X-Git-Url: https://git.creatis.insa-lyon.fr/pubgit/?a=blobdiff_plain;f=src%2FgdcmHeader.cxx;h=faa95f7bccd4fcffd175453a3da4f9426066f316;hb=e0202ea1c0b395871073c56f06fbcd789d2056de;hp=856e6f4504812369a343cc3e094e4969057ca2a1;hpb=1aa2bb207fb6b9ce78f35e7547dd4c770e516adb;p=gdcm.git diff --git a/src/gdcmHeader.cxx b/src/gdcmHeader.cxx index 856e6f45..faa95f7b 100644 --- a/src/gdcmHeader.cxx +++ b/src/gdcmHeader.cxx @@ -1,52 +1,60 @@ +// gdcmHeader.cxx + #include "gdcm.h" -extern "C" { -#include "glib.h" -} #include +#include // For nthos: #ifdef _MSC_VER #include #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); - } - }; -} +#define HEADER_LENGTH_TO_READ 256 // on ne lit plus que le debut +#define _MaxSizeLoadElementValue_ 1024 // longueur au dela de laquelle on ne charge plus les valeurs //FIXME: this looks dirty to me... + #define str2num(str, typeNum) *((typeNum *)(str)) +// str est un pointeur dans un tableau de caractères, qui doit contenir, +// à cet endroit la, la représentation binaire d'un entier (16 ou 32 bits) +// je veux récupérer ça ... dans un entier. +// s'il y a une autre solution, évitant des cast et les indirections, +// je suis preneur + VRHT * gdcmHeader::dicom_vr = (VRHT*)0; -gdcmDictSet* gdcmHeader::Dicts = new gdcmDictSet(); void gdcmHeader::Initialise(void) { if (!gdcmHeader::dicom_vr) InitVRDict(); - RefPubDict = gdcmHeader::Dicts->GetDefaultPublicDict(); + Dicts = new gdcmDictSet(); + RefPubDict = Dicts->GetDefaultPubDict(); RefShaDict = (gdcmDict*)0; } -gdcmHeader::gdcmHeader (const char* InFilename) { - filename = InFilename; - Initialise(); - fp=fopen(InFilename,"rw"); - dbg.Error(!fp, "gdcmHeader::gdcmHeader cannot open file", InFilename); - ParseHeader(); + +gdcmHeader::gdcmHeader(const char *InFilename, bool exception_on_error) + throw(gdcmFileError) { + SetMaxSizeLoadElementValue(_MaxSizeLoadElementValue_); + filename = InFilename; + Initialise(); + fp=fopen(InFilename,"rb"); + if(exception_on_error) { + if(!fp) + throw gdcmFileError("gdcmHeader::gdcmHeader(const char *, bool)"); + } + else + dbg.Error(!fp, "gdcmHeader::gdcmHeader cannot open file", InFilename); + ParseHeader(); + AddAndDefaultElements(); } + gdcmHeader::~gdcmHeader (void) { fclose(fp); return; @@ -58,62 +66,54 @@ 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; } /** * \ingroup gdcmHeader - * \brief La seule maniere sure que l'on aie pour determiner - * si on est en LITTLE_ENDIAN, BIG-ENDIAN, - * BAD-LITTLE-ENDIAN, BAD-BIG-ENDIAN - * est de trouver l'element qui donne la longueur d'un 'GROUP' - * (on sait que la longueur de cet element vaut 0x00000004) - * et de regarder comment cette longueur est codee en memoire - * - * Le probleme vient de ce que parfois, il n'y en a pas ... - * - * On fait alors le pari qu'on a a faire a du LITTLE_ENDIAN propre. - * (Ce qui est la norme -pas respectee- depuis ACR-NEMA) - * Si ce n'est pas le cas, on ne peut rien faire. + * \brief Discover what the swap code is (among little endian, big endian, + * bad little endian, bad big endian). * - * (il faudrait avoir des fonctions auxquelles - * on passe le code Swap en parametre, pour faire des essais 'manuels') */ void gdcmHeader::CheckSwap() { + // The only guaranted way of finding the swap code is to find a + // group tag since we know it's length has to be of four bytes i.e. + // 0x00000004. Finding the swap code in then straigthforward. Trouble + // occurs when we can't find such group... guint32 s; guint32 x=4; // x : pour ntohs bool net2host; // true when HostByteOrder is the same as NetworkByteOrder int lgrLue; char * entCur; - char deb[HEADER_LENGHT_TO_READ]; + char deb[HEADER_LENGTH_TO_READ]; // First, compare HostByteOrder and NetworkByteOrder in order to // determine if we shall need to swap bytes (i.e. the Endian type). @@ -123,16 +123,16 @@ void gdcmHeader::CheckSwap() net2host = false; // The easiest case is the one of a DICOM header, since it possesses a - // file preamble where it suffice to look for the sting "DICM". - lgrLue = fread(deb, 1, HEADER_LENGHT_TO_READ, fp); + // file preamble where it suffice to look for the string "DICM". + lgrLue = fread(deb, 1, HEADER_LENGTH_TO_READ, fp); 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 +153,21 @@ 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"); } @@ -189,36 +189,34 @@ 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 : dbg.Verbose(0, "gdcmHeader::CheckSwap:", - "ACE/NEMA unfound swap info (time to raise bets)"); + "ACR/NEMA unfound swap info (time to raise bets)"); } // 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 - if (! net2host ) sw = 0; else @@ -226,56 +224,46 @@ 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; +} + +void gdcmHeader::GetPixels(size_t lgrTotale, void* _Pixels) { + size_t pixelsOffset; + pixelsOffset = GetPixelOffset(); + fseek(fp, pixelsOffset, SEEK_SET); + fread(_Pixels, 1, lgrTotale, fp); +} + + + /** * \ingroup gdcmHeader - * \brief recupere la longueur d'un champ DICOM. - * Preconditions: - * 1/ le fichier doit deja avoir ete ouvert, - * 2/ CheckSwap() doit avoir ete appele - * 3/ la partie 'group' ainsi que la partie 'elem' - * de l'acr_element doivent avoir ete lues. - * - * ACR-NEMA : we allways get - * GroupNumber (2 Octets) - * ElementNumber (2 Octets) - * ElementSize (4 Octets) - * DICOM en implicit Value Representation : - * GroupNumber (2 Octets) - * ElementNumber (2 Octets) - * ElementSize (4 Octets) - * - * DICOM en explicit Value Representation : - * GroupNumber (2 Octets) - * ElementNumber (2 Octets) - * ValueRepresentation (2 Octets) - * ElementSize (2 Octets) - * - * ATTENTION : dans le cas ou ValueRepresentation = OB, OW, SQ, UN - * GroupNumber (2 Octets) - * ElementNumber (2 Octets) - * ValueRepresentation (2 Octets) - * zone reservee (2 Octets) - * ElementSize (4 Octets) - * - * @param sw code swap - * @param skippedLength pointeur sur nombre d'octets que l'on a saute qd - * la lecture est finie - * @param longueurLue pointeur sur longueur (en nombre d'octets) - * effectivement lue - * @return longueur retenue pour le champ + * \brief Find the value representation of the current tag. */ - void gdcmHeader::FindVR( ElValue *ElVal) { - char VR[3]; - int lgrLue; - long PositionOnEntry = ftell(fp); - 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 @@ -283,64 +271,366 @@ 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; + + 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++ implementations 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; - // 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); + // 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. - 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 ImplicitVRLittleEndian one. + * + * @return True when ImplicitVRLittleEndian found. False in all other cases. + */ +bool gdcmHeader::IsImplicitVRLittleEndianTransferSyntax(void) { + ElValue* Element = PubElVals.GetElementByNumber(0x0002, 0x0010); + if ( !Element ) + return false; + LoadElementValueSafe(Element); + string Transfer = Element->GetValue(); + if ( Transfer == "1.2.840.10008.1.2" ) + return true; + return false; +} + +/** + * \ingroup gdcmHeader + * \brief Determines if the Transfer Syntax was allready encountered + * and if it corresponds to a ExplicitVRLittleEndian one. + * + * @return True when ExplicitVRLittleEndian found. False in all other cases. + */ +bool gdcmHeader::IsExplicitVRLittleEndianTransferSyntax(void) { + ElValue* Element = PubElVals.GetElementByNumber(0x0002, 0x0010); + if ( !Element ) + return false; + LoadElementValueSafe(Element); + string Transfer = Element->GetValue(); + if ( Transfer == "1.2.840.10008.1.2.1" ) + return true; + return false; +} + +/** + * \ingroup gdcmHeader + * \brief Determines if the Transfer Syntax was allready encountered + * and if it corresponds to a DeflatedExplicitVRLittleEndian one. + * + * @return True when DeflatedExplicitVRLittleEndian found. False in all other cases. + */ +bool gdcmHeader::IsDeflatedExplicitVRLittleEndianTransferSyntax(void) { + ElValue* Element = PubElVals.GetElementByNumber(0x0002, 0x0010); + if ( !Element ) + return false; + LoadElementValueSafe(Element); + string Transfer = Element->GetValue(); + if ( Transfer == "1.2.840.10008.1.2.1.99" ) + return true; + return false; +} + + +/** + * \ingroup gdcmHeader + * \brief Determines if the Transfer Syntax was allready encountered + * and if it corresponds to a Explicit VR Big Endian one. + * + * @return True when big endian found. False in all other cases. + */ +bool gdcmHeader::IsExplicitVRBigEndianTransferSyntax(void) { + ElValue* Element = PubElVals.GetElementByNumber(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; +} + + +/** + * \ingroup gdcmHeader + * \brief Determines if the Transfer Syntax was allready encountered + * and if it corresponds to a JPEGBaseLineProcess1 one. + * + * @return True when JPEGBaseLineProcess1found. False in all other cases. + */ +bool gdcmHeader::IsJPEGBaseLineProcess1TransferSyntax(void) { + ElValue* Element = PubElVals.GetElementByNumber(0x0002, 0x0010); + if ( !Element ) + return false; + LoadElementValueSafe(Element); + string Transfer = Element->GetValue(); + if ( Transfer == "1.2.840.10008.1.2.4.50" ) + return true; + return false; +} + +/** + * \ingroup gdcmHeader + * \brief Determines if the Transfer Syntax was allready encountered + * and if it corresponds to a JPEGExtendedProcess2-4 one. + * + * @return True when JPEGExtendedProcess2-4 found. False in all other cases. + */ +bool gdcmHeader::IsJPEGExtendedProcess2_4TransferSyntax(void) { + ElValue* Element = PubElVals.GetElementByNumber(0x0002, 0x0010); + if ( !Element ) + return false; + LoadElementValueSafe(Element); + string Transfer = Element->GetValue(); + if ( Transfer == "1.2.840.10008.1.2.4.51" ) + return true; + return false; +} + + +/** + * \ingroup gdcmHeader + * \brief Determines if the Transfer Syntax was allready encountered + * and if it corresponds to a JPEGExtendeProcess3-5 one. + * + * @return True when JPEGExtendedProcess3-5 found. False in all other cases. + */ +bool gdcmHeader::IsJPEGExtendedProcess3_5TransferSyntax(void) { + ElValue* Element = PubElVals.GetElementByNumber(0x0002, 0x0010); + if ( !Element ) + return false; + LoadElementValueSafe(Element); + string Transfer = Element->GetValue(); + if ( Transfer == "1.2.840.10008.1.2.4.52" ) + return true; + return false; +} + +/** + * \ingroup gdcmHeader + * \brief Determines if the Transfer Syntax was allready encountered + * and if it corresponds to a JPEGSpectralSelectionProcess6-8 one. + * + * @return True when JPEGSpectralSelectionProcess6-8 found. False in all other cases. + */ +bool gdcmHeader::IsJPEGSpectralSelectionProcess6_8TransferSyntax(void) { + ElValue* Element = PubElVals.GetElementByNumber(0x0002, 0x0010); + if ( !Element ) + return false; + LoadElementValueSafe(Element); + string Transfer = Element->GetValue(); + if ( Transfer == "1.2.840.10008.1.2.4.53" ) + return true; + return false; +} + +// +// Euhhhhhhh +// Il y en a encore DIX-SEPT, comme ça. +// Il faudrait trouver qq chose + rusé ... +// +// --> probablement TOUS les supprimer (Eric dixit) +// + + +void gdcmHeader::FixFoundLength(ElValue * ElVal, guint32 FoundLength) { + // Heuristic: a final fix. + if ( FoundLength == 0xffffffff) + FoundLength = 0; + ElVal->SetLength(FoundLength); +} + +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(); + if (errno == 1) + return 0; + 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."); + errno = 1; + return 0; + } + if ( n == 0xe0dd ) + FoundSequenceDelimiter = true; + else if ( n != 0xe000) { + dbg.Verbose(1, "gdcmHeader::FindLengthOB: ", + "wrong element for an item sequence."); + errno = 1; + return 0; + } + 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) { - guint32 length32; +void gdcmHeader::FindLength(ElValue * ElVal) { + 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); + 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. - length32 = ReadInt32(); + + // 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 consists 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 == 0x000) && (length16 == 0x0400) ) { + if ( ! IsExplicitVRBigEndianTransferSyntax() ) { + dbg.Verbose(0, "gdcmHeader::FindLength", "not explicit VR"); + errno = 1; + return; + } + 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 = GetDictEntryByKey(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 ( 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 +658,8 @@ 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) ) @@ -378,35 +667,93 @@ guint16 gdcmHeader::SwapShort(guint16 a) { return (a); } -void gdcmHeader::SkipElementValue(ElValue * ElVal) { +void gdcmHeader::SkipBytes(guint32 NBytes) { //FIXME don't dump the returned value - (void)fseek(fp, (long)ElVal->GetLength(), SEEK_CUR); + (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 Loads the element content if it's length is not bigger + * than the value specified with + * gdcmHeader::SetMaxSizeLoadElementValue() + */ 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" ) { + // + // Ne faudrait-il pas au contraire trouver immediatement + // une maniere 'propre' de traiter les sequences (vr = SQ) + // car commencer par les ignorer risque de conduire a qq chose + // qui pourrait ne pas etre generalisable + // + 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; // modif sauvage JPR + // On charge la longueur du groupe + // quand l'element 0x0000 est présent ! + + 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; + } + + // The elements whose length is bigger than the specified upper bound + // are not loaded. Instead we leave a short notice of the offset of + // the element content and it's length. + if (length > MaxSizeLoadElementValue) { + ostringstream s; + s << "gdcm::NotLoaded."; + s << " Address:" << (long)ElVal->GetOffset(); + s << " 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(); @@ -423,32 +770,48 @@ void gdcmHeader::LoadElementValue(ElValue * ElVal) { } // FIXME The exact size should be length if we move to strings or whatever - char* NewValue = (char*)g_malloc(length+1); + char* NewValue = (char*)malloc(length+1); if( !NewValue) { dbg.Verbose(1, "LoadElementValue: Failed to allocate NewValue"); return; } 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); - Error::FileReadError(fp, "gdcmHeader::LoadElementValue"); + free(NewValue); + dbg.Verbose(1, "gdcmHeader::LoadElementValue","unread element value"); 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"); + errno = 0; + if ( item_read != 1 ) { + dbg.Verbose(1, "gdcmHeader::ReadInt16", " File read error"); + errno = 1; + return 0; + } g = SwapShort(g); return g; } @@ -457,12 +820,62 @@ 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"); + errno = 0; + if ( item_read != 1 ) { + dbg.Verbose(1, "gdcmHeader::ReadInt32", " File read error"); + errno = 1; + return 0; + } g = SwapLong(g); return g; } +/** + * \ingroup gdcmHeader + * \brief Build a new Element Value from all the low level arguments. + * Check for existence of dictionary entry, and build + * a default one when absent. + * @param Group group of the underlying DictEntry + * @param Elem element of the underlying DictEntry + */ +ElValue* gdcmHeader::NewElValueByKey(guint16 Group, guint16 Elem) { + // Find out if the tag we encountered is in the dictionaries: + gdcmDictEntry * NewTag = GetDictEntryByKey(Group, Elem); + if (!NewTag) + NewTag = new gdcmDictEntry(Group, Elem); + + ElValue* NewElVal = new ElValue(NewTag); + if (!NewElVal) { + dbg.Verbose(1, "gdcmHeader::NewElValueByKey", + "failed to allocate ElValue"); + return (ElValue*)0; + } + return NewElVal; +} + +/** + * \ingroup gdcmHeader + * \brief Build a new Element Value from all the low level arguments. + * Check for existence of dictionary entry, and build + * a default one when absent. + * @param Name Name of the underlying DictEntry + */ +ElValue* gdcmHeader::NewElValueByName(string Name) { + + gdcmDictEntry * NewTag = GetDictEntryByName(Name); + if (!NewTag) + NewTag = new gdcmDictEntry(0xffff, 0xffff, "LO", "Unknown", Name); + + ElValue* NewElVal = new ElValue(NewTag); + if (!NewElVal) { + dbg.Verbose(1, "gdcmHeader::ObtainElValueByName", + "failed to allocate ElValue"); + return (ElValue*)0; + } + return NewElVal; +} + + /** * \ingroup gdcmHeader * \brief Read the next tag without loading it's value @@ -474,42 +887,38 @@ ElValue * gdcmHeader::ReadNextElement(void) { guint16 n; ElValue * NewElVal; - try { - g = ReadInt16(); - n = ReadInt16(); - } - catch ( Error::FileReadError ) { + g = ReadInt16(); + n = ReadInt16(); + if (errno == 1) // 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", "Unknown"); - - NewElVal = new ElValue(NewTag); - if (!NewElVal) { - dbg.Verbose(1, "ReadNextElement: failed to allocate ElValue"); - return (ElValue*)0; - } - + + NewElVal = NewElValueByKey(g, n); FindVR(NewElVal); FindLength(NewElVal); + if (errno == 1) + // 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 ) @@ -525,8 +934,9 @@ bool gdcmHeader::IsAnInteger(guint16 group, guint16 element, return true; if ( (group == 0x0028) && (element == 0x0005) ) - // This tag is retained from ACR/NEMA - // CHECKME Why should "Image Dimensions" be a single integer ? + // The "Image Dimensions" tag is retained from ACR/NEMA and contains + // the number of dimensions of the contained object (1 for Signal, + // 2 for Image, 3 for Volume, 4 for Sequence). return true; if ( (group == 0x0028) && (element == 0x0200) ) @@ -551,8 +961,8 @@ size_t gdcmHeader::GetPixelOffset(void) { guint16 grPixel; guint16 numPixel; string ImageLocation = GetPubElValByName("Image Location"); - if ( ImageLocation == "UNFOUND" ) { - grPixel = 0x7FE0; + if ( ImageLocation == "gdcm::Unfound" ) { + grPixel = 0x7fe0; } else { grPixel = (guint16) atoi( ImageLocation.c_str() ); } @@ -562,26 +972,63 @@ size_t gdcmHeader::GetPixelOffset(void) { numPixel = 0x1010; else numPixel = 0x0010; - ElValue* PixelElement = PubElVals.GetElement(grPixel, numPixel); + ElValue* PixelElement = PubElVals.GetElementByNumber(grPixel, numPixel); if (PixelElement) return PixelElement->GetOffset(); else return 0; } -gdcmDictEntry * gdcmHeader::IsInDicts(guint32 group, guint32 element) { +/** + * \ingroup gdcmHeader + * \brief Searches both the public and the shadow dictionary (when they + * exist) for the presence of the DictEntry with given + * group and element. The public dictionary has precedence on the + * shadow one. + * @param group group of the searched DictEntry + * @param element element of the searched DictEntry + * @return Corresponding DictEntry when it exists, NULL otherwise. + */ +gdcmDictEntry * gdcmHeader::GetDictEntryByKey(guint16 group, guint16 element) { gdcmDictEntry * found = (gdcmDictEntry*)0; if (!RefPubDict && !RefShaDict) { - //FIXME build a default dictionary ! - printf("FIXME in gdcmHeader::IsInDicts\n"); + dbg.Verbose(0, "FIXME in gdcmHeader::GetDictEntry", + "we SHOULD have a default dictionary"); } if (RefPubDict) { - found = RefPubDict->GetTag(group, element); + found = RefPubDict->GetTagByKey(group, element); if (found) return found; } if (RefShaDict) { - found = RefShaDict->GetTag(group, element); + found = RefShaDict->GetTagByKey(group, element); + if (found) + return found; + } + return found; +} + +/** + * \ingroup gdcmHeader + * \brief Searches both the public and the shadow dictionary (when they + * exist) for the presence of the DictEntry with given name. + * The public dictionary has precedence on the shadow one. + * @earam Name name of the searched DictEntry + * @return Corresponding DictEntry when it exists, NULL otherwise. + */ +gdcmDictEntry * gdcmHeader::GetDictEntryByName(string Name) { + gdcmDictEntry * found = (gdcmDictEntry*)0; + if (!RefPubDict && !RefShaDict) { + dbg.Verbose(0, "FIXME in gdcmHeader::GetDictEntry", + "we SHOULD have a default dictionary"); + } + if (RefPubDict) { + found = RefPubDict->GetTagByName(Name); + if (found) + return found; + } + if (RefShaDict) { + found = RefShaDict->GetTagByName(Name); if (found) return found; } @@ -589,18 +1036,138 @@ gdcmDictEntry * gdcmHeader::IsInDicts(guint32 group, guint32 element) { } string gdcmHeader::GetPubElValByNumber(guint16 group, guint16 element) { - return PubElVals.GetElValue(group, element); + return PubElVals.GetElValueByNumber(group, element); +} + +string gdcmHeader::GetPubElValRepByNumber(guint16 group, guint16 element) { + ElValue* elem = PubElVals.GetElementByNumber(group, element); + if ( !elem ) + return "gdcm::Unfound"; + return elem->GetVR(); } string gdcmHeader::GetPubElValByName(string TagName) { - return PubElVals.GetElValue(TagName); + return PubElVals.GetElValueByName(TagName); +} + +string gdcmHeader::GetPubElValRepByName(string TagName) { + ElValue* elem = PubElVals.GetElementByName(TagName); + if ( !elem ) + return "gdcm::Unfound"; + return elem->GetVR(); +} + +string gdcmHeader::GetShaElValByNumber(guint16 group, guint16 element) { + return ShaElVals.GetElValueByNumber(group, element); +} + +string gdcmHeader::GetShaElValRepByNumber(guint16 group, guint16 element) { + ElValue* elem = ShaElVals.GetElementByNumber(group, element); + if ( !elem ) + return "gdcm::Unfound"; + return elem->GetVR(); +} + +string gdcmHeader::GetShaElValByName(string TagName) { + return ShaElVals.GetElValueByName(TagName); +} + +string gdcmHeader::GetShaElValRepByName(string TagName) { + ElValue* elem = ShaElVals.GetElementByName(TagName); + if ( !elem ) + return "gdcm::Unfound"; + return elem->GetVR(); +} + +string gdcmHeader::GetElValByNumber(guint16 group, guint16 element) { + string pub = GetPubElValByNumber(group, element); + if (pub.length()) + return pub; + return GetShaElValByNumber(group, element); +} + +string gdcmHeader::GetElValRepByNumber(guint16 group, guint16 element) { + string pub = GetPubElValRepByNumber(group, element); + if (pub.length()) + return pub; + return GetShaElValRepByNumber(group, element); +} + +string gdcmHeader::GetElValByName(string TagName) { + string pub = GetPubElValByName(TagName); + if (pub.length()) + return pub; + return GetShaElValByName(TagName); +} + +string gdcmHeader::GetElValRepByName(string TagName) { + string pub = GetPubElValRepByName(TagName); + if (pub.length()) + return pub; + return GetShaElValRepByName(TagName); } /** * \ingroup gdcmHeader - * \brief Parses the header of the file but does NOT load element values. + * \brief Accesses an existing ElValue in the PubElVals of this instance + * through it's (group, element) and modifies it's content with + * the given value. + * @param content new value to substitute with + * @param group group of the ElVal to modify + * @param element element of the ElVal to modify */ -void gdcmHeader::ParseHeader(void) { +int gdcmHeader::SetPubElValByNumber(string content, guint16 group, + guint16 element) +{ + //CLEANME TagKey key = gdcmDictEntry::TranslateToKey(group, element); + //CLEANME PubElVals.tagHt[key]->SetValue(content); + return ( PubElVals.SetElValueByNumber (content, group, element) ); +} + +/** + * \ingroup gdcmHeader + * \brief Accesses an existing ElValue in the PubElVals of this instance + * through tag name and modifies it's content with the given value. + * @param content new value to substitute with + * @param TagName name of the tag to be modified + */ +int gdcmHeader::SetPubElValByName(string content, string TagName) { + //CLEANME TagKey key = gdcmDictEntry::TranslateToKey(group, element); + //CLEANME PubElVals.tagHt[key]->SetValue(content); + return ( PubElVals.SetElValueByName (content, TagName) ); +} + +/** + * \ingroup gdcmHeader + * \brief Accesses an existing ElValue in the ShaElVals of this instance + * through it's (group, element) and modifies it's content with + * the given value. + * @param content new value to substitute with + * @param group group of the ElVal to modify + * @param element element of the ElVal to modify + */ +int gdcmHeader::SetShaElValByNumber(string content, + guint16 group, guint16 element) +{ + return ( ShaElVals.SetElValueByNumber (content, group, element) ); +} + +/** + * \ingroup gdcmHeader + * \brief Accesses an existing ElValue in the ShaElVals of this instance + * through tag name and modifies it's content with the given value. + * @param content new value to substitute with + * @param TagName name of the tag to be modified + */ +int gdcmHeader::SetShaElValByName(string content, string TagName) { + return ( ShaElVals.SetElValueByName (content, TagName) ); +} + +/** + * \ingroup gdcmHeader + * \brief Parses the header of the file but WITHOUT loading element values. + */ +void gdcmHeader::ParseHeader(bool exception_on_error) throw(gdcmFormatError) { ElValue * newElValue = (ElValue *)0; rewind(fp); @@ -611,16 +1178,45 @@ void gdcmHeader::ParseHeader(void) { } } +/** + * \ingroup gdcmHeader + * \brief Once the header is parsed add some gdcm convenience/helper elements + * in the ElValSet. For example add: + * - gdcmImageType which is an entry containing a short for the + * type of image and whose value ranges in + * I8 (unsigned 8 bit image) + * I16 (unsigned 8 bit image) + * IS16 (signed 8 bit image) + * - gdcmXsize, gdcmYsize, gdcmZsize whose values are respectively + * the ones of the official DICOM fields Rows, Columns and Planes. + */ +void gdcmHeader::AddAndDefaultElements(void) { + ElValue* NewEntry = (ElValue*)0; + + NewEntry = NewElValueByName("gdcmXSize"); + NewEntry->SetValue(GetElValByName("Rows")); + PubElVals.Add(NewEntry); + + NewEntry = NewElValueByName("gdcmYSize"); + NewEntry->SetValue(GetElValByName("Columns")); + PubElVals.Add(NewEntry); + + NewEntry = NewElValueByName("gdcmZSize"); + NewEntry->SetValue(GetElValByName("Planes")); + PubElVals.Add(NewEntry); +} + /** * \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); + rewind(fp); TagElValueHT ht = PubElVals.GetTagHt(); - for (TagElValueHT::iterator tag = ht.begin(); tag != ht.end(); ++tag) + for (TagElValueHT::iterator tag = ht.begin(); tag != ht.end(); ++tag) { LoadElementValue(tag->second); + } } void gdcmHeader::PrintPubElVal(ostream & os) {