Make tests optional that depend on pydot and pandas (optional installs).

[pp-mo]

Failed on these in my amazon instance VM.

We could probably do with an "official" list of what installs developers should use?
See also #478 -- but possibly this should have gone on discussion groups instead??

[pp-mo]

Re: recent change (pp-mo@d9451c3)
Generalised as now want to make other tests 'optional' too (specifically for #453)

[rhattersley]

The generalised version is certainly clever, but I'm not sure it's worth it since there's still quite a lot of boilerplate left in the client module.

How about if the client code (e.g. iris.tests.test_pandas) just did:

# Since pandas is an optional dependency, we disable all these tests
# if it's not installed.
try:
    import pandas
    import iris.pandas
    skip_pandas = lambda fn: fn
except ImportError:
    skip = unittest.skip('Test(s) require pandas, which is not installed.')
    skip_pandas = lambda fn: skip(fn)

I'd have more time for the generalised version if it could be used more like:

pandas, skip_pandas = tests.skip_module('pandas')
if pandas is not None:
    import iris.pandas

[pp-mo]

Thanks @rhattersley . But you're reading my mind now... That's just the thought process I've been going through since pushing this (which I now wish I hadn't).
There may even still be an idiom here that's as good as a definition. I'm looking into it..

[pp-mo]

The generalised version is certainly clever, but I'm not sure it's worth it since there's still quite a lot of boilerplate left in the client module.

I'm tending to agree, and I think it's more obvious to leave the workings visible in the test file.
Changes here now done that way : pp-mo@88919fd

[rhattersley]

Could this be simplified to just unittest.skipIf(...) with a condition clause based on dot.DOT_AVAILABLE?

[rhattersley]

Or perhaps ditch DOT_AVAILABLE altogether since it's only used here and just test the executable path directly.

[pp-mo]

That's a nasty catch in the new licence header checks
(cf. https://travis-ci.org/SciTools/iris/builds/7167947)
It fails if you don't have spaces around the dash in "YY1Y - YY2Y", and doesn't tell you which file is to blame..

[pp-mo]

Thanks @rhattersley, @esc24 .
I'm hoping this latest addresses all comments so far.
Please re-review.

[rhattersley]

Please re-review.

👍 Looks good enough for me (at cfdd8483dc559e969ddd7fa26795dc5908deb9a6). Please squash and I'll merge.

[pp-mo]

@rhattersley Please squash and I'll merge.

Thanks, done that. All re-tested, should be ok this time.

[rhattersley]

Thanks @pp-mo. 😄

[esc24]

Great stuff @pp-mo. These kind of contributions, although not the most exciting features to work on, are really important.