1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2008/01/04 13:32:01 $
7 Version: $Revision: 1.378 $
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)
287 if(RecCode.find("ACRNEMA_LIBIDO") == 0 || // any version
288 RecCode.find("CANRME_AILIBOD") == 0) // for brain-damaged softwares
289 // with "little-endian strings"
291 Filetype = ACR_LIBIDO;
292 std::string rows = GetEntryString(0x0028, 0x0010);
293 std::string columns = GetEntryString(0x0028, 0x0011);
294 SetEntryString(columns, 0x0028, 0x0010);
295 SetEntryString(rows , 0x0028, 0x0011);
297 // --- End of ACR-LibIDO kludge ---
303 * \brief Adds a new element we want to load anyway
304 * @param group Group number of the target tag.
305 * @param elem Element number of the target tag.
307 void Document::AddForceLoadElement (uint16_t group, uint16_t elem)
312 UserForceLoadList.push_back(el);
315 * \brief Get the public dictionary used
317 Dict *Document::GetPubDict()
323 * \brief Get the shadow dictionary used
325 Dict *Document::GetShaDict()
331 * \brief Set the shadow dictionary used
332 * @param dict dictionary to use in shadow
334 bool Document::SetShaDict(Dict *dict)
341 * \brief Set the shadow dictionary used
342 * @param dictName name of the dictionary to use in shadow
344 bool Document::SetShaDict(DictKey const &dictName)
346 RefShaDict = Global::GetDicts()->GetDict(dictName);
351 * \brief This predicate tells us whether or not the current Document
352 * was properly parsed and contains at least *one* Dicom Element
353 * (and nothing more, sorry).
354 * @return false when we're 150 % sure it's NOT a Dicom/Acr file,
357 bool Document::IsParsable()
359 if ( Filetype == Unknown )
361 gdcmWarningMacro( "Wrong filetype for " << GetFileName());
367 gdcmWarningMacro( "No tag in internal hash table.");
374 * \brief This predicate tells us whether or not the current Document
375 * was properly parsed and contains at least *one* Dicom Element
376 * (and nothing more, sorry).
377 * @return false when we're 150 % sure it's NOT a Dicom/Acr file,
380 bool Document::IsReadable()
386 * \brief Predicate for dicom version 3 file.
387 * @return True when the file is a dicom version 3.
389 bool Document::IsDicomV3()
391 // Checking if Transfer Syntax exists is enough
392 // Anyway, it's too late check if the 'Preamble' was found ...
393 // And ... would it be a rich idea to check ?
394 // (some 'no Preamble' DICOM images exist !)
395 return GetDocEntry(0x0002, 0x0010) != NULL;
399 * \brief Predicate for Papyrus file
400 * Dedicated to whomsoever it may concern
401 * @return True when the file is a Papyrus file.
403 bool Document::IsPapyrus()
405 // check for Papyrus private Sequence
406 DocEntry *e = GetDocEntry(0x0041, 0x1050);
409 // check if it's actually a Sequence
410 if ( !dynamic_cast<SeqEntry*>(e) )
416 * \brief returns the File Type
417 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
418 * @return the FileType code
420 FileType Document::GetFileType()
426 * \brief Accessor to the Transfer Syntax (when present) of the
427 * current document (it internally handles reading the
428 * value from disk when only parsing occured).
429 * @return The encountered Transfer Syntax of the current document, if DICOM.
430 * GDCM_UNKNOWN for ACR-NEMA files (or broken headers ...)
432 std::string Document::GetTransferSyntax()
434 DocEntry *entry = GetDocEntry(0x0002, 0x0010);
440 // The entry might be present but not loaded (parsing and loading
441 // happen at different stages): try loading and proceed with check...
444 // (parsing and loading happen at the very same stage!)
445 //LoadDocEntrySafe(entry); //JPRx
446 if (DataEntry *dataEntry = dynamic_cast<DataEntry *>(entry) )
448 std::string transfer = dataEntry->GetString();
449 // The actual transfer (as read from disk) might be padded. We
450 // first need to remove the potential padding. We can make the
451 // weak assumption that padding was not executed with digits...
452 if ( transfer.length() == 0 )
454 // for brain damaged headers
455 gdcmWarningMacro( "Transfer Syntax has length = 0.");
458 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
460 transfer.erase(transfer.length()-1, 1);
461 if ( transfer.length() == 0 )
463 // for brain damaged headers
464 gdcmWarningMacro( "Transfer Syntax contains no valid character.");
474 * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS
475 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
477 std::string Document::GetTransferSyntaxName()
479 // use the TS (TS : Transfer Syntax)
480 std::string transferSyntax = GetEntryString(0x0002,0x0010);
482 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
484 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
485 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
486 return "Uncompressed ACR-NEMA";
488 if ( transferSyntax == GDCM_UNFOUND )
490 gdcmDebugMacro( "Unfound Transfer Syntax (0002,0010)");
491 return "Uncompressed ACR-NEMA";
494 // we do it only when we need it
495 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
497 // Global::GetTS() is a global static you shall never try to delete it!
501 // --------------- Swap Code ------------------
503 * \brief Swaps the bytes so they agree with the processor order
504 * @return The properly swaped 16 bits integer.
506 uint16_t Document::SwapShort(uint16_t a)
508 if ( SwapCode == 4321 || SwapCode == 2143 )
510 //a = ((( a << 8 ) & 0xff00 ) | (( a >> 8 ) & 0x00ff ) );
512 a = ( a << 8 ) | ( a >> 8 );
518 * \brief Swaps back the bytes of 4-byte long integer accordingly to
520 * @return The properly swaped 32 bits integer.
522 uint32_t Document::SwapLong(uint32_t a)
529 // a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
530 // ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
532 a=( ( a<<24) | ((a<<8) & 0x00ff0000) |
533 ((a>>8) & 0x0000ff00) | (a>>24) );
536 // a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
537 a=( (a<<16) | (a>>16) );
540 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
543 gdcmErrorMacro( "Unexpected swap code:" << SwapCode );
550 * \brief Swaps back the bytes of 8-byte long 'double' accordingly to
552 * @return The properly swaped 64 bits double.
554 double Document::SwapDouble(double a)
558 // There were no 'double' at ACR-NEMA time.
559 // We just have to deal with 'straight Little Endian' and
560 // 'straight Big Endian'
565 char *beg = (char *)&a;
568 for (unsigned int i = 0; i<7; i++)
579 gdcmErrorMacro( "Unexpected swap code:" << SwapCode );
586 // -----------------File I/O ---------------
588 * \brief Tries to open the file Document::Filename and
589 * checks the preamble when existing,
590 * or if the file starts with an ACR-NEMA look-like element.
591 * @return The FILE pointer on success, 0 on failure.
593 std::ifstream *Document::OpenFile()
595 HasDCMPreamble = false;
596 if (Filename.length() == 0)
603 gdcmDebugMacro( "File already open: " << Filename.c_str());
607 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
610 // Don't user gdcmErrorMacro :
611 // a spurious message will appear when you use, for instance
612 // gdcm::FileHelper *fh = new gdcm::FileHelper( outputFileName );
613 // to create outputFileName.
615 // FIXME : if the upper comment is still usefull
616 // --> the constructor is not so good ...
618 gdcmWarningMacro( "Cannot open file: " << Filename.c_str());
622 //exit(1); // No function is allowed to leave the application instead
623 // of warning the caller
627 Fp->read((char*)&zero, (size_t)2);
635 Fp->seekg(126L, std::ios::cur); // Once per Document
636 char dicm[4]; // = {' ',' ',' ',' '};
637 Fp->read(dicm, (size_t)4);
644 if ( memcmp(dicm, "DICM", 4) == 0 )
646 HasDCMPreamble = true;
650 //-- Broken ACR or DICOM (?) with no Preamble; may start with a Shadow Group --
651 // FIXME : We cannot be sure the preable is only zeroes..
652 // (see ACUSON-24-YBR_FULL-RLE.dcm )
654 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
655 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
656 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
657 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 ||
658 zero == 0x0028 || 0x2800 // worse : some ACR-NEMA like files
659 // start 00028 group ?!?
662 std::string msg = Util::Format(
663 "ACR/DICOM starting by 0x(%04x) at the beginning of the file\n", zero);
664 // FIXME : is it a Warning message, or a Debug message?
665 gdcmWarningMacro( msg.c_str() );
669 // -- Neither ACR/No Preamble Dicom nor DICOMV3 file
671 // Don't user Warning nor Error, not to pollute the output
672 // while directory recursive parsing ...
673 gdcmDebugMacro( "Neither ACR/No Preamble Dicom nor DICOMV3 file: "
674 << Filename.c_str());
679 * \brief closes the file
680 * @return TRUE if the close was successfull
682 bool Document::CloseFile()
694 * \brief Writes in a file all the Entries (Dicom Elements)
695 * @param fp file pointer on an already open file (actually: Output File Stream)
696 * @param filetype Type of the File to be written
697 * (ACR-NEMA, ExplicitVR, ImplicitVR)
699 void Document::WriteContent(std::ofstream *fp, FileType filetype, bool, bool)
701 // Skip if user wants to write an ACR-NEMA file
703 if ( filetype == ImplicitVR || filetype == ExplicitVR ||
704 filetype == JPEG || filetype == JPEG2000 )
706 // writing Dicom File Preamble
707 char filePreamble[128];
708 memset(filePreamble, 0, 128);
709 fp->write(filePreamble, 128);
710 fp->write("DICM", 4);
713 * \todo rewrite later, if really usefull
714 * - 'Group Length' element is optional in DICOM
715 * - but un-updated odd groups lengthes can causes pb
718 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
719 * UpdateGroupLength(false,filetype);
720 * if ( filetype == ACR)
721 * UpdateGroupLength(true,ACR);
723 * --> Computing group length for groups with embeded Sequences
724 * --> was too much tricky / we were [in a hurry / too lazy]
725 * --> We don't write the element 0x0000 (group length)
727 // This one is recursive
728 // false : outside MetaElements
729 // false : outside Sequence
730 ElementSet::WriteContent(fp, filetype, false, false);
733 // -----------------------------------------
736 * \brief Loads (from disk) the element content
737 * when a string is not suitable
738 * @param group group number of the Entry
739 * @param elem element number of the Entry
741 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
743 // Search the corresponding DocEntry
744 DocEntry *docEntry = GetDocEntry(group, elem);
747 gdcmDebugMacro(std::hex << group << "|" << elem
748 << " doesn't exist" );
751 DataEntry *dataEntry = dynamic_cast<DataEntry *>(docEntry);
754 gdcmWarningMacro(std::hex << group << "|" << elem
755 << " is NOT a DataEntry");
758 LoadEntryBinArea(dataEntry);
762 * \brief Loads (from disk) the element content
763 * when a string is not suitable
764 * @param entry Entry whose binArea is going to be loaded
766 void Document::LoadEntryBinArea(DataEntry *entry)
768 if( entry->GetBinArea() )
775 //size_t o =(size_t)entry->GetOffset();
776 Fp->seekg((size_t)entry->GetOffset(), std::ios::beg); // FIXME : for each DataEntry !
778 size_t l = entry->GetLength();
779 uint8_t *data = new uint8_t[l];
782 gdcmWarningMacro( "Cannot allocate DataEntry content for : "
783 << std::hex << entry->GetGroup()
784 << "|" << entry->GetElement() );
789 Fp->read((char*)data, l);
790 if ( Fp->fail() || Fp->eof() )
793 entry->SetState(DataEntry::STATE_UNREAD);
797 // Swap the data content if necessary
799 unsigned short vrLgth =
800 Global::GetVR()->GetAtomicElementLength(entry->GetVR());
802 // FIXME : trouble expected if we read an ... OW Entry (LUT, etc ..)
803 // if( entry->GetVR() == "OW" )
814 uint16_t *data16 = (uint16_t *)data;
815 for(i=0;i<l/vrLgth;i++)
816 data16[i] = SwapShort(data16[i]);
821 uint32_t *data32 = (uint32_t *)data;
822 for(i=0;i<l/vrLgth;i++)
823 data32[i] = SwapLong(data32[i]);
828 double *data64 = (double *)data;
829 for(i=0;i<l/vrLgth;i++)
830 data64[i] = SwapDouble(data64[i]);
835 entry->SetBinArea(data);
837 if ( openFile ) // The file is left in the state (open/close) it was at entrance
842 * \brief Loads the element while preserving the current
843 * underlying file position indicator as opposed to
844 * LoadDocEntry that modifies it
845 * \note seems to be unused!.
846 * @param entry DocEntry whose value will be loaded.
848 //void Document::LoadDocEntrySafe(DocEntry *entry)
852 // long PositionOnEntry = Fp->tellg(); // LoadDocEntrySafe is not used
853 // LoadDocEntry(entry);
854 // Fp->seekg(PositionOnEntry, std::ios::beg); // LoadDocEntrySafe is not used
859 * \brief Compares two documents, according to DicomDir rules
860 * \warning Does NOT work with ACR-NEMA files
861 * \todo Find a trick to solve the pb (use RET fields ?)
862 * @param document to compare with current one
863 * @return true if 'smaller'
865 bool Document::operator<(Document &document)
868 std::string s1 = GetEntryString(0x0010,0x0010);
869 std::string s2 = document.GetEntryString(0x0010,0x0010);
881 s1 = GetEntryString(0x0010,0x0020);
882 s2 = document.GetEntryString(0x0010,0x0020);
893 // Study Instance UID
894 s1 = GetEntryString(0x0020,0x000d);
895 s2 = document.GetEntryString(0x0020,0x000d);
906 // Serie Instance UID
907 s1 = GetEntryString(0x0020,0x000e);
908 s2 = document.GetEntryString(0x0020,0x000e);
923 //-----------------------------------------------------------------------------
927 * \brief Reads a given length of bytes
928 * (in order to avoid to many CPU time-consuming fread-s)
929 * @param l length to read
931 void Document::ReadBegBuffer(size_t l)
934 Fp->read (BegBuffer, (size_t)l);
937 throw FormatError( "Document::ReadBegBuffer()", " file error." );
941 throw FormatError( "Document::ReadBegBuffer()", "EOF." );
943 PtrBegBuffer = BegBuffer;
944 CurrentOffsetPosition+=l;
947 * \brief Reads a supposed to be 16 Bits integer
948 * (swaps it depending on processor endianness)
951 uint16_t Document::ReadInt16()
955 Fp->read ((char*)&g, (size_t)2);
958 throw FormatError( "Document::ReadInt16()", " file error." );
962 throw FormatError( "Document::ReadInt16()", "EOF." );
969 * \brief Gets from BegBuffer a supposed to be 16 Bits integer
970 * (swaps it depending on processor endianness)
973 uint16_t Document::GetInt16()
975 uint16_t g = *((uint16_t*)PtrBegBuffer);
981 * \brief Reads a supposed to be 32 Bits integer
982 * (swaps it depending on processor endianness)
985 uint32_t Document::ReadInt32()
989 Fp->read ((char*)&g, (size_t)4);
992 throw FormatError( "Document::ReadInt32()", " file error." );
996 throw FormatError( "Document::ReadInt32()", "EOF." );
1003 * \brief Gets from BegBuffer a supposed to be 32 Bits integer
1004 * (swaps it depending on processor endianness)
1005 * @return read value
1007 uint32_t Document::GetInt32()
1009 uint32_t g = *((uint32_t*)PtrBegBuffer);
1016 * \brief Re-computes the length of the Dicom group 0002.
1018 int Document::ComputeGroup0002Length( )
1023 int groupLength = 0;
1024 bool found0002 = false;
1026 // for each zero-level Tag in the DCM Header
1027 DocEntry *entry = GetFirstEntry();
1030 gr = entry->GetGroup();
1036 if ( entry->GetElement() != 0x0000 )
1038 vr = entry->GetVR();
1040 //if ( (vr == "OB")||(vr == "OW")||(vr == "UT")||(vr == "SQ"))
1041 // (no SQ, OW, OL, UT in group 0x0002;)
1044 // explicit VR AND (OB, OW, OL, SQ, UT, UN) : 4 more bytes
1047 groupLength += 2 + 2 + 4 + entry->GetLength();
1050 else if (found0002 )
1053 entry = GetNextEntry();
1059 * \brief CallStartMethod
1061 void Document::CallStartMethod()
1065 CommandManager::ExecuteCommand(this,CMD_STARTPROGRESS);
1069 * \brief CallProgressMethod
1071 void Document::CallProgressMethod()
1073 CommandManager::ExecuteCommand(this,CMD_PROGRESS);
1077 * \brief CallEndMethod
1079 void Document::CallEndMethod()
1082 CommandManager::ExecuteCommand(this,CMD_ENDPROGRESS);
1085 //-----------------------------------------------------------------------------
1088 * \brief Loads all the needed Dictionaries
1089 * \warning NOT end user intended method !
1091 void Document::Initialize()
1093 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
1099 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
1100 * @param set DocEntrySet we are going to parse ('zero level' or a SQItem)
1101 * @param offset start of parsing
1102 * @param l_max length to parse (meaningless when we are in 'delimitor mode')
1103 * @param delim_mode : whether we are in 'delimitor mode' (l=0xffffff) or not
1105 bool Document::ParseDES(DocEntrySet *set, long offset,
1106 long l_max, bool delim_mode)
1108 DocEntry *newDocEntry;
1109 DataEntry *newDataEntry;
1110 SeqEntry *newSeqEntry;
1112 bool used; // will be set to false when something wrong happens to an Entry.
1113 // (Entry will then be deleted)
1114 bool delim_mode_intern = delim_mode;
1116 bool successParseDES = true;
1117 gdcmDebugMacro( "Enter in ParseDES, delim-mode " << delim_mode
1118 << " at offset " << std::hex << "0x(" << offset << ")" );
1122 ///\todo FIXME : On 64 bits processors, tellg gives unexpected results after a while ?
1123 /// Probabely a bug in gdcm code somewhere (some memory erased ?)
1125 // Uncomment to track the bug
1127 if( Debug::GetDebugFlag() )
1128 std::cout << std::dec <<"(long)(Fp->tellg()) " << (long)(Fp->tellg()) // in Debug mode
1129 << std::hex << " 0x(" <<(long)(Fp->tellg()) << ")" << std::endl;
1132 // if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) // Once per DocEntry
1133 if ( !delim_mode ) // 'and then' doesn't exist in C++ :-(
1134 if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per DocEntry, when no delim mode
1139 newDocEntry = ReadNextDocEntry( );
1146 // Uncoment this cerr line to be able to 'follow' the DocEntries
1147 // when something *very* strange happens
1148 if( Debug::GetDebugFlag() )
1149 std::cerr<<newDocEntry->GetKey()<<" "<<newDocEntry->GetVR()<<std::endl;
1151 // an Item Starter found elsewhere but in the first position
1152 // of a SeqEntry means previous entry was a Sequence
1153 // but we didn't get it (private Sequence + Implicit VR)
1154 // we have to backtrack.
1155 if ( !first && newDocEntry->IsItemStarter() )
1157 // Debug message within the method !
1158 newDocEntry = Backtrack(newDocEntry, set);
1162 PreviousDocEntry = newDocEntry;
1166 newDataEntry = dynamic_cast<DataEntry*>(newDocEntry);
1170 //////////////////////////// DataEntry
1172 //vr = newDocEntry->GetVR(); // useless ?
1174 if ( !set->AddEntry( newDataEntry ) )
1176 gdcmDebugMacro( "in ParseDES : cannot add a DataEntry "
1177 << newDataEntry->GetKey()
1178 << " (at offset : 0x("
1179 << newDataEntry->GetOffset() << ") )" );
1184 newDataEntry->Delete();
1185 // Load only if we can add (not a duplicate key)
1186 LoadDocEntry( newDataEntry );
1188 if ( newDataEntry->GetElement() == 0x0000 ) // if on group length
1190 if ( newDataEntry->GetGroup()%2 != 0 ) // if Shadow Group
1192 if ( LoadMode & LD_NOSHADOW ) // if user asked to skip shad.gr
1194 std::string strLgrGroup = newDataEntry->GetString();
1197 //if ( newDataEntry->IsUnfound() ) /?!? JPR
1199 lgrGroup = atoi(strLgrGroup.c_str());
1200 Fp->seekg(lgrGroup, std::ios::cur); // Once per Shadow group, when NOSHADOW
1201 RemoveEntry( newDocEntry ); // Remove and delete
1208 bool delimitor = newDataEntry->IsItemDelimitor();
1209 bool outOfBounds = false;
1211 if ( ((long)(Fp->tellg())-offset) >= l_max ) //Once per DataEntry when no delim mode
1214 // 'and then', 'or else' don't exist in C++ :-(
1215 // if ( (delimitor) ||
1216 // (!delim_mode && ((long)(Fp->tellg())-offset) >= l_max) ) // Once per DataEntry
1218 if ( delimitor || outOfBounds )
1221 newDocEntry->Delete();
1225 // Just to make sure we are at the beginning of next entry.
1226 SkipToNextDocEntry(newDocEntry); // FIXME : once per DocEntry, segfault if commented out
1230 /////////////////////// SeqEntry : VR = "SQ"
1232 unsigned long l = newDocEntry->GetReadLength();
1233 if ( l != 0 ) // don't mess the delim_mode for 'zero-length sequence'
1235 if ( l == 0xffffffff )
1237 delim_mode_intern = true;
1241 delim_mode_intern = false;
1245 if ( (LoadMode & LD_NOSHADOWSEQ) && ! delim_mode_intern )
1247 // User asked to skip SeQuences *only* if they belong to Shadow Group
1248 if ( newDocEntry->GetGroup()%2 != 0 )
1250 Fp->seekg( l, std::ios::cur); // once per SQITEM, when NOSHADOWSEQ
1251 newDocEntry->Delete(); // Delete, not in the set
1255 if ( (LoadMode & LD_NOSEQ) && ! delim_mode_intern )
1257 // User asked to skip *any* SeQuence
1258 Fp->seekg( l, std::ios::cur); // Once per SQ, when NOSEQ
1259 newDocEntry->Delete(); // Delete, not in the set
1262 // delay the dynamic cast as late as possible
1263 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1265 // no other way to create the Delimitor ...
1266 newSeqEntry->SetDelimitorMode( delim_mode_intern );
1268 // At the top of the hierarchy, stands a Document. When "set"
1269 // is a Document, then we are building the first depth level.
1270 // Hence the SeqEntry we are building simply has a depth
1272 if ( set == this ) // ( dynamic_cast< Document* > ( set ) )
1274 newSeqEntry->SetDepthLevel( 1 );
1276 // But when "set" is already a SQItem, we are building a nested
1277 // sequence, and hence the depth level of the new SeqEntry
1278 // we are building, is one level deeper:
1280 // time waste hunting
1281 else if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1283 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 res = ParseSQ( newSeqEntry,
1295 newDocEntry->GetOffset(),
1296 l, delim_mode_intern);
1298 gdcmDebugMacro( "Exit from ParseSQ, delim " << delim_mode_intern << " -->return : " << res);
1304 gdcmDebugMacro( "in ParseDES : ParseSQ failed " << newSeqEntry->GetKey()
1305 << " (at offset : 0x(" << std::hex
1306 << newSeqEntry->GetOffset() << ") )" );
1308 successParseDES = false;
1309 newDocEntry = BacktrackSQtoOB(newSeqEntry, set);
1310 set->AddEntry( newDocEntry); /// \todo fixme : the whole process is not done
1311 /// (check SQItem length, when the DataElement is inside a Sequence)
1320 if ( !set->AddEntry( newSeqEntry ) )
1322 gdcmWarningMacro( "in ParseDES : cannot add a SeqEntry "
1323 << newSeqEntry->GetKey()
1324 << " (at offset : 0x("
1325 << newSeqEntry->GetOffset() << ") )" );
1330 newDocEntry->Delete();
1333 // if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) // Once per SeqEntry
1335 if ( !delim_mode ) // 'and then' doesn't exist in C++ :-(
1336 if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per SeqEntry when no delim mode
1340 newDocEntry->Delete();
1343 } // end SeqEntry : VR = "SQ"
1347 newDocEntry->Delete();
1351 if (UnexpectedEOF) // some terminator was missing
1357 gdcmDebugMacro( "Exit from ParseDES, delim-mode " << delim_mode );
1359 return successParseDES;
1363 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1364 * @return false if expected fff0,e000 not found
1366 bool Document::ParseSQ( SeqEntry *seqEntry,
1367 long offset, long l_max, bool delim_mode)
1369 int SQItemNumber = 0;
1371 long offsetStartCurrentSQItem = offset;
1375 // the first time, we read the fff0,e000 of the first SQItem
1376 DocEntry *newDocEntry = ReadNextDocEntry();
1380 // The most frequent is when a SQ terminator is missing (?!?)
1381 gdcmWarningMacro("in ParseSQ : should never get here!");
1382 UnexpectedEOF = true;
1387 if ( newDocEntry->IsSequenceDelimitor() )
1389 seqEntry->SetDelimitationItem( newDocEntry );
1390 newDocEntry->Delete();
1394 else // ! delim_mode
1396 if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per SQItem when no delim mode
1398 newDocEntry->Delete();
1402 // create the current SQItem
1403 SQItem *itemSQ = SQItem::New( seqEntry->GetDepthLevel() );
1404 unsigned int l = newDocEntry->GetReadLength();
1406 if ( l == 0xffffffff )
1415 // avoid infinite loop when Bad assumption was made on illegal 'unknown length' UN //JPRx
1417 if (offsetStartCurrentSQItem <= OffsetOfPreviousParseDES)
1419 gdcmWarningMacro("Bad assumption was made on illegal 'unknown length' UN!" << std::endl <<
1420 "OffsetOfPreviousParseDES " << std::hex << OffsetOfPreviousParseDES
1421 << " offsetStartCurrentSQItem " << offsetStartCurrentSQItem);
1422 /// \todo when "Bad assumption (SQ) on illegal 'unknown length' UN", Backtrack again + try OB
1427 OffsetOfPreviousParseDES = offsetStartCurrentSQItem;
1430 // fill up the current SQItem, starting at the beginning of fff0,e000
1432 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg); // Once per SQItem
1433 bool res = ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
1434 offsetStartCurrentSQItem = Fp->tellg(); // Once per SQItem
1436 seqEntry->AddSQItem( itemSQ, SQItemNumber );
1438 newDocEntry->Delete();
1440 if (!res) // to avoid extra -useless- work when illegal UN stands for OB ...
1442 gdcmDebugMacro("in ParseSQ : ParseDES failed");
1445 //if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max ) //JPRx
1446 if ( !delim_mode && (offsetStartCurrentSQItem-offset ) >= l_max )
1455 * \brief When a private Sequence + Implicit VR
1456 * (or UN + undefined length element) is encountered
1457 * we cannot guess it's a Sequence till we find the first
1458 * Item Starter. We then backtrack to do the job.
1459 * @param docEntry Item Starter that warned us
1460 * @param set DocEntrySet (ElementSet/SQItem) the DocEntry will belong
1462 DocEntry *Document::Backtrack(DocEntry *docEntry, DocEntrySet *set)
1464 // delete the Item Starter, built erroneously out of any Sequence
1465 // it's not yet in the HTable/chained list
1468 // Get all info we can from PreviousDocEntry
1469 uint16_t group = PreviousDocEntry->GetGroup();
1470 uint16_t elem = PreviousDocEntry->GetElement();
1471 uint32_t lgt = PreviousDocEntry->GetLength();
1472 long offset = PreviousDocEntry->GetOffset();
1474 gdcmDebugMacro( "Backtrack :" << std::hex << group
1476 << " at offset 0x(" <<offset << ")" );
1478 set->RemoveEntry( PreviousDocEntry );
1480 // forge the Seq Entry
1481 DocEntry *newEntry = NewSeqEntry(group, elem);
1482 newEntry->SetLength(lgt);
1483 newEntry->SetOffset(offset);
1485 // Move back to the beginning of the Sequence
1487 Fp->seekg(offset, std::ios::beg); // Only for Shadow Implicit VR SQ
1488 return newEntry; // It will be added where it has to be!
1493 * \brief When a private OB + Implicit VR
1494 * (or UN + undefined length element) is encountered
1495 * we made first a bad assumption -it should be a Sequence-
1496 * When we realize we were wrong, we backtrack.
1497 * WARNING : it will probabely fail if the element
1498 * is embedded within a sequence
1499 * @param docEntry Item Starter that warned us
1500 * @param set DocEntrySet (ElementSet/SQItem) the DocEntry will belong
1503 DataEntry *Document::BacktrackSQtoOB(SeqEntry *docEntry, DocEntrySet *set)
1505 // Get all info we can from SeqEntry
1507 uint16_t group = docEntry->GetGroup();
1508 uint16_t elem = docEntry->GetElement();
1509 uint32_t lgt = docEntry->GetLength();
1510 long offset = docEntry->GetOffset();
1512 gdcmDebugMacro( "BacktrackSQtoOB(" << std::hex << group
1514 << " at offset 0x(" <<offset << ")" );
1516 //set->RemoveEntry( docEntry );
1517 docEntry->Delete(); // not yet included into a DocEntrySet !
1519 // forge the OB Entry
1520 DataEntry *newEntry = NewDataEntry(group, elem, "OB");
1521 newEntry->SetOffset(offset);
1523 // Move back to the beginning of the Element
1524 Fp->seekg(offset, std::ios::beg);
1526 uint32_t lengthOB = FindDocEntryLengthOBOrOW();// for encapsulation of encoded pixel
1528 newEntry->SetReadLength(lengthOB);
1529 newEntry->SetLength(lengthOB);
1531 Fp->seekg(offset, std::ios::beg);
1532 LoadDocEntry(newEntry, true); // true : forceload
1534 /// \todo fixme : it will fail if the element
1535 /// is embedded within a sequence (sequence length checking is not done)
1537 return newEntry; // It will be added (ElementSet/SQItem) where it has to be!
1541 * \brief Loads (or not) the element content depending if its length exceeds
1542 * or not the value specified with Document::SetMaxSizeLoadEntry()
1543 * @param entry Header Entry (Dicom Element) to be dealt with
1544 * @param forceLoad whether you want to force loading of 'long' elements
1546 void Document::LoadDocEntry(DocEntry *entry, bool forceLoad)
1548 uint16_t group = entry->GetGroup();
1549 uint16_t elem = entry->GetElement();
1550 const VRKey &vr = entry->GetVR();
1551 uint32_t length = entry->GetLength();
1553 // Fp->seekg((long)entry->GetOffset(), std::ios::beg); // JPRx
1555 // A SeQuence "contains" a set of Elements.
1556 // (fffe e000) tells us an Element is beginning
1557 // (fffe e00d) tells us an Element just ended
1558 // (fffe e0dd) tells us the current SeQuence just ended
1559 // (fffe 0000) is an 'impossible' tag value,
1560 // found in MR-PHILIPS-16-Multi-Seq.dcm
1562 if ( (group == 0xfffe && elem != 0x0000 ) || vr == "SQ" )
1564 // NO more value field for SQ !
1568 DataEntry *dataEntryPtr = dynamic_cast< DataEntry* >(entry);
1574 // When the length is zero things are easy:
1577 dataEntryPtr->SetBinArea(NULL,true);
1581 // The elements whose length is bigger than the specified upper bound
1586 if (length > MaxSizeLoadEntry)
1588 dataEntryPtr->SetBinArea(NULL,true);
1589 dataEntryPtr->SetState(DataEntry::STATE_NOTLOADED);
1591 // to be sure we are at the end of the value ...
1592 // Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1593 // std::ios::beg); //JPRx
1598 /// \todo: a method that *doesn't* load anything (maybe with MaxSizeLoadEntry=0 ?)
1599 /// + a ForceLoad call on the +/- 20 'usefull' fields
1600 /// Allow user to tell the fields he wants to ForceLoad
1601 /// during initial stage.
1602 /// Later, a GetString or GetBinArea will load the value from disk, if not loaded
1603 /// + a method that load *everything* that's not yet loaded
1605 LoadEntryBinArea(dataEntryPtr); // last one, not to erase length !
1609 * \brief Find the value Length of the passed Doc Entry
1610 * @param entry Header Entry whose length of the value shall be loaded.
1612 void Document::FindDocEntryLength( DocEntry *entry )
1613 throw ( FormatError )
1615 const VRKey &vr = entry->GetVR();
1617 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1622 // For some images, length of UN elements is coded on 2 bytes (instead of 4)
1623 // There are *not* readable !
1624 // You can make a quick and dirty patch, commenting out
1626 // in the following line.
1627 // (the 'straight' images will no longer be readable ...)
1629 if ( vr == "OB" || vr == "OW" || vr == "OL" || vr == "SQ" || vr == "UT"
1630 || vr == "UN" || changeFromUN == true)
1632 changeFromUN = false;
1633 // The following reserved two bytes (see PS 3.5-2003, section
1634 // "7.1.2 Data element structure with explicit vr", p 27) must be
1635 // skipped before proceeding on reading the length on 4 bytes.
1637 //Fp->seekg( 2L, std::ios::cur); // Once per OB,OW,OL,UT,UN,SQ DocEntry
1638 uint32_t length32 = ReadInt32(); // Once per OB,OW,OL,UT,UN,SQ DocEntry
1639 CurrentOffsetPosition+=4;
1640 if ( (vr == "OB" || vr == "OW" || vr == "OL") && length32 == 0xffffffff )
1645 lengthOB = FindDocEntryLengthOBOrOW();// for encapsulation of encoded pixel
1647 catch ( FormatUnexpected )
1649 // Computing the length failed (this happens with broken
1650 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1651 // chance to get the pixels by deciding the element goes
1652 // until the end of the file. Hence we artificially fix the
1653 // the length and proceed.
1654 gdcmWarningMacro( " Computing the length failed for " <<
1655 entry->GetKey() <<" in " <<GetFileName());
1657 long currentPosition = Fp->tellg(); // Only for gdcm-JPEG-LossLess3a.dcm-like
1658 Fp->seekg(0L,std::ios::end); // Only for gdcm-JPEG-LossLess3a.dcm-like
1660 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition; // Only for gdcm-JPEG-LossLess3a.dcm-like
1661 Fp->seekg(currentPosition, std::ios::beg); // Only for gdcm-JPEG-LossLess3a.dcm-like
1663 entry->SetReadLength(lengthUntilEOF);
1664 entry->SetLength(lengthUntilEOF);
1667 entry->SetReadLength(lengthOB);
1668 entry->SetLength(lengthOB);
1671 FixDocEntryFoundLength(entry, length32);
1674 // Length is encoded on 2 bytes.
1675 //length16 = ReadInt16();
1676 length16 = GetInt16();
1677 // 0xffff means that we deal with 'No Length' Sequence
1678 // or 'No Length' SQItem
1679 if ( length16 == 0xffff)
1683 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1688 // Either implicit VR or a non DICOM conformal (see note below) explicit
1689 // VR that ommited the VR of (at least) this element. Farts happen.
1690 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1691 // on Data elements "Implicit and Explicit VR Data Elements shall
1692 // not coexist in a Data Set and Data Sets nested within it".]
1693 // Length is on 4 bytes.
1695 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1696 // even if Transfer Syntax is 'Implicit VR ...'
1697 // --> Except for 'Implicit VR Big Endian Transfer Syntax GE Private'
1698 // where Group 0x0002 is *also* encoded in Implicit VR !
1700 FixDocEntryFoundLength( entry, GetInt32() /*ReadInt32()*/ );
1706 * \brief Find the Length till the next sequence delimiter
1707 * \warning NOT end user intended method !
1710 uint32_t Document::FindDocEntryLengthOBOrOW()
1711 throw( FormatUnexpected )
1713 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1715 long positionOnEntry = Fp->tellg(); // Only for OB,OW DataElements
1717 bool foundSequenceDelimiter = false;
1718 uint32_t totalLength = 0;
1720 while ( !foundSequenceDelimiter )
1726 { ///\todo make sure there is never OL encoded pixel data!
1728 //group = ReadInt16(); // Once per fragment (if any) of OB,OW DataElements
1729 //elem = ReadInt16(); // Once per fragment (if any) of OB,OW DataElements
1730 ReadBegBuffer(4); // Once per fragment (if any) of OB,OW DataElements
1732 catch ( FormatError )
1734 throw FormatError("Unexpected end of file encountered during ",
1735 "Document::FindDocEntryLengthOBOrOW()");
1740 // We have to decount the group and element we just read
1742 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
1745 "Neither an Item tag nor a Sequence delimiter tag on :"
1746 << std::hex << group << "|" << elem << ") Pos. on entry was 0x(" <<positionOnEntry<< ") "
1749 Fp->seekg(positionOnEntry, std::ios::beg); // Once per fragment (if any) of OB,OW DataElements
1750 throw FormatUnexpected(
1751 "Neither an Item tag nor a Sequence delimiter tag.");
1753 if ( elem == 0xe0dd )
1755 foundSequenceDelimiter = true;
1757 uint32_t itemLength = ReadInt32(); // Once per fragment (if any) of OB,OW DataElements
1758 // We add 4 bytes since we just read the ItemLength with ReadInt32
1759 totalLength += itemLength + 4;
1760 SkipBytes(itemLength);
1762 if ( foundSequenceDelimiter )
1767 Fp->seekg( positionOnEntry, std::ios::beg); // Only once for OB,OW DataElements
1772 * \brief Find the Value Representation of the current Dicom Element.
1773 * @return Value Representation of the current Entry
1775 VRKey Document::FindDocEntryVR()
1777 if ( Filetype != ExplicitVR )
1779 return GDCM_VRUNKNOWN;
1782 // Delimiters (0xfffe), are not explicit VR ...
1783 if ( CurrentGroup == 0xfffe )
1784 return GDCM_VRUNKNOWN;
1786 //long positionOnEntry;
1787 //if( Debug::GetWarningFlag() )
1788 // positionOnEntry = Fp->tellg(); // Only in Warning Mode
1790 // Warning: we believe this is explicit VR (Value Representation) because
1791 // we used a heuristic that found "UL" in the first tag and/or
1792 // 'Transfer Syntax' told us it is.
1793 // Alas this doesn't guarantee that all the tags will be in explicit VR.
1794 // In some cases one finds implicit VR tags mixed within an explicit VR file
1796 // 'Normaly' the only case is : group 0002 Explicit, and other groups Implicit
1798 // Hence we make sure the present tag is in explicit VR and try to fix things
1799 // if it happens not to be the case.
1802 //Fp->read(&(vr[0]),(size_t)2);
1803 vr[0] = *PtrBegBuffer++;
1804 vr[1] = *PtrBegBuffer++;
1806 //if ( !CheckDocEntryVR(vr) ) // avoid useless function call
1807 if ( !Global::GetVR()->IsValidVR(vr) )
1810 gdcmWarningMacro( "Unknown VR " << vr.GetHexaRepresentation() << std::hex
1811 << " at offset : 0x(" << CurrentOffsetPosition-4
1812 << ") for group " << std::hex << CurrentGroup );
1814 //Fp->seekg(positionOnEntry, std::ios::beg); //JPRx
1815 //Fp->seekg((long)-2, std::ios::cur);// only for unrecognized VR (?!?)
1816 //see :MR_Philips_Intera_PrivateSequenceExplicitVR.dcm
1818 return GDCM_VRUNKNOWN;
1824 * \brief Check the correspondance between the VR of the header entry
1825 * and the taken VR. If they are different, the header entry is
1826 * updated with the new VR.
1827 * @param vr Dicom Value Representation
1828 * @return false if the VR is incorrect or if the VR isn't referenced
1829 * otherwise, it returns true
1831 bool Document::CheckDocEntryVR(const VRKey &vr)
1833 return Global::GetVR()->IsValidVR(vr);
1837 * \brief Skip a given Header Entry
1838 * @param entry entry to skip
1840 void Document::SkipDocEntry(DocEntry *entry)
1842 SkipBytes(entry->GetLength());
1846 * \brief Skips to the beginning of the next Header Entry
1847 * @param currentDocEntry entry to skip
1849 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1851 long l = currentDocEntry->GetReadLength();
1852 if ( (uint32_t) l == (uint32_t)-1 ) // length = 0xffff shouldn't appear here ...
1853 // ... but PMS imagers happen !
1855 Fp->seekg((size_t)(currentDocEntry->GetOffset()), std::ios::beg); //FIXME :each DocEntry
1856 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1858 Fp->seekg( l,std::ios::cur); //FIXME :each DocEntry
1863 * \brief When the length of an element value is obviously wrong (because
1864 * the parser went Jabberwocky) one can hope improving things by
1865 * applying some heuristics.
1866 * @param entry entry to check
1867 * @param foundLength first assumption about length (before bug fix, or set to zero if =0xffffffff)
1869 void Document::FixDocEntryFoundLength(DocEntry *entry,
1870 uint32_t foundLength)
1872 entry->SetReadLength( foundLength );// will be updated only if a bug is found
1874 if ( foundLength == 0xffffffff)
1877 //entry->SetLength(foundLength);
1878 entry->SetLength(0);
1879 return; // return ASAP; don't waist time on useless tests
1882 uint16_t gr = entry->GetGroup();
1883 uint16_t elem = entry->GetElement();
1885 if ( foundLength % 2)
1887 gdcmWarningMacro( "Warning : Tag (" << std::hex << gr << "|" << elem << ") with uneven length "
1888 << std::dec << foundLength << " 0x(" << std::hex << foundLength << ") "
1889 //<< " at offset x(" << offset << ")"
1893 //////// Fix for some naughty General Electric images.
1894 // Allthough not recent many such GE corrupted images are still present
1895 // on Creatis hard disks. Hence this fix shall remain when such images
1896 // are no longer in use (we are talking a few years, here)...
1897 // Note: XMedCon probably uses such a trick since it is able to read
1898 // those pesky GE images ...
1899 if ( foundLength == 13)
1901 // Only happens for this length !
1902 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1905 entry->SetReadLength(10); // a bug is to be fixed !?
1909 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1910 // Occurence of such images is quite low (unless one leaves close to a
1911 // 'Leonardo' source. Hence, one might consider commenting out the
1912 // following fix on efficiency reasons.
1913 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1915 // Ideally we should check we are in Explicit and double check
1916 // that VR=UL... this is done properly in gdcm2
1917 if( foundLength == 6 )
1919 gdcmWarningMacro( "Replacing Length from 6 into 4" );
1921 entry->SetReadLength(4); // a bug is to be fixed !
1923 else if ( foundLength%4 )
1925 gdcmErrorMacro( "This looks like to a buggy Siemens DICOM file."
1926 "The length of this tag seems to be wrong" );
1930 else if ( entry->GetVR() == "SQ" )
1932 foundLength = 0; // ReadLength is unchanged
1935 //////// We encountered a 'delimiter' element i.e. a tag of the form
1936 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1937 // taken into account.
1938 else if ( gr == 0xfffe )
1940 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1941 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1942 // causes extra troubles...
1943 if ( elem != 0x0000 )
1949 foundLength=12; // to skip the mess that follows this bugged Tag !
1952 entry->SetLength(foundLength);
1956 * \brief Apply some heuristics to predict whether the considered
1957 * element value contains/represents an integer or not.
1958 * @param entry The element value on which to apply the predicate.
1959 * @return The result of the heuristical predicate.
1961 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1963 uint16_t elem = entry->GetElement();
1964 uint16_t group = entry->GetGroup();
1965 const VRKey &vr = entry->GetVR();
1966 uint32_t length = entry->GetLength();
1968 // When we have some semantics on the element we just read, and if we
1969 // a priori know we are dealing with an integer, then we shall be
1970 // able to swap its element value properly.
1971 if ( elem == 0 ) // This is the group length of the group
1979 // Although this should never happen, still some images have a
1980 // corrupted group length [e.g. have a glance at offset x(8336) of
1981 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm.
1982 // Since for dicom compliant and well behaved headers, the present
1983 // test is useless (and might even look a bit paranoid), when we
1984 // encounter such an ill-formed image, we simply display a warning
1985 // message and proceed on parsing (while crossing fingers).
1986 long filePosition = Fp->tellg(); // Only when elem 0x0000 length is not 4 (?!?)
1988 gdcmWarningMacro( "Erroneous Group Length element length on : ("
1989 << std::hex << group << " , " << elem
1990 << ") -before- position x(" << filePosition << ")"
1991 << "lgt : " << length );
1995 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
2003 * \brief Discover what the swap code is (among little endian, big endian,
2004 * bad little endian, bad big endian).
2006 * @return false when we are absolutely sure
2007 * it's neither ACR-NEMA nor DICOM
2008 * true when we hope ours assuptions are OK
2010 bool Document::CheckSwap()
2017 // First, compare HostByteOrder and NetworkByteOrder in order to
2018 // determine if we shall need to swap bytes (i.e. the Endian type).
2019 bool net2host = Util::IsCurrentProcessorBigEndian();
2021 // The easiest case is the one of a 'true' DICOM header, we just have
2022 // to look for the string "DICM" inside the file preamble.
2025 char *entCur = deb + 128;
2026 if ( memcmp(entCur, "DICM", (size_t)4) == 0 )
2028 gdcmDebugMacro( "Looks like DICOM Version3 (preamble + DCM)" );
2030 // Group 0002 should always be VR, and the first element 0000
2031 // Let's be carefull (so many wrong headers ...)
2032 // and determine the value representation (VR) :
2033 // Let's skip to the first element (0002,0000) and check there if we find
2034 // "UL" - or "OB" if the 1st one is (0002,0001) -,
2035 // in which case we (almost) know it is explicit VR.
2036 // WARNING: if it happens to be implicit VR then what we will read
2037 // is the length of the group. If this ascii representation of this
2038 // length happens to be "UL" then we shall believe it is explicit VR.
2039 // We need to skip :
2040 // * the 128 bytes of File Preamble (often padded with zeroes),
2041 // * the 4 bytes of "DICM" string,
2042 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
2043 // i.e. a total of 136 bytes.
2046 // group 0x0002 *is always* Explicit VR Sometimes,
2047 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
2048 // *Implicit* VR (see former 'gdcmData/icone.dcm')
2050 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
2051 memcmp(entCur, "OB", (size_t)2) == 0 ||
2052 memcmp(entCur, "UI", (size_t)2) == 0 ||
2053 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
2054 // when Write DCM *adds*
2056 // Use Document::dicom_vr to test all the possibilities
2057 // instead of just checking for UL, OB and UI !? group 0000
2059 Filetype = ExplicitVR;
2060 gdcmDebugMacro( "Group 0002 : Explicit Value Representation");
2064 Filetype = ImplicitVR;
2065 gdcmWarningMacro( "Group 0002 :Not an explicit Value Representation;"
2066 << "Looks like a bugged Header!");
2069 // Here, we assume that the file IS kosher Dicom !
2070 // (The meta elements - group 0x0002 - ARE little endian !)
2074 gdcmDebugMacro( "HostByteOrder != NetworkByteOrder, SwapCode = 4321");
2079 gdcmDebugMacro( "HostByteOrder = NetworkByteOrder, SwapCode = 1234");
2082 // Position the file position indicator at first tag
2083 // (i.e. after the file preamble and the "DICM" string).
2085 Fp->seekg ( 132L, std::ios::beg); // Once per Document
2086 CurrentOffsetPosition = 132;
2088 } // ------------------------------- End of DicomV3 ----------------
2090 // Alas, this is not a DicomV3 file and whatever happens there is no file
2091 // preamble. We can reset the file position indicator to where the data
2092 // is (i.e. the beginning of the file).
2094 gdcmWarningMacro( "Not a Kosher DICOM Version3 file (no preamble)");
2096 Fp->seekg(0, std::ios::beg); // Once per ACR-NEMA Document
2097 CurrentOffsetPosition = 0;
2098 // Let's check 'No Preamble Dicom File' :
2099 // Should start with group 0x0002
2100 // and be Explicit Value Representation
2102 s16 = *((uint16_t *)(deb));
2115 if ( SwapCode != 0 )
2117 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
2118 memcmp(entCur, "OB", (size_t)2) == 0 ||
2119 memcmp(entCur, "UI", (size_t)2) == 0 ||
2120 memcmp(entCur, "SH", (size_t)2) == 0 ||
2121 memcmp(entCur, "AE", (size_t)2) == 0 ||
2122 memcmp(entCur, "OB", (size_t)2) == 0 )
2124 Filetype = ExplicitVR; // FIXME : not enough to say it's Explicit
2125 // Wait untill reading Transfer Syntax
2126 gdcmDebugMacro( "Group 0002 : Explicit Value Representation");
2130 // ------------------------------- End of 'No Preamble' DicomV3 -------------
2132 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2133 // By clean we mean that the length of the first group is written down.
2134 // If this is the case and since the length of the first group HAS to be
2135 // four (bytes), then determining the proper swap code is straightforward.
2138 // We assume the array of char we are considering contains the binary
2139 // representation of a 32 bits integer. Hence the following dirty
2141 s32 = *((uint32_t *)(entCur));
2161 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2162 // It is time for despaired wild guesses.
2163 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2164 // i.e. the 'group length' element is not present :
2166 // check the supposed-to-be 'group number'
2167 // in ( 0x0001 .. 0x0008 )
2168 // to determine ' SwapCode' value .
2169 // Only 0 or 4321 will be possible
2170 // (no oportunity to check for the formerly well known
2171 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2172 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc-3, 4, ..., 8-)
2173 // the file IS NOT ACR-NEMA nor DICOM V3
2174 // Find a trick to tell it the caller...
2176 s16 = *((uint16_t *)(deb));
2178 gdcmDebugMacro("not a DicomV3 nor a 'clean' ACR/NEMA;"
2179 << " (->despaired wild guesses !)");
2208 s16 = *((uint16_t *)(deb));
2211 s16 = *((uint16_t *)(deb+2));
2213 Fp->seekg ( 0L, std::ios::beg); // Once per Document
2214 CurrentOffsetPosition = 0;
2215 switch(s16) // try an other trick!
2216 // -> to be able to decode 0029|1010 DataElement
2217 // -> and be not less cleaver than dcmdump ;-)
2226 gdcmWarningMacro("ACR/NEMA unfound swap info (Hopeless !)");
2230 // Check if next 2 bytes are a VR
2231 // Probabely something more time-consuming exists with std::string
2232 const char VRvalues[] = "AEASATCSDADTFLFDISLOLTPNSHSLSSSTTMUIULUSUTOBOWOLOFATUNSQRT";
2234 const char *pt = VRvalues;
2235 for (int i=0;i<nbVal;i++)
2237 if(*(deb+4) == *pt++) {
2238 if(*(deb+5) == *pt++) {
2239 Filetype = ExplicitVR;
2247 Filetype = ImplicitVR;
2254 * \brief Change the Byte Swap code.
2256 void Document::SwitchByteSwapCode()
2258 gdcmDebugMacro( "Switching Byte Swap code from "<< SwapCode
2259 << " at: 0x" << std::hex << Fp->tellg() ); // Only when DEBUG
2260 if ( SwapCode == 1234 )
2264 else if ( SwapCode == 4321 )
2268 else if ( SwapCode == 3412 )
2272 else if ( SwapCode == 2143 )
2276 gdcmDebugMacro( " Into: "<< SwapCode );
2280 * \brief during parsing, Header Elements too long are not loaded in memory
2281 * @param newSize new size
2283 void Document::SetMaxSizeLoadEntry(long newSize)
2289 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2291 MaxSizeLoadEntry = 0xffffffff;
2294 MaxSizeLoadEntry = newSize;
2298 * \brief Read the next tag WITHOUT loading it's value
2299 * (read the 'Group Number', the 'Element Number',
2300 * gets the Dict Entry
2301 * gets the VR, gets the length, gets the offset value)
2302 * @return On succes : the newly created DocEntry, NULL on failure.
2304 DocEntry *Document::ReadNextDocEntry()
2308 ReadBegBuffer(8); // Avoid to many time consuming freads
2309 //CurrentGroup = ReadInt16();
2310 //CurrentElem = ReadInt16();
2312 catch ( FormatError )
2314 // We reached the EOF (or an error occured) therefore
2315 // header parsing has to be considered as finished.
2319 changeFromUN = false;
2320 CurrentGroup = GetInt16();
2321 CurrentElem = GetInt16();
2323 // In 'true DICOM' files Group 0002 is always little endian
2324 if ( HasDCMPreamble )
2326 if ( !Group0002Parsed && CurrentGroup != 0x0002) // avoid calling a function when useless
2327 HandleOutOfGroup0002(CurrentGroup, CurrentElem);
2329 // Sometimes file contains groups of tags with reversed endianess.
2330 HandleBrokenEndian(CurrentGroup, CurrentElem);
2333 VRKey vr = FindDocEntryVR();
2336 if ( vr == GDCM_VRUNKNOWN )
2338 if ( CurrentElem == 0x0000 ) // Group Length
2340 realVR = "UL"; // must be UL
2342 else if (CurrentGroup == 0xfffe) // Don't get DictEntry for Delimitors
2347 // Was commented out in order not to generate 'Shadow Groups' where some
2348 // Data Elements are Explicit VR and some other ones Implicit VR
2349 // -> Better we fix the problem at Write time
2351 else if (CurrentGroup%2 == 1 )
2353 if (CurrentElem >= 0x0010 && CurrentElem <=0x00ff )
2355 // DICOM PS 3-5 7.8.1 a) states that :
2356 // Private Creator Data Elements numbered (gggg,0010-00FF) (gggg is odd)
2357 // attributes have to be LO (Long String) and the VM shall be equal to 1
2360 // Seems not to be true
2361 // Still in gdcmtk, David Clunnie disagrees, Marco Eichelberg says it's OK ...
2362 // We let it for a while?
2363 //(We should check length==4, for more security, but we don't have it yet !)
2365 else if ( CurrentElem == 0x0001)
2367 realVR = "UL"; // Private Group Length To End
2369 else // check the private dictionary for shadow elements when Implicit VR!
2371 DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem);
2374 realVR = dictEntry->GetVR();
2375 dictEntry->Unregister(); // GetDictEntry registered it
2382 DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem);//only when ImplicitVR
2385 realVR = dictEntry->GetVR();
2386 dictEntry->Unregister(); // GetDictEntry registered it
2391 // if UN found, let's check the dictionary, and trust it!
2392 // (maybe a private dictionary exists?)
2393 else if (vr == "UN")
2395 DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem);
2398 realVR = dictEntry->GetVR();
2399 dictEntry->Unregister(); // GetDictEntry registered it
2401 // for VR = "UN", length is always stored on 4 bytes.
2402 // remember this info, in order not to crash later
2404 /// \todo : fixme If inside a vr = "UN" DataElement (but SQ according to a private dictionnary)
2405 /// there is some more vr = "UN" DataElements, it will probabely fail.
2406 /// --> find a -non time consuming- trick to store changeFromUN info at DataElement level,
2407 /// not at the Document level.
2414 //if ( Global::GetVR()->IsVROfSequence(realVR) )
2417 newEntry = NewSeqEntry(CurrentGroup, CurrentElem);
2421 newEntry = NewDataEntry(CurrentGroup, CurrentElem, realVR);
2422 static_cast<DataEntry *>(newEntry)->SetState(DataEntry::STATE_NOTLOADED);
2425 if ( vr == GDCM_VRUNKNOWN )
2427 if ( Filetype == ExplicitVR )
2429 // We thought this was explicit VR, but we end up with an
2430 // implicit VR tag. Let's backtrack.
2432 //if ( newEntry->GetGroup() != 0xfffe )
2433 if (CurrentGroup != 0xfffe)
2435 int offset = Fp->tellg();//Only when heuristic for Explicit/Implicit was wrong
2437 gdcmWarningMacro("Entry (" << newEntry->GetKey() << ") at x("
2438 << std::hex << offset << ") should be Explicit VR");
2441 newEntry->SetImplicitVR();
2446 FindDocEntryLength(newEntry);
2448 catch ( FormatError )
2455 newEntry->SetOffset(Fp->tellg()); // for each DocEntry
2460 * \brief Handle broken private tag from Philips NTSCAN
2461 * where the endianess is being switched to BigEndian
2462 * for no apparent reason
2465 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
2467 // for strange PMS Gyroscan Intera images
2468 // Item 'starter' has a tag : 0x3f3f,0x3f00, for no apparent reason
2470 // --- Feel free to remove this test *on your own coy of gdcm*
2471 // if you are sure you'll never face this problem.
2473 if ((group == 0x3f3f) && (elem == 0x3f00))
2475 // start endian swap mark for group found
2476 gdcmDebugMacro( " delimiter 0x3f3f found." );
2482 // --- End of removable code
2484 // Endian reversion.
2485 // Some files contain groups of tags with reversed endianess.
2486 static int reversedEndian = 0;
2487 // try to fix endian switching in the middle of headers
2488 if ((group == 0xfeff) && (elem == 0x00e0))
2490 // start endian swap mark for group found
2491 gdcmDebugMacro( "Start endian swap mark found." );
2493 SwitchByteSwapCode();
2498 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2500 // end of reversed endian group
2501 gdcmDebugMacro( "End of reversed endian." );
2503 SwitchByteSwapCode();
2505 else if (group == 0xfeff && elem == 0xdde0)
2507 // reversed Sequence Terminator found
2508 // probabely a bug in the header !
2509 // Do what you want, it breaks !
2511 //SwitchByteSwapCode();
2512 gdcmWarningMacro( "Should never get here! reversed Sequence Terminator!" );
2517 else if (group == 0xfffe && elem == 0xe0dd)
2519 gdcmDebugMacro( "Straight Sequence Terminator." );
2524 * \brief Group 0002 is always coded Little Endian
2525 * whatever Transfer Syntax is
2528 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2530 // Endian reversion.
2531 // Some files contain groups of tags with reversed endianess.
2533 Group0002Parsed = true;
2534 // we just came out of group 0002
2535 // if Transfer Syntax is Big Endian we have to change CheckSwap
2537 std::string ts = GetTransferSyntax();
2538 TS::SpecialType s = Global::GetTS()->GetSpecialTransferSyntax(ts);
2540 // Group 0002 is always 'Explicit ...'
2541 // even when Transfer Syntax says 'Implicit ..."
2543 if ( s == TS::ImplicitVRLittleEndian
2545 s == TS::ImplicitVRBigEndianPrivateGE
2548 Filetype = ImplicitVR;
2551 // FIXME Strangely, this works with
2552 //'Implicit VR BigEndian Transfer Syntax' (GE Private)
2554 // --> Probabely normal, since we considered we never have
2555 // to trust manufacturers.
2556 // (we often find 'Implicit VR' tag,
2557 // even when Transfer Syntax tells us it's Explicit ...
2559 // NEVER trust the meta elements!
2560 // (see what ezDICOM does ...)
2563 if ( s == TS::ExplicitVRBigEndian )
2565 gdcmDebugMacro("Transfer Syntax Name = ["
2566 << GetTransferSyntaxName() << "]" );
2567 SwitchByteSwapCode();
2568 group = SwapShort(group);
2569 elem = SwapShort(elem);
2572 //-- Broken ACR may start with a Shadow Group --
2573 // worse : some ACR-NEMA like files start 00028 group ?!?
2574 if ( !( (group >= 0x0001 && group <= 0x0008) || group == 0x0028 ) )
2576 // We trust what we see.
2577 SwitchByteSwapCode();
2578 group = SwapShort(group);
2579 elem = SwapShort(elem);
2580 // not what we where told (by meta elements) !
2581 gdcmDebugMacro("Transfer Syntax Name = ["
2582 << GetTransferSyntaxName() << "]" );
2585 /// \todo find a trick to warn user and stop processing
2587 if ( s == TS::DeflatedExplicitVRLittleEndian)
2589 gdcmWarningMacro("Transfer Syntax ["
2590 << GetTransferSyntaxName() << "] :"
2591 << " not yet dealt with ");
2595 // The following shouldn't occur very often
2596 // Let's check at the very end.
2598 if ( ts == GDCM_UNKNOWN )
2600 gdcmDebugMacro("True DICOM File, with NO Transfer Syntax (?!) " );
2604 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2606 gdcmWarningMacro("True DICOM File, with illegal Transfer Syntax: ["
2612 //-----------------------------------------------------------------------------
2615 //-----------------------------------------------------------------------------
2616 } // end namespace gdcm