Martin Yeo

So excited to see this in action (_hopefully_ I can be some use in helping)!

@stephenworsley fancy updating the branch and we can restart the review process?

From @SciTools/peloton Thank you to the stale bot for reminding us about this. We've had a look at CF conventions, and 3 years on it doesn't look like cell methods...

Several stakeholders raised this as needing to be fixed at a recent 'surgery'.

Note that I have a potential solution to the memory benchmarks being noisy, but it needs more exploration: #5981

We should now be in a good position to add this without missing a spot: https://github.com/SciTools/cf-units/blob/539dbccd5597b73c1785bbd31d51e4c304e7ffc1/cf_units/tests/test_coding_standards.py#L121-L122

> N.B. [pytest doctests --> Python standard-library doctest module](https://docs.pytest.org/en/7.1.x/how-to/doctest.html#how-to-run-doctests). Sphinx interpretation is probably [the same](https://www.sphinx-doc.org/en/master/usage/extensions/doctest.html#directive-doctest) , maybe not totally ? https://github.com/SciTools/iris/issues/6110

Hashing arrays may be a solution, so long as the hashing is tolerant of minor changes.

From discussion with @HGWright and @ESadek-MO and @trexfeathers: string representation of NumPy arrays is not expected to be stable - there is no 'contract' for this, so long as a...

@pp-mo to provide tolerance while still providing human readability, I think we would need to write our own parser, perhaps based on [`numpy.fromstring()`](https://numpy.org/doc/stable/reference/generated/numpy.fromstring.html) but supporting N dimensions. To provide the...