- Tidied LICENSE information
- Added a s3_nc_cfa_split.py program to split a large netCDF file into smaller netCDF subarray files and produce a CFA-netCDF master array file.
- Bugfixes to the splitter.
- Bug fix for indexing.
- Changed the name of the template config file and the name in the README.md file to match the code.
- Allow an environment variable "S3_NC_CONFIG" to be set to point to the config file.
- Fixed bug in previous file that prevented writing CFA sub-array files (!)
- Update the s3_nc_cfa_agg.py program so that it is compatible with more models and Datasets in CMIP6. This relates mostly to the way the time dimension is recorded, and the calendar type.
- Changed the way that the indexing for unequal sized partitions is calculated. It is now (potentially) slower, but more robust.
- Added support for reading unequal sized partitions. These may occur in files written by the s3_nc_cfa_agg.py program.
- s3nc_cfa_agg.py now uses FileManager.request_file rather than FileManager._open. More elegant and API focused.
- FileManager.request_file is now compatible with passing globs into it as the filename parameter.
- Fixed a problem where a BytesIO buffer was being passed by reference rather than copied, leading to a "file operation on unopened file" error.
- Corrected install procedure in README.
- Corrected bug in test_s3Dataset_read.
- Fixed unreleased file for Datasets on disk
- Fixed incorrect parsing for CFA 0.4
- complete rewrite
- v0.5 CFA
- partition matrix represented internally by netCDF Dataset
- user can supply sub array size when creating variable
- cacheless operation, except for read of very large files
- intelligent memory handling
- excellent sparse-array handling
- complete compliance with netCDF4 API interface