Grassroots DICOM read-only mirror. Only for Pull Request.

The idea is to use for VR long long instead long (previous VR update in #85 int to long does not work on many platforms, size of long is varying and can be easily same as int type before), also tell compiler about VR's type with enum:long long (C++11).

File gdcmTagToType.h have to be re-generated, xls file was updated

Edit 2: done so far, note: many lines are from generated file

Edit 1: VS 2013 x64 compiler errors: `12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(84): warning C4309: 'initializing' : truncation of constant value 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(93): warning C4309: 'initializing' : truncation of constant value 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(313): error C2766: explicit specialization; 'gdcm::VRToEncoding<0>' has already been defined 12> c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(304) : see previous definition of 'VRToEncoding<0>' 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(313): error C2766: explicit specialization; 'gdcm::VRToType<0>' has already been defined 12> c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(304) : see previous definition of 'VRToType<0>' 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(356): error C2196: case value '0' already used 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(360): error C2196: case value '0' already used 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(369): error C2196: case value '1' already used 12> gdcmDataSet.cxx 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(84): warning C4309: 'initializing' : truncation of constant value 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(93): warning C4309: 'initializing' : truncation of constant value 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(313): error C2766: explicit specialization; 'gdcm::VRToEncoding<0>' has already been defined 12> c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(304) : see previous definition of 'VRToEncoding<0>' 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(313): error C2766: explicit specialization; 'gdcm::VRToType<0>' has already been defined 12> c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(304) : see previous definition of 'VRToType<0>' 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(356): error C2196: case value '0' already used 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(360): error C2196: case value '0' already used 12>c:\users\mi\desktop\gdcm\source\datastructureandencodingdefinition\gdcmVR.h(369): error C2196: case value '1' already used 12> gdcmExplicitDataElement.cxx

<...>`

Have found that an interventional workspot from Philips writes the Slope/Intercept values inside the XRay3DReconstructionSequence. The DCS doesn't give word about this decision.

Indeed, in the XRay3DReconstructionImageIOD, the multi-frame functional group doesn't define the Pixel Value Transformation as mandatory, but user option.

This PR adds support to these specific DICOM files.

Bug: I have RTStruct slices located in [-102.134, -99.1341, -96.1341] along z-axis and the CT slices are located in [-102, -99, -96]. So the SOPInstanceUID was empty because the difference between both was >0.01.

Solution: I propose to take into account the slice thickness (tag 0x0018, 0x0050) instead of the value of 0.01 if this tag is present.

Add a swig typemap for (const char *, gdcm::VL) that takes a byte string (i.e. byte in Python 3 and str in Python 2) to make it possible to use DataElement::SetByteValue() and CSAElement::SetByteValue() from Python 3.

We have came across a strange error, occurring on older version of xcode only (clang 8.0.0) when debug build is used, compiling with -O2 doesn't manifest this bug.

It shows up both in GDCM build and in ITK build, see https://open.cdash.org/viewTest.php?onlyfailed&buildid=6236162 , https://open.cdash.org/viewTest.php?buildid=6230072

The mechanism of the bug is that static function GetTag() of Element class is executed with different template parameters then the calling class when called from inside of templated memer function. Solution is to call Tag(Group,Element) directly. @seanm

• MITRA LINKED ATTIBUTES 1.0 0031,xx20 was listed as IS instead of LO
• several MITRA OBJECT UTF8 ATTRIBUTES 1.0 entries were OB instead of PN
• many MITRA PRESENTATION 1.0 entries were missing
• most of the above, even when present, were missing name attributes

Previously, the inter slice value was based on SpacingBetweenSlice tag. But prefer to compute inter slice value from the difference of z positions of the 2 first slices. Extracted from https://stackoverflow.com/questions/37730772/get-distance-between-slices-in-dicom/41848360#41848360

Follow https://github.com/malaterre/GDCM/pull/125

Using CMAKE_INTERPROCEDURAL_OPTIMIZATION:BOOL=ON allows the enhanced linker/compiler interactions to detect violations of the "one definition rule" where object symbols in different compilation units with different definitions results in undefined behavior.

The last two patches in this commit address the found missing name mangling for the 3 jpeglib variants (8,12,16bit) where some symbols were duplicated (which means that only one of them was retained in normal non-interprocedual linkage)

When compiling on Windows with MinGW, a prefix of "lib" is automatically prepended to the JNI library name. However, the library loader in gdcmJNI.java assumes there is no lib prefix if running on Windows, causing the library to not be found.

This modification manually disables the "lib" prefix when compiling for Windows.

This patch addresses compiler warnings on OSx (Apple LLVM version 8.0.0 (clang-800.0.38)) concerning undefined behavior when shifting negative numbers:

warning: shifting a negative signed value is undefined [-Wshift-negative-value]

Some Lossless JPEG, Nonhierarchical (Processes 14, Transfer Syntax UID 1.2.840.10008.1.2.4.57) from particular scanners do contain some non-0 value in Se and Ah. Although they are not used, this fact makes impossible to import such DICOMs. Lesser check makes sure even such DICOMs are supported.