Cf-python

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

**2017-03-17**

* Fixed bug when accessing PP file whose format/endian/word-size has
been specified

----

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

**2017-03-14**

* Can specify 'pp' or 'PP' in um option to `cf.read`

----

-----------

**2017-02-24**

* Changed weights in calculation of variance to reliability weights
(from frequency weights). This not only scientifically better, but
faster, too.

----

-----------

**2017-02-22**

* Rounded datetime to time-since conversions to the nearest
microsecond, to reflect the accuracy of netCDF4.netcdftime
* Removed import tests from setup.py
* New option --um to ``cfa``, ``cfdump``
* New parameter um to `cf.read`

----

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

**2017-01-31**

* Rounded datetime to time-since conversions to the nearest
microsecond, to reflect the accuracy of netCDF4.netcdftime
* Fix for netCDF4.__version__ > 1.2.4 do to with datetime.calendar
*handle with care*

----

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

**2016-09-21**

* Added --build-id to LDFLAGS in umread Makefile, for sake of RPM
builds (otherwise fails when building debuginfo RPM). Pull request
16, thanks to Klaus Zimmermann.
* Improved test handling. Pull request 21, thanks to Klaus
Zimmermann.
* Removed udunits2 database. This removes the modified version of the
udunits2 database in order to avoid redundancies, possible version
incompatibilities, and license questions. The modifications are
instead carried out programmatically in units.py. Pull request 20,
thanks to Klaus Zimmermann.

----