V3.0

This version brings two compatibility breaks:

• CliMAF is no more compatible with python2
• CliMAF module html has been renamed into chtml to avoid conflicts with python standard package html
• Because CliMAF now uses xarray as sole netcdf package, users on shared machine must export OPENBLAS_NUM_THREADS=1 or another sensible value

The other improvements and bugfixes are:

• CliMAF now runs on new IPSL clusters 'spirit1' and "spirit2' and at TGCC on machine 'irene' and 'irene-amd' (including tests suite), all CliMAF examples have been adapted; also, the default format set for CDO outputs is now NetDCF4.
• function fds now sets the dataset's frequency using file's global attribute 'frequency'
• dataset's method cdataset.check has been reworked, accepts flags for setting what to check, and provides a consistent return value (which can be None if a decision cannot be made)
• function ds can resolve period=* if it is the only wildcard; triggered by setting classes.auto_resolve to True
• default is now to deactivate optimization for CMIP6 data indexing, which seems buggy
• msftyz data for IPSLCM6 and CMIP6 are automatically fixed for their degenerated dimension, provided environment variable CLIMAF_FIX_IPSL_CMIP6_MSFTYZ is set (to any value)
• when dealing with netcdf files without time bounds, it must be specified when the project is defined using option 'use_frequency=True'. 'frequency' must also be a facet of the project and it must be initialized using cdef. 'frequency' currently support the following values: 'mon' (month), 'MS' (month), 'D' (day), 'H' (hour), 'M' (minute).
• A project can now be derived from an existing one (facets, dataloc).

Bugs :

• operators regrid and regridn take care of discarding any ancillary variable in input data (such as area with sos), in order that CDO succeeds in regriding
• fix on processing period=last_XX for a CMIP5 case
• fix bugs in find_files (for case period = fx), in clist

Technical :

• clean-up of netcdf libraries use : xarray becomes the sole package used; module anynetcdf is discarded and only functions provided by module nectdfbasics are used throughout CliMAF code, except for a very few calls to xarray (in driver.cread and classes.cdatset.check)
• remove dependency to module 'cesemp_env' in bin/climaf
• ease installation by embarking epstopdf
• Engine mcdo.py now applies first the selection on date, in order to take advantage of much improved CDO/NetCDF performance on this operation

V2.0.2

This version comes with bugfixes and improvements:

• optimization of the find of CMIP6 data when some facets are wildcards
• new glob function for cdataset
• cache improvements and add of the cost computation
• ESMValTool can now be called from CliMAF

V2.0.1

This is a light versions with:

• changes to process multiple ensembles
• robustness improvment
• small fixes

V2.0.0

This version brings full compatibility with both python2 and python3. It also break compatibility with former versions.
In particular, the operator slice has been renamed into cslice_average

Several other changes have been done including:

• additional cache implementation for the results of cvalues
• changes in module location of some functions and variables
• add options in dataloc to define more precisely locations

All the changes are described here:
https://climaf.readthedocs.io/en/v2.0.0/news.html

Version 1.2.13

This new version comes with several changes including :

• change in cache structure to limit the number of i-nodes
• modules climaf.site_settings and climaf.clogging have been moved to env.site_settings and env.clogging to allow an easier access to the variables contained their in scripts
• a new test base is available
  All the changes are described here: https://climaf.readthedocs.io/en/v1.2.13/news.html

It will be the last version available only in python2.

V1.2.12

Change doc.

Version 0.5.0

Introduces the ability to define project's attributes, and adds a large number of improvements; see "what's new" in doc

Version 0.4.0

First version presented 8 and 10 april 2014