Cubeprinter adopt

[pp-mo]

🚀 Pull Request

Description

Followon from #3998
Final small formatting changes
Unsurprisingly, I had to squash + rebase from #3998 to get this testing.

Closes : #4211

Consult Iris pull request check list

[pp-mo]

Surprisingly, this caused no obvious problems.
So now I'm now going to crowbar-replace the old with the new styles, and see what happens ...

[pp-mo]

22 failures is really not that bad.
I think, at that rate, that I will simply adjust the reference files.
-- whereas I was considering a "clever" testing strategy more akin to the 'tolerant comparison' techniques used above.

From discussion with @bjlittle I don't think we need that -- simpler to just change the references.
Likewise, decided to remove the additional colons in scalar-coord and attribute outputs, as again it doesn't need to be that close to the existing results.
So let's have colons in the section headings, but not elsewhere.

[pp-mo]

I don't understand the latest docstest fail.
When I run the doctests locally, it insists that the result of printing a cube.metadata._asdict() shows as an OrderedDict.
And I can confirm that it looks like that, e.g.

>>> print(Cube([0]).metadata._asdict())
OrderedDict([('standard_name', None), ('long_name', None), ('var_name', None), ('units', Unit('unknown')), ('attributes', {}), ('cell_methods', ())])
>>> 

For me it's doing that in my test env too, and with older versions of Iris.

But in the CI testing it prints out as an ordinary dict.
And that is in spite of my running the tests via nox : nox -s doctest-3.7
-- which surely ought to take care of any environment differences ??
Possibly it is a lockfiles thing ?

UPDATE:
It is a Python 37/38 difference :

$ conda activate py37
$ python
Python 3.7.10 | packaged by conda-forge | (default, Feb 19 2021, 16:07:37) 
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from collections import namedtupl 
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ImportError: cannot import name 'namedtupl' from 'collections' (/tmp/persistent/miniconda3/envs/py37/lib/python3.7/collections/__init__.py)
>>> from collections import namedtuple
>>> nt = namedtuple('nt', ['a','b'])
>>> print(nt(a=1, b=2)._asdict())
OrderedDict([('a', 1), ('b', 2)])
>>> 
$ 
$ conda activate py38
$ python
Python 3.8.10 | packaged by conda-forge | (default, May 11 2021, 07:01:05) 
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from collections import namedtuple
>>> nt = namedtuple('nt', ['a','b'])
>>> print(nt(a=1, b=2)._asdict())
{'a': 1, 'b': 2}
>>> 
$ 

So, I needed to run nox -s doctest-3.8 and not 3.7.
Which is what cirrus does.

[pp-mo]

Yay, done at last.
@bjlittle very grateful for your eyes on this ..

[trexfeathers]

Followon from #3998

Does this supercede #3998?

[pp-mo]

Does this supercede #3998?

Absolutely, yes.
( Sorry, sure I posted this earlier but it vanished? perhaps failed to hit the button )

[pp-mo]

Thanks for looking @bjlittle 👀
Just now spotted conflicts + fixing that ...

[pp-mo]

Hit that spurious, intermittent failure of the animation test again.
Now fixed.

Unfortunately I re-spun it and so have lost the reference.
But there is a problem lurking there somewhere...
I think it is nothing to do with this PR, but also it did not go away with rebasing, so I guess we'll see it again sometime (on master)

[pp-mo]

Merged in the re-ordering from #4251 that ensures attributes come last, after cell-methods.
Fixed resulting test changes.
All passing now.
@bjlittle ready when you are !

[bjlittle]

@pp-mo Great stuff 👍

Thanks for diligently wading through this unpleasant mire of Cube.summary refactorisation, the result is a vast improvement IMHO.

Just a couple of minor comments to service, then we can bank it 🍻