Fix 76

[eteq]

This fixes #76 by adjusting the from_name mocking to turn on or off depending on whether a test is remote-data or not. I can confirm via the method of unplugging my ethernet cord that after this the remote_data tests fail if the internet is off.

[cdeil]

The mock is being called from docs/conf.py leading to this error:
https://travis-ci.org/astropy/astroplan/jobs/78631829#L649

See #96

[cdeil]

There's also this fail!?
https://travis-ci.org/astropy/astroplan/jobs/78631828#L402

[eteq]

@cdeil - that second failure is a slightly hard-to-parse indicator that a warning isn't being raised that should be. Investigating now.

As for the second, thanks for the tip there - I didn't realize the docs used the mock... Will fix.

[cdeil]

As I said in #96 ... I'm not actually sure any more the Sphinx build needs the mock ...

[eteq]

Ok, will check that and just remove if that's the case.

[eteq]

Turns out this is fixing a problem that I created in #90: it turns out that the way we're doing conftest.py, I was turning off a lot of the built-in astropy behavior. The latest commits turn it back on, and also fix a problem that had snuck in because this stuff was off.

So it's possible this will fail somewhere in some apparently unrelated place because it has substantial testing side-effects

But I think this now fixes #96 simply by not doing the mocking (it appears to not be necessary right now - I built the docs without warnings without an internet connection)

[eteq]

FYI, @jberlanga, as an aside: this is why I think you were seeing weird changes in the beahvior of summer_triangle.rst in #87 : between when you made it and when you rebased, the testing infrastructure (uintentionally) changed. So that's why I'm now skipping a lot more lines then you did there.

[cdeil]

@eteq – I find this quite complicated to understand.

Could you please add a few sentences explaining the logic of what is happening in conftest.py, e.g. in a docstring at the top of the file?

Is it as simple as this?

• Tests that use from_name should not be marked with @remote_data
• If tests are run with --remote-data, the normal from_name is called which accesses the internet.
• If tests are run without --remote-data, the mock from_name is called which which does not access the internet.

[eteq]

Alright, the tests passed one commit back, and the last commit adds an explanation as @cdeil requested, so this should be good to go.

To answer your sepcific quetsion here, @cdeil, the idea after this change is that something with from_name should have two tests: if remote_data is marked, then the test will run actually accessing the internet, but if it is not, then the mock will be used. So you'd write a remote_data test specifically to test the internet-based functionality (i.e. actually looking up some location that's not in the mocks), while the rest would be for when you just want from_name as a convenient way to create a FixedTarget or the like (especially for the docs we'll want this once the doc tests are up and running smoothly).

[cdeil]

@eteq – Hmm.

In your setup you say that the test should be duplicated (once with, once without the @remote_data decorator), but the test that's of interest here astroplan/target.py::astroplan.target.FixedTarget.from_name, currently isn't duplicated.

I'll try to come up with a solution that doesn't require duplicating the test in #103.
If that doesn't work, I can duplicate the test there, so that the mock actually is tested.