Nctoolkit

This is a major release with breaking changes.

A critical change is that accessing dataset attributes will force evaluation, i.e. `ds.run()` will occur if you access an attribute such as `ds.times`. This makes behaviour more consistent with what new users would expect.

This should not have a major impact on any coding workflows.

The `cdo_command` method has been changed so that it no longer checks the validity of the command before calling CDO. Checks can now be run using the `check` argument.

Some improvements have been made to method internals for `pub_plot`.

This is a minor release with no breaking changes.

The `pub_plot` method now accepts `fig` as an argument to make it easier to manipulate figures.

Improvements have been made to some internals.

This is a bug-fix release to ensure nctoolkit runs from scripts in macOS. This changes internal behaviour on macOS, so that multiprocess instead of multiprocessing is used for handling parallel processing.

This release contains a new method for producing high quality static plots. Yuri Artioli of Plymouth Marine Laboratory contributed the core code to this new `pub_plot` method.

This is a quick release to make the package fully compatible with Pandas v2.0.0. Changes in the pandas api were causing some add/subtract etc. methods to fail for daily data. This is now fixed.

A couple of improvements have been made to internals.

This is a major(ish) release with some breaking changes related to plotting.

On pypi, cartopy has been switched to an optional dependency because it was causing installation difficulties for some users. You can now do a "complete" installation using pip to get all optional dependencies::

$ pip install nctoolkit[complete]

This does not impact the conda version, which will behave as before.

Support is now available for Python 3.11.

File paths with spaces are now supported.