1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/28 10:34:28 $
7 Version: $Revision: 1.216 $
9 Copyright (c) CREATIS (Centre de Recherche et d'Applications en Traitement de
10 l'Image). All rights reserved. See Doc/License.txt or
11 http://www.creatis.insa-lyon.fr/Public/Gdcm/License.html for details.
13 This software is distributed WITHOUT ANY WARRANTY; without even
14 the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
15 PURPOSE. See the above copyright notices for more information.
17 =========================================================================*/
19 #include "gdcmDocument.h"
20 #include "gdcmValEntry.h"
21 #include "gdcmBinEntry.h"
22 #include "gdcmSeqEntry.h"
23 #include "gdcmGlobal.h"
25 #include "gdcmDebug.h"
27 #include "gdcmDictSet.h"
28 #include "gdcmDocEntrySet.h"
29 #include "gdcmSQItem.h"
35 #if defined(_MSC_VER) || defined(__BORLANDC__) || defined(__MINGW32__)
39 #ifdef CMAKE_HAVE_NETINET_IN_H
40 #include <netinet/in.h>
46 //-----------------------------------------------------------------------------
47 // Refer to Document::CheckSwap()
48 //const unsigned int Document::HEADER_LENGTH_TO_READ = 256;
50 // Refer to Document::SetMaxSizeLoadEntry()
51 const unsigned int Document::MAX_SIZE_LOAD_ELEMENT_VALUE = 0xfff; // 4096
52 const unsigned int Document::MAX_SIZE_PRINT_ELEMENT_VALUE = 0x7fffffff;
54 //-----------------------------------------------------------------------------
55 // Constructor / Destructor
59 * @param filename 'Document' (File or DicomDir) to be opened for parsing
61 Document::Document( std::string const &filename ) : ElementSet(-1)
63 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
73 Group0002Parsed = false;
75 gdcmVerboseMacro( "Starting parsing of file: " << Filename.c_str());
76 // Fp->seekg( 0, std::ios::beg);
78 Fp->seekg(0, std::ios::end);
79 long lgt = Fp->tellg();
81 Fp->seekg( 0, std::ios::beg);
84 long beg = Fp->tellg();
87 ParseDES( this, beg, lgt, false); // Loading is done during parsing
89 Fp->seekg( 0, std::ios::beg);
91 // Load 'non string' values
93 std::string PhotometricInterpretation = GetEntryValue(0x0028,0x0004);
94 if( PhotometricInterpretation == "PALETTE COLOR " )
96 LoadEntryBinArea(0x0028,0x1200); // gray LUT
98 /// The tags refered by the three following lines used to be CORRECTLY
99 /// defined as having an US Value Representation in the public
100 /// dictionary. BUT the semantics implied by the three following
101 /// lines state that the corresponding tag contents are in fact
102 /// the ones of a BinEntry.
103 /// In order to fix things "Quick and Dirty" the dictionary was
104 /// altered on PURPOSE but now contains a WRONG value.
105 /// In order to fix things and restore the dictionary to its
106 /// correct value, one needs to decided of the semantics by deciding
107 /// whether the following tags are either:
108 /// - multivaluated US, and hence loaded as ValEntry, but afterwards
109 /// also used as BinEntry, which requires the proper conversion,
110 /// - OW, and hence loaded as BinEntry, but afterwards also used
111 /// as ValEntry, which requires the proper conversion.
112 LoadEntryBinArea(0x0028,0x1201); // R LUT
113 LoadEntryBinArea(0x0028,0x1202); // G LUT
114 LoadEntryBinArea(0x0028,0x1203); // B LUT
116 // Segmented Red Palette Color LUT Data
117 LoadEntryBinArea(0x0028,0x1221);
118 // Segmented Green Palette Color LUT Data
119 LoadEntryBinArea(0x0028,0x1222);
120 // Segmented Blue Palette Color LUT Data
121 LoadEntryBinArea(0x0028,0x1223);
123 //FIXME later : how to use it?
124 LoadEntryBinArea(0x0028,0x3006); //LUT Data (CTX dependent)
128 // --------------------------------------------------------------
129 // Specific code to allow gdcm to read ACR-LibIDO formated images
130 // Note: ACR-LibIDO is an extension of the ACR standard that was
131 // used at CREATIS. For the time being (say a couple years)
132 // we keep this kludge to allow a smooth move to gdcm for
133 // CREATIS developpers (sorry folks).
135 // if recognition code tells us we deal with a LibIDO image
136 // we switch lineNumber and columnNumber
139 RecCode = GetEntryValue(0x0008, 0x0010); // recognition code (RET)
140 if (RecCode == "ACRNEMA_LIBIDO_1.1" ||
141 RecCode == "CANRME_AILIBOD1_1." ) // for brain-damaged softwares
142 // with "little-endian strings"
144 Filetype = ACR_LIBIDO;
145 std::string rows = GetEntryValue(0x0028, 0x0010);
146 std::string columns = GetEntryValue(0x0028, 0x0011);
147 SetValEntry(columns, 0x0028, 0x0010);
148 SetValEntry(rows , 0x0028, 0x0011);
150 // ----------------- End of ACR-LibIDO kludge ------------------
154 * \brief This default constructor doesn't parse the file. You should
155 * then invoke \ref Document::SetFileName and then the parsing.
157 Document::Document() : ElementSet(-1)
161 SetMaxSizeLoadEntry(MAX_SIZE_LOAD_ELEMENT_VALUE);
164 Filetype = ExplicitVR;
165 Group0002Parsed = false;
169 * \brief Canonical destructor.
171 Document::~Document ()
177 //-----------------------------------------------------------------------------
181 * \brief Prints The Dict Entries of THE public Dicom Dictionary
182 * @param os ostream to print to
185 void Document::PrintPubDict(std::ostream &os)
187 RefPubDict->SetPrintLevel(PrintLevel);
188 RefPubDict->Print(os);
192 * \brief Prints The Dict Entries of THE shadow Dicom Dictionary
193 * @param os ostream to print to
196 void Document::PrintShaDict(std::ostream &os)
198 RefShaDict->SetPrintLevel(PrintLevel);
199 RefShaDict->Print(os);
202 //-----------------------------------------------------------------------------
205 * \brief Get the public dictionary used
207 Dict *Document::GetPubDict()
213 * \brief Get the shadow dictionary used
215 Dict *Document::GetShaDict()
221 * \brief Set the shadow dictionary used
222 * @param dict dictionary to use in shadow
224 bool Document::SetShaDict(Dict *dict)
231 * \brief Set the shadow dictionary used
232 * @param dictName name of the dictionary to use in shadow
234 bool Document::SetShaDict(DictKey const &dictName)
236 RefShaDict = Global::GetDicts()->GetDict(dictName);
241 * \brief This predicate, based on hopefully reasonable heuristics,
242 * decides whether or not the current Document was properly parsed
243 * and contains the mandatory information for being considered as
244 * a well formed and usable Dicom/Acr File.
245 * @return true when Document is the one of a reasonable Dicom/Acr file,
248 bool Document::IsReadable()
250 if( Filetype == Unknown)
252 gdcmVerboseMacro( "Wrong filetype");
258 gdcmVerboseMacro( "No tag in internal hash table.");
266 * \brief Accessor to the Transfer Syntax (when present) of the
267 * current document (it internally handles reading the
268 * value from disk when only parsing occured).
269 * @return The encountered Transfer Syntax of the current document.
271 std::string Document::GetTransferSyntax()
273 DocEntry *entry = GetDocEntry(0x0002, 0x0010);
279 // The entry might be present but not loaded (parsing and loading
280 // happen at different stages): try loading and proceed with check...
281 LoadDocEntrySafe(entry);
282 if (ValEntry *valEntry = dynamic_cast< ValEntry* >(entry) )
284 std::string transfer = valEntry->GetValue();
285 // The actual transfer (as read from disk) might be padded. We
286 // first need to remove the potential padding. We can make the
287 // weak assumption that padding was not executed with digits...
288 if ( transfer.length() == 0 )
290 // for brain damaged headers
293 while ( !isdigit((unsigned char)transfer[transfer.length()-1]) )
295 transfer.erase(transfer.length()-1, 1);
303 * \brief Predicate for dicom version 3 file.
304 * @return True when the file is a dicom version 3.
306 bool Document::IsDicomV3()
308 // Checking if Transfer Syntax exists is enough
309 // Anyway, it's to late check if the 'Preamble' was found ...
310 // And ... would it be a rich idea to check ?
311 // (some 'no Preamble' DICOM images exist !)
312 return GetDocEntry(0x0002, 0x0010) != NULL;
316 * \brief Predicate for Papyrus file
317 * Dedicated to whomsoever it may concern
318 * @return True when the file is a Papyrus file.
320 bool Document::IsPapyrus()
322 // check for Papyrus private Sequence
323 DocEntry *e = GetDocEntry(0x0041, 0x1050);
326 // check if it's actually a Sequence
327 if ( !dynamic_cast<SeqEntry*>(e) )
333 * \brief returns the File Type
334 * (ACR, ACR_LIBIDO, ExplicitVR, ImplicitVR, Unknown)
335 * @return the FileType code
337 FileType Document::GetFileType()
343 * \brief Tries to open the file \ref Document::Filename and
344 * checks the preamble when existing.
345 * @return The FILE pointer on success.
347 std::ifstream *Document::OpenFile()
349 HasDCMPreamble = false;
350 if (Filename.length() == 0)
357 gdcmVerboseMacro( "File already open: " << Filename.c_str());
361 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
364 gdcmDebugMacro( "Cannot open file: " << Filename.c_str());
371 Fp->read((char*)&zero, (size_t)2);
378 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
380 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
381 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
382 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
383 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
386 = Util::Format("ACR/DICOM with no preamble: (%04x)\n", zero);
387 gdcmVerboseMacro( msg.c_str() );
392 Fp->seekg(126L, std::ios::cur);
394 Fp->read(dicm, (size_t)4);
400 if( memcmp(dicm, "DICM", 4) == 0 )
402 HasDCMPreamble = true;
407 gdcmVerboseMacro( "Not DICOM/ACR (missing preamble)" << Filename.c_str());
413 * \brief closes the file
414 * @return TRUE if the close was successfull
416 bool Document::CloseFile()
424 return true; //FIXME how do we detect a non-closed ifstream ?
428 * \brief Writes in a file all the Header Entries (Dicom Elements)
429 * @param fp file pointer on an already open file (actually: Output File Stream)
430 * @param filetype Type of the File to be written
431 * (ACR-NEMA, ExplicitVR, ImplicitVR)
432 * @return Always true.
434 void Document::WriteContent(std::ofstream *fp, FileType filetype)
436 // \TODO move the following lines (and a lot of others, to be written)
437 // to a future function CheckAndCorrectHeader
439 // (necessary if user wants to write a DICOM V3 file
440 // starting from an ACR-NEMA (V2) Header
442 if ( filetype == ImplicitVR || filetype == ExplicitVR )
444 // writing Dicom File Preamble
445 char filePreamble[128];
446 memset(filePreamble, 0, 128);
447 fp->write(filePreamble, 128);
448 fp->write("DICM", 4);
452 * \todo rewrite later, if really usefull
453 * - 'Group Length' element is optional in DICOM
454 * - but un-updated odd groups lengthes can causes pb
457 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
458 * UpdateGroupLength(false,filetype);
459 * if ( filetype == ACR)
460 * UpdateGroupLength(true,ACR);
463 ElementSet::WriteContent(fp, filetype); // This one is recursive
466 //-----------------------------------------------------------------------------
469 * \brief Loads (from disk) the element content
470 * when a string is not suitable
471 * @param group group number of the Entry
472 * @param elem element number of the Entry
474 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
476 // Search the corresponding DocEntry
477 DocEntry *docElement = GetDocEntry(group, elem);
481 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
485 LoadEntryBinArea(binElement);
489 * \brief Loads (from disk) the element content
490 * when a string is not suitable
491 * @param elem Entry whose binArea is going to be loaded
493 void Document::LoadEntryBinArea(BinEntry *elem)
495 if(elem->GetBinArea())
502 size_t o =(size_t)elem->GetOffset();
503 Fp->seekg(o, std::ios::beg);
505 size_t l = elem->GetLength();
506 uint8_t *a = new uint8_t[l];
509 gdcmVerboseMacro( "Cannot allocate BinEntry content");
513 /// \todo check the result
514 Fp->read((char*)a, l);
515 if( Fp->fail() || Fp->eof())
528 * \brief Sets a 'non string' value to a given Dicom Element
529 * @param area area containing the 'non string' value
530 * @param group Group number of the searched Dicom Element
531 * @param elem Element number of the searched Dicom Element
534 /*bool Document::SetEntryBinArea(uint8_t *area,
535 uint16_t group, uint16_t elem)
537 DocEntry *currentEntry = GetDocEntry(group, elem);
543 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(currentEntry) )
545 binEntry->SetBinArea( area );
554 * \brief Loads the element while preserving the current
555 * underlying file position indicator as opposed to
556 * LoadDocEntry that modifies it.
557 * @param entry Header Entry whose value will be loaded.
560 void Document::LoadDocEntrySafe(DocEntry *entry)
564 long PositionOnEntry = Fp->tellg();
566 Fp->seekg(PositionOnEntry, std::ios::beg);
571 * \brief Swaps back the bytes of 4-byte long integer accordingly to
573 * @return The properly swaped 32 bits integer.
575 uint32_t Document::SwapLong(uint32_t a)
582 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
583 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
586 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
589 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
592 gdcmErrorMacro( "Unset swap code:" << SwapCode );
599 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
601 * @return The properly unswaped 32 bits integer.
603 uint32_t Document::UnswapLong(uint32_t a)
609 * \brief Swaps the bytes so they agree with the processor order
610 * @return The properly swaped 16 bits integer.
612 uint16_t Document::SwapShort(uint16_t a)
614 if ( SwapCode == 4321 || SwapCode == 2143 )
616 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
622 * \brief Unswaps the bytes so they agree with the processor order
623 * @return The properly unswaped 16 bits integer.
625 uint16_t Document::UnswapShort(uint16_t a)
630 //-----------------------------------------------------------------------------
634 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
635 * @return length of the parsed set.
637 void Document::ParseDES(DocEntrySet *set, long offset,
638 long l_max, bool delim_mode)
640 DocEntry *newDocEntry = 0;
641 ValEntry *newValEntry;
642 BinEntry *newBinEntry;
643 SeqEntry *newSeqEntry;
649 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
655 newDocEntry = ReadNextDocEntry( );
662 vr = newDocEntry->GetVR();
663 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
664 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
665 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
667 if ( newValEntry || newBinEntry )
671 if ( Filetype == ExplicitVR &&
672 !Global::GetVR()->IsVROfBinaryRepresentable(vr) )
674 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
675 gdcmVerboseMacro( std::hex << newDocEntry->GetGroup()
676 << "|" << newDocEntry->GetElement()
677 << " : Neither Valentry, nor BinEntry."
678 "Probably unknown VR.");
681 //////////////////// BinEntry or UNKOWN VR:
682 // When "this" is a Document the Key is simply of the
683 // form ( group, elem )...
684 if ( dynamic_cast< Document* > ( set ) )
686 newBinEntry->SetKey( newBinEntry->GetKey() );
688 // but when "this" is a SQItem, we are inserting this new
689 // valEntry in a sequence item, and the key has the
690 // generalized form (refer to \ref BaseTagKey):
691 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
693 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
694 + newBinEntry->GetKey() );
697 LoadDocEntry( newBinEntry );
698 if( !set->AddEntry( newBinEntry ) )
700 //Expect big troubles if here
701 //delete newBinEntry;
707 /////////////////////// ValEntry
708 // When "set" is a Document, then we are at the top of the
709 // hierarchy and the Key is simply of the form ( group, elem )...
710 if ( dynamic_cast< Document* > ( set ) )
712 newValEntry->SetKey( newValEntry->GetKey() );
714 // ...but when "set" is a SQItem, we are inserting this new
715 // valEntry in a sequence item. Hence the key has the
716 // generalized form (refer to \ref BaseTagKey):
717 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
719 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
720 + newValEntry->GetKey() );
723 LoadDocEntry( newValEntry );
724 bool delimitor=newValEntry->IsItemDelimitor();
725 if( !set->AddEntry( newValEntry ) )
727 // If here expect big troubles
728 //delete newValEntry; //otherwise mem leak
738 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
746 // Just to make sure we are at the beginning of next entry.
747 SkipToNextDocEntry(newDocEntry);
752 unsigned long l = newDocEntry->GetReadLength();
753 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
755 if ( l == 0xffffffff )
764 // no other way to create it ...
765 newSeqEntry->SetDelimitorMode( delim_mode );
767 // At the top of the hierarchy, stands a Document. When "set"
768 // is a Document, then we are building the first depth level.
769 // Hence the SeqEntry we are building simply has a depth
771 if (/*Document *dummy =*/ dynamic_cast< Document* > ( set ) )
774 newSeqEntry->SetDepthLevel( 1 );
775 newSeqEntry->SetKey( newSeqEntry->GetKey() );
777 // But when "set" is already a SQItem, we are building a nested
778 // sequence, and hence the depth level of the new SeqEntry
779 // we are building, is one level deeper:
780 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
782 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
783 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
784 + newSeqEntry->GetKey() );
788 { // Don't try to parse zero-length sequences
789 ParseSQ( newSeqEntry,
790 newDocEntry->GetOffset(),
793 if( !set->AddEntry( newSeqEntry ) )
797 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
811 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
812 * @return parsed length for this level
814 void Document::ParseSQ( SeqEntry *seqEntry,
815 long offset, long l_max, bool delim_mode)
817 int SQItemNumber = 0;
819 long offsetStartCurrentSQItem = offset;
823 // the first time, we read the fff0,e000 of the first SQItem
824 DocEntry *newDocEntry = ReadNextDocEntry();
828 // FIXME Should warn user
833 if ( newDocEntry->IsSequenceDelimitor() )
835 seqEntry->SetDelimitationItem( newDocEntry );
839 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
844 // create the current SQItem
845 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
846 std::ostringstream newBase;
847 newBase << seqEntry->GetKey()
851 itemSQ->SetBaseTagKey( newBase.str() );
852 unsigned int l = newDocEntry->GetReadLength();
854 if ( l == 0xffffffff )
863 // when we're here, element fffe,e000 is already passed.
864 // it's lost for the SQItem we're going to process !!
866 //ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
867 //delete newDocEntry; // FIXME well ... it's too late to use it !
869 // Let's try :------------
870 // remove fff0,e000, created out of the SQItem
872 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg);
873 // fill up the current SQItem, starting at the beginning of fff0,e000
874 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
875 offsetStartCurrentSQItem = Fp->tellg();
876 // end try -----------------
878 seqEntry->AddSQItem( itemSQ, SQItemNumber );
880 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
888 * \brief Loads the element content if its length doesn't exceed
889 * the value specified with Document::SetMaxSizeLoadEntry()
890 * @param entry Header Entry (Dicom Element) to be dealt with
892 void Document::LoadDocEntry(DocEntry *entry)
894 uint16_t group = entry->GetGroup();
895 std::string vr = entry->GetVR();
896 uint32_t length = entry->GetLength();
898 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
900 // A SeQuence "contains" a set of Elements.
901 // (fffe e000) tells us an Element is beginning
902 // (fffe e00d) tells us an Element just ended
903 // (fffe e0dd) tells us the current SeQuence just ended
904 if( group == 0xfffe )
906 // NO more value field for SQ !
910 // When the length is zero things are easy:
913 ((ValEntry *)entry)->SetValue("");
917 // The elements whose length is bigger than the specified upper bound
918 // are not loaded. Instead we leave a short notice of the offset of
919 // the element content and it's length.
921 std::ostringstream s;
922 if (length > MaxSizeLoadEntry)
924 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
926 //s << "gdcm::NotLoaded (BinEntry)";
928 s << " Address:" << (long)entry->GetOffset();
929 s << " Length:" << entry->GetLength();
930 s << " x(" << std::hex << entry->GetLength() << ")";
931 binEntryPtr->SetValue(s.str());
933 // Be carefull : a BinEntry IS_A ValEntry ...
934 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
936 // s << "gdcm::NotLoaded. (ValEntry)";
938 s << " Address:" << (long)entry->GetOffset();
939 s << " Length:" << entry->GetLength();
940 s << " x(" << std::hex << entry->GetLength() << ")";
941 valEntryPtr->SetValue(s.str());
946 gdcmErrorMacro( "MaxSizeLoadEntry exceeded, neither a BinEntry "
947 << "nor a ValEntry ?! Should never print that !" );
950 // to be sure we are at the end of the value ...
951 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
956 // When we find a BinEntry not very much can be done :
957 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
960 binEntryPtr->SetValue(s.str());
961 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
965 /// \todo Any compacter code suggested (?)
966 if ( IsDocEntryAnInteger(entry) )
970 // When short integer(s) are expected, read and convert the following
971 // n *two characters properly i.e. consider them as short integers as
972 // opposed to strings.
973 // Elements with Value Multiplicity > 1
974 // contain a set of integers (not a single one)
975 if (vr == "US" || vr == "SS")
978 NewInt = ReadInt16();
982 for (int i=1; i < nbInt; i++)
985 NewInt = ReadInt16();
990 // See above comment on multiple integers (mutatis mutandis).
991 else if (vr == "UL" || vr == "SL")
994 NewInt = ReadInt32();
998 for (int i=1; i < nbInt; i++)
1001 NewInt = ReadInt32();
1006 #ifdef GDCM_NO_ANSI_STRING_STREAM
1007 s << std::ends; // to avoid oddities on Solaris
1008 #endif //GDCM_NO_ANSI_STRING_STREAM
1010 ((ValEntry *)entry)->SetValue(s.str());
1014 // FIXME: We need an additional byte for storing \0 that is not on disk
1015 char *str = new char[length+1];
1016 Fp->read(str, (size_t)length);
1017 str[length] = '\0'; //this is only useful when length is odd
1018 // Special DicomString call to properly handle \0 and even length
1019 std::string newValue;
1022 newValue = Util::DicomString(str, length+1);
1023 gdcmVerboseMacro("Warning: bad length: " << length <<
1024 ",For string :" << newValue.c_str());
1025 // Since we change the length of string update it length
1026 //entry->SetReadLength(length+1);
1030 newValue = Util::DicomString(str, length);
1034 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1036 if ( Fp->fail() || Fp->eof())
1038 gdcmVerboseMacro("Unread element value");
1039 valEntry->SetValue(GDCM_UNREAD);
1045 // Because of correspondance with the VR dic
1046 valEntry->SetValue(newValue);
1050 valEntry->SetValue(newValue);
1055 gdcmErrorMacro( "Should have a ValEntry, here !");
1061 * \brief Find the value Length of the passed Header Entry
1062 * @param entry Header Entry whose length of the value shall be loaded.
1064 void Document::FindDocEntryLength( DocEntry *entry )
1065 throw ( FormatError )
1067 std::string vr = entry->GetVR();
1070 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1072 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1074 // The following reserved two bytes (see PS 3.5-2003, section
1075 // "7.1.2 Data element structure with explicit vr", p 27) must be
1076 // skipped before proceeding on reading the length on 4 bytes.
1077 Fp->seekg( 2L, std::ios::cur);
1078 uint32_t length32 = ReadInt32();
1080 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1085 lengthOB = FindDocEntryLengthOBOrOW();
1087 catch ( FormatUnexpected )
1089 // Computing the length failed (this happens with broken
1090 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1091 // chance to get the pixels by deciding the element goes
1092 // until the end of the file. Hence we artificially fix the
1093 // the length and proceed.
1094 long currentPosition = Fp->tellg();
1095 Fp->seekg(0L,std::ios::end);
1097 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1098 Fp->seekg(currentPosition, std::ios::beg);
1100 entry->SetReadLength(lengthUntilEOF);
1101 entry->SetLength(lengthUntilEOF);
1104 entry->SetReadLength(lengthOB);
1105 entry->SetLength(lengthOB);
1108 FixDocEntryFoundLength(entry, length32);
1112 // Length is encoded on 2 bytes.
1113 length16 = ReadInt16();
1115 // FIXME : This heuristic supposes that the first group following
1116 // group 0002 *has* and element 0000.
1117 // BUT ... Element 0000 is optionnal :-(
1120 // Fixed using : HandleOutOfGroup0002()
1121 // (first hereafter strategy ...)
1123 // We can tell the current file is encoded in big endian (like
1124 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1125 // and it's value is the one of the encoding of a big endian file.
1126 // In order to deal with such big endian encoded files, we have
1127 // (at least) two strategies:
1128 // * when we load the "Transfer Syntax" tag with value of big endian
1129 // encoding, we raise the proper flags. Then we wait for the end
1130 // of the META group (0x0002) among which is "Transfer Syntax",
1131 // before switching the swap code to big endian. We have to postpone
1132 // the switching of the swap code since the META group is fully encoded
1133 // in little endian, and big endian coding only starts at the next
1134 // group. The corresponding code can be hard to analyse and adds
1135 // many additional unnecessary tests for regular tags.
1136 // * the second strategy consists in waiting for trouble, that shall
1137 // appear when we find the first group with big endian encoding. This
1138 // is easy to detect since the length of a "Group Length" tag (the
1139 // ones with zero as element number) has to be of 4 (0x0004). When we
1140 // encounter 1024 (0x0400) chances are the encoding changed and we
1141 // found a group with big endian encoding.
1142 //---> Unfortunately, element 0000 is optional.
1143 //---> This will not work when missing!
1144 // We shall use this second strategy. In order to make sure that we
1145 // can interpret the presence of an apparently big endian encoded
1146 // length of a "Group Length" without committing a big mistake, we
1147 // add an additional check: we look in the already parsed elements
1148 // for the presence of a "Transfer Syntax" whose value has to be "big
1149 // endian encoding". When this is the case, chances are we have got our
1150 // hands on a big endian encoded file: we switch the swap code to
1151 // big endian and proceed...
1153 // if ( element == 0x0000 && length16 == 0x0400 )
1155 // std::string ts = GetTransferSyntax();
1156 // if ( Global::GetTS()->GetSpecialTransferSyntax(ts)
1157 // != TS::ExplicitVRBigEndian )
1159 // throw FormatError( "Document::FindDocEntryLength()",
1160 // " not explicit VR." );
1164 // SwitchByteSwapCode();
1166 // // Restore the unproperly loaded values i.e. the group, the element
1167 // // and the dictionary entry depending on them.
1168 // uint16_t correctGroup = SwapShort( entry->GetGroup() );
1169 // uint16_t correctElem = SwapShort( entry->GetElement() );
1170 // DictEntry *newTag = GetDictEntry( correctGroup, correctElem );
1173 // // This correct tag is not in the dictionary. Create a new one.
1174 // newTag = NewVirtualDictEntry(correctGroup, correctElem);
1176 // // FIXME this can create a memory leaks on the old entry that be
1177 // // left unreferenced.
1178 // entry->SetDictEntry( newTag );
1181 // 0xffff means that we deal with 'No Length' Sequence
1182 // or 'No Length' SQItem
1183 if ( length16 == 0xffff)
1187 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1192 // Either implicit VR or a non DICOM conformal (see note below) explicit
1193 // VR that ommited the VR of (at least) this element. Farts happen.
1194 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1195 // on Data elements "Implicit and Explicit VR Data Elements shall
1196 // not coexist in a Data Set and Data Sets nested within it".]
1197 // Length is on 4 bytes.
1199 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1200 // even if Transfer Syntax is 'Implicit VR ...'
1202 FixDocEntryFoundLength( entry, ReadInt32() );
1208 * \brief Find the Value Representation of the current Dicom Element.
1209 * @return Value Representation of the current Entry
1211 std::string Document::FindDocEntryVR()
1213 if ( Filetype != ExplicitVR )
1214 return GDCM_UNKNOWN;
1216 long positionOnEntry = Fp->tellg();
1217 // Warning: we believe this is explicit VR (Value Representation) because
1218 // we used a heuristic that found "UL" in the first tag. Alas this
1219 // doesn't guarantee that all the tags will be in explicit VR. In some
1220 // cases (see e-film filtered files) one finds implicit VR tags mixed
1221 // within an explicit VR file. Hence we make sure the present tag
1222 // is in explicit VR and try to fix things if it happens not to be
1226 Fp->read (vr, (size_t)2);
1229 if( !CheckDocEntryVR(vr) )
1231 Fp->seekg(positionOnEntry, std::ios::beg);
1232 return GDCM_UNKNOWN;
1238 * \brief Check the correspondance between the VR of the header entry
1239 * and the taken VR. If they are different, the header entry is
1240 * updated with the new VR.
1241 * @param vr Dicom Value Representation
1242 * @return false if the VR is incorrect of if the VR isn't referenced
1243 * otherwise, it returns true
1245 bool Document::CheckDocEntryVR(VRKey vr)
1247 // CLEANME searching the dicom_vr at each occurence is expensive.
1248 // PostPone this test in an optional integrity check at the end
1249 // of parsing or only in debug mode.
1250 if ( !Global::GetVR()->IsValidVR(vr) )
1257 * \brief Get the transformed value of the header entry. The VR value
1258 * is used to define the transformation to operate on the value
1259 * \warning NOT end user intended method !
1260 * @param entry entry to tranform
1261 * @return Transformed entry value
1263 std::string Document::GetDocEntryValue(DocEntry *entry)
1265 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1267 std::string val = ((ValEntry *)entry)->GetValue();
1268 std::string vr = entry->GetVR();
1269 uint32_t length = entry->GetLength();
1270 std::ostringstream s;
1273 // When short integer(s) are expected, read and convert the following
1274 // n * 2 bytes properly i.e. as a multivaluated strings
1275 // (each single value is separated fromthe next one by '\'
1276 // as usual for standard multivaluated filels
1277 // Elements with Value Multiplicity > 1
1278 // contain a set of short integers (not a single one)
1280 if( vr == "US" || vr == "SS" )
1285 for (int i=0; i < nbInt; i++)
1291 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1292 newInt16 = SwapShort( newInt16 );
1297 // When integer(s) are expected, read and convert the following
1298 // n * 4 bytes properly i.e. as a multivaluated strings
1299 // (each single value is separated fromthe next one by '\'
1300 // as usual for standard multivaluated filels
1301 // Elements with Value Multiplicity > 1
1302 // contain a set of integers (not a single one)
1303 else if( vr == "UL" || vr == "SL" )
1308 for (int i=0; i < nbInt; i++)
1314 newInt32 = ( val[4*i+0] & 0xFF )
1315 + (( val[4*i+1] & 0xFF ) << 8 )
1316 + (( val[4*i+2] & 0xFF ) << 16 )
1317 + (( val[4*i+3] & 0xFF ) << 24 );
1318 newInt32 = SwapLong( newInt32 );
1322 #ifdef GDCM_NO_ANSI_STRING_STREAM
1323 s << std::ends; // to avoid oddities on Solaris
1324 #endif //GDCM_NO_ANSI_STRING_STREAM
1328 return ((ValEntry *)entry)->GetValue();
1332 * \brief Get the reverse transformed value of the header entry. The VR
1333 * value is used to define the reverse transformation to operate on
1335 * \warning NOT end user intended method !
1336 * @param entry Entry to reverse transform
1337 * @return Reverse transformed entry value
1339 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1341 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1343 std::string vr = entry->GetVR();
1344 std::vector<std::string> tokens;
1345 std::ostringstream s;
1347 if ( vr == "US" || vr == "SS" )
1351 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1352 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1353 for (unsigned int i=0; i<tokens.size(); i++)
1355 newInt16 = atoi(tokens[i].c_str());
1356 s << ( newInt16 & 0xFF )
1357 << (( newInt16 >> 8 ) & 0xFF );
1361 if ( vr == "UL" || vr == "SL")
1365 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1366 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1367 for (unsigned int i=0; i<tokens.size();i++)
1369 newInt32 = atoi(tokens[i].c_str());
1370 s << (char)( newInt32 & 0xFF )
1371 << (char)(( newInt32 >> 8 ) & 0xFF )
1372 << (char)(( newInt32 >> 16 ) & 0xFF )
1373 << (char)(( newInt32 >> 24 ) & 0xFF );
1378 #ifdef GDCM_NO_ANSI_STRING_STREAM
1379 s << std::ends; // to avoid oddities on Solaris
1380 #endif //GDCM_NO_ANSI_STRING_STREAM
1384 return ((ValEntry *)entry)->GetValue();
1388 * \brief Skip a given Header Entry
1389 * \warning NOT end user intended method !
1390 * @param entry entry to skip
1392 void Document::SkipDocEntry(DocEntry *entry)
1394 SkipBytes(entry->GetLength());
1398 * \brief Skips to the begining of the next Header Entry
1399 * \warning NOT end user intended method !
1400 * @param currentDocEntry entry to skip
1402 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1404 Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg);
1405 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1406 Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur);
1410 * \brief When the length of an element value is obviously wrong (because
1411 * the parser went Jabberwocky) one can hope improving things by
1412 * applying some heuristics.
1413 * @param entry entry to check
1414 * @param foundLength first assumption about length
1416 void Document::FixDocEntryFoundLength(DocEntry *entry,
1417 uint32_t foundLength)
1419 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1420 if ( foundLength == 0xffffffff)
1425 uint16_t gr = entry->GetGroup();
1426 uint16_t elem = entry->GetElement();
1428 if ( foundLength % 2)
1430 gdcmVerboseMacro( "Warning : Tag with uneven length " << foundLength
1431 << " in x(" << std::hex << gr << "," << elem <<")");
1434 //////// Fix for some naughty General Electric images.
1435 // Allthough not recent many such GE corrupted images are still present
1436 // on Creatis hard disks. Hence this fix shall remain when such images
1437 // are no longer in use (we are talking a few years, here)...
1438 // Note: XMedCom probably uses such a trick since it is able to read
1439 // those pesky GE images ...
1440 if ( foundLength == 13)
1442 // Only happens for this length !
1443 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1446 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
1450 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1451 // Occurence of such images is quite low (unless one leaves close to a
1452 // 'Leonardo' source. Hence, one might consider commenting out the
1453 // following fix on efficiency reasons.
1454 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1457 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
1460 else if ( entry->GetVR() == "SQ" )
1462 foundLength = 0; // ReadLength is unchanged
1465 //////// We encountered a 'delimiter' element i.e. a tag of the form
1466 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1467 // taken into account.
1468 else if( gr == 0xfffe )
1470 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1471 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1472 // causes extra troubles...
1473 if( entry->GetElement() != 0x0000 )
1479 entry->SetLength(foundLength);
1483 * \brief Apply some heuristics to predict whether the considered
1484 * element value contains/represents an integer or not.
1485 * @param entry The element value on which to apply the predicate.
1486 * @return The result of the heuristical predicate.
1488 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1490 uint16_t elem = entry->GetElement();
1491 uint16_t group = entry->GetGroup();
1492 const std::string &vr = entry->GetVR();
1493 uint32_t length = entry->GetLength();
1495 // When we have some semantics on the element we just read, and if we
1496 // a priori know we are dealing with an integer, then we shall be
1497 // able to swap it's element value properly.
1498 if ( elem == 0 ) // This is the group length of the group
1506 // Allthough this should never happen, still some images have a
1507 // corrupted group length [e.g. have a glance at offset x(8336) of
1508 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
1509 // Since for dicom compliant and well behaved headers, the present
1510 // test is useless (and might even look a bit paranoid), when we
1511 // encounter such an ill-formed image, we simply display a warning
1512 // message and proceed on parsing (while crossing fingers).
1513 long filePosition = Fp->tellg();
1514 gdcmVerboseMacro( "Erroneous Group Length element length on : ("
1515 << std::hex << group << " , " << elem
1516 << ") -before- position x(" << filePosition << ")"
1517 << "lgt : " << length );
1521 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1529 * \brief Find the Length till the next sequence delimiter
1530 * \warning NOT end user intended method !
1534 uint32_t Document::FindDocEntryLengthOBOrOW()
1535 throw( FormatUnexpected )
1537 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1538 long positionOnEntry = Fp->tellg();
1539 bool foundSequenceDelimiter = false;
1540 uint32_t totalLength = 0;
1542 while ( !foundSequenceDelimiter )
1548 group = ReadInt16();
1551 catch ( FormatError )
1553 throw FormatError("Unexpected end of file encountered during ",
1554 "Document::FindDocEntryLengthOBOrOW()");
1557 // We have to decount the group and element we just read
1560 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
1562 long filePosition = Fp->tellg();
1563 gdcmVerboseMacro( "Neither an Item tag nor a Sequence delimiter tag on :"
1564 << std::hex << group << " , " << elem
1565 << ") -before- position x(" << filePosition << ")" );
1567 Fp->seekg(positionOnEntry, std::ios::beg);
1568 throw FormatUnexpected( "Neither an Item tag nor a Sequence delimiter tag.");
1571 if ( elem == 0xe0dd )
1573 foundSequenceDelimiter = true;
1576 uint32_t itemLength = ReadInt32();
1577 // We add 4 bytes since we just read the ItemLength with ReadInt32
1578 totalLength += itemLength + 4;
1579 SkipBytes(itemLength);
1581 if ( foundSequenceDelimiter )
1586 Fp->seekg( positionOnEntry, std::ios::beg);
1591 * \brief Reads a supposed to be 16 Bits integer
1592 * (swaps it depending on processor endianity)
1593 * @return read value
1595 uint16_t Document::ReadInt16()
1596 throw( FormatError )
1599 Fp->read ((char*)&g, (size_t)2);
1602 throw FormatError( "Document::ReadInt16()", " file error." );
1606 throw FormatError( "Document::ReadInt16()", "EOF." );
1613 * \brief Reads a supposed to be 32 Bits integer
1614 * (swaps it depending on processor endianity)
1615 * @return read value
1617 uint32_t Document::ReadInt32()
1618 throw( FormatError )
1621 Fp->read ((char*)&g, (size_t)4);
1624 throw FormatError( "Document::ReadInt32()", " file error." );
1628 throw FormatError( "Document::ReadInt32()", "EOF." );
1635 * \brief skips bytes inside the source file
1636 * \warning NOT end user intended method !
1639 void Document::SkipBytes(uint32_t nBytes)
1641 //FIXME don't dump the returned value
1642 Fp->seekg((long)nBytes, std::ios::cur);
1646 * \brief Loads all the needed Dictionaries
1647 * \warning NOT end user intended method !
1649 void Document::Initialize()
1651 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
1657 * \brief Discover what the swap code is (among little endian, big endian,
1658 * bad little endian, bad big endian).
1660 * @return false when we are absolutely sure
1661 * it's neither ACR-NEMA nor DICOM
1662 * true when we hope ours assuptions are OK
1664 bool Document::CheckSwap()
1666 // The only guaranted way of finding the swap code is to find a
1667 // group tag since we know it's length has to be of four bytes i.e.
1668 // 0x00000004. Finding the swap code in then straigthforward. Trouble
1669 // occurs when we can't find such group...
1671 uint32_t x = 4; // x : for ntohs
1672 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
1678 // First, compare HostByteOrder and NetworkByteOrder in order to
1679 // determine if we shall need to swap bytes (i.e. the Endian type).
1680 if ( x == ntohs(x) )
1689 // The easiest case is the one of a 'true' DICOM header, we just have
1690 // to look for the string "DICM" inside the file preamble.
1693 char *entCur = deb + 128;
1694 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
1696 gdcmVerboseMacro( "Looks like DICOM Version3 (preamble + DCM)" );
1698 // Group 0002 should always be VR, and the first element 0000
1699 // Let's be carefull (so many wrong headers ...)
1700 // and determine the value representation (VR) :
1701 // Let's skip to the first element (0002,0000) and check there if we find
1702 // "UL" - or "OB" if the 1st one is (0002,0001) -,
1703 // in which case we (almost) know it is explicit VR.
1704 // WARNING: if it happens to be implicit VR then what we will read
1705 // is the length of the group. If this ascii representation of this
1706 // length happens to be "UL" then we shall believe it is explicit VR.
1707 // We need to skip :
1708 // * the 128 bytes of File Preamble (often padded with zeroes),
1709 // * the 4 bytes of "DICM" string,
1710 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
1711 // i.e. a total of 136 bytes.
1714 // group 0x0002 *is always* Explicit VR Sometimes ,
1715 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
1716 // *Implicit* VR (see former 'gdcmData/icone.dcm')
1718 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
1719 memcmp(entCur, "OB", (size_t)2) == 0 ||
1720 memcmp(entCur, "UI", (size_t)2) == 0 ||
1721 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
1722 // when Write DCM *adds*
1724 // Use Document::dicom_vr to test all the possibilities
1725 // instead of just checking for UL, OB and UI !? group 0000
1727 Filetype = ExplicitVR;
1728 gdcmVerboseMacro( "Group 0002 : Explicit Value Representation");
1732 Filetype = ImplicitVR;
1733 gdcmVerboseMacro( "Group 0002 :Not an explicit Value Representation;"
1734 << "Looks like a bugged Header!");
1740 gdcmVerboseMacro( "HostByteOrder != NetworkByteOrder");
1745 gdcmVerboseMacro( "HostByteOrder = NetworkByteOrder");
1748 // Position the file position indicator at first tag
1749 // (i.e. after the file preamble and the "DICM" string).
1750 Fp->seekg(0, std::ios::beg);
1751 Fp->seekg ( 132L, std::ios::beg);
1755 // Alas, this is not a DicomV3 file and whatever happens there is no file
1756 // preamble. We can reset the file position indicator to where the data
1757 // is (i.e. the beginning of the file).
1758 gdcmVerboseMacro( "Not a DICOM Version3 file");
1759 Fp->seekg(0, std::ios::beg);
1761 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
1762 // By clean we mean that the length of the first tag is written down.
1763 // If this is the case and since the length of the first group HAS to be
1764 // four (bytes), then determining the proper swap code is straightforward.
1767 // We assume the array of char we are considering contains the binary
1768 // representation of a 32 bits integer. Hence the following dirty
1770 s32 = *((uint32_t *)(entCur));
1791 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
1792 // It is time for despaired wild guesses.
1793 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
1794 // i.e. the 'group length' element is not present :
1796 // check the supposed-to-be 'group number'
1797 // in ( 0x0001 .. 0x0008 )
1798 // to determine ' SwapCode' value .
1799 // Only 0 or 4321 will be possible
1800 // (no oportunity to check for the formerly well known
1801 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
1802 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
1803 // the file IS NOT ACR-NEMA nor DICOM V3
1804 // Find a trick to tell it the caller...
1806 s16 = *((uint16_t *)(deb));
1833 gdcmVerboseMacro( "ACR/NEMA unfound swap info (Really hopeless !)");
1841 * \brief Change the Byte Swap code.
1843 void Document::SwitchByteSwapCode()
1845 gdcmVerboseMacro( "Switching Byte Swap code from "<< SwapCode);
1846 if ( SwapCode == 1234 )
1850 else if ( SwapCode == 4321 )
1854 else if ( SwapCode == 3412 )
1858 else if ( SwapCode == 2143 )
1865 * \brief during parsing, Header Elements too long are not loaded in memory
1868 void Document::SetMaxSizeLoadEntry(long newSize)
1874 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
1876 MaxSizeLoadEntry = 0xffffffff;
1879 MaxSizeLoadEntry = newSize;
1884 * \brief Header Elements too long will not be printed
1885 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
1888 void Document::SetMaxSizePrintEntry(long newSize)
1894 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
1896 MaxSizePrintEntry = 0xffffffff;
1899 MaxSizePrintEntry = newSize;
1905 * \brief Handle broken private tag from Philips NTSCAN
1906 * where the endianess is being switch to BigEndian for no
1910 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
1912 // Endian reversion. Some files contain groups of tags with reversed endianess.
1913 static int reversedEndian = 0;
1914 // try to fix endian switching in the middle of headers
1915 if ((group == 0xfeff) && (elem == 0x00e0))
1917 // start endian swap mark for group found
1919 SwitchByteSwapCode();
1924 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
1926 // end of reversed endian group
1928 SwitchByteSwapCode();
1933 * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS
1934 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
1936 std::string Document::GetTransferSyntaxName()
1938 // use the TS (TS : Transfer Syntax)
1939 std::string transferSyntax = GetEntryValue(0x0002,0x0010);
1941 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
1943 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
1944 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
1945 return "Uncompressed ACR-NEMA";
1947 if ( transferSyntax == GDCM_UNFOUND )
1949 gdcmVerboseMacro( "Unfound Transfer Syntax (0002,0010)");
1950 return "Uncompressed ACR-NEMA";
1953 // we do it only when we need it
1954 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
1956 // Global::GetTS() is a global static you shall never try to delete it!
1961 * \brief Group 0002 is always coded Little Endian
1962 * whatever Transfer Syntax is
1965 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
1967 // Endian reversion. Some files contain groups of tags with reversed endianess.
1968 if ( !Group0002Parsed && group != 0x0002)
1970 Group0002Parsed = true;
1971 // we just came out of group 0002
1972 // if Transfer syntax is Big Endian we have to change CheckSwap
1974 std::string ts = GetTransferSyntax();
1975 if ( !Global::GetTS()->IsTransferSyntax(ts) )
1977 gdcmVerboseMacro("True DICOM File, with NO Tansfer Syntax: " << ts );
1981 // Group 0002 is always 'Explicit ...' enven when Transfer Syntax says 'Implicit ..."
1983 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ImplicitVRLittleEndian )
1985 Filetype = ImplicitVR;
1988 // FIXME Strangely, this works with
1989 //'Implicit VR Transfer Syntax (GE Private)
1990 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ExplicitVRBigEndian )
1992 gdcmVerboseMacro("Transfer Syntax Name = ["
1993 << GetTransferSyntaxName() << "]" );
1994 SwitchByteSwapCode();
1995 group = SwapShort(group);
1996 elem = SwapShort(elem);
2002 * \brief Read the next tag but WITHOUT loading it's value
2003 * (read the 'Group Number', the 'Element Number',
2004 * gets the Dict Entry
2005 * gets the VR, gets the length, gets the offset value)
2006 * @return On succes the newly created DocEntry, NULL on failure.
2008 DocEntry *Document::ReadNextDocEntry()
2015 group = ReadInt16();
2018 catch ( FormatError e )
2020 // We reached the EOF (or an error occured) therefore
2021 // header parsing has to be considered as finished.
2026 // Sometimes file contains groups of tags with reversed endianess.
2027 HandleBrokenEndian(group, elem);
2029 // In 'true DICOM' files Group 0002 is always little endian
2030 if ( HasDCMPreamble )
2031 HandleOutOfGroup0002(group, elem);
2033 std::string vr = FindDocEntryVR();
2034 std::string realVR = vr;
2036 if( vr == GDCM_UNKNOWN)
2038 DictEntry *dictEntry = GetDictEntry(group,elem);
2040 realVR = dictEntry->GetVR();
2044 if( Global::GetVR()->IsVROfSequence(realVR) )
2045 newEntry = NewSeqEntry(group, elem);
2046 else if( Global::GetVR()->IsVROfStringRepresentable(realVR) )
2047 newEntry = NewValEntry(group, elem,vr);
2049 newEntry = NewBinEntry(group, elem,vr);
2051 if( vr == GDCM_UNKNOWN )
2053 if( Filetype == ExplicitVR )
2055 // We thought this was explicit VR, but we end up with an
2056 // implicit VR tag. Let's backtrack.
2057 if ( newEntry->GetGroup() != 0xfffe )
2060 msg = Util::Format("Entry (%04x,%04x) should be Explicit VR\n",
2061 newEntry->GetGroup(), newEntry->GetElement());
2062 gdcmVerboseMacro( msg.c_str() );
2065 newEntry->SetImplicitVR();
2070 FindDocEntryLength(newEntry);
2072 catch ( FormatError e )
2080 newEntry->SetOffset(Fp->tellg());
2085 //GenerateFreeTagKeyInGroup? What was it designed for ?!?
2087 * \brief Generate a free TagKey i.e. a TagKey that is not present
2088 * in the TagHt dictionary.
2089 * @param group The generated tag must belong to this group.
2090 * @return The element of tag with given group which is fee.
2092 //uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2094 // for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2096 // TagKey key = DictEntry::TranslateToKey(group, elem);
2097 // if (TagHT.count(key) == 0)
2102 // return UINT32_MAX;
2106 * \brief Compares two documents, according to \ref DicomDir rules
2107 * \warning Does NOT work with ACR-NEMA files
2108 * \todo Find a trick to solve the pb (use RET fields ?)
2110 * @return true if 'smaller'
2112 bool Document::operator<(Document &document)
2115 std::string s1 = GetEntryValue(0x0010,0x0010);
2116 std::string s2 = document.GetEntryValue(0x0010,0x0010);
2128 s1 = GetEntryValue(0x0010,0x0020);
2129 s2 = document.GetEntryValue(0x0010,0x0020);
2140 // Study Instance UID
2141 s1 = GetEntryValue(0x0020,0x000d);
2142 s2 = document.GetEntryValue(0x0020,0x000d);
2153 // Serie Instance UID
2154 s1 = GetEntryValue(0x0020,0x000e);
2155 s2 = document.GetEntryValue(0x0020,0x000e);
2171 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
2172 * @param filetype Type of the File to be written
2174 int Document::ComputeGroup0002Length( FileType filetype )
2179 int groupLength = 0;
2180 bool found0002 = false;
2182 // for each zero-level Tag in the DCM Header
2183 DocEntry *entry = GetFirstEntry();
2186 gr = entry->GetGroup();
2192 if( entry->GetElement() != 0x0000 )
2194 vr = entry->GetVR();
2196 if( filetype == ExplicitVR )
2198 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
2200 // explicit VR AND OB, OW, SQ : 4 more bytes
2204 groupLength += 2 + 2 + 4 + entry->GetLength();
2207 else if (found0002 )
2210 entry = GetNextEntry();
2216 * \brief Walk recursively the given \ref DocEntrySet, and feed
2217 * the given hash table (\ref TagDocEntryHT) with all the
2218 * \ref DocEntry (Dicom entries) encountered.
2219 * This method does the job for \ref BuildFlatHashTable.
2220 * @param builtHT Where to collect all the \ref DocEntry encountered
2221 * when recursively walking the given set.
2222 * @param set The structure to be traversed (recursively).
2224 /*void Document::BuildFlatHashTableRecurse( TagDocEntryHT &builtHT,
2227 if (ElementSet *elementSet = dynamic_cast< ElementSet* > ( set ) )
2229 TagDocEntryHT const ¤tHT = elementSet->GetTagHT();
2230 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2231 i != currentHT.end();
2234 DocEntry *entry = i->second;
2235 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2237 const ListSQItem& items = seqEntry->GetSQItems();
2238 for( ListSQItem::const_iterator item = items.begin();
2239 item != items.end();
2242 BuildFlatHashTableRecurse( builtHT, *item );
2246 builtHT[entry->GetKey()] = entry;
2251 if (SQItem *SQItemSet = dynamic_cast< SQItem* > ( set ) )
2253 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
2254 for (ListDocEntry::const_iterator i = currentList.begin();
2255 i != currentList.end();
2258 DocEntry *entry = *i;
2259 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2261 const ListSQItem& items = seqEntry->GetSQItems();
2262 for( ListSQItem::const_iterator item = items.begin();
2263 item != items.end();
2266 BuildFlatHashTableRecurse( builtHT, *item );
2270 builtHT[entry->GetKey()] = entry;
2277 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2280 * The structure used by a Document (through \ref ElementSet),
2281 * in order to hold the parsed entries of a Dicom header, is a recursive
2282 * one. This is due to the fact that the sequences (when present)
2283 * can be nested. Additionaly, the sequence items (represented in
2284 * gdcm as \ref SQItem) add an extra complexity to the data
2285 * structure. Hence, a gdcm user whishing to visit all the entries of
2286 * a Dicom header will need to dig in the gdcm internals (which
2287 * implies exposing all the internal data structures to the API).
2288 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2289 * recursively builds a temporary hash table, which holds all the
2290 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
2292 * \warning Of course there is NO integrity constrain between the
2293 * returned \ref TagDocEntryHT and the \ref ElementSet used
2294 * to build it. Hence if the underlying \ref ElementSet is
2295 * altered, then it is the caller responsability to invoke
2296 * \ref BuildFlatHashTable again...
2297 * @return The flat std::map<> we juste build.
2299 /*TagDocEntryHT *Document::BuildFlatHashTable()
2301 TagDocEntryHT *FlatHT = new TagDocEntryHT;
2302 BuildFlatHashTableRecurse( *FlatHT, this );
2306 } // end namespace gdcm
2308 //-----------------------------------------------------------------------------