1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2008/01/02 10:48:52 $
7 Version: $Revision: 1.377 $
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 void 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 gdcmDebugMacro( "Enter in ParseDES, delim-mode " << delim_mode
1117 << " at offset " << std::hex << "0x(" << offset << ")" );
1121 ///\todo FIXME : On 64 bits processors, tellg gives unexpected results after a while ?
1122 /// Probabely a bug in gdcm code somewhere (some memory erased ?)
1124 // Uncomment to track the bug
1126 if( Debug::GetDebugFlag() )
1127 std::cout << std::dec <<"(long)(Fp->tellg()) " << (long)(Fp->tellg()) // in Debug mode
1128 << std::hex << " 0x(" <<(long)(Fp->tellg()) << ")" << std::endl;
1131 // if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) // Once per DocEntry
1132 if ( !delim_mode ) // 'and then' doesn't exist in C++ :-(
1133 if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per DocEntry, when no delim mode
1138 newDocEntry = ReadNextDocEntry( );
1145 // Uncoment this cerr line to be able to 'follow' the DocEntries
1146 // when something *very* strange happens
1147 if( Debug::GetDebugFlag() )
1148 std::cerr<<newDocEntry->GetKey()<<" "<<newDocEntry->GetVR()<<std::endl;
1150 // an Item Starter found elsewhere but in the first position
1151 // of a SeqEntry means previous entry was a Sequence
1152 // but we didn't get it (private Sequence + Implicit VR)
1153 // we have to backtrack.
1154 if ( !first && newDocEntry->IsItemStarter() )
1156 // Debug message within the method !
1157 newDocEntry = Backtrack(newDocEntry, set);
1161 PreviousDocEntry = newDocEntry;
1165 newDataEntry = dynamic_cast<DataEntry*>(newDocEntry);
1169 //////////////////////////// DataEntry
1171 //vr = newDocEntry->GetVR(); // useless ?
1173 if ( !set->AddEntry( newDataEntry ) )
1175 gdcmDebugMacro( "in ParseDES : cannot add a DataEntry "
1176 << newDataEntry->GetKey()
1177 << " (at offset : 0x("
1178 << newDataEntry->GetOffset() << ") )" );
1183 newDataEntry->Delete();
1184 // Load only if we can add (not a duplicate key)
1185 LoadDocEntry( newDataEntry );
1187 if ( newDataEntry->GetElement() == 0x0000 ) // if on group length
1189 if ( newDataEntry->GetGroup()%2 != 0 ) // if Shadow Group
1191 if ( LoadMode & LD_NOSHADOW ) // if user asked to skip shad.gr
1193 std::string strLgrGroup = newDataEntry->GetString();
1196 //if ( newDataEntry->IsUnfound() ) /?!? JPR
1198 lgrGroup = atoi(strLgrGroup.c_str());
1199 Fp->seekg(lgrGroup, std::ios::cur); // Once per Shadow group, when NOSHADOW
1200 RemoveEntry( newDocEntry ); // Remove and delete
1207 bool delimitor = newDataEntry->IsItemDelimitor();
1208 bool outOfBounds = false;
1210 if ( ((long)(Fp->tellg())-offset) >= l_max ) //Once per DataEntry when no delim mode
1213 // 'and then', 'or else' don't exist in C++ :-(
1214 // if ( (delimitor) ||
1215 // (!delim_mode && ((long)(Fp->tellg())-offset) >= l_max) ) // Once per DataEntry
1217 if ( delimitor || outOfBounds )
1220 newDocEntry->Delete();
1224 // Just to make sure we are at the beginning of next entry.
1225 SkipToNextDocEntry(newDocEntry); // FIXME : once per DocEntry, segfault if commented out
1229 /////////////////////// SeqEntry : VR = "SQ"
1231 unsigned long l = newDocEntry->GetReadLength();
1232 if ( l != 0 ) // don't mess the delim_mode for 'zero-length sequence'
1234 if ( l == 0xffffffff )
1236 delim_mode_intern = true;
1240 delim_mode_intern = false;
1244 if ( (LoadMode & LD_NOSHADOWSEQ) && ! delim_mode_intern )
1246 // User asked to skip SeQuences *only* if they belong to Shadow Group
1247 if ( newDocEntry->GetGroup()%2 != 0 )
1249 Fp->seekg( l, std::ios::cur); // once per SQITEM, when NOSHADOWSEQ
1250 newDocEntry->Delete(); // Delete, not in the set
1254 if ( (LoadMode & LD_NOSEQ) && ! delim_mode_intern )
1256 // User asked to skip *any* SeQuence
1257 Fp->seekg( l, std::ios::cur); // Once per SQ, when NOSEQ
1258 newDocEntry->Delete(); // Delete, not in the set
1261 // delay the dynamic cast as late as possible
1262 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
1264 // no other way to create the Delimitor ...
1265 newSeqEntry->SetDelimitorMode( delim_mode_intern );
1267 // At the top of the hierarchy, stands a Document. When "set"
1268 // is a Document, then we are building the first depth level.
1269 // Hence the SeqEntry we are building simply has a depth
1271 if ( set == this ) // ( dynamic_cast< Document* > ( set ) )
1273 newSeqEntry->SetDepthLevel( 1 );
1275 // But when "set" is already a SQItem, we are building a nested
1276 // sequence, and hence the depth level of the new SeqEntry
1277 // we are building, is one level deeper:
1279 // time waste hunting
1280 else if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
1282 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
1286 { // Don't try to parse zero-length sequences
1288 gdcmDebugMacro( "Entry in ParseSQ, delim " << delim_mode_intern
1289 << " at offset 0x(" << std::hex
1290 << newDocEntry->GetOffset() << ")");
1292 bool res = ParseSQ( newSeqEntry,
1293 newDocEntry->GetOffset(),
1294 l, delim_mode_intern);
1296 gdcmDebugMacro( "Exit from ParseSQ, delim " << delim_mode_intern << " -->return : " << res);
1298 if ( !set->AddEntry( newSeqEntry ) )
1300 gdcmWarningMacro( "in ParseDES : cannot add a SeqEntry "
1301 << newSeqEntry->GetKey()
1302 << " (at offset : 0x("
1303 << newSeqEntry->GetOffset() << ") )" );
1308 newDocEntry->Delete();
1311 // if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max) // Once per SeqEntry
1313 if ( !delim_mode ) // 'and then' doesn't exist in C++ :-(
1314 if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per SeqEntry when no delim mode
1318 newDocEntry->Delete();
1321 } // end SeqEntry : VR = "SQ"
1325 newDocEntry->Delete();
1329 if (UnexpectedEOF) // some terminator was missing
1332 gdcmDebugMacro( "Exit from ParseDES, delim-mode " << delim_mode );
1336 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
1337 * @return false if expected fff0,e000 not found
1339 bool Document::ParseSQ( SeqEntry *seqEntry,
1340 long offset, long l_max, bool delim_mode)
1342 int SQItemNumber = 0;
1344 long offsetStartCurrentSQItem = offset;
1348 // the first time, we read the fff0,e000 of the first SQItem
1349 DocEntry *newDocEntry = ReadNextDocEntry();
1353 // The most frequent is when a SQ terminator is missing (?!?)
1354 gdcmWarningMacro("in ParseSQ : should never get here!");
1355 UnexpectedEOF = true;
1360 if ( newDocEntry->IsSequenceDelimitor() )
1362 seqEntry->SetDelimitationItem( newDocEntry );
1363 newDocEntry->Delete();
1367 else // ! delim_mode
1369 if ( ((long)(Fp->tellg())-offset) >= l_max) // Once per SQItem when no delim mode
1371 newDocEntry->Delete();
1375 // create the current SQItem
1376 SQItem *itemSQ = SQItem::New( seqEntry->GetDepthLevel() );
1377 unsigned int l = newDocEntry->GetReadLength();
1379 if ( l == 0xffffffff )
1388 // avoid infinite loop when Bad assumption was made on illegal 'unknown length' UN //JPRx
1390 if (offsetStartCurrentSQItem <= OffsetOfPreviousParseDES)
1392 gdcmWarningMacro("Bad assumption was made on illegal 'unknown length' UN!");
1393 gdcmWarningMacro("OffsetOfPreviousParseDES " << std::hex << OffsetOfPreviousParseDES
1394 << " offsetStartCurrentSQItem " << offsetStartCurrentSQItem);
1395 /// \todo when "Bad assumption (SQ) on illegal 'unknown length' UN", Backtrack again + try OB
1400 OffsetOfPreviousParseDES = offsetStartCurrentSQItem;
1403 // fill up the current SQItem, starting at the beginning of fff0,e000
1405 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg); // Once per SQItem
1406 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
1407 offsetStartCurrentSQItem = Fp->tellg(); // Once per SQItem
1409 seqEntry->AddSQItem( itemSQ, SQItemNumber );
1411 newDocEntry->Delete();
1413 //if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max ) //JPRx
1414 if ( !delim_mode && (offsetStartCurrentSQItem-offset ) >= l_max )
1423 * \brief When a private Sequence + Implicit VR is encountered
1424 * we cannot guess it's a Sequence till we find the first
1425 * Item Starter. We then backtrack to do the job.
1426 * @param docEntry Item Starter that warned us
1428 DocEntry *Document::Backtrack(DocEntry *docEntry, DocEntrySet *set)
1430 // delete the Item Starter, built erroneously out of any Sequence
1431 // it's not yet in the HTable/chained list
1434 // Get all info we can from PreviousDocEntry
1435 uint16_t group = PreviousDocEntry->GetGroup();
1436 uint16_t elem = PreviousDocEntry->GetElement();
1437 uint32_t lgt = PreviousDocEntry->GetLength();
1438 long offset = PreviousDocEntry->GetOffset();
1440 gdcmDebugMacro( "Backtrack :" << std::hex << group
1442 << " at offset 0x(" <<offset << ")" );
1444 set->RemoveEntry( PreviousDocEntry );
1446 // forge the Seq Entry
1447 DocEntry *newEntry = NewSeqEntry(group, elem);
1448 newEntry->SetLength(lgt);
1449 newEntry->SetOffset(offset);
1451 // Move back to the beginning of the Sequence
1453 Fp->seekg(offset, std::ios::beg); // Only for Shadow Implicit VR SQ
1454 return newEntry; // It will added where it has to be!
1458 * \brief Loads (or not) the element content depending if its length exceeds
1459 * or not the value specified with Document::SetMaxSizeLoadEntry()
1460 * @param entry Header Entry (Dicom Element) to be dealt with
1461 * @param forceLoad whether you want to force loading of 'long' elements
1463 void Document::LoadDocEntry(DocEntry *entry, bool forceLoad)
1465 uint16_t group = entry->GetGroup();
1466 uint16_t elem = entry->GetElement();
1467 const VRKey &vr = entry->GetVR();
1468 uint32_t length = entry->GetLength();
1470 // Fp->seekg((long)entry->GetOffset(), std::ios::beg); // JPRx
1472 // A SeQuence "contains" a set of Elements.
1473 // (fffe e000) tells us an Element is beginning
1474 // (fffe e00d) tells us an Element just ended
1475 // (fffe e0dd) tells us the current SeQuence just ended
1476 // (fffe 0000) is an 'impossible' tag value,
1477 // found in MR-PHILIPS-16-Multi-Seq.dcm
1479 if ( (group == 0xfffe && elem != 0x0000 ) || vr == "SQ" )
1481 // NO more value field for SQ !
1485 DataEntry *dataEntryPtr = dynamic_cast< DataEntry* >(entry);
1491 // When the length is zero things are easy:
1494 dataEntryPtr->SetBinArea(NULL,true);
1498 // The elements whose length is bigger than the specified upper bound
1503 if (length > MaxSizeLoadEntry)
1505 dataEntryPtr->SetBinArea(NULL,true);
1506 dataEntryPtr->SetState(DataEntry::STATE_NOTLOADED);
1508 // to be sure we are at the end of the value ...
1509 // Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
1510 // std::ios::beg); //JPRx
1515 /// \todo: a method that *doesn't* load anything (maybe with MaxSizeLoadEntry=0 ?)
1516 /// + a ForceLoad call on the +/- 20 'usefull' fields
1517 /// Allow user to tell the fields he wants to ForceLoad
1518 /// during initial stage.
1519 /// Later, a GetString or GetBinArea will load the value from disk, if not loaded
1520 /// + a method that load *everything* that's not yet loaded
1522 LoadEntryBinArea(dataEntryPtr); // last one, not to erase length !
1526 * \brief Find the value Length of the passed Doc Entry
1527 * @param entry Header Entry whose length of the value shall be loaded.
1529 void Document::FindDocEntryLength( DocEntry *entry )
1530 throw ( FormatError )
1532 const VRKey &vr = entry->GetVR();
1534 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1539 // For some images, length of UN elements is coded on 2 bytes (instead of 4)
1540 // There are *not* readable !
1541 // You can make a quick and dirty patch, commenting out
1543 // in the following line.
1544 // (the 'straight' images will no longer be readable ...)
1546 if ( vr == "OB" || vr == "OW" || vr == "OL" || vr == "SQ" || vr == "UT"
1547 || vr == "UN" || changeFromUN == true)
1549 changeFromUN = false;
1550 // The following reserved two bytes (see PS 3.5-2003, section
1551 // "7.1.2 Data element structure with explicit vr", p 27) must be
1552 // skipped before proceeding on reading the length on 4 bytes.
1554 //Fp->seekg( 2L, std::ios::cur); // Once per OB,OW,OL,UT,UN,SQ DocEntry
1555 uint32_t length32 = ReadInt32(); // Once per OB,OW,OL,UT,UN,SQ DocEntry
1556 CurrentOffsetPosition+=4;
1557 if ( (vr == "OB" || vr == "OW" || vr == "OL") && length32 == 0xffffffff )
1562 lengthOB = FindDocEntryLengthOBOrOW();// for encapsulation of encoded pixel
1564 catch ( FormatUnexpected )
1566 // Computing the length failed (this happens with broken
1567 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1568 // chance to get the pixels by deciding the element goes
1569 // until the end of the file. Hence we artificially fix the
1570 // the length and proceed.
1571 gdcmWarningMacro( " Computing the length failed for " <<
1572 entry->GetKey() <<" in " <<GetFileName());
1574 long currentPosition = Fp->tellg(); // Only for gdcm-JPEG-LossLess3a.dcm-like
1575 Fp->seekg(0L,std::ios::end); // Only for gdcm-JPEG-LossLess3a.dcm-like
1577 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition; // Only for gdcm-JPEG-LossLess3a.dcm-like
1578 Fp->seekg(currentPosition, std::ios::beg); // Only for gdcm-JPEG-LossLess3a.dcm-like
1580 entry->SetReadLength(lengthUntilEOF);
1581 entry->SetLength(lengthUntilEOF);
1584 entry->SetReadLength(lengthOB);
1585 entry->SetLength(lengthOB);
1588 FixDocEntryFoundLength(entry, length32);
1591 // Length is encoded on 2 bytes.
1592 //length16 = ReadInt16();
1593 length16 = GetInt16();
1594 // 0xffff means that we deal with 'No Length' Sequence
1595 // or 'No Length' SQItem
1596 if ( length16 == 0xffff)
1600 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1605 // Either implicit VR or a non DICOM conformal (see note below) explicit
1606 // VR that ommited the VR of (at least) this element. Farts happen.
1607 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1608 // on Data elements "Implicit and Explicit VR Data Elements shall
1609 // not coexist in a Data Set and Data Sets nested within it".]
1610 // Length is on 4 bytes.
1612 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1613 // even if Transfer Syntax is 'Implicit VR ...'
1614 // --> Except for 'Implicit VR Big Endian Transfer Syntax GE Private'
1615 // where Group 0x0002 is *also* encoded in Implicit VR !
1617 FixDocEntryFoundLength( entry, GetInt32() /*ReadInt32()*/ );
1623 * \brief Find the Length till the next sequence delimiter
1624 * \warning NOT end user intended method !
1627 uint32_t Document::FindDocEntryLengthOBOrOW()
1628 throw( FormatUnexpected )
1630 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1632 long positionOnEntry = Fp->tellg(); // Only for OB,OW DataElements
1634 bool foundSequenceDelimiter = false;
1635 uint32_t totalLength = 0;
1637 while ( !foundSequenceDelimiter )
1643 { ///\todo make sure there is never OL encoded pixel data!
1645 //group = ReadInt16(); // Once per fragment (if any) of OB,OW DataElements
1646 //elem = ReadInt16(); // Once per fragment (if any) of OB,OW DataElements
1647 ReadBegBuffer(4); // Once per fragment (if any) of OB,OW DataElements
1649 catch ( FormatError )
1651 throw FormatError("Unexpected end of file encountered during ",
1652 "Document::FindDocEntryLengthOBOrOW()");
1657 // We have to decount the group and element we just read
1659 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
1662 "Neither an Item tag nor a Sequence delimiter tag on :"
1663 << std::hex << group << " , " << elem
1666 Fp->seekg(positionOnEntry, std::ios::beg); // Once per fragment (if any) of OB,OW DataElements
1667 throw FormatUnexpected(
1668 "Neither an Item tag nor a Sequence delimiter tag.");
1670 if ( elem == 0xe0dd )
1672 foundSequenceDelimiter = true;
1674 uint32_t itemLength = ReadInt32(); // Once per fragment (if any) of OB,OW DataElements
1675 // We add 4 bytes since we just read the ItemLength with ReadInt32
1676 totalLength += itemLength + 4;
1677 SkipBytes(itemLength);
1679 if ( foundSequenceDelimiter )
1684 Fp->seekg( positionOnEntry, std::ios::beg); // Only once for OB,OW DataElements
1689 * \brief Find the Value Representation of the current Dicom Element.
1690 * @return Value Representation of the current Entry
1692 VRKey Document::FindDocEntryVR()
1694 if ( Filetype != ExplicitVR )
1696 return GDCM_VRUNKNOWN;
1699 // Delimiters (0xfffe), are not explicit VR ...
1700 if ( CurrentGroup == 0xfffe )
1701 return GDCM_VRUNKNOWN;
1703 //long positionOnEntry;
1704 //if( Debug::GetWarningFlag() )
1705 // positionOnEntry = Fp->tellg(); // Only in Warning Mode
1707 // Warning: we believe this is explicit VR (Value Representation) because
1708 // we used a heuristic that found "UL" in the first tag and/or
1709 // 'Transfer Syntax' told us it is.
1710 // Alas this doesn't guarantee that all the tags will be in explicit VR.
1711 // In some cases one finds implicit VR tags mixed within an explicit VR file
1713 // 'Normaly' the only case is : group 0002 Explicit, and other groups Implicit
1715 // Hence we make sure the present tag is in explicit VR and try to fix things
1716 // if it happens not to be the case.
1719 //Fp->read(&(vr[0]),(size_t)2);
1720 vr[0] = *PtrBegBuffer++;
1721 vr[1] = *PtrBegBuffer++;
1723 //if ( !CheckDocEntryVR(vr) ) // avoid useless function call
1724 if ( !Global::GetVR()->IsValidVR(vr) )
1727 gdcmWarningMacro( "Unknown VR " << vr.GetHexaRepresentation()
1728 << " at offset : 0x(" << CurrentOffsetPosition-4
1729 << ") for group " << std::hex << CurrentGroup );
1731 //Fp->seekg(positionOnEntry, std::ios::beg); //JPRx
1732 //Fp->seekg((long)-2, std::ios::cur);// only for unrecognized VR (?!?)
1733 //see :MR_Philips_Intera_PrivateSequenceExplicitVR.dcm
1735 return GDCM_VRUNKNOWN;
1741 * \brief Check the correspondance between the VR of the header entry
1742 * and the taken VR. If they are different, the header entry is
1743 * updated with the new VR.
1744 * @param vr Dicom Value Representation
1745 * @return false if the VR is incorrect or if the VR isn't referenced
1746 * otherwise, it returns true
1748 bool Document::CheckDocEntryVR(const VRKey &vr)
1750 return Global::GetVR()->IsValidVR(vr);
1754 * \brief Skip a given Header Entry
1755 * @param entry entry to skip
1757 void Document::SkipDocEntry(DocEntry *entry)
1759 SkipBytes(entry->GetLength());
1763 * \brief Skips to the beginning of the next Header Entry
1764 * @param currentDocEntry entry to skip
1766 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1768 long l = currentDocEntry->GetReadLength();
1769 if ( (uint32_t) l == (uint32_t)-1 ) // length = 0xffff shouldn't appear here ...
1770 // ... but PMS imagers happen !
1772 Fp->seekg((size_t)(currentDocEntry->GetOffset()), std::ios::beg); //FIXME :each DocEntry
1773 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1775 Fp->seekg( l,std::ios::cur); //FIXME :each DocEntry
1780 * \brief When the length of an element value is obviously wrong (because
1781 * the parser went Jabberwocky) one can hope improving things by
1782 * applying some heuristics.
1783 * @param entry entry to check
1784 * @param foundLength first assumption about length (before bug fix, or set to zero if =0xffffffff)
1786 void Document::FixDocEntryFoundLength(DocEntry *entry,
1787 uint32_t foundLength)
1789 entry->SetReadLength( foundLength );// will be updated only if a bug is found
1791 if ( foundLength == 0xffffffff)
1794 //entry->SetLength(foundLength);
1795 entry->SetLength(0);
1796 return; // return ASAP; don't waist time on useless tests
1799 uint16_t gr = entry->GetGroup();
1800 uint16_t elem = entry->GetElement();
1802 if ( foundLength % 2)
1804 gdcmWarningMacro( "Warning : Tag (" << std::hex << gr << "|" << elem << ") with uneven length "
1805 << std::dec << foundLength << " 0x(" << std::hex << foundLength << ") "
1806 //<< " at offset x(" << offset << ")"
1810 //////// Fix for some naughty General Electric images.
1811 // Allthough not recent many such GE corrupted images are still present
1812 // on Creatis hard disks. Hence this fix shall remain when such images
1813 // are no longer in use (we are talking a few years, here)...
1814 // Note: XMedCon probably uses such a trick since it is able to read
1815 // those pesky GE images ...
1816 if ( foundLength == 13)
1818 // Only happens for this length !
1819 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1822 entry->SetReadLength(10); // a bug is to be fixed !?
1826 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1827 // Occurence of such images is quite low (unless one leaves close to a
1828 // 'Leonardo' source. Hence, one might consider commenting out the
1829 // following fix on efficiency reasons.
1830 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1832 // Ideally we should check we are in Explicit and double check
1833 // that VR=UL... this is done properly in gdcm2
1834 if( foundLength == 6 )
1836 gdcmWarningMacro( "Replacing Length from 6 into 4" );
1838 entry->SetReadLength(4); // a bug is to be fixed !
1840 else if ( foundLength%4 )
1842 gdcmErrorMacro( "This looks like to a buggy Siemens DICOM file."
1843 "The length of this tag seems to be wrong" );
1847 else if ( entry->GetVR() == "SQ" )
1849 foundLength = 0; // ReadLength is unchanged
1852 //////// We encountered a 'delimiter' element i.e. a tag of the form
1853 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1854 // taken into account.
1855 else if ( gr == 0xfffe )
1857 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1858 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1859 // causes extra troubles...
1860 if ( elem != 0x0000 )
1866 foundLength=12; // to skip the mess that follows this bugged Tag !
1869 entry->SetLength(foundLength);
1873 * \brief Apply some heuristics to predict whether the considered
1874 * element value contains/represents an integer or not.
1875 * @param entry The element value on which to apply the predicate.
1876 * @return The result of the heuristical predicate.
1878 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1880 uint16_t elem = entry->GetElement();
1881 uint16_t group = entry->GetGroup();
1882 const VRKey &vr = entry->GetVR();
1883 uint32_t length = entry->GetLength();
1885 // When we have some semantics on the element we just read, and if we
1886 // a priori know we are dealing with an integer, then we shall be
1887 // able to swap its element value properly.
1888 if ( elem == 0 ) // This is the group length of the group
1896 // Although this should never happen, still some images have a
1897 // corrupted group length [e.g. have a glance at offset x(8336) of
1898 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm.
1899 // Since for dicom compliant and well behaved headers, the present
1900 // test is useless (and might even look a bit paranoid), when we
1901 // encounter such an ill-formed image, we simply display a warning
1902 // message and proceed on parsing (while crossing fingers).
1903 long filePosition = Fp->tellg(); // Only when elem 0x0000 length is not 4 (?!?)
1905 gdcmWarningMacro( "Erroneous Group Length element length on : ("
1906 << std::hex << group << " , " << elem
1907 << ") -before- position x(" << filePosition << ")"
1908 << "lgt : " << length );
1912 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1920 * \brief Discover what the swap code is (among little endian, big endian,
1921 * bad little endian, bad big endian).
1923 * @return false when we are absolutely sure
1924 * it's neither ACR-NEMA nor DICOM
1925 * true when we hope ours assuptions are OK
1927 bool Document::CheckSwap()
1934 // First, compare HostByteOrder and NetworkByteOrder in order to
1935 // determine if we shall need to swap bytes (i.e. the Endian type).
1936 bool net2host = Util::IsCurrentProcessorBigEndian();
1938 // The easiest case is the one of a 'true' DICOM header, we just have
1939 // to look for the string "DICM" inside the file preamble.
1942 char *entCur = deb + 128;
1943 if ( memcmp(entCur, "DICM", (size_t)4) == 0 )
1945 gdcmDebugMacro( "Looks like DICOM Version3 (preamble + DCM)" );
1947 // Group 0002 should always be VR, and the first element 0000
1948 // Let's be carefull (so many wrong headers ...)
1949 // and determine the value representation (VR) :
1950 // Let's skip to the first element (0002,0000) and check there if we find
1951 // "UL" - or "OB" if the 1st one is (0002,0001) -,
1952 // in which case we (almost) know it is explicit VR.
1953 // WARNING: if it happens to be implicit VR then what we will read
1954 // is the length of the group. If this ascii representation of this
1955 // length happens to be "UL" then we shall believe it is explicit VR.
1956 // We need to skip :
1957 // * the 128 bytes of File Preamble (often padded with zeroes),
1958 // * the 4 bytes of "DICM" string,
1959 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
1960 // i.e. a total of 136 bytes.
1963 // group 0x0002 *is always* Explicit VR Sometimes,
1964 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
1965 // *Implicit* VR (see former 'gdcmData/icone.dcm')
1967 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
1968 memcmp(entCur, "OB", (size_t)2) == 0 ||
1969 memcmp(entCur, "UI", (size_t)2) == 0 ||
1970 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
1971 // when Write DCM *adds*
1973 // Use Document::dicom_vr to test all the possibilities
1974 // instead of just checking for UL, OB and UI !? group 0000
1976 Filetype = ExplicitVR;
1977 gdcmDebugMacro( "Group 0002 : Explicit Value Representation");
1981 Filetype = ImplicitVR;
1982 gdcmWarningMacro( "Group 0002 :Not an explicit Value Representation;"
1983 << "Looks like a bugged Header!");
1986 // Here, we assume that the file IS kosher Dicom !
1987 // (The meta elements - group 0x0002 - ARE little endian !)
1991 gdcmDebugMacro( "HostByteOrder != NetworkByteOrder, SwapCode = 4321");
1996 gdcmDebugMacro( "HostByteOrder = NetworkByteOrder, SwapCode = 1234");
1999 // Position the file position indicator at first tag
2000 // (i.e. after the file preamble and the "DICM" string).
2002 Fp->seekg ( 132L, std::ios::beg); // Once per Document
2003 CurrentOffsetPosition = 132;
2005 } // ------------------------------- End of DicomV3 ----------------
2007 // Alas, this is not a DicomV3 file and whatever happens there is no file
2008 // preamble. We can reset the file position indicator to where the data
2009 // is (i.e. the beginning of the file).
2011 gdcmWarningMacro( "Not a Kosher DICOM Version3 file (no preamble)");
2013 Fp->seekg(0, std::ios::beg); // Once per ACR-NEMA Document
2014 CurrentOffsetPosition = 0;
2015 // Let's check 'No Preamble Dicom File' :
2016 // Should start with group 0x0002
2017 // and be Explicit Value Representation
2019 s16 = *((uint16_t *)(deb));
2032 if ( SwapCode != 0 )
2034 if ( memcmp(entCur, "UL", (size_t)2) == 0 ||
2035 memcmp(entCur, "OB", (size_t)2) == 0 ||
2036 memcmp(entCur, "UI", (size_t)2) == 0 ||
2037 memcmp(entCur, "SH", (size_t)2) == 0 ||
2038 memcmp(entCur, "AE", (size_t)2) == 0 ||
2039 memcmp(entCur, "OB", (size_t)2) == 0 )
2041 Filetype = ExplicitVR; // FIXME : not enough to say it's Explicit
2042 // Wait untill reading Transfer Syntax
2043 gdcmDebugMacro( "Group 0002 : Explicit Value Representation");
2047 // ------------------------------- End of 'No Preamble' DicomV3 -------------
2049 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
2050 // By clean we mean that the length of the first group is written down.
2051 // If this is the case and since the length of the first group HAS to be
2052 // four (bytes), then determining the proper swap code is straightforward.
2055 // We assume the array of char we are considering contains the binary
2056 // representation of a 32 bits integer. Hence the following dirty
2058 s32 = *((uint32_t *)(entCur));
2078 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
2079 // It is time for despaired wild guesses.
2080 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
2081 // i.e. the 'group length' element is not present :
2083 // check the supposed-to-be 'group number'
2084 // in ( 0x0001 .. 0x0008 )
2085 // to determine ' SwapCode' value .
2086 // Only 0 or 4321 will be possible
2087 // (no oportunity to check for the formerly well known
2088 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
2089 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc-3, 4, ..., 8-)
2090 // the file IS NOT ACR-NEMA nor DICOM V3
2091 // Find a trick to tell it the caller...
2093 s16 = *((uint16_t *)(deb));
2095 gdcmDebugMacro("not a DicomV3 nor a 'clean' ACR/NEMA;"
2096 << " (->despaired wild guesses !)");
2125 s16 = *((uint16_t *)(deb));
2128 s16 = *((uint16_t *)(deb+2));
2130 Fp->seekg ( 0L, std::ios::beg); // Once per Document
2131 CurrentOffsetPosition = 0;
2132 switch(s16) // try an other trick!
2133 // -> to be able to decode 0029|1010 DataElement
2134 // -> and be not less cleaver than dcmdump ;-)
2143 gdcmWarningMacro("ACR/NEMA unfound swap info (Hopeless !)");
2147 // Check if next 2 bytes are a VR
2148 // Probabely something more time-consuming exists with std::string
2149 const char VRvalues[] = "AEASATCSDADTFLFDISLOLTPNSHSLSSSTTMUIULUSUTOBOWOLOFATUNSQRT";
2151 const char *pt = VRvalues;
2152 for (int i=0;i<nbVal;i++)
2154 if(*(deb+4) == *pt++) {
2155 if(*(deb+5) == *pt++) {
2156 Filetype = ExplicitVR;
2164 Filetype = ImplicitVR;
2171 * \brief Change the Byte Swap code.
2173 void Document::SwitchByteSwapCode()
2175 gdcmDebugMacro( "Switching Byte Swap code from "<< SwapCode
2176 << " at: 0x" << std::hex << Fp->tellg() ); // Only when DEBUG
2177 if ( SwapCode == 1234 )
2181 else if ( SwapCode == 4321 )
2185 else if ( SwapCode == 3412 )
2189 else if ( SwapCode == 2143 )
2193 gdcmDebugMacro( " Into: "<< SwapCode );
2197 * \brief during parsing, Header Elements too long are not loaded in memory
2198 * @param newSize new size
2200 void Document::SetMaxSizeLoadEntry(long newSize)
2206 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
2208 MaxSizeLoadEntry = 0xffffffff;
2211 MaxSizeLoadEntry = newSize;
2215 * \brief Read the next tag WITHOUT loading it's value
2216 * (read the 'Group Number', the 'Element Number',
2217 * gets the Dict Entry
2218 * gets the VR, gets the length, gets the offset value)
2219 * @return On succes : the newly created DocEntry, NULL on failure.
2221 DocEntry *Document::ReadNextDocEntry()
2225 ReadBegBuffer(8); // Avoid to many time consuming freads
2226 //CurrentGroup = ReadInt16();
2227 //CurrentElem = ReadInt16();
2229 catch ( FormatError )
2231 // We reached the EOF (or an error occured) therefore
2232 // header parsing has to be considered as finished.
2236 changeFromUN = false;
2237 CurrentGroup = GetInt16();
2238 CurrentElem = GetInt16();
2240 // In 'true DICOM' files Group 0002 is always little endian
2241 if ( HasDCMPreamble )
2243 if ( !Group0002Parsed && CurrentGroup != 0x0002) // avoid calling a function when useless
2244 HandleOutOfGroup0002(CurrentGroup, CurrentElem);
2246 // Sometimes file contains groups of tags with reversed endianess.
2247 HandleBrokenEndian(CurrentGroup, CurrentElem);
2250 VRKey vr = FindDocEntryVR();
2253 if ( vr == GDCM_VRUNKNOWN )
2255 if ( CurrentElem == 0x0000 ) // Group Length
2257 realVR = "UL"; // must be UL
2259 else if (CurrentGroup == 0xfffe) // Don't get DictEntry for Delimitors
2264 // Was commented out in order not to generate 'Shadow Groups' where some
2265 // Data Elements are Explicit VR and some other ones Implicit VR
2266 // -> Better we fix the problem at Write time
2268 else if (CurrentGroup%2 == 1 )
2270 if (CurrentElem >= 0x0010 && CurrentElem <=0x00ff )
2272 // DICOM PS 3-5 7.8.1 a) states that :
2273 // Private Creator Data Elements numbered (gggg,0010-00FF) (gggg is odd)
2274 // attributes have to be LO (Long String) and the VM shall be equal to 1
2277 // Seems not to be true
2278 // Still in gdcmtk, David Clunnie disagrees, Marco Eichelberg says it's OK ...
2279 // We let it for a while?
2280 //(We should check length==4, for more security, but we don't have it yet !)
2282 else if ( CurrentElem == 0x0001)
2284 realVR = "UL"; // Private Group Length To End
2286 else // check the private dictionary for shadow elements when Implicit VR!
2288 DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem);
2291 realVR = dictEntry->GetVR();
2292 dictEntry->Unregister(); // GetDictEntry registered it
2299 DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem);//only when ImplicitVR
2302 realVR = dictEntry->GetVR();
2303 dictEntry->Unregister(); // GetDictEntry registered it
2308 // if UN found, let's check the dictionary, and trust it!
2309 // (maybe a private dictionary exists?)
2310 else if (vr == "UN")
2312 DictEntry *dictEntry = GetDictEntry(CurrentGroup,CurrentElem);
2315 realVR = dictEntry->GetVR();
2316 dictEntry->Unregister(); // GetDictEntry registered it
2318 // for VR = "UN", length is always stored on 4 bytes.
2319 // remember this info, in order not to crash later
2321 /// \todo : fixme If inside a vr = "UN" DataElement (but SQ according to a private dictionnary)
2322 /// there is some more vr = "UN" DataElements, it will probabely fail.
2323 /// --> find a -non time consuming- trick to store changeFromUN info at DataElement level,
2324 /// not at the Document level.
2331 //if ( Global::GetVR()->IsVROfSequence(realVR) )
2334 newEntry = NewSeqEntry(CurrentGroup, CurrentElem);
2338 newEntry = NewDataEntry(CurrentGroup, CurrentElem, realVR);
2339 static_cast<DataEntry *>(newEntry)->SetState(DataEntry::STATE_NOTLOADED);
2342 if ( vr == GDCM_VRUNKNOWN )
2344 if ( Filetype == ExplicitVR )
2346 // We thought this was explicit VR, but we end up with an
2347 // implicit VR tag. Let's backtrack.
2349 //if ( newEntry->GetGroup() != 0xfffe )
2350 if (CurrentGroup != 0xfffe)
2352 int offset = Fp->tellg();//Only when heuristic for Explicit/Implicit was wrong
2354 gdcmWarningMacro("Entry (" << newEntry->GetKey() << ") at x("
2355 << std::hex << offset << ") should be Explicit VR");
2358 newEntry->SetImplicitVR();
2363 FindDocEntryLength(newEntry);
2365 catch ( FormatError )
2372 newEntry->SetOffset(Fp->tellg()); // for each DocEntry
2377 * \brief Handle broken private tag from Philips NTSCAN
2378 * where the endianess is being switched to BigEndian
2379 * for no apparent reason
2382 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
2384 // for strange PMS Gyroscan Intera images
2385 // Item 'starter' has a tag : 0x3f3f,0x3f00, for no apparent reason
2387 // --- Feel free to remove this test *on your own coy of gdcm*
2388 // if you are sure you'll never face this problem.
2390 if ((group == 0x3f3f) && (elem == 0x3f00))
2392 // start endian swap mark for group found
2393 gdcmDebugMacro( " delimiter 0x3f3f found." );
2399 // --- End of removable code
2401 // Endian reversion.
2402 // Some files contain groups of tags with reversed endianess.
2403 static int reversedEndian = 0;
2404 // try to fix endian switching in the middle of headers
2405 if ((group == 0xfeff) && (elem == 0x00e0))
2407 // start endian swap mark for group found
2408 gdcmDebugMacro( "Start endian swap mark found." );
2410 SwitchByteSwapCode();
2415 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
2417 // end of reversed endian group
2418 gdcmDebugMacro( "End of reversed endian." );
2420 SwitchByteSwapCode();
2422 else if (group == 0xfeff && elem == 0xdde0)
2424 // reversed Sequence Terminator found
2425 // probabely a bug in the header !
2426 // Do what you want, it breaks !
2428 //SwitchByteSwapCode();
2429 gdcmWarningMacro( "Should never get here! reversed Sequence Terminator!" );
2434 else if (group == 0xfffe && elem == 0xe0dd)
2436 gdcmDebugMacro( "Straight Sequence Terminator." );
2441 * \brief Group 0002 is always coded Little Endian
2442 * whatever Transfer Syntax is
2445 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
2447 // Endian reversion.
2448 // Some files contain groups of tags with reversed endianess.
2450 Group0002Parsed = true;
2451 // we just came out of group 0002
2452 // if Transfer Syntax is Big Endian we have to change CheckSwap
2454 std::string ts = GetTransferSyntax();
2455 TS::SpecialType s = Global::GetTS()->GetSpecialTransferSyntax(ts);
2457 // Group 0002 is always 'Explicit ...'
2458 // even when Transfer Syntax says 'Implicit ..."
2460 if ( s == TS::ImplicitVRLittleEndian
2462 s == TS::ImplicitVRBigEndianPrivateGE
2465 Filetype = ImplicitVR;
2468 // FIXME Strangely, this works with
2469 //'Implicit VR BigEndian Transfer Syntax' (GE Private)
2471 // --> Probabely normal, since we considered we never have
2472 // to trust manufacturers.
2473 // (we often find 'Implicit VR' tag,
2474 // even when Transfer Syntax tells us it's Explicit ...
2476 // NEVER trust the meta elements!
2477 // (see what ezDICOM does ...)
2480 if ( s == TS::ExplicitVRBigEndian )
2482 gdcmDebugMacro("Transfer Syntax Name = ["
2483 << GetTransferSyntaxName() << "]" );
2484 SwitchByteSwapCode();
2485 group = SwapShort(group);
2486 elem = SwapShort(elem);
2489 //-- Broken ACR may start with a Shadow Group --
2490 // worse : some ACR-NEMA like files start 00028 group ?!?
2491 if ( !( (group >= 0x0001 && group <= 0x0008) || group == 0x0028 ) )
2493 // We trust what we see.
2494 SwitchByteSwapCode();
2495 group = SwapShort(group);
2496 elem = SwapShort(elem);
2497 // not what we where told (by meta elements) !
2498 gdcmDebugMacro("Transfer Syntax Name = ["
2499 << GetTransferSyntaxName() << "]" );
2502 /// \todo find a trick to warn user and stop processing
2504 if ( s == TS::DeflatedExplicitVRLittleEndian)
2506 gdcmWarningMacro("Transfer Syntax ["
2507 << GetTransferSyntaxName() << "] :"
2508 << " not yet dealt with ");
2512 // The following shouldn't occur very often
2513 // Let's check at the very end.
2515 if ( ts == GDCM_UNKNOWN )
2517 gdcmDebugMacro("True DICOM File, with NO Transfer Syntax (?!) " );
2521 if ( !Global::GetTS()->IsTransferSyntax(ts) )
2523 gdcmWarningMacro("True DICOM File, with illegal Transfer Syntax: ["
2529 //-----------------------------------------------------------------------------
2532 //-----------------------------------------------------------------------------
2533 } // end namespace gdcm