Cf-python

-------------

**2019-10-01**

* Updated description in ``setup.py``

----

version 3.0.0 (*first Python 3 release*)
----------------------------------------

**2019-10-01**

* Complete refactor for Python 3, including some API changes.

Scripts written for version 2.x but running under version 3.x should
either work as expected, or provide informative error messages on
the new API usage. However, it is advised that the outputs of older
scripts be checked when running with Python 3 versions of the cf
library.
* Deprecated ``cfdump`` (its functionality is now included in
``cfa``).

----

version 2.3.8 (*last Python 2 release*)
---------------------------------------

**2019-10-07**

* In `cf.write`, can set ``single=False`` to mean ``double=True``, and
vice versa.
* Fixed bug in `cf.aggregate` - removed overly strict test on
dimension coordinate bounds.
* Fixed bug in `cf.read` that set the climatology attribute to True
when there are no bounds.
* Fixed bug in `cf.write` when writing missing values (set_fill was
off, now on)

----

-------------

**2019-04-04**

* Changed calculation of chunksize in parallel case to avoid potential
problems and introduced a new method `cf.SET_PERFORMANCE` to tune
the chunksize and the fraction of memory to keep free.

----

-------------

**2019-03-27**

* Fix bug in creating a during cell method during a field collapse.

----

-------------

**2019-03-05**

* Allow failure to compile to go through with a warning, rather than
failing to install. if this happens, reading a PP/UM file will
result in "Exception: Can't determine format of file test2.pp"
* Fixed bug in `cf.Field.convolution_filter` giving false error over
units.

----

-------------

**2018-12-10**

* `cf.Field.regridc` now compares the units of the source and
destination grids and converts between them if possible or raises an
error if they are not equivalent.

----

-------------

**2018-11-07**

* Fixed bug in `cf.Field.regridc` that caused it to fail when
regridding a multidimensional field along only one dimension.
* Fixed bug which in which the default logarithm is base 10, rather
than base e