X-Git-Url: https://git.creatis.insa-lyon.fr/pubgit/?a=blobdiff_plain;f=src%2FgdcmHeader.cxx;h=a2bece5180e74a3a48decff566b0f06c739246c6;hb=f66abc53491ab130c2dbea483ba241edf6da2e88;hp=856e6f4504812369a343cc3e094e4969057ca2a1;hpb=1aa2bb207fb6b9ce78f35e7547dd4c770e516adb;p=gdcm.git diff --git a/src/gdcmHeader.cxx b/src/gdcmHeader.cxx index 856e6f45..a2bece51 100644 --- a/src/gdcmHeader.cxx +++ b/src/gdcmHeader.cxx @@ -1,7 +1,4 @@ #include "gdcm.h" -extern "C" { -#include "glib.h" -} #include // For nthos: #ifdef _MSC_VER @@ -9,6 +6,7 @@ extern "C" { #else #include #endif +#include // for isalpha #include #include #include "gdcmUtil.h" @@ -40,6 +38,7 @@ void gdcmHeader::Initialise(void) { } gdcmHeader::gdcmHeader (const char* InFilename) { + SetMaxSizeLoadElementValue(1024); filename = InFilename; Initialise(); fp=fopen(InFilename,"rw"); @@ -58,32 +57,32 @@ void gdcmHeader::InitVRDict (void) { return; } VRHT *vr = new VRHT; - (*vr)["AE"] = "Application Entity"; // 16 car max - (*vr)["AS"] = "Age String"; // 4 car fixe - (*vr)["AT"] = "Attribute Tag"; // 2 unsigned short int - (*vr)["CS"] = "Code String"; // 16 car max - (*vr)["DA"] = "Date"; // 8 car fixe - (*vr)["DS"] = "Decimal String"; // Decimal codé Binaire 16 max - (*vr)["DT"] = "Date Time"; // 26 car max - (*vr)["FL"] = "Floating Point Single"; // 4 octets IEEE 754:1985 - (*vr)["FD"] = "Floating Point Double"; // 8 octets IEEE 754:1985 - (*vr)["IS"] = "Integer String"; // en format externe 12 max - (*vr)["LO"] = "Long String"; // 64 octets max - (*vr)["LT"] = "Long Text"; // 10240 max - (*vr)["OB"] = "Other Byte String"; - (*vr)["OW"] = "Other Word String"; - (*vr)["PN"] = "Person Name"; - (*vr)["SH"] = "Short String"; // 16 car max - (*vr)["SL"] = "Signed Long"; + (*vr)["AE"] = "Application Entity"; // At most 16 bytes + (*vr)["AS"] = "Age String"; // Exactly 4 bytes + (*vr)["AT"] = "Attribute Tag"; // 2 16-bit unsigned short integers + (*vr)["CS"] = "Code String"; // At most 16 bytes + (*vr)["DA"] = "Date"; // Exactly 8 bytes + (*vr)["DS"] = "Decimal String"; // At most 16 bytes + (*vr)["DT"] = "Date Time"; // At most 26 bytes + (*vr)["FL"] = "Floating Point Single"; // 32-bit IEEE 754:1985 float + (*vr)["FD"] = "Floating Point Double"; // 64-bit IEEE 754:1985 double + (*vr)["IS"] = "Integer String"; // At most 12 bytes + (*vr)["LO"] = "Long String"; // At most 64 chars + (*vr)["LT"] = "Long Text"; // At most 10240 chars + (*vr)["OB"] = "Other Byte String"; // String of bytes (vr independant) + (*vr)["OW"] = "Other Word String"; // String of 16-bit words (vr dep) + (*vr)["PN"] = "Person Name"; // At most 64 chars + (*vr)["SH"] = "Short String"; // At most 16 chars + (*vr)["SL"] = "Signed Long"; // Exactly 4 bytes (*vr)["SQ"] = "Sequence of Items"; // Not Applicable - (*vr)["SS"] = "Signed Short"; // 2 octets - (*vr)["ST"] = "Short Text"; // 1024 car max - (*vr)["TM"] = "Time"; // 16 car max - (*vr)["UI"] = "Unique Identifier"; // 64 car max - (*vr)["UN"] = "Unknown"; - (*vr)["UT"] = "Unlimited Text"; // 2 puissance 32 -1 car max - (*vr)["UL"] = "Unsigned Long "; // 4 octets fixe - (*vr)["US"] = "Unsigned Short "; // 2 octets fixe + (*vr)["SS"] = "Signed Short"; // Exactly 2 bytes + (*vr)["ST"] = "Short Text"; // At most 1024 chars + (*vr)["TM"] = "Time"; // At most 16 bytes + (*vr)["UI"] = "Unique Identifier"; // At most 64 bytes + (*vr)["UL"] = "Unsigned Long "; // Exactly 4 bytes + (*vr)["UN"] = "Unknown"; // Any length of bytes + (*vr)["US"] = "Unsigned Short "; // Exactly 2 bytes + (*vr)["UT"] = "Unlimited Text"; // At most 2^32 -1 chars dicom_vr = vr; } @@ -129,10 +128,10 @@ void gdcmHeader::CheckSwap() entCur = deb + 128; if(memcmp(entCur, "DICM", (size_t)4) == 0) { filetype = TrueDicom; - dbg.Verbose(0, "gdcmHeader::CheckSwap:", "looks like DICOM Version3"); + dbg.Verbose(1, "gdcmHeader::CheckSwap:", "looks like DICOM Version3"); } else { filetype = Unknown; - dbg.Verbose(0, "gdcmHeader::CheckSwap:", "not a DICOM Version3 file"); + dbg.Verbose(1, "gdcmHeader::CheckSwap:", "not a DICOM Version3 file"); } if(filetype == TrueDicom) { @@ -153,21 +152,21 @@ void gdcmHeader::CheckSwap() entCur = deb + 136; if(memcmp(entCur, "UL", (size_t)2) == 0) { filetype = ExplicitVR; - dbg.Verbose(0, "gdcmHeader::CheckSwap:", + dbg.Verbose(1, "gdcmHeader::CheckSwap:", "explicit Value Representation"); } else { filetype = ImplicitVR; - dbg.Verbose(0, "gdcmHeader::CheckSwap:", + dbg.Verbose(1, "gdcmHeader::CheckSwap:", "not an explicit Value Representation"); } if (net2host) { sw = 4321; - dbg.Verbose(0, "gdcmHeader::CheckSwap:", + dbg.Verbose(1, "gdcmHeader::CheckSwap:", "HostByteOrder != NetworkByteOrder"); } else { sw = 0; - dbg.Verbose(0, "gdcmHeader::CheckSwap:", + dbg.Verbose(1, "gdcmHeader::CheckSwap:", "HostByteOrder = NetworkByteOrder"); } @@ -189,23 +188,23 @@ void gdcmHeader::CheckSwap() // four (bytes), then determining the proper swap code is straightforward. entCur = deb + 4; - s = str2num(entCur, int); + s = str2num(entCur, guint32); switch (s) { case 0x00040000 : - sw=3412; + sw = 3412; filetype = ACR; return; case 0x04000000 : - sw=4321; + sw = 4321; filetype = ACR; return; case 0x00000400 : - sw=2143; + sw = 2143; filetype = ACR; return; case 0x00000004 : - sw=0; + sw = 0; filetype = ACR; return; default : @@ -215,7 +214,7 @@ void gdcmHeader::CheckSwap() // We are out of luck. It is not a DicomV3 nor a 'clean' ACR/NEMA file. // It is time for despaired wild guesses. So, let's assume this file - // happens to be 'dirty' ACR/NEMA, i.e. the length of the group it + // happens to be 'dirty' ACR/NEMA, i.e. the length of the group is // not present. Then the only info we have is the net2host one. //FIXME Si c'est du RAW, ca degagera + tard @@ -226,6 +225,24 @@ void gdcmHeader::CheckSwap() return; } +void gdcmHeader::SwitchSwapToBigEndian(void) { + dbg.Verbose(1, "gdcmHeader::FindLength", "Switching to BigEndian mode."); + if ( sw == 0 ) { + sw = 4321; + return; + } + if ( sw == 4321 ) { + sw = 0; + return; + } + if ( sw == 3412 ) { + sw = 2143; + return; + } + if ( sw == 2143 ) + sw = 3412; +} + /** * \ingroup gdcmHeader * \brief recupere la longueur d'un champ DICOM. @@ -267,15 +284,9 @@ void gdcmHeader::CheckSwap() void gdcmHeader::FindVR( ElValue *ElVal) { char VR[3]; + string vr; int lgrLue; long PositionOnEntry = ftell(fp); - - if (filetype != ExplicitVR) - return; - - lgrLue=fread (&VR, (size_t)2,(size_t)1, fp); - VR[2]=0; - // Warning: we believe this is explicit VR (Value Representation) because // we used a heuristic that found "UL" in the first tag. Alas this // doesn't guarantee that all the tags will be in explicit VR. In some @@ -283,64 +294,167 @@ void gdcmHeader::FindVR( ElValue *ElVal) { // within an explicit VR file. Hence we make sure the present tag // is in explicit VR and try to fix things if it happens not to be // the case. + bool RealExplicit = true; + + if (filetype != ExplicitVR) + return; - // FIXME There should be only one occurence returned. Avoid the - // first extraction by calling proper method. - VRAtr FoundVR = dicom_vr->find(string(VR))->first; - if ( ! FoundVR.empty()) { - ElVal->SetVR(FoundVR); + lgrLue=fread (&VR, (size_t)2,(size_t)1, fp); + VR[2]=0; + vr = string(VR); + + // Assume we are reading a falsely explicit VR file i.e. we reached + // a tag where we expect reading a VR but are in fact we read the + // first to bytes of the length. Then we will interogate (through find) + // the dicom_vr dictionary with oddities like "\004\0" which crashes + // both GCC and VC++ implentations of the STL map. Hence when the + // expected VR read happens to be non-ascii characters we consider + // we hit falsely explicit VR tag. + + if ( (!isalpha(VR[0])) && (!isalpha(VR[1])) ) + RealExplicit = false; + + // CLEANME searching the dicom_vr at each occurence is expensive. + // PostPone this test in an optional integrity check at the end + // of parsing or only in debug mode. + if ( RealExplicit && !dicom_vr->count(vr) ) + RealExplicit = false; + + if ( RealExplicit ) { + if ( ElVal->IsVrUnknown() ) + ElVal->SetVR(vr); return; } // We thought this was explicit VR, but we end up with an // implicit VR tag. Let's backtrack. - ElVal->SetVR("Implicit"); + dbg.Verbose(1, "gdcmHeader::FindVR:", "Falsely explicit vr file"); fseek(fp, PositionOnEntry, SEEK_SET); + // When this element is known in the dictionary we shall use, e.g. for + // the semantics (see the usage of IsAnInteger), the vr proposed by the + // dictionary entry. Still we have to flag the element as implicit since + // we know now our assumption on expliciteness is not furfilled. + // avoid . + if ( ElVal->IsVrUnknown() ) + ElVal->SetVR("Implicit"); + ElVal->SetImplicitVr(); +} + +/** + * \ingroup gdcmHeader + * \brief Determines if the Transfer Syntax was allready encountered + * and if it corresponds to a Big Endian one. + * + * @return True when big endian found. False in all other cases. + */ +bool gdcmHeader::IsBigEndianTransferSyntax(void) { + ElValue* Element = PubElVals.GetElement(0x0002, 0x0010); + if ( !Element ) + return false; + LoadElementValueSafe(Element); + string Transfer = Element->GetValue(); + if ( Transfer == "1.2.840.10008.1.2.2" ) + return true; + return false; +} + +void gdcmHeader::FixFoundLength(ElValue * ElVal, guint32 FoudLength) { + // Heuristic: a final fix. + if ( FoudLength == 0xffffffff) + FoudLength = 0; + ElVal->SetLength(FoudLength); } void gdcmHeader::FindLength( ElValue * ElVal) { - guint32 length32; + guint16 element = ElVal->GetElement(); + string vr = ElVal->GetVR(); guint16 length16; - if (filetype == ExplicitVR) { - string vr = ElVal->GetVR(); - if ( (vr != "Implicit") - && ( (vr=="OB") || (vr=="OW") || (vr=="SQ") || (vr=="UN") ) ) { - - // The following two bytes are reserved, so we skip them, - // and we proceed on reading the length on 4 bytes. - fseek(fp, 2L,SEEK_CUR); - length32 = ReadInt32(); - - } else { - // Length is encoded on 2 bytes. - length16 = ReadInt16(); - - if ( length16 == 0xffff) { - length32 = 0; - } else { - length32 = length16; + if ( (filetype == ExplicitVR) && ! ElVal->IsImplicitVr() ) { + + if ( (vr=="OB") || (vr=="OW") || (vr=="SQ") || (vr=="UN") ) { + // The following reserved two bytes (see PS 3.5-2001, section + // 7.1.2 Data element structure with explicit vr p27) must be + // skipped before proceeding on reading the length on 4 bytes. + fseek(fp, 2L, SEEK_CUR); + FixFoundLength(ElVal, ReadInt32()); + return; + } + + // Length is encoded on 2 bytes. + length16 = ReadInt16(); + + // We can tell the current file is encoded in big endian (like + // Data/US-RGB-8-epicard) when we find the "Transfer Syntax" tag + // and it's value is the one of the encoding of a bie endian file. + // In order to deal with such big endian encoded files, we have + // (at least) two strategies: + // * when we load the "Transfer Syntax" tag with value of big endian + // encoding, we raise the proper flags. Then we wait for the end + // of the META group (0x0002) among which is "Transfer Syntax", + // before switching the swap code to big endian. We have to postpone + // the switching of the swap code since the META group is fully encoded + // in little endian, and big endian coding only starts at the next + // group. The corresponding code can be hard to analyse and adds + // many additional unnecessary tests for regular tags. + // * the second strategy consist to wait for trouble, that shall appear + // when we find the first group with big endian encoding. This is + // easy to detect since the length of a "Group Length" tag (the + // ones with zero as element number) has to be of 4 (0x0004). When we + // encouter 1024 (0x0400) chances are the encoding changed and we + // found a group with big endian encoding. + // We shall use this second strategy. In order make sure that we + // can interpret the presence of an apparently big endian encoded + // length of a "Group Length" without committing a big mistake, we + // add an additional check: we look in the allready parsed elements + // for the presence of a "Transfer Syntax" whose value has to be "big + // endian encoding". When this is the case, chances are we got our + // hands on a big endian encoded file: we switch the swap code to + // big endian and proceed... + if ( (element == 0) && (length16 == 1024) ) { + if ( ! IsBigEndianTransferSyntax() ) + throw Error::FileReadError(fp, "gdcmHeader::FindLength"); + length16 = 4; + SwitchSwapToBigEndian(); + // Restore the unproperly loaded values i.e. the group, the element + // and the dictionary entry depending on them. + guint16 CorrectGroup = SwapShort(ElVal->GetGroup()); + guint16 CorrectElem = SwapShort(ElVal->GetElement()); + gdcmDictEntry * NewTag = IsInDicts(CorrectGroup, CorrectElem); + if (!NewTag) { + // This correct tag is not in the dictionary. Create a new one. + NewTag = new gdcmDictEntry(CorrectGroup, CorrectElem); } + // FIXME this can create a memory leaks on the old entry that be + // left unreferenced. + ElVal->SetDictEntry(NewTag); } - } else { - // Either implicit VR or an explicit VR that (at least for this - // element) lied a little bit. Length is on 4 bytes. - length32 = ReadInt32(); + + // Heuristic: well some files are really ill-formed. + if ( length16 == 0xffff) { + length16 = 0; + dbg.Verbose(0, "gdcmHeader::FindLength", + "Erroneous element length fixed."); + } + FixFoundLength(ElVal, (guint32)length16); + return; } - - // Traitement des curiosites sur la longueur - if ( length32 == 0xffffffff) - length32=0; - - ElVal->SetLength(length32); -} + // Either implicit VR or a non DICOM conformal (see not below) explicit + // VR that ommited the VR of (at least) this element. Farts happen. + // [Note: according to the part 5, PS 3.5-2001, section 7.1 p25 + // on Data elements "Implicit and Explicit VR Data Elements shall + // not coexist in a Data Set and Data Sets nested within it".] + // Length is on 4 bytes. + FixFoundLength(ElVal, ReadInt32()); +} /** * \ingroup gdcmHeader - * \brief remet les octets dans un ordre compatible avec celui du processeur - - * @return longueur retenue pour le champ + * \brief Swaps back the bytes of 4-byte long integer accordingly to + * processor order. + * + * @return The suggested integer. */ guint32 gdcmHeader::SwapLong(guint32 a) { // FIXME: il pourrait y avoir un pb pour les entiers negatifs ... @@ -368,9 +482,8 @@ guint32 gdcmHeader::SwapLong(guint32 a) { /** * \ingroup gdcmHeader - * \brief remet les octets dans un ordre compatible avec celui du processeur - - * @return longueur retenue pour le champ + * \brief Swaps the bytes so they agree with the processor order + * @return The properly swaped 16 bits integer. */ guint16 gdcmHeader::SwapShort(guint16 a) { if ( (sw==4321) || (sw==2143) ) @@ -383,30 +496,78 @@ void gdcmHeader::SkipElementValue(ElValue * ElVal) { (void)fseek(fp, (long)ElVal->GetLength(), SEEK_CUR); } +void gdcmHeader::SetMaxSizeLoadElementValue(long NewSize) { + if (NewSize < 0) + return; + if ((guint32)NewSize >= (guint32)0xffffffff) { + MaxSizeLoadElementValue = 0xffffffff; + return; + } + MaxSizeLoadElementValue = NewSize; +} + +/** + * \ingroup gdcmHeader + * \brief Loads the element if it's size is not to big. + * @param ElVal Element whose value shall be loaded. + * @param MaxSize Size treshold above which the element value is not + * loaded in memory. The element value is allways loaded + * when MaxSize is equal to UINT32_MAX. + * @return + */ void gdcmHeader::LoadElementValue(ElValue * ElVal) { size_t item_read; guint16 group = ElVal->GetGroup(); guint16 elem = ElVal->GetElement(); string vr = ElVal->GetVR(); guint32 length = ElVal->GetLength(); + bool SkipLoad = false; + fseek(fp, (long)ElVal->GetOffset(), SEEK_SET); // Sequences not treated yet ! - if( vr == "SQ" ) { + if( vr == "SQ" ) + SkipLoad = true; + + // Heuristic : a sequence "contains" a set of tags (called items). It looks + // like the last tag of a sequence (the one that terminates the sequence) + // has a group of 0xfffe (with a dummy length). + if( group == 0xfffe ) + SkipLoad = true; + + // The group length doesn't represent data to be loaded in memory, since + // each element of the group shall be loaded individualy. + if( elem == 0 ) + SkipLoad = true; + + if ( SkipLoad ) { + // FIXME the following skip is not necessary SkipElementValue(ElVal); ElVal->SetLength(0); + ElVal->SetValue("gdcm::Skipped"); return; } - // A sequence "contains" a set of tags (called items). It looks like - // the last tag of a sequence (the one that terminates the sequence) - // has a group of 0xfffe (with a dummy length). - if( group == 0xfffe) { - SkipElementValue(ElVal); - ElVal->SetLength(0); + + // When the length is zero things are easy: + if ( length == 0 ) { + ElVal->SetValue(""); + return; + } + + // Values bigger than specified are not loaded. + if (length > MaxSizeLoadElementValue) { + ostringstream s; + s << "gdcm::NotLoaded."; + s << " Address:" << (long)ElVal->GetOffset(); + s << " Length:" << ElVal->GetLength(); + //mesg += " Length:" + ElVal->GetLength(); + ElVal->SetValue(s.str()); return; } - if ( IsAnInteger(group, elem, vr, length) ) { + // When an integer is expected, read and convert the following two or + // four bytes properly i.e. as an integer as opposed to a string. + if ( IsAnInteger(ElVal) ) { guint32 NewInt; if( length == 2 ) { NewInt = ReadInt16(); @@ -430,8 +591,6 @@ void gdcmHeader::LoadElementValue(ElValue * ElVal) { } NewValue[length]= 0; - // FIXME les elements trop long (seuil a fixer a la main) ne devraient - // pas etre charge's !!!! Voir TODO. item_read = fread(NewValue, (size_t)length, (size_t)1, fp); if ( item_read != 1 ) { g_free(NewValue); @@ -442,6 +601,20 @@ void gdcmHeader::LoadElementValue(ElValue * ElVal) { ElVal->SetValue(NewValue); } +/** + * \ingroup gdcmHeader + * \brief Loads the element while preserving the current + * underlying file position indicator as opposed to + * to LoadElementValue that modifies it. + * @param ElVal Element whose value shall be loaded. + * @return + */ +void gdcmHeader::LoadElementValueSafe(ElValue * ElVal) { + long PositionOnEntry = ftell(fp); + LoadElementValue(ElVal); + fseek(fp, PositionOnEntry, SEEK_SET); +} + guint16 gdcmHeader::ReadInt16(void) { guint16 g; @@ -487,7 +660,7 @@ ElValue * gdcmHeader::ReadNextElement(void) { // Find out if the tag we encountered is in the dictionaries: gdcmDictEntry * NewTag = IsInDicts(g, n); if (!NewTag) - NewTag = new gdcmDictEntry(g, n, "Unknown", "Unknown", "Unknown"); + NewTag = new gdcmDictEntry(g, n); NewElVal = new ElValue(NewTag); if (!NewElVal) { @@ -496,20 +669,29 @@ ElValue * gdcmHeader::ReadNextElement(void) { } FindVR(NewElVal); - FindLength(NewElVal); + try { FindLength(NewElVal); } + catch ( Error::FileReadError ) { // Call it quits + return (ElValue *)0; + } NewElVal->SetOffset(ftell(fp)); return NewElVal; } -bool gdcmHeader::IsAnInteger(guint16 group, guint16 element, - string vr, guint32 length ) { - // When we have some semantics on the element we just read, and we - // a priori now we are dealing with an integer, then we can swap it's - // element value properly. +bool gdcmHeader::IsAnInteger(ElValue * ElVal) { + guint16 group = ElVal->GetGroup(); + guint16 element = ElVal->GetElement(); + string vr = ElVal->GetVR(); + guint32 length = ElVal->GetLength(); + + // When we have some semantics on the element we just read, and if we + // a priori know we are dealing with an integer, then we shall be + // able to swap it's element value properly. if ( element == 0 ) { // This is the group length of the group - if (length != 4) - dbg.Error("gdcmHeader::ShouldBeSwaped", "should be four"); - return true; + if (length == 4) + return true; + else + dbg.Error("gdcmHeader::IsAnInteger", + "Erroneous Group Length element length."); } if ( group % 2 != 0 )