets the offset value)
2223 * @return On succes : the newly created DocEntry, NULL on failure.
2225 DocEntry *Document::ReadNextDocEntry()
2229 ReadBegBuffer(8); // Avoid to many time consuming freads
2230 //CurrentGroup = ReadInt16();
2231 //CurrentElem = ReadInt16();
2233 catch ( FormatError )
2235 // We reached the EOF (or an error occured) therefore
2236 // header parsing has to be considered as finished.
2240 changeFromUN = false;
2241 CurrentGroup = GetInt16();
2242 CurrentElem = GetInt16();
2244 // In 'true DICOM' files Group 0002 is always little endian
2245 if ( HasDCMPreamble )
2247 if ( !Group0002Parsed && CurrentGroup != 0x0002) // avoid calling a function when useless
2248 HandleOutOfGroup0002(CurrentGroup, CurrentElem);
2250 // Sometimes file contains groups of tags with reversed endianess.
2251 HandleBrokenEndian(CurrentGroup, CurrentElem);
2254 VRKey vr = FindDocEntryVR();
2257 if ( vr == GDCM_VRUNKNOWN )
2259 if ( CurrentElem == 0x0000 ) // Group Length
2261 realVR = "UL"; // must be UL
2263 else if (CurrentGroup == 0xfffe) // Don't get DictEntry for Delimitors
2268 // Was commented out in order not to generate 'Shadow Groups' where some
2269 // Data Elements are Explicit VR and some other ones Implicit VR
2270 // -> Better we fix the problem at Write time
2272 else if (CurrentGroup%2 == 1 )
2274 if (CurrentElem >= 0x0010 && CurrentElem <=0x00ff )
2276 // DICOM PS 3-5 7.8.1 a) states that :
2277 // Private Creator Data Elements numbered (gggg,0010-00FF) (gggg is odd)
2278 // attributes have to be LO (Long String) and the VM shall be equal to 1
2281 // Seems not to be true
2282 // Still in gdcmtk, David Clunnie disagrees, Marco Eichelberg says it's OK ...
2283 // We let it for a while?
2284 //(We should check length==4, for more security, but we don't have it yet !)
2286 else if ( CurrentElem == 0x0001)
2288 realVR = "UL"; // Private Group Length To End
2290 else // check the private dictionary for shadow elements when Implicit VR!
2292 DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem);
2295 realVR = dictEntry->GetVR();
2296 dictEntry->Unregister(); // GetDictEntry registered it
2303 DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem);//only when ImplicitVR
2306 realVR = dictEntry->GetVR();
2307 dictEntry->Unregister(); // GetDictEntry registered it
2312 // if UN found, let's check the dictionary, and trust it!
2313 // (maybe a private dictionary exists?)
2314 else if (vr == "UN")
2316 DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem);
2319 realVR = dictEntry->GetVR();
2320 dictEntry->Unregister(); // GetDictEntry registered it
2322 // for VR = "UN", length is always stored on 4 bytes.
2323 // remember this info, in order not to crash later
2325 /// \todo : fixme If inside a vr = "UN" DataElement (but SQ according to a private dictionnary)
2326 /// there is some more vr = "UN" DataElements, it will probabely fail.
2327 /// --> find a -non time consuming- trick to store changeFromUN info at DataElement level,
2328 /// not at the Document level.
2335 //if ( Global::GetVR()->IsVROfSequence(realVR) )
2338 newEntry = NewSeqEntry(CurrentGroup, CurrentElem);
2342 newEntry = NewDataEntry(CurrentGroup, CurrentElem, realVR);
2343 static_cast<DataEntry *>(newEntry)->SetState(DataEntry::STATE_NOTLOADED);
2346 if ( vr == GDCM_VRUNKNOWN )
2348 if ( Filetype == ExplicitVR )
2350 // We thought this was explicit VR, but we end up with an
2351 // implicit VR tag. Let's backtrack.
2353 //if ( newEntry->GetGroup() != 0xfffe )
2354 if (CurrentGroup != 0xfffe)
2356 int offset = Fp->tellg();//Only when heuristic for Explicit/Implicit was wrong
2358 gdcmWarningMacro("Entry (" << newEntry->GetKey() << ") at x("
2359 << std::hex << offset << ") should be Explicit VR");
2362 newEntry->SetImplicitVR();
2367 FindDocEntryLength(newEntry);
2369 catch ( FormatError )
2376 newEntry->SetOffset(Fp->tellg()); // for each DocEntry
2381 * \brief Handle broken private tag from Philips NTSCAN
2382 * where the endianess is being switched to BigEndian
2383 * for no apparent reason
2386 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
2388 // for strange PMS Gyroscan Intera images
2389 // Item 'starter' has a tag : 0x3f3f,0x3f00, for no apparent reason
2391 // --- Feel free to remove this test *on your own coy of gdcm*
2392 // if you are sure you'll never face this problem.
2394 if ((group == 0x3f3f) && (elem == 0x3f00))
2396 // start endian swap mark for group found
2397 gdcmDebugMacro( " delimiter 0x3f3f found." );
2403 // --- End of removable code
2405 // Endian reversion.
2406 // Some files contain groups of tags with reversed endianess.
2407 static int reversedEndian = 0;
2408 // try to fix endian switching in the middle of headers
2409 if ((group == 0xfeff) && (elem == 0x00e0))
2411 // start endian swap mark for group found
2412 gdcmDebugMacro( "Start endian swap mark found." );
2414 SwitchByteSwapCode();
2419 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2421 // end of reversed endian group
2422 gdcmDebugMacro( "End of reversed endian." );
2424 SwitchByteSwapCode();
2426 else if (group == 0xfeff && elem == 0xdde0)
2428 // reversed Sequence Terminator found
2429 // probabely a bug in the header !
2430 // Do what you want, it breaks !
2432 //SwitchByteSwapCode();
2433 gdcmWarningMacro( "Should never get here! reversed Sequence Terminator!" );
2438 else if (group == 0xfffe && elem == 0xe0dd)
2440 gdcmDebugMacro( "Straight Sequence Terminator." );
2445 * \brief Group 0002 is always coded Little Endian
2446 * whatever Transfer Syntax is
2449 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2451 // Endian reversion.
2452 // Some files contain groups of tags with reversed endianess.
2454 Group0002Parsed = true;
2455 // we just came out of group 0002
2456 // if Transfer Syntax is Big Endian we have to change CheckSwap
2458 std::string ts = GetTransferSyntax();
2459 TS::SpecialType s = Global::GetTS()->GetSpecialTransferSyntax(ts);
2461 // Group 0002 is always 'Explicit ...'
2462 // even when Transfer Syntax says 'Implicit ..."
2464 if ( s == TS::ImplicitVRLittleEndian
2466 s == TS::ImplicitVRBigEndianPrivateGE
2469 Filetype = ImplicitVR;
2472 // FIXME Strangely, this works with
2473 //'Implicit VR BigEndian Transfer Syntax' (GE Private)
2475 // --> Probabely normal, since we considered we never have
2476 // to trust manufacturers.
2477 // (we often find 'Implicit VR' tag,
2478 // even when Transfer Syntax tells us it's Explicit ...
2480 // NEVER trust the meta elements!
2481 // (see what ezDICOM does ...)
2484 if ( s == TS::ExplicitVRBigEndian )
2486 gdcmDebugMacro("Transfer Syntax Name = ["
2487 << GetTransferSyntaxName() << "]" );
2488 SwitchByteSwapCode();
2489 group = SwapShort(group);
2490 elem = SwapShort(elem);
2493 //-- Broken ACR may start with a Shadow Group --
2494 // worse : some ACR-NEMA like files start 00028 group ?!?
2495 if ( !( (group >= 0x0001 && group <= 0x0008) || group == 0x0028 ) )
2497 // We trust what we see.
2498 SwitchByteSwapCode();
2499 group = SwapShort(group);
2500 elem = SwapShort(elem);
2501 // not what we where told (by meta elements) !
2502 gdcmDebugMacro("Transfer Syntax Name = ["
2503 << GetTransferSyntaxName() << "]" );
2506 /// \todo find a trick to warn user and stop processing
2508 if ( s == TS::DeflatedExplicitVRLittleEndian)
2510 gdcmWarningMacro("Transfer Syntax ["
2511 << GetTransferSyntaxName() << "] :"
2512 << " not yet dealt with ");
2516 // The following shouldn't occur very often
2517 // Let's check at the very end.
2519 if ( ts == GDCM_UNKNOWN )
2521 gdcmDebugMacro("True DICOM File, with NO Transfer Syntax (?!) " );
2525 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2527 gdcmWarningMacro("True DICOM File, with illegal Transfer Syntax: ["
2533 //-----------------------------------------------------------------------------
2536 //-----------------------------------------------------------------------------
2537 } // end namespace gdcm