- introducing zlib.dll from conda forge libzlib and removing z*.lib files fixes #104
- fixing tests removing np.NaN to work with latest numpy version.
- updating zlib windows libs and h files to 1.3.1 fixes #100
- fixing bug when writing an object column filtered so that the first row has not index 0
- conda forge of 0.4.8 failed due to missing file in source distribution, adding file to manifest and releasing again, otherwise no changes.
- Updated code to work with Cython 3.0.0, earlier versions no longer supported.
- Including dll files in pyreadr folder and putting them as package data for windows, this solves issue #72. For unix no package data meaning no dll files but also no .c translated files in the wheel.
- added toml file, fixes #88
- Added Citation
- fixed signature #84
- fixed numpy.float deprecation
- fixed truncation of files when writing on windows
- Introducing new pandas data types Float64Dtype and Float32Dtype when writing. This requires pandas > 1.2.0. Python 3.6 taken out of wheels as there is no 1.2.0 version.
- fixed bug when translating datetime to string when writing.
- Added support for pathlib, solves #70
- removed warnings about np.bool and np.object
- fixing seg fault with linux wheels #62
- Implemented reading dataframe rownames as pandas index
- Implementing reading matrices, arrays, tables, solves #5.
- librdata source updated to commit 8623be2c626392028cd75c5fc45a90e1f70d97b9
- Fixed sharing flags for opening files on windows
- Added download_file function
- using std=gnu99 from env_vars.sh in wheels to solve #56
- Added license to setup.py
- Solved #48
- Updated librdata source to commit: 7188fa54b1894da24ceb8ccefd0a62113a38497c This improves reading of altrep objects and gives a better error when encountering S4 objects solves issues: #36, #35, #30
- Implemented reading files with lzma compression
- Not producing wheels for python 3.5 anymore
- introduces gzip compression as an option (#41)
- fixes bug #40
- Removed the limit on column bytes length
- Reading dates vectors correctly (libradata issue #24.
- Librdata src updated to commit ba28c3ba1bb224901b873ce477282f61a51c567e
- corrected bug when writing file not being able to delete them with python on windows.
- raising error if trying to run setup.py on windows 32 bit.
- reading files with bzip2 compression
- changed compilation on windows to mingw so that bzip can work (not working with mvsc). This also will help keeping the same source as librdata instead of manually doing the changes required for the sources to work with mvsc.
- write is able to cope with mixed integer pandas types
- Windows dlls are now copied into the package folder instead of python root
- expanding user for list objects and write functions
- expanding user (~) when reading files, checking that file exists.
- updated librdata to commit a29cba35bc167b2f60c90a7cafa5630fbe54b053 to correct bug reported at the end of issue #3
- pandas added as install requirement
- static linking iconv on mac.
- librdata updated to 5987b140875eab59bf876ed18f2e5344484fe376 implemented the ability to read vectors of type Date.
- librdata source updated to 615dee09955b318ce128e0bfcc9d50aa3c9b7b7c fixes #10 and reading files produced on windows with non ascii characters in content.
- Manifest file changed to include windows dll and those are taken from setup.py for unix, this fixes installation error messages on macos when using brew.
- librdata source updated to b7ca1252b670d60a6e8d3c8ce6fcc8ab3b43d6ab fixes #3.
- setup.py fixed to build on conda-forge.
- conda forge package released
- fixed distinguishing between empty string and NaN/NaT/None when writing.
- librdata src updated to latest sources as 2019.01.19
- fixed distinguishing between empty string and NA when reading.
- writing support added.
- First stable release.