1 /*=========================================================================
4 Module: $RCSfile: gdcmDocument.cxx,v $
6 Date: $Date: 2005/01/26 17:17:31 $
7 Version: $Revision: 1.214 $
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()
350 HasDCMPreamble = false;
351 if (Filename.length() == 0)
358 gdcmVerboseMacro( "File already open: " << Filename.c_str());
362 Fp = new std::ifstream(Filename.c_str(), std::ios::in | std::ios::binary);
365 gdcmDebugMacro( "Cannot open file: " << Filename.c_str());
372 Fp->read((char*)&zero, (size_t)2);
379 //ACR -- or DICOM with no Preamble; may start with a Shadow Group --
381 zero == 0x0001 || zero == 0x0100 || zero == 0x0002 || zero == 0x0200 ||
382 zero == 0x0003 || zero == 0x0300 || zero == 0x0004 || zero == 0x0400 ||
383 zero == 0x0005 || zero == 0x0500 || zero == 0x0006 || zero == 0x0600 ||
384 zero == 0x0007 || zero == 0x0700 || zero == 0x0008 || zero == 0x0800 )
387 = Util::Format("ACR/DICOM with no preamble: (%04x)\n", zero);
388 gdcmVerboseMacro( msg.c_str() );
393 Fp->seekg(126L, std::ios::cur);
395 Fp->read(dicm, (size_t)4);
401 if( memcmp(dicm, "DICM", 4) == 0 )
403 HasDCMPreamble = true;
408 gdcmVerboseMacro( "Not DICOM/ACR (missing preamble)" << Filename.c_str());
414 * \brief closes the file
415 * @return TRUE if the close was successfull
417 bool Document::CloseFile()
425 return true; //FIXME how do we detect a non-closed ifstream ?
429 * \brief Writes in a file all the Header Entries (Dicom Elements)
430 * @param fp file pointer on an already open file (actually: Output File Stream)
431 * @param filetype Type of the File to be written
432 * (ACR-NEMA, ExplicitVR, ImplicitVR)
433 * @return Always true.
435 void Document::WriteContent(std::ofstream *fp, FileType filetype)
437 // \TODO move the following lines (and a lot of others, to be written)
438 // to a future function CheckAndCorrectHeader
440 // (necessary if user wants to write a DICOM V3 file
441 // starting from an ACR-NEMA (V2) Header
443 if ( filetype == ImplicitVR || filetype == ExplicitVR )
445 // writing Dicom File Preamble
446 char filePreamble[128];
447 memset(filePreamble, 0, 128);
448 fp->write(filePreamble, 128);
449 fp->write("DICM", 4);
453 * \todo rewrite later, if really usefull
454 * - 'Group Length' element is optional in DICOM
455 * - but un-updated odd groups lengthes can causes pb
458 * if ( (filetype == ImplicitVR) || (filetype == ExplicitVR) )
459 * UpdateGroupLength(false,filetype);
460 * if ( filetype == ACR)
461 * UpdateGroupLength(true,ACR);
464 ElementSet::WriteContent(fp, filetype); // This one is recursive
467 //-----------------------------------------------------------------------------
470 * \brief Loads (from disk) the element content
471 * when a string is not suitable
472 * @param group group number of the Entry
473 * @param elem element number of the Entry
475 void Document::LoadEntryBinArea(uint16_t group, uint16_t elem)
477 // Search the corresponding DocEntry
478 DocEntry *docElement = GetDocEntry(group, elem);
482 BinEntry *binElement = dynamic_cast<BinEntry *>(docElement);
486 LoadEntryBinArea(binElement);
490 * \brief Loads (from disk) the element content
491 * when a string is not suitable
492 * @param elem Entry whose binArea is going to be loaded
494 void Document::LoadEntryBinArea(BinEntry *elem)
496 if(elem->GetBinArea())
503 size_t o =(size_t)elem->GetOffset();
504 Fp->seekg(o, std::ios::beg);
506 size_t l = elem->GetLength();
507 uint8_t *a = new uint8_t[l];
510 gdcmVerboseMacro( "Cannot allocate BinEntry content");
514 /// \todo check the result
515 Fp->read((char*)a, l);
516 if( Fp->fail() || Fp->eof())
529 * \brief Sets a 'non string' value to a given Dicom Element
530 * @param area area containing the 'non string' value
531 * @param group Group number of the searched Dicom Element
532 * @param elem Element number of the searched Dicom Element
535 /*bool Document::SetEntryBinArea(uint8_t *area,
536 uint16_t group, uint16_t elem)
538 DocEntry *currentEntry = GetDocEntry(group, elem);
544 if ( BinEntry *binEntry = dynamic_cast<BinEntry*>(currentEntry) )
546 binEntry->SetBinArea( area );
555 * \brief Loads the element while preserving the current
556 * underlying file position indicator as opposed to
557 * LoadDocEntry that modifies it.
558 * @param entry Header Entry whose value will be loaded.
561 void Document::LoadDocEntrySafe(DocEntry *entry)
565 long PositionOnEntry = Fp->tellg();
567 Fp->seekg(PositionOnEntry, std::ios::beg);
572 * \brief Swaps back the bytes of 4-byte long integer accordingly to
574 * @return The properly swaped 32 bits integer.
576 uint32_t Document::SwapLong(uint32_t a)
583 a=( ((a<<24) & 0xff000000) | ((a<<8) & 0x00ff0000) |
584 ((a>>8) & 0x0000ff00) | ((a>>24) & 0x000000ff) );
587 a=( ((a<<16) & 0xffff0000) | ((a>>16) & 0x0000ffff) );
590 a=( ((a<< 8) & 0xff00ff00) | ((a>>8) & 0x00ff00ff) );
593 gdcmErrorMacro( "Unset swap code:" << SwapCode );
600 * \brief Unswaps back the bytes of 4-byte long integer accordingly to
602 * @return The properly unswaped 32 bits integer.
604 uint32_t Document::UnswapLong(uint32_t a)
610 * \brief Swaps the bytes so they agree with the processor order
611 * @return The properly swaped 16 bits integer.
613 uint16_t Document::SwapShort(uint16_t a)
615 if ( SwapCode == 4321 || SwapCode == 2143 )
617 a = ((( a << 8 ) & 0x0ff00 ) | (( a >> 8 ) & 0x00ff ) );
623 * \brief Unswaps the bytes so they agree with the processor order
624 * @return The properly unswaped 16 bits integer.
626 uint16_t Document::UnswapShort(uint16_t a)
631 //-----------------------------------------------------------------------------
635 * \brief Parses a DocEntrySet (Zero-level DocEntries or SQ Item DocEntries)
636 * @return length of the parsed set.
638 void Document::ParseDES(DocEntrySet *set, long offset,
639 long l_max, bool delim_mode)
641 DocEntry *newDocEntry = 0;
642 ValEntry *newValEntry;
643 BinEntry *newBinEntry;
644 SeqEntry *newSeqEntry;
650 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
656 newDocEntry = ReadNextDocEntry( );
663 vr = newDocEntry->GetVR();
664 newValEntry = dynamic_cast<ValEntry*>(newDocEntry);
665 newBinEntry = dynamic_cast<BinEntry*>(newDocEntry);
666 newSeqEntry = dynamic_cast<SeqEntry*>(newDocEntry);
668 if ( newValEntry || newBinEntry )
672 if ( Filetype == ExplicitVR &&
673 !Global::GetVR()->IsVROfBinaryRepresentable(vr) )
675 ////// Neither ValEntry NOR BinEntry: should mean UNKOWN VR
676 gdcmVerboseMacro( std::hex << newDocEntry->GetGroup()
677 << "|" << newDocEntry->GetElement()
678 << " : Neither Valentry, nor BinEntry."
679 "Probably unknown VR.");
682 //////////////////// BinEntry or UNKOWN VR:
683 // When "this" is a Document the Key is simply of the
684 // form ( group, elem )...
685 if ( dynamic_cast< Document* > ( set ) )
687 newBinEntry->SetKey( newBinEntry->GetKey() );
689 // but when "this" is a SQItem, we are inserting this new
690 // valEntry in a sequence item, and the key has the
691 // generalized form (refer to \ref BaseTagKey):
692 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
694 newBinEntry->SetKey( parentSQItem->GetBaseTagKey()
695 + newBinEntry->GetKey() );
698 LoadDocEntry( newBinEntry );
699 if( !set->AddEntry( newBinEntry ) )
701 //Expect big troubles if here
702 //delete newBinEntry;
708 /////////////////////// ValEntry
709 // When "set" is a Document, then we are at the top of the
710 // hierarchy and the Key is simply of the form ( group, elem )...
711 if ( dynamic_cast< Document* > ( set ) )
713 newValEntry->SetKey( newValEntry->GetKey() );
715 // ...but when "set" is a SQItem, we are inserting this new
716 // valEntry in a sequence item. Hence the key has the
717 // generalized form (refer to \ref BaseTagKey):
718 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
720 newValEntry->SetKey( parentSQItem->GetBaseTagKey()
721 + newValEntry->GetKey() );
724 LoadDocEntry( newValEntry );
725 bool delimitor=newValEntry->IsItemDelimitor();
726 if( !set->AddEntry( newValEntry ) )
728 // If here expect big troubles
729 //delete newValEntry; //otherwise mem leak
739 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
747 // Just to make sure we are at the beginning of next entry.
748 SkipToNextDocEntry(newDocEntry);
753 unsigned long l = newDocEntry->GetReadLength();
754 if ( l != 0 ) // don't mess the delim_mode for zero-length sequence
756 if ( l == 0xffffffff )
765 // no other way to create it ...
766 newSeqEntry->SetDelimitorMode( delim_mode );
768 // At the top of the hierarchy, stands a Document. When "set"
769 // is a Document, then we are building the first depth level.
770 // Hence the SeqEntry we are building simply has a depth
772 if (/*Document *dummy =*/ dynamic_cast< Document* > ( set ) )
775 newSeqEntry->SetDepthLevel( 1 );
776 newSeqEntry->SetKey( newSeqEntry->GetKey() );
778 // But when "set" is already a SQItem, we are building a nested
779 // sequence, and hence the depth level of the new SeqEntry
780 // we are building, is one level deeper:
781 if (SQItem *parentSQItem = dynamic_cast< SQItem* > ( set ) )
783 newSeqEntry->SetDepthLevel( parentSQItem->GetDepthLevel() + 1 );
784 newSeqEntry->SetKey( parentSQItem->GetBaseTagKey()
785 + newSeqEntry->GetKey() );
789 { // Don't try to parse zero-length sequences
790 ParseSQ( newSeqEntry,
791 newDocEntry->GetOffset(),
794 if( !set->AddEntry( newSeqEntry ) )
798 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
812 * \brief Parses a Sequence ( SeqEntry after SeqEntry)
813 * @return parsed length for this level
815 void Document::ParseSQ( SeqEntry *seqEntry,
816 long offset, long l_max, bool delim_mode)
818 int SQItemNumber = 0;
820 long offsetStartCurrentSQItem = offset;
824 // the first time, we read the fff0,e000 of the first SQItem
825 DocEntry *newDocEntry = ReadNextDocEntry();
829 // FIXME Should warn user
834 if ( newDocEntry->IsSequenceDelimitor() )
836 seqEntry->SetDelimitationItem( newDocEntry );
840 if ( !delim_mode && ((long)(Fp->tellg())-offset) >= l_max)
845 // create the current SQItem
846 SQItem *itemSQ = new SQItem( seqEntry->GetDepthLevel() );
847 std::ostringstream newBase;
848 newBase << seqEntry->GetKey()
852 itemSQ->SetBaseTagKey( newBase.str() );
853 unsigned int l = newDocEntry->GetReadLength();
855 if ( l == 0xffffffff )
864 // when we're here, element fffe,e000 is already passed.
865 // it's lost for the SQItem we're going to process !!
867 //ParseDES(itemSQ, newDocEntry->GetOffset(), l, dlm_mod);
868 //delete newDocEntry; // FIXME well ... it's too late to use it !
870 // Let's try :------------
871 // remove fff0,e000, created out of the SQItem
873 Fp->seekg(offsetStartCurrentSQItem, std::ios::beg);
874 // fill up the current SQItem, starting at the beginning of fff0,e000
875 ParseDES(itemSQ, offsetStartCurrentSQItem, l+8, dlm_mod);
876 offsetStartCurrentSQItem = Fp->tellg();
877 // end try -----------------
879 seqEntry->AddSQItem( itemSQ, SQItemNumber );
881 if ( !delim_mode && ((long)(Fp->tellg())-offset ) >= l_max )
889 * \brief Loads the element content if its length doesn't exceed
890 * the value specified with Document::SetMaxSizeLoadEntry()
891 * @param entry Header Entry (Dicom Element) to be dealt with
893 void Document::LoadDocEntry(DocEntry *entry)
895 uint16_t group = entry->GetGroup();
896 std::string vr = entry->GetVR();
897 uint32_t length = entry->GetLength();
899 Fp->seekg((long)entry->GetOffset(), std::ios::beg);
901 // A SeQuence "contains" a set of Elements.
902 // (fffe e000) tells us an Element is beginning
903 // (fffe e00d) tells us an Element just ended
904 // (fffe e0dd) tells us the current SeQuence just ended
905 if( group == 0xfffe )
907 // NO more value field for SQ !
911 // When the length is zero things are easy:
914 ((ValEntry *)entry)->SetValue("");
918 // The elements whose length is bigger than the specified upper bound
919 // are not loaded. Instead we leave a short notice of the offset of
920 // the element content and it's length.
922 std::ostringstream s;
923 if (length > MaxSizeLoadEntry)
925 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
927 //s << "gdcm::NotLoaded (BinEntry)";
929 s << " Address:" << (long)entry->GetOffset();
930 s << " Length:" << entry->GetLength();
931 s << " x(" << std::hex << entry->GetLength() << ")";
932 binEntryPtr->SetValue(s.str());
934 // Be carefull : a BinEntry IS_A ValEntry ...
935 else if (ValEntry *valEntryPtr = dynamic_cast< ValEntry* >(entry) )
937 // s << "gdcm::NotLoaded. (ValEntry)";
939 s << " Address:" << (long)entry->GetOffset();
940 s << " Length:" << entry->GetLength();
941 s << " x(" << std::hex << entry->GetLength() << ")";
942 valEntryPtr->SetValue(s.str());
947 gdcmErrorMacro( "MaxSizeLoadEntry exceeded, neither a BinEntry "
948 << "nor a ValEntry ?! Should never print that !" );
951 // to be sure we are at the end of the value ...
952 Fp->seekg((long)entry->GetOffset()+(long)entry->GetLength(),
957 // When we find a BinEntry not very much can be done :
958 if (BinEntry *binEntryPtr = dynamic_cast< BinEntry* >(entry) )
961 binEntryPtr->SetValue(s.str());
962 LoadEntryBinArea(binEntryPtr); // last one, not to erase length !
966 /// \todo Any compacter code suggested (?)
967 if ( IsDocEntryAnInteger(entry) )
971 // When short integer(s) are expected, read and convert the following
972 // n *two characters properly i.e. consider them as short integers as
973 // opposed to strings.
974 // Elements with Value Multiplicity > 1
975 // contain a set of integers (not a single one)
976 if (vr == "US" || vr == "SS")
979 NewInt = ReadInt16();
983 for (int i=1; i < nbInt; i++)
986 NewInt = ReadInt16();
991 // See above comment on multiple integers (mutatis mutandis).
992 else if (vr == "UL" || vr == "SL")
995 NewInt = ReadInt32();
999 for (int i=1; i < nbInt; i++)
1002 NewInt = ReadInt32();
1007 #ifdef GDCM_NO_ANSI_STRING_STREAM
1008 s << std::ends; // to avoid oddities on Solaris
1009 #endif //GDCM_NO_ANSI_STRING_STREAM
1011 ((ValEntry *)entry)->SetValue(s.str());
1015 // FIXME: We need an additional byte for storing \0 that is not on disk
1016 char *str = new char[length+1];
1017 Fp->read(str, (size_t)length);
1018 str[length] = '\0'; //this is only useful when length is odd
1019 // Special DicomString call to properly handle \0 and even length
1020 std::string newValue;
1023 newValue = Util::DicomString(str, length+1);
1024 gdcmVerboseMacro("Warning: bad length: " << length <<
1025 ",For string :" << newValue.c_str());
1026 // Since we change the length of string update it length
1027 //entry->SetReadLength(length+1);
1031 newValue = Util::DicomString(str, length);
1035 if ( ValEntry *valEntry = dynamic_cast<ValEntry* >(entry) )
1037 if ( Fp->fail() || Fp->eof())
1039 gdcmVerboseMacro("Unread element value");
1040 valEntry->SetValue(GDCM_UNREAD);
1046 // Because of correspondance with the VR dic
1047 valEntry->SetValue(newValue);
1051 valEntry->SetValue(newValue);
1056 gdcmErrorMacro( "Should have a ValEntry, here !");
1062 * \brief Find the value Length of the passed Header Entry
1063 * @param entry Header Entry whose length of the value shall be loaded.
1065 void Document::FindDocEntryLength( DocEntry *entry )
1066 throw ( FormatError )
1068 std::string vr = entry->GetVR();
1071 if ( Filetype == ExplicitVR && !entry->IsImplicitVR() )
1073 if ( vr == "OB" || vr == "OW" || vr == "SQ" || vr == "UN" )
1075 // The following reserved two bytes (see PS 3.5-2003, section
1076 // "7.1.2 Data element structure with explicit vr", p 27) must be
1077 // skipped before proceeding on reading the length on 4 bytes.
1078 Fp->seekg( 2L, std::ios::cur);
1079 uint32_t length32 = ReadInt32();
1081 if ( (vr == "OB" || vr == "OW") && length32 == 0xffffffff )
1086 lengthOB = FindDocEntryLengthOBOrOW();
1088 catch ( FormatUnexpected )
1090 // Computing the length failed (this happens with broken
1091 // files like gdcm-JPEG-LossLess3a.dcm). We still have a
1092 // chance to get the pixels by deciding the element goes
1093 // until the end of the file. Hence we artificially fix the
1094 // the length and proceed.
1095 long currentPosition = Fp->tellg();
1096 Fp->seekg(0L,std::ios::end);
1098 long lengthUntilEOF = (long)(Fp->tellg())-currentPosition;
1099 Fp->seekg(currentPosition, std::ios::beg);
1101 entry->SetReadLength(lengthUntilEOF);
1102 entry->SetLength(lengthUntilEOF);
1105 entry->SetReadLength(lengthOB);
1106 entry->SetLength(lengthOB);
1109 FixDocEntryFoundLength(entry, length32);
1113 // Length is encoded on 2 bytes.
1114 length16 = ReadInt16();
1116 // FIXME : This heuristic supposes that the first group following
1117 // group 0002 *has* and element 0000.
1118 // BUT ... Element 0000 is optionnal :-(
1121 // Fixed using : HandleOutOfGroup0002()
1122 // (first hereafter strategy ...)
1124 // We can tell the current file is encoded in big endian (like
1125 // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag
1126 // and it's value is the one of the encoding of a big endian file.
1127 // In order to deal with such big endian encoded files, we have
1128 // (at least) two strategies:
1129 // * when we load the "Transfer Syntax" tag with value of big endian
1130 // encoding, we raise the proper flags. Then we wait for the end
1131 // of the META group (0x0002) among which is "Transfer Syntax",
1132 // before switching the swap code to big endian. We have to postpone
1133 // the switching of the swap code since the META group is fully encoded
1134 // in little endian, and big endian coding only starts at the next
1135 // group. The corresponding code can be hard to analyse and adds
1136 // many additional unnecessary tests for regular tags.
1137 // * the second strategy consists in waiting for trouble, that shall
1138 // appear when we find the first group with big endian encoding. This
1139 // is easy to detect since the length of a "Group Length" tag (the
1140 // ones with zero as element number) has to be of 4 (0x0004). When we
1141 // encounter 1024 (0x0400) chances are the encoding changed and we
1142 // found a group with big endian encoding.
1143 //---> Unfortunately, element 0000 is optional.
1144 //---> This will not work when missing!
1145 // We shall use this second strategy. In order to make sure that we
1146 // can interpret the presence of an apparently big endian encoded
1147 // length of a "Group Length" without committing a big mistake, we
1148 // add an additional check: we look in the already parsed elements
1149 // for the presence of a "Transfer Syntax" whose value has to be "big
1150 // endian encoding". When this is the case, chances are we have got our
1151 // hands on a big endian encoded file: we switch the swap code to
1152 // big endian and proceed...
1154 // if ( element == 0x0000 && length16 == 0x0400 )
1156 // std::string ts = GetTransferSyntax();
1157 // if ( Global::GetTS()->GetSpecialTransferSyntax(ts)
1158 // != TS::ExplicitVRBigEndian )
1160 // throw FormatError( "Document::FindDocEntryLength()",
1161 // " not explicit VR." );
1165 // SwitchByteSwapCode();
1167 // // Restore the unproperly loaded values i.e. the group, the element
1168 // // and the dictionary entry depending on them.
1169 // uint16_t correctGroup = SwapShort( entry->GetGroup() );
1170 // uint16_t correctElem = SwapShort( entry->GetElement() );
1171 // DictEntry *newTag = GetDictEntry( correctGroup, correctElem );
1174 // // This correct tag is not in the dictionary. Create a new one.
1175 // newTag = NewVirtualDictEntry(correctGroup, correctElem);
1177 // // FIXME this can create a memory leaks on the old entry that be
1178 // // left unreferenced.
1179 // entry->SetDictEntry( newTag );
1182 // 0xffff means that we deal with 'No Length' Sequence
1183 // or 'No Length' SQItem
1184 if ( length16 == 0xffff)
1188 FixDocEntryFoundLength( entry, (uint32_t)length16 );
1193 // Either implicit VR or a non DICOM conformal (see note below) explicit
1194 // VR that ommited the VR of (at least) this element. Farts happen.
1195 // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25
1196 // on Data elements "Implicit and Explicit VR Data Elements shall
1197 // not coexist in a Data Set and Data Sets nested within it".]
1198 // Length is on 4 bytes.
1200 // Well ... group 0002 is always coded in 'Explicit VR Litle Endian'
1201 // even if Transfer Syntax is 'Implicit VR ...'
1203 FixDocEntryFoundLength( entry, ReadInt32() );
1209 * \brief Find the Value Representation of the current Dicom Element.
1210 * @return Value Representation of the current Entry
1212 std::string Document::FindDocEntryVR()
1214 if ( Filetype != ExplicitVR )
1215 return GDCM_UNKNOWN;
1217 long positionOnEntry = Fp->tellg();
1218 // Warning: we believe this is explicit VR (Value Representation) because
1219 // we used a heuristic that found "UL" in the first tag. Alas this
1220 // doesn't guarantee that all the tags will be in explicit VR. In some
1221 // cases (see e-film filtered files) one finds implicit VR tags mixed
1222 // within an explicit VR file. Hence we make sure the present tag
1223 // is in explicit VR and try to fix things if it happens not to be
1227 Fp->read (vr, (size_t)2);
1230 if( !CheckDocEntryVR(vr) )
1232 Fp->seekg(positionOnEntry, std::ios::beg);
1233 return GDCM_UNKNOWN;
1239 * \brief Check the correspondance between the VR of the header entry
1240 * and the taken VR. If they are different, the header entry is
1241 * updated with the new VR.
1242 * @param vr Dicom Value Representation
1243 * @return false if the VR is incorrect of if the VR isn't referenced
1244 * otherwise, it returns true
1246 bool Document::CheckDocEntryVR(VRKey vr)
1248 // CLEANME searching the dicom_vr at each occurence is expensive.
1249 // PostPone this test in an optional integrity check at the end
1250 // of parsing or only in debug mode.
1251 if ( !Global::GetVR()->IsValidVR(vr) )
1258 * \brief Get the transformed value of the header entry. The VR value
1259 * is used to define the transformation to operate on the value
1260 * \warning NOT end user intended method !
1261 * @param entry entry to tranform
1262 * @return Transformed entry value
1264 std::string Document::GetDocEntryValue(DocEntry *entry)
1266 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1268 std::string val = ((ValEntry *)entry)->GetValue();
1269 std::string vr = entry->GetVR();
1270 uint32_t length = entry->GetLength();
1271 std::ostringstream s;
1274 // When short integer(s) are expected, read and convert the following
1275 // n * 2 bytes properly i.e. as a multivaluated strings
1276 // (each single value is separated fromthe next one by '\'
1277 // as usual for standard multivaluated filels
1278 // Elements with Value Multiplicity > 1
1279 // contain a set of short integers (not a single one)
1281 if( vr == "US" || vr == "SS" )
1286 for (int i=0; i < nbInt; i++)
1292 newInt16 = ( val[2*i+0] & 0xFF ) + ( ( val[2*i+1] & 0xFF ) << 8);
1293 newInt16 = SwapShort( newInt16 );
1298 // When integer(s) are expected, read and convert the following
1299 // n * 4 bytes properly i.e. as a multivaluated strings
1300 // (each single value is separated fromthe next one by '\'
1301 // as usual for standard multivaluated filels
1302 // Elements with Value Multiplicity > 1
1303 // contain a set of integers (not a single one)
1304 else if( vr == "UL" || vr == "SL" )
1309 for (int i=0; i < nbInt; i++)
1315 newInt32 = ( val[4*i+0] & 0xFF )
1316 + (( val[4*i+1] & 0xFF ) << 8 )
1317 + (( val[4*i+2] & 0xFF ) << 16 )
1318 + (( val[4*i+3] & 0xFF ) << 24 );
1319 newInt32 = SwapLong( newInt32 );
1323 #ifdef GDCM_NO_ANSI_STRING_STREAM
1324 s << std::ends; // to avoid oddities on Solaris
1325 #endif //GDCM_NO_ANSI_STRING_STREAM
1329 return ((ValEntry *)entry)->GetValue();
1333 * \brief Get the reverse transformed value of the header entry. The VR
1334 * value is used to define the reverse transformation to operate on
1336 * \warning NOT end user intended method !
1337 * @param entry Entry to reverse transform
1338 * @return Reverse transformed entry value
1340 std::string Document::GetDocEntryUnvalue(DocEntry *entry)
1342 if ( IsDocEntryAnInteger(entry) && entry->IsImplicitVR() )
1344 std::string vr = entry->GetVR();
1345 std::vector<std::string> tokens;
1346 std::ostringstream s;
1348 if ( vr == "US" || vr == "SS" )
1352 tokens.erase( tokens.begin(), tokens.end()); // clean any previous value
1353 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1354 for (unsigned int i=0; i<tokens.size(); i++)
1356 newInt16 = atoi(tokens[i].c_str());
1357 s << ( newInt16 & 0xFF )
1358 << (( newInt16 >> 8 ) & 0xFF );
1362 if ( vr == "UL" || vr == "SL")
1366 tokens.erase(tokens.begin(),tokens.end()); // clean any previous value
1367 Util::Tokenize (((ValEntry *)entry)->GetValue(), tokens, "\\");
1368 for (unsigned int i=0; i<tokens.size();i++)
1370 newInt32 = atoi(tokens[i].c_str());
1371 s << (char)( newInt32 & 0xFF )
1372 << (char)(( newInt32 >> 8 ) & 0xFF )
1373 << (char)(( newInt32 >> 16 ) & 0xFF )
1374 << (char)(( newInt32 >> 24 ) & 0xFF );
1379 #ifdef GDCM_NO_ANSI_STRING_STREAM
1380 s << std::ends; // to avoid oddities on Solaris
1381 #endif //GDCM_NO_ANSI_STRING_STREAM
1385 return ((ValEntry *)entry)->GetValue();
1389 * \brief Skip a given Header Entry
1390 * \warning NOT end user intended method !
1391 * @param entry entry to skip
1393 void Document::SkipDocEntry(DocEntry *entry)
1395 SkipBytes(entry->GetLength());
1399 * \brief Skips to the begining of the next Header Entry
1400 * \warning NOT end user intended method !
1401 * @param currentDocEntry entry to skip
1403 void Document::SkipToNextDocEntry(DocEntry *currentDocEntry)
1405 Fp->seekg((long)(currentDocEntry->GetOffset()), std::ios::beg);
1406 if (currentDocEntry->GetGroup() != 0xfffe) // for fffe pb
1407 Fp->seekg( (long)(currentDocEntry->GetReadLength()),std::ios::cur);
1411 * \brief When the length of an element value is obviously wrong (because
1412 * the parser went Jabberwocky) one can hope improving things by
1413 * applying some heuristics.
1414 * @param entry entry to check
1415 * @param foundLength first assumption about length
1417 void Document::FixDocEntryFoundLength(DocEntry *entry,
1418 uint32_t foundLength)
1420 entry->SetReadLength( foundLength ); // will be updated only if a bug is found
1421 if ( foundLength == 0xffffffff)
1426 uint16_t gr = entry->GetGroup();
1427 uint16_t elem = entry->GetElement();
1429 if ( foundLength % 2)
1431 gdcmVerboseMacro( "Warning : Tag with uneven length " << foundLength
1432 << " in x(" << std::hex << gr << "," << elem <<")");
1435 //////// Fix for some naughty General Electric images.
1436 // Allthough not recent many such GE corrupted images are still present
1437 // on Creatis hard disks. Hence this fix shall remain when such images
1438 // are no longer in use (we are talking a few years, here)...
1439 // Note: XMedCom probably uses such a trick since it is able to read
1440 // those pesky GE images ...
1441 if ( foundLength == 13)
1443 // Only happens for this length !
1444 if ( gr != 0x0008 || ( elem != 0x0070 && elem != 0x0080 ) )
1447 entry->SetReadLength(10); /// \todo a bug is to be fixed !?
1451 //////// Fix for some brain-dead 'Leonardo' Siemens images.
1452 // Occurence of such images is quite low (unless one leaves close to a
1453 // 'Leonardo' source. Hence, one might consider commenting out the
1454 // following fix on efficiency reasons.
1455 else if ( gr == 0x0009 && ( elem == 0x1113 || elem == 0x1114 ) )
1458 entry->SetReadLength(4); /// \todo a bug is to be fixed !?
1461 else if ( entry->GetVR() == "SQ" )
1463 foundLength = 0; // ReadLength is unchanged
1466 //////// We encountered a 'delimiter' element i.e. a tag of the form
1467 // "fffe|xxxx" which is just a marker. Delimiters length should not be
1468 // taken into account.
1469 else if( gr == 0xfffe )
1471 // According to the norm, fffe|0000 shouldn't exist. BUT the Philips
1472 // image gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm happens to
1473 // causes extra troubles...
1474 if( entry->GetElement() != 0x0000 )
1480 entry->SetLength(foundLength);
1484 * \brief Apply some heuristics to predict whether the considered
1485 * element value contains/represents an integer or not.
1486 * @param entry The element value on which to apply the predicate.
1487 * @return The result of the heuristical predicate.
1489 bool Document::IsDocEntryAnInteger(DocEntry *entry)
1491 uint16_t elem = entry->GetElement();
1492 uint16_t group = entry->GetGroup();
1493 const std::string &vr = entry->GetVR();
1494 uint32_t length = entry->GetLength();
1496 // When we have some semantics on the element we just read, and if we
1497 // a priori know we are dealing with an integer, then we shall be
1498 // able to swap it's element value properly.
1499 if ( elem == 0 ) // This is the group length of the group
1507 // Allthough this should never happen, still some images have a
1508 // corrupted group length [e.g. have a glance at offset x(8336) of
1509 // gdcmData/gdcm-MR-PHILIPS-16-Multi-Seq.dcm].
1510 // Since for dicom compliant and well behaved headers, the present
1511 // test is useless (and might even look a bit paranoid), when we
1512 // encounter such an ill-formed image, we simply display a warning
1513 // message and proceed on parsing (while crossing fingers).
1514 long filePosition = Fp->tellg();
1515 gdcmVerboseMacro( "Erroneous Group Length element length on : ("
1516 << std::hex << group << " , " << elem
1517 << ") -before- position x(" << filePosition << ")"
1518 << "lgt : " << length );
1522 if ( vr == "UL" || vr == "US" || vr == "SL" || vr == "SS" )
1530 * \brief Find the Length till the next sequence delimiter
1531 * \warning NOT end user intended method !
1535 uint32_t Document::FindDocEntryLengthOBOrOW()
1536 throw( FormatUnexpected )
1538 // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data.
1539 long positionOnEntry = Fp->tellg();
1540 bool foundSequenceDelimiter = false;
1541 uint32_t totalLength = 0;
1543 while ( !foundSequenceDelimiter )
1549 group = ReadInt16();
1552 catch ( FormatError )
1554 throw FormatError("Unexpected end of file encountered during ",
1555 "Document::FindDocEntryLengthOBOrOW()");
1558 // We have to decount the group and element we just read
1561 if ( group != 0xfffe || ( ( elem != 0xe0dd ) && ( elem != 0xe000 ) ) )
1563 long filePosition = Fp->tellg();
1564 gdcmVerboseMacro( "Neither an Item tag nor a Sequence delimiter tag on :"
1565 << std::hex << group << " , " << elem
1566 << ") -before- position x(" << filePosition << ")" );
1568 Fp->seekg(positionOnEntry, std::ios::beg);
1569 throw FormatUnexpected( "Neither an Item tag nor a Sequence delimiter tag.");
1572 if ( elem == 0xe0dd )
1574 foundSequenceDelimiter = true;
1577 uint32_t itemLength = ReadInt32();
1578 // We add 4 bytes since we just read the ItemLength with ReadInt32
1579 totalLength += itemLength + 4;
1580 SkipBytes(itemLength);
1582 if ( foundSequenceDelimiter )
1587 Fp->seekg( positionOnEntry, std::ios::beg);
1592 * \brief Reads a supposed to be 16 Bits integer
1593 * (swaps it depending on processor endianity)
1594 * @return read value
1596 uint16_t Document::ReadInt16()
1597 throw( FormatError )
1600 Fp->read ((char*)&g, (size_t)2);
1603 throw FormatError( "Document::ReadInt16()", " file error." );
1607 throw FormatError( "Document::ReadInt16()", "EOF." );
1614 * \brief Reads a supposed to be 32 Bits integer
1615 * (swaps it depending on processor endianity)
1616 * @return read value
1618 uint32_t Document::ReadInt32()
1619 throw( FormatError )
1622 Fp->read ((char*)&g, (size_t)4);
1625 throw FormatError( "Document::ReadInt32()", " file error." );
1629 throw FormatError( "Document::ReadInt32()", "EOF." );
1636 * \brief skips bytes inside the source file
1637 * \warning NOT end user intended method !
1640 void Document::SkipBytes(uint32_t nBytes)
1642 //FIXME don't dump the returned value
1643 Fp->seekg((long)nBytes, std::ios::cur);
1647 * \brief Loads all the needed Dictionaries
1648 * \warning NOT end user intended method !
1650 void Document::Initialize()
1652 RefPubDict = Global::GetDicts()->GetDefaultPubDict();
1658 * \brief Discover what the swap code is (among little endian, big endian,
1659 * bad little endian, bad big endian).
1661 * @return false when we are absolutely sure
1662 * it's neither ACR-NEMA nor DICOM
1663 * true when we hope ours assuptions are OK
1665 bool Document::CheckSwap()
1667 // The only guaranted way of finding the swap code is to find a
1668 // group tag since we know it's length has to be of four bytes i.e.
1669 // 0x00000004. Finding the swap code in then straigthforward. Trouble
1670 // occurs when we can't find such group...
1672 uint32_t x = 4; // x : for ntohs
1673 bool net2host; // true when HostByteOrder is the same as NetworkByteOrder
1679 // First, compare HostByteOrder and NetworkByteOrder in order to
1680 // determine if we shall need to swap bytes (i.e. the Endian type).
1681 if ( x == ntohs(x) )
1690 // The easiest case is the one of a 'true' DICOM header, we just have
1691 // to look for the string "DICM" inside the file preamble.
1694 char *entCur = deb + 128;
1695 if( memcmp(entCur, "DICM", (size_t)4) == 0 )
1697 gdcmVerboseMacro( "Looks like DICOM Version3 (preamble + DCM)" );
1699 // Group 0002 should always be VR, and the first element 0000
1700 // Let's be carefull (so many wrong headers ...)
1701 // and determine the value representation (VR) :
1702 // Let's skip to the first element (0002,0000) and check there if we find
1703 // "UL" - or "OB" if the 1st one is (0002,0001) -,
1704 // in which case we (almost) know it is explicit VR.
1705 // WARNING: if it happens to be implicit VR then what we will read
1706 // is the length of the group. If this ascii representation of this
1707 // length happens to be "UL" then we shall believe it is explicit VR.
1708 // We need to skip :
1709 // * the 128 bytes of File Preamble (often padded with zeroes),
1710 // * the 4 bytes of "DICM" string,
1711 // * the 4 bytes of the first tag (0002, 0000),or (0002, 0001)
1712 // i.e. a total of 136 bytes.
1715 // group 0x0002 *is always* Explicit VR Sometimes ,
1716 // even if elem 0002,0010 (Transfer Syntax) tells us the file is
1717 // *Implicit* VR (see former 'gdcmData/icone.dcm')
1719 if( memcmp(entCur, "UL", (size_t)2) == 0 ||
1720 memcmp(entCur, "OB", (size_t)2) == 0 ||
1721 memcmp(entCur, "UI", (size_t)2) == 0 ||
1722 memcmp(entCur, "CS", (size_t)2) == 0 ) // CS, to remove later
1723 // when Write DCM *adds*
1725 // Use Document::dicom_vr to test all the possibilities
1726 // instead of just checking for UL, OB and UI !? group 0000
1728 Filetype = ExplicitVR;
1729 gdcmVerboseMacro( "Group 0002 : Explicit Value Representation");
1733 Filetype = ImplicitVR;
1734 gdcmVerboseMacro( "Group 0002 :Not an explicit Value Representation;"
1735 << "Looks like a bugged Header!");
1741 gdcmVerboseMacro( "HostByteOrder != NetworkByteOrder");
1746 gdcmVerboseMacro( "HostByteOrder = NetworkByteOrder");
1749 // Position the file position indicator at first tag
1750 // (i.e. after the file preamble and the "DICM" string).
1751 Fp->seekg(0, std::ios::beg);
1752 Fp->seekg ( 132L, std::ios::beg);
1756 // Alas, this is not a DicomV3 file and whatever happens there is no file
1757 // preamble. We can reset the file position indicator to where the data
1758 // is (i.e. the beginning of the file).
1759 gdcmVerboseMacro( "Not a DICOM Version3 file");
1760 Fp->seekg(0, std::ios::beg);
1762 // Our next best chance would be to be considering a 'clean' ACR/NEMA file.
1763 // By clean we mean that the length of the first tag is written down.
1764 // If this is the case and since the length of the first group HAS to be
1765 // four (bytes), then determining the proper swap code is straightforward.
1768 // We assume the array of char we are considering contains the binary
1769 // representation of a 32 bits integer. Hence the following dirty
1771 s32 = *((uint32_t *)(entCur));
1792 // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file.
1793 // It is time for despaired wild guesses.
1794 // So, let's check if this file wouldn't happen to be 'dirty' ACR/NEMA,
1795 // i.e. the 'group length' element is not present :
1797 // check the supposed-to-be 'group number'
1798 // in ( 0x0001 .. 0x0008 )
1799 // to determine ' SwapCode' value .
1800 // Only 0 or 4321 will be possible
1801 // (no oportunity to check for the formerly well known
1802 // ACR-NEMA 'Bad Big Endian' or 'Bad Little Endian'
1803 // if unsuccessfull (i.e. neither 0x0002 nor 0x0200 etc -3, 4, ..., 8-)
1804 // the file IS NOT ACR-NEMA nor DICOM V3
1805 // Find a trick to tell it the caller...
1807 s16 = *((uint16_t *)(deb));
1834 gdcmVerboseMacro( "ACR/NEMA unfound swap info (Really hopeless !)");
1842 * \brief Change the Byte Swap code.
1844 void Document::SwitchByteSwapCode()
1846 gdcmVerboseMacro( "Switching Byte Swap code from "<< SwapCode);
1847 if ( SwapCode == 1234 )
1851 else if ( SwapCode == 4321 )
1855 else if ( SwapCode == 3412 )
1859 else if ( SwapCode == 2143 )
1866 * \brief during parsing, Header Elements too long are not loaded in memory
1869 void Document::SetMaxSizeLoadEntry(long newSize)
1875 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
1877 MaxSizeLoadEntry = 0xffffffff;
1880 MaxSizeLoadEntry = newSize;
1885 * \brief Header Elements too long will not be printed
1886 * \todo See comments of \ref Document::MAX_SIZE_PRINT_ELEMENT_VALUE
1889 void Document::SetMaxSizePrintEntry(long newSize)
1895 if ((uint32_t)newSize >= (uint32_t)0xffffffff )
1897 MaxSizePrintEntry = 0xffffffff;
1900 MaxSizePrintEntry = newSize;
1906 * \brief Handle broken private tag from Philips NTSCAN
1907 * where the endianess is being switch to BigEndian for no
1911 void Document::HandleBrokenEndian(uint16_t &group, uint16_t &elem)
1913 // Endian reversion. Some files contain groups of tags with reversed endianess.
1914 static int reversedEndian = 0;
1915 // try to fix endian switching in the middle of headers
1916 if ((group == 0xfeff) && (elem == 0x00e0))
1918 // start endian swap mark for group found
1920 SwitchByteSwapCode();
1925 else if (group == 0xfffe && elem == 0xe00d && reversedEndian)
1927 // end of reversed endian group
1929 SwitchByteSwapCode();
1934 * \brief Accesses the info from 0002,0010 : Transfer Syntax and TS
1935 * @return The full Transfer Syntax Name (as opposed to Transfer Syntax UID)
1937 std::string Document::GetTransferSyntaxName()
1939 // use the TS (TS : Transfer Syntax)
1940 std::string transferSyntax = GetEntryValue(0x0002,0x0010);
1942 if ( (transferSyntax.find(GDCM_NOTLOADED) < transferSyntax.length()) )
1944 gdcmErrorMacro( "Transfer Syntax not loaded. " << std::endl
1945 << "Better you increase MAX_SIZE_LOAD_ELEMENT_VALUE" );
1946 return "Uncompressed ACR-NEMA";
1948 if ( transferSyntax == GDCM_UNFOUND )
1950 gdcmVerboseMacro( "Unfound Transfer Syntax (0002,0010)");
1951 return "Uncompressed ACR-NEMA";
1954 // we do it only when we need it
1955 const TSKey &tsName = Global::GetTS()->GetValue( transferSyntax );
1957 // Global::GetTS() is a global static you shall never try to delete it!
1962 * \brief Group 0002 is always coded Little Endian
1963 * whatever Transfer Syntax is
1966 void Document::HandleOutOfGroup0002(uint16_t &group, uint16_t &elem)
1968 // Endian reversion. Some files contain groups of tags with reversed endianess.
1969 if ( !Group0002Parsed && group != 0x0002)
1971 Group0002Parsed = true;
1972 // we just came out of group 0002
1973 // if Transfer syntax is Big Endian we have to change CheckSwap
1975 std::string ts = GetTransferSyntax();
1976 if ( !Global::GetTS()->IsTransferSyntax(ts) )
1978 gdcmVerboseMacro("True DICOM File, with NO Tansfer Syntax: " << ts );
1982 // Group 0002 is always 'Explicit ...' enven when Transfer Syntax says 'Implicit ..."
1984 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ImplicitVRLittleEndian )
1986 Filetype = ImplicitVR;
1989 // FIXME Strangely, this works with
1990 //'Implicit VR Transfer Syntax (GE Private)
1991 if ( Global::GetTS()->GetSpecialTransferSyntax(ts) == TS::ExplicitVRBigEndian )
1993 gdcmVerboseMacro("Transfer Syntax Name = ["
1994 << GetTransferSyntaxName() << "]" );
1995 SwitchByteSwapCode();
1996 group = SwapShort(group);
1997 elem = SwapShort(elem);
2003 * \brief Read the next tag but WITHOUT loading it's value
2004 * (read the 'Group Number', the 'Element Number',
2005 * gets the Dict Entry
2006 * gets the VR, gets the length, gets the offset value)
2007 * @return On succes the newly created DocEntry, NULL on failure.
2009 DocEntry *Document::ReadNextDocEntry()
2016 group = ReadInt16();
2019 catch ( FormatError e )
2021 // We reached the EOF (or an error occured) therefore
2022 // header parsing has to be considered as finished.
2027 // Sometimes file contains groups of tags with reversed endianess.
2028 HandleBrokenEndian(group, elem);
2030 // In 'true DICOM' files Group 0002 is always little endian
2031 if ( HasDCMPreamble )
2032 HandleOutOfGroup0002(group, elem);
2034 std::string vr = FindDocEntryVR();
2035 std::string realVR = vr;
2037 if( vr == GDCM_UNKNOWN)
2039 DictEntry *dictEntry = GetDictEntry(group,elem);
2041 realVR = dictEntry->GetVR();
2045 if( Global::GetVR()->IsVROfSequence(realVR) )
2046 newEntry = NewSeqEntry(group, elem);
2047 else if( Global::GetVR()->IsVROfStringRepresentable(realVR) )
2048 newEntry = NewValEntry(group, elem,vr);
2050 newEntry = NewBinEntry(group, elem,vr);
2052 if( vr == GDCM_UNKNOWN )
2054 if( Filetype == ExplicitVR )
2056 // We thought this was explicit VR, but we end up with an
2057 // implicit VR tag. Let's backtrack.
2058 if ( newEntry->GetGroup() != 0xfffe )
2061 msg = Util::Format("Entry (%04x,%04x) should be Explicit VR\n",
2062 newEntry->GetGroup(), newEntry->GetElement());
2063 gdcmVerboseMacro( msg.c_str() );
2066 newEntry->SetImplicitVR();
2071 FindDocEntryLength(newEntry);
2073 catch ( FormatError e )
2081 newEntry->SetOffset(Fp->tellg());
2086 //GenerateFreeTagKeyInGroup? What was it designed for ?!?
2088 * \brief Generate a free TagKey i.e. a TagKey that is not present
2089 * in the TagHt dictionary.
2090 * @param group The generated tag must belong to this group.
2091 * @return The element of tag with given group which is fee.
2093 //uint32_t Document::GenerateFreeTagKeyInGroup(uint16_t group)
2095 // for (uint32_t elem = 0; elem < UINT32_MAX; elem++)
2097 // TagKey key = DictEntry::TranslateToKey(group, elem);
2098 // if (TagHT.count(key) == 0)
2103 // return UINT32_MAX;
2107 * \brief Compares two documents, according to \ref DicomDir rules
2108 * \warning Does NOT work with ACR-NEMA files
2109 * \todo Find a trick to solve the pb (use RET fields ?)
2111 * @return true if 'smaller'
2113 bool Document::operator<(Document &document)
2116 std::string s1 = GetEntryValue(0x0010,0x0010);
2117 std::string s2 = document.GetEntryValue(0x0010,0x0010);
2129 s1 = GetEntryValue(0x0010,0x0020);
2130 s2 = document.GetEntryValue(0x0010,0x0020);
2141 // Study Instance UID
2142 s1 = GetEntryValue(0x0020,0x000d);
2143 s2 = document.GetEntryValue(0x0020,0x000d);
2154 // Serie Instance UID
2155 s1 = GetEntryValue(0x0020,0x000e);
2156 s2 = document.GetEntryValue(0x0020,0x000e);
2172 * \brief Re-computes the length of a ACR-NEMA/Dicom group from a DcmHeader
2173 * @param filetype Type of the File to be written
2175 int Document::ComputeGroup0002Length( FileType filetype )
2180 int groupLength = 0;
2181 bool found0002 = false;
2183 // for each zero-level Tag in the DCM Header
2184 DocEntry *entry = GetFirstEntry();
2187 gr = entry->GetGroup();
2192 vr = entry->GetVR();
2194 if (filetype == ExplicitVR)
2196 if ( (vr == "OB") || (vr == "OW") || (vr == "SQ") )
2198 groupLength += 4; // explicit VR AND OB, OW, SQ : 4 more bytes
2201 groupLength += 2 + 2 + 4 + entry->GetLength();
2203 else if (found0002 )
2206 entry = GetNextEntry();
2212 * \brief Walk recursively the given \ref DocEntrySet, and feed
2213 * the given hash table (\ref TagDocEntryHT) with all the
2214 * \ref DocEntry (Dicom entries) encountered.
2215 * This method does the job for \ref BuildFlatHashTable.
2216 * @param builtHT Where to collect all the \ref DocEntry encountered
2217 * when recursively walking the given set.
2218 * @param set The structure to be traversed (recursively).
2220 /*void Document::BuildFlatHashTableRecurse( TagDocEntryHT &builtHT,
2223 if (ElementSet *elementSet = dynamic_cast< ElementSet* > ( set ) )
2225 TagDocEntryHT const ¤tHT = elementSet->GetTagHT();
2226 for( TagDocEntryHT::const_iterator i = currentHT.begin();
2227 i != currentHT.end();
2230 DocEntry *entry = i->second;
2231 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2233 const ListSQItem& items = seqEntry->GetSQItems();
2234 for( ListSQItem::const_iterator item = items.begin();
2235 item != items.end();
2238 BuildFlatHashTableRecurse( builtHT, *item );
2242 builtHT[entry->GetKey()] = entry;
2247 if (SQItem *SQItemSet = dynamic_cast< SQItem* > ( set ) )
2249 const ListDocEntry& currentList = SQItemSet->GetDocEntries();
2250 for (ListDocEntry::const_iterator i = currentList.begin();
2251 i != currentList.end();
2254 DocEntry *entry = *i;
2255 if ( SeqEntry *seqEntry = dynamic_cast<SeqEntry*>(entry) )
2257 const ListSQItem& items = seqEntry->GetSQItems();
2258 for( ListSQItem::const_iterator item = items.begin();
2259 item != items.end();
2262 BuildFlatHashTableRecurse( builtHT, *item );
2266 builtHT[entry->GetKey()] = entry;
2273 * \brief Build a \ref TagDocEntryHT (i.e. a std::map<>) from the current
2276 * The structure used by a Document (through \ref ElementSet),
2277 * in order to hold the parsed entries of a Dicom header, is a recursive
2278 * one. This is due to the fact that the sequences (when present)
2279 * can be nested. Additionaly, the sequence items (represented in
2280 * gdcm as \ref SQItem) add an extra complexity to the data
2281 * structure. Hence, a gdcm user whishing to visit all the entries of
2282 * a Dicom header will need to dig in the gdcm internals (which
2283 * implies exposing all the internal data structures to the API).
2284 * In order to avoid this burden to the user, \ref BuildFlatHashTable
2285 * recursively builds a temporary hash table, which holds all the
2286 * Dicom entries in a flat structure (a \ref TagDocEntryHT i.e. a
2288 * \warning Of course there is NO integrity constrain between the
2289 * returned \ref TagDocEntryHT and the \ref ElementSet used
2290 * to build it. Hence if the underlying \ref ElementSet is
2291 * altered, then it is the caller responsability to invoke
2292 * \ref BuildFlatHashTable again...
2293 * @return The flat std::map<> we juste build.
2295 /*TagDocEntryHT *Document::BuildFlatHashTable()
2297 TagDocEntryHT *FlatHT = new TagDocEntryHT;
2298 BuildFlatHashTableRecurse( *FlatHT, this );
2302 } // end namespace gdcm
2304 //-----------------------------------------------------------------------------