Minor update

• Encapsulated h5pp variables in exported/installed files to avoid pollution of users variables. Every variable now starts with H5PP_
• Corrected travis-ci testing with by specifiying preferred conda libs

Minor update

• Fixed -lpthread in static linking linking (whole-archive)
• Clarified readme
• Moved pthread linking to hdf5 find package

Improved build and linking on older compilers

• Improved compatibility with older compilers. Now working with GNU GCC 7 and newer, Clang 6 and newer.

• HDF5 now works with version 1.8

• Spdlog now includes headers correctly from conda-forge version.

• CMake configure-time time checks for headers such as , and <experimental/type_traits>

New download method
h5pp is now available through anaconda: https://anaconda.org/DavidAce/h5pp
Install with conda install -c davidace h5pp

h5pp v0.4.0 - Performance improvements

Higlights
Copy-free writing of std::complex and pod-structs

Changelog:

• Avoid internal copy of std::complex and struct types.
• Better checks to see if row-major conversion is necessary.
• Interpret C-style data as row-major directly
• Allow empty datasets
• Expanded tests to cover new Scalar2 and Scalar3 types
• Improved logging for debugging

Added Scalar2 and Scalar3 support

h5pp can now write Scalar2, Scalar3 types, which are simply structs with the pattern struct{scalar x; scalar y;} or struct{scalar x; scalar y; scalar z; }, where "scalar" is any arithmetic type (int, double, uint, long double, etc). As long as your type has these three members you should be able to write it with h5pp.

First packaged release

First release with packaged .deb file