Cf-python

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

**2012-11-27**

* Assignment to a field's data array with metadata-aware broadcasting,
assigning to subspaces, assignment where data meets conditions,
assignment to unmasked elements, etc. (setitem method)
* Proper treatment of the missing data mask, including metadata-aware
assignment (setmask method)
* Proper treatment of ancillary data.
* Ancillary data and transforms are subspaced with their parent field.
* Much faster aggregation algorithm (with thanks to Jonathan
Gregory). Also aggregates fields transforms, ancillary variables and
flags.

----

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

**2012-10-01**

* Restructured documentation and package code files.
* Large Amounts of Massive Arrays (LAMA) functionality.
* Metadata-aware field manipulation and combination with
metadata-aware broadcasting.
* Better treatment of cell measures.
* Slightly faster aggregation algorithm (a much improved one is in
development).
* API changes for clarity.
* Bug fixes.
* Added 'TEMPDIR' to the `cf.CONSTANTS` dictionary
* This is a snapshot of the trunk at revision r195.

----

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

**2012-09-19**

* Loads of exciting improvements - mainly LAMA functionality,
metadata-aware field manipulation and documentation.
* This is a snapshot of the trunk at revision r185. A proper vn0.9.5
release is imminent.

----

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

**2012-04-17**

* General bug fixes and code restructure

----

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

**2012-03-15**

* A proper treatment of units using the Udunits C library and the
extra time functionality provided by the netCDF4 package.
* A command line script to do CF-netCDF to CF-netCDF via cf-python.

----

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

**2012-02-08**

* Objects renamed in line with the CF data model: `cf.Space` becomes
`cf.Field` and `cf.Grid` becomes `cf.Space`.
* Field aggregation using the CF aggregation rules is available when
reading fields from disk and on fields in memory. The data of a
field resulting from aggregation are stored as a collection of the
data from the component fields and so, as before, may be file
pointers, arrays in memory or a mixture of these two forms.
* Units, missing data flags, dimension order, dimension direction and
packing flags may all be different between data components and are
conformed at the time of data access.
* Files in UK Met Office PP format may now be read into CF fields.
* A command line script for PP to CF-netCDF file conversion is
provided.

----