X-Git-Url: https://git.creatis.insa-lyon.fr/pubgit/?a=blobdiff_plain;f=src%2FgdcmHeader.cxx;h=d3e99c2062ff83f549af69c9ff94646fdd29910b;hb=8afb2999a90aefbce7662c43fe19dad237d49331;hp=dc6bbf41a7ee6ca995746d9a3baad681f57b3ca1;hpb=e8be02129ece7a8c7dd90307a8c83e1ef5d157b6;p=gdcm.git diff --git a/src/gdcmHeader.cxx b/src/gdcmHeader.cxx index dc6bbf41..d3e99c20 100644 --- a/src/gdcmHeader.cxx +++ b/src/gdcmHeader.cxx @@ -1,7 +1,4 @@ -#include "gdcmlib.h" -extern "C" { -#include "glib.h" -} +#include "gdcm.h" #include // For nthos: #ifdef _MSC_VER @@ -9,12 +6,24 @@ extern "C" { #else #include #endif +#include // for isalpha #include #include #include "gdcmUtil.h" #define HEADER_LENGHT_TO_READ 256 // on ne lit plus que le debut +namespace Error { + struct FileReadError { + FileReadError(FILE* fp, const char* Mesg) { + if (feof(fp)) + dbg.Verbose(1, "EOF encountered :", Mesg); + if (ferror(fp)) + dbg.Verbose(1, "Error on reading :", Mesg); + } + }; +} + //FIXME: this looks dirty to me... #define str2num(str, typeNum) *((typeNum *)(str)) @@ -24,22 +33,21 @@ gdcmDictSet* gdcmHeader::Dicts = new gdcmDictSet(); void gdcmHeader::Initialise(void) { if (!gdcmHeader::dicom_vr) InitVRDict(); - PixelPosition = (size_t)0; - PixelsTrouves = false; RefPubDict = gdcmHeader::Dicts->GetDefaultPublicDict(); RefShaDict = (gdcmDict*)0; } -gdcmHeader::gdcmHeader (char* InFilename) { +gdcmHeader::gdcmHeader (const char* InFilename) { + SetMaxSizeLoadElementValue(1024); filename = InFilename; Initialise(); fp=fopen(InFilename,"rw"); dbg.Error(!fp, "gdcmHeader::gdcmHeader cannot open file", InFilename); - BuildHeader(); - fclose(fp); + ParseHeader(); } gdcmHeader::~gdcmHeader (void) { + fclose(fp); return; } @@ -49,32 +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; } @@ -120,10 +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) { @@ -144,21 +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"); } @@ -180,23 +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 : @@ -206,7 +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 @@ -217,6 +225,25 @@ void gdcmHeader::CheckSwap() return; } +void gdcmHeader::SwitchSwapToBigEndian(void) { + dbg.Verbose(1, "gdcmHeader::SwitchSwapToBigEndian", + "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. @@ -256,17 +283,14 @@ void gdcmHeader::CheckSwap() * @return longueur retenue pour le champ */ -void gdcmHeader::FindVR( ElValue *pleCourant) { - char VR[3]; - int lgrLue; - long PositionOnEntry = ftell(fp); - +void gdcmHeader::FindVR( ElValue *ElVal) { if (filetype != ExplicitVR) return; - lgrLue=fread (&VR, (size_t)2,(size_t)1, fp); - VR[2]=0; - + char VR[3]; + string vr; + int lgrLue; + long PositionOnEntry = ftell(fp); // 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 @@ -274,69 +298,221 @@ void gdcmHeader::FindVR( ElValue *pleCourant) { // 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. - - // 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()) { - pleCourant->SetVR(FoundVR); + bool RealExplicit = true; + + 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() ) { + // When not a dictionary entry, we can safely overwrite the vr. + ElVal->SetVR(vr); + return; + } + if ( ElVal->GetVR() == vr ) { + // The vr we just read and the dictionary agree. Nothing to do. + return; + } + // The vr present in the file and the dictionary disagree. We assume + // the file writer knew best and use the vr of the file. Since it would + // be unwise to overwrite the vr of a dictionary (since it would + // compromise it's next user), we need to clone the actual DictEntry + // and change the vr for the read one. + gdcmDictEntry* NewTag = new gdcmDictEntry(ElVal->GetGroup(), + ElVal->GetElement(), + vr, + "FIXME", + ElVal->GetName()); + ElVal->SetDictEntry(NewTag); return; } // We thought this was explicit VR, but we end up with an // implicit VR tag. Let's backtrack. - pleCourant->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(); } -void gdcmHeader::FindLength( ElValue *pleCourant) { - int lgrLue; - guint32 l_gr; - unsigned short int l_gr_2; - - string vr = pleCourant->GetVR(); +/** + * \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); +} + +guint32 gdcmHeader::FindLengthOB(void) { + // See PS 3.5-2001, section A.4 p. 49 on encapsulation of encoded pixel data. + guint16 g; + guint16 n; + long PositionOnEntry = ftell(fp); + bool FoundSequenceDelimiter = false; + guint32 TotalLength = 0; + guint32 ItemLength; + + while ( ! FoundSequenceDelimiter) { + g = ReadInt16(); + n = ReadInt16(); + TotalLength += 4; // We even have to decount the group and element + if ( g != 0xfffe ) { + dbg.Verbose(1, "gdcmHeader::FindLengthOB: ", + "wrong group for an item sequence."); + throw Error::FileReadError(fp, "gdcmHeader::FindLengthOB"); + } + if ( n == 0xe0dd ) + FoundSequenceDelimiter = true; + else if ( n != 0xe000) { + dbg.Verbose(1, "gdcmHeader::FindLengthOB: ", + "wrong element for an item sequence."); + throw Error::FileReadError(fp, "gdcmHeader::FindLengthOB"); + } + ItemLength = ReadInt32(); + TotalLength += ItemLength + 4; // We add 4 bytes since we just read + // the ItemLength with ReadInt32 + SkipBytes(ItemLength); + } + fseek(fp, PositionOnEntry, SEEK_SET); + return TotalLength; +} + +void gdcmHeader::FindLength(ElValue * ElVal) { + guint16 element = ElVal->GetElement(); + string vr = ElVal->GetVR(); + guint16 length16; - if ( (filetype == ExplicitVR) && (vr != "Implicit") ) { - if ( ( 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); - lgrLue=fread (&l_gr, (size_t)4,(size_t)1, fp); - l_gr = SwapLong((guint32)l_gr); - - } else { - // Length is encoded on 2 bytes. - lgrLue=fread (&l_gr_2, (size_t)2,(size_t)1, fp); - - l_gr_2 = SwapShort((unsigned short)l_gr_2); - - if ( l_gr_2 == 0xffff) { - l_gr = 0; - } else { - l_gr = l_gr_2; + 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); + guint32 length32 = ReadInt32(); + if ( (vr == "OB") && (length32 == 0xffffffff) ) { + ElVal->SetLength(FindLengthOB()); + return; } + FixFoundLength(ElVal, length32); + return; } - } else { - // Either implicit VR or an explicit VR that (at least for this - // element) lied a little bit. Length is on 4 bytes. - lgrLue=fread (&l_gr, (size_t)4,(size_t)1, fp); - l_gr= SwapLong((long)l_gr); + + // 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 big 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 in waiting 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); + } + + // 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 ( l_gr == 0xffffffff) - l_gr=0; - - pleCourant->SetLength(l_gr); + // 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 ... @@ -364,131 +540,220 @@ 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. */ -short int gdcmHeader::SwapShort(short int a) { +guint16 gdcmHeader::SwapShort(guint16 a) { if ( (sw==4321) || (sw==2143) ) a =(((a<<8) & 0x0ff00) | ((a>>8)&0x00ff)); return (a); } +void gdcmHeader::SkipBytes(guint32 NBytes) { + //FIXME don't dump the returned value + (void)fseek(fp, (long)NBytes, SEEK_CUR); +} + +void gdcmHeader::SkipElementValue(ElValue * ElVal) { + SkipBytes(ElVal->GetLength()); +} + +void gdcmHeader::SetMaxSizeLoadElementValue(long NewSize) { + if (NewSize < 0) + return; + if ((guint32)NewSize >= (guint32)0xffffffff) { + MaxSizeLoadElementValue = 0xffffffff; + return; + } + MaxSizeLoadElementValue = NewSize; +} + /** * \ingroup gdcmHeader - * \brief lit le dicom_element suivant. - * (le fichier doit deja avoir ete ouvert, - * _IdAcrCheckSwap(ID_DCM_HDR *e) avoir ete appele) - * @param e ID_DCM_HDR dans lequel effectuer la recherche. - * @param sw code swap. - * @return En cas de succes, 1 - * 0 en cas d'echec. + * \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 */ - -ElValue * gdcmHeader::ReadNextElement(void) { - guint16 g; - guint16 n; - guint32 l; - size_t lgrLue; - ElValue * nouvDcmElem; - - // ------------------------- Lecture Num group : g - lgrLue=fread (&g, (size_t)2,(size_t)1, fp); +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); - if (feof(fp)) { - dbg.Verbose(1, "ReadNextElement: EOF encountered"); - return (NULL); + // Sequences not treated yet ! + 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; } - if (ferror(fp)){ - dbg.Verbose(1, "ReadNextElement: failed to read NumGr"); - return (NULL); + + // 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 (sw) g= SwapShort(((short)g)); - - // ------------------------- Lecture Num Elem : n - lgrLue=fread (&n, (size_t)2,(size_t)1, fp); + // 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(); + } else if( length == 4 ) { + NewInt = ReadInt32(); + } else + dbg.Error(true, "LoadElementValue: Inconsistency when reading Int."); + + //FIXME: make the following an util fonction + ostringstream s; + s << NewInt; + ElVal->SetValue(s.str()); + return; + } - if (feof(fp)) { - dbg.Verbose(1, "ReadNextElement: EOF encountered"); - return (NULL); + // FIXME The exact size should be length if we move to strings or whatever + char* NewValue = (char*)g_malloc(length+1); + if( !NewValue) { + dbg.Verbose(1, "LoadElementValue: Failed to allocate NewValue"); + return; } - if (ferror(fp)){ - dbg.Verbose(1, "ReadNextElement: failed to read NumElem"); - return (NULL); + NewValue[length]= 0; + + item_read = fread(NewValue, (size_t)length, (size_t)1, fp); + if ( item_read != 1 ) { + g_free(NewValue); + Error::FileReadError(fp, "gdcmHeader::LoadElementValue"); + ElVal->SetValue("gdcm::UnRead"); + return; } + 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; + size_t item_read; + item_read = fread (&g, (size_t)2,(size_t)1, fp); + if ( item_read != 1 ) + throw Error::FileReadError(fp, "gdcmHeader::ReadInt16"); + g = SwapShort(g); + return g; +} + +guint32 gdcmHeader::ReadInt32(void) { + guint32 g; + size_t item_read; + item_read = fread (&g, (size_t)4,(size_t)1, fp); + if ( item_read != 1 ) + throw Error::FileReadError(fp, "gdcmHeader::ReadInt32"); + g = SwapLong(g); + return g; +} + +/** + * \ingroup gdcmHeader + * \brief Read the next tag without loading it's value + * @return On succes the newly created ElValue, NULL on failure. + */ + +ElValue * gdcmHeader::ReadNextElement(void) { + guint16 g; + guint16 n; + ElValue * NewElVal; - if(sw) n= SwapShort(((short)n)); + try { + g = ReadInt16(); + n = ReadInt16(); + } + catch ( Error::FileReadError ) { + // We reached the EOF (or an error occured) and header parsing + // has to be considered as finished. + return (ElValue *)0; + } // 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", "Unkown"); + NewTag = new gdcmDictEntry(g, n); - nouvDcmElem = new ElValue(NewTag); - if (!nouvDcmElem) { + NewElVal = new ElValue(NewTag); + if (!NewElVal) { dbg.Verbose(1, "ReadNextElement: failed to allocate ElValue"); - return(NULL); + return (ElValue*)0; } - FindVR(nouvDcmElem); - FindLength(nouvDcmElem); - nouvDcmElem->SetOffset(ftell(fp)); - l = nouvDcmElem->GetLength(); - - //FIXMEif(!memcmp( VR,"SQ",(size_t)2 )) { // ca annonce une SEQUENCE d'items ?! - //FIXME l_gr=0; // on lira donc les items de la sequence - //FIXME} - //FIXMEreturn(l_gr); - - // Une sequence contient un ensemble de group element repetes n fois - // et g=fffe indique la fin (contient une longueur bidon). - if(g==0xfffe) l=0; // pour sauter les indicateurs de 'SQ' - - - // ------------------------- Lecture Valeur element - - // FIXME The exact size should be l if we move to strings or whatever - // CLEAN ME NEWValue used to be nouvDcmElem->valeurElem - char* NewValue = (char*)g_malloc(l+1); - if(NewValue) { - NewValue[l]= 0; - } else { - return (NULL); + FindVR(NewElVal); + try { FindLength(NewElVal); } + catch ( Error::FileReadError ) { // Call it quits + return (ElValue *)0; } - - // FIXME les elements trop long (seuil a fixer a la main) ne devraient - // pas etre charge's !!!! Voir TODO. - lgrLue=fread (NewValue, (size_t)l,(size_t)1, fp); - - if ( IsAnInteger(g, n, NewTag->GetVR(), l) ) { - // CLEANME THe following is really UGLY ! - if( l == 4 ) { - *(guint32 *) NewValue = SwapLong ((*(guint32 *) NewValue)); - } else { - if( l == 2 ) - *(guint16 *) NewValue = SwapShort ((*(guint16 *)NewValue)); - } - //FIXME: don't we have to distinguish guin16 and guint32 - //FIXME: make the following an util fonction - ostringstream s; - s << *(guint32 *) NewValue; - nouvDcmElem->value = s.str(); - g_free(NewValue); - } else - nouvDcmElem->value = NewValue; - return nouvDcmElem; + 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 ) @@ -517,78 +782,35 @@ bool gdcmHeader::IsAnInteger(guint16 group, guint16 element, /** * \ingroup gdcmHeader - * \brief If we encountered the offset of the pixels in the file - * (Pixel Data) then keep the info aside. + * \brief Recover the offset (from the beginning of the file) of the pixels. */ -void gdcmHeader::SetAsidePixelData(ElValue* elem) { - // They are two cases : - // * the pixel data (i.e. the image or the volume) is pointed by it's - // default official tag (0x7fe0,0x0010), - // * the writer of this file decided to put the image "address" (i.e the - // offset from the begining of the file) at a different tag. - // Then the "Pixel Data" offset might be found by indirection through - // the "Image Location" tag (0x0028, 0x0200). In other terms the Image - // Location tag contains the group where the "Pixel Data" offset is and - // inside this group the element is conventionally at element 0x0010 - // (when the norm is respected). - // - // Hence getting our hands on the Pixel Data is a two stage process: - // 1/ * find if the "Pixel Data" tag exists. - // * if it does not exist, look for the "Pixel Location" tag. - // 2/ look at the proper tag ("Pixel Data" or "Pixel Location" when - // it exists) what the offset is. - cout << "aaaaaaaaaaaaaaaaaaaaa"; - // PubElVals.PrintByName(cout); - ostringstream val; - val << hex << GetPubElValByName("Image Location"); - cout << GetPubElValByName("Image Location") << endl; - cout <GetGroup(); - n = elem->GetElement(); - if (!grPixelTrouve) { // on n a pas encore trouve les pixels - if (g > 0x0028) { - if (n > 0x0200 || g == 0x7FE0 ) { // on a depasse (28,200) - grPixel = 0x7FE0; - numPixel = 0x0010; - grPixelTrouve = true; - } - } else { // on est sur (28,200) - if (g == 0x0028) { - if (n == 0x0200) { - grPixelTrouve = 1; - char* NewValue = (char*)g_malloc(elem->GetLength()+1); - // FIXME: not very elegant conversion - for(int i=0;i<4;i++) - *((char*)(&grPixel)+i) = *(NewValue+i); - elem->SetValue(NewValue); - - if (grPixel != 0x7FE0) // Vieux pb Philips - numPixel = 0x1010; // encore utile ?? - else - numPixel = 0x0010; - } - } - } - } else { // on vient de trouver les pixels - if (g == grPixel) { - if (n == numPixel) { - PixelPosition = elem->Offset; - PixelsTrouves = true; - dbg.Verbose(0, "gdcmHeader::SetAsidePixelData:", - "Pixel data found"); - } - } +size_t gdcmHeader::GetPixelOffset(void) { + // If this file complies with the norm we should encounter the + // "Image Location" tag (0x0028, 0x0200). This tag contains the + // the group that contains the pixel data (hence the "Pixel Data" + // is found by indirection through the "Image Location"). + // Inside the group pointed by "Image Location" the searched element + // is conventionally the element 0x0010 (when the norm is respected). + // When the "Image Location" is absent we default to group 0x7fe0. + guint16 grPixel; + guint16 numPixel; + string ImageLocation = GetPubElValByName("Image Location"); + if ( ImageLocation == "UNFOUND" ) { + grPixel = 0x7FE0; + } else { + grPixel = (guint16) atoi( ImageLocation.c_str() ); } + if (grPixel != 0x7fe0) + // FIXME is this still necessary ? + // Now, this looks like an old dirty fix for Philips imager + numPixel = 0x1010; + else + numPixel = 0x0010; + ElValue* PixelElement = PubElVals.GetElement(grPixel, numPixel); + if (PixelElement) + return PixelElement->GetOffset(); + else + return 0; } gdcmDictEntry * gdcmHeader::IsInDicts(guint32 group, guint32 element) { @@ -619,21 +841,30 @@ string gdcmHeader::GetPubElValByName(string TagName) { } /** - * \ingroup gdcmHeader - * \brief renvoie un pointeur sur le ID_DCM_HDR correspondant au fichier - * @param filename Nom du fichier ACR / LibIDO / DICOM - * @return le ID_DCM_HDR + * \ingroup gdcmHeader + * \brief Parses the header of the file but does NOT load element values. */ - -void gdcmHeader::BuildHeader(void) { +void gdcmHeader::ParseHeader(void) { ElValue * newElValue = (ElValue *)0; rewind(fp); CheckSwap(); while ( (newElValue = ReadNextElement()) ) { + SkipElementValue(newElValue); PubElVals.Add(newElValue); } - SetAsidePixelData((ElValue*)0); +} + +/** + * \ingroup gdcmHeader + * \brief Loads the element values of all the elements present in the + * public tag based hash table. + */ +void gdcmHeader::LoadElements(void) { + rewind(fp); + TagElValueHT ht = PubElVals.GetTagHt(); + for (TagElValueHT::iterator tag = ht.begin(); tag != ht.end(); ++tag) + LoadElementValue(tag->second); } void gdcmHeader::PrintPubElVal(ostream & os) {