These are chat archives for astropy/astropy

12th
Jul 2017
Christoph Deil
@cdeil
Jul 12 2017 11:27
We have a .. plot:: plot.py in side a mydocs.rst file, in a project using the Astropy affiliated package setup. Does anyone know how to skip that plot.py file from being executed as a doctest? (it is and we do want it executed as part of the docs build).
matrixbot
@matrixbot
Jul 12 2017 11:28
Cadair wont it always be executed as part of the doc build
Cadair on collection because it's a sphinx directive
Christoph Deil
@cdeil
Jul 12 2017 11:32
It is and we want it executed as part of the Sphinx build.
The problem is that it is executed as part of the pytest doctests, which we don't want, and that is failing: https://gist.github.com/cdeil/e17584e17ef4e0b93723b78ea0e1f464
If someone has time to have a quick look and help out, please look here: https://github.com/hipspy/hips/pull/68#pullrequestreview-49458410
matrixbot
@matrixbot
Jul 12 2017 11:37
Cadair ah ok
Cadair I have no idea about the doctests :)
Cadair anyone going to euroscipy this year?
Christoph Deil
@cdeil
Jul 12 2017 11:45
I think I found a way: add doctest_norecursedirs = docs/plot.py in setup.cfg.
If there's a better way and someone sees this later, please comment on that PR I linked to.
For me it's probably no for Euroscipy this year.
matrixbot
@matrixbot
Jul 12 2017 14:10
Cadair bsipocz: astrofrog and I were talking about writing a little web app to montior astropy dev builds of affiliated packages
Cadair my idea was to have something that explicitly reports from travis (just for the dev build) to a web app
Cadair which then displays the status
bsipocz sounds great
Cadair so it would know what astropy commit was run against and when the run was
bsipocz would it shop around in the affiliates' own travis builds, or have one of its own?
Cadair shop around
bsipocz although I still think it's not really our job/task, but the maintainers!
Cadair well it puts no requirement on you to do anything with the info
Cadair it's just a status page of affilated packages vs astropy masterr
Cadair so you can see if something has gone horribly wrong
Cadair lol
bsipocz (but that's just the usual rant of mine as I got rather upset when I saw cron builds failing for weeks, but didn't get reported back upstream, yet the users will probably blaim it on us astropy to break that package..)
matrixbot
@matrixbot
Jul 12 2017 14:30
bsipocz @cdeil - have you tried to add .. doctest-skip:: ?
bsipocz it may not work when you have a separate plot.py, but in astropy/docs/nddata/utils.rst it works whe n yhe plot directive has the code in the main rst file
bsipocz an alternative solution could be to set up the sphinx-gallery examples, we have some of those skipped, too. But probably that's a big overkill
matrixbot
@matrixbot
Jul 12 2017 14:45
Cadair bsipocz: sure, my main issue with that frameattribute thing was lack of time to react
Cadair and you can slow it all the hell down by doing RC after RC or you can have a look at a dashboard thing to see if all seems ok
bsipocz I don't mean you. That was indeed a massive mistake mostly from my side. I should have checked the changelog before merging, assumed it's a bugfix
Cadair sure, I think it was supposed to be a bug fix
bsipocz yes, but the issue is that getting in more bug and docs fixes is still very tempting, even after the last RC
Cadair I mean the only way to do it is to have "release 24-48 hours after last merge"
Cadair but I would not like doing that for sunpy
matrixbot
@matrixbot
Jul 12 2017 14:53
bsipocz yeap. just checked, the sdist was made ~20 hours after merging the frameattribute PR, and ~3 hours after merging the last meaningless fixup PR (but those very last ones were really just to cleanup the release, xfail some more tests, etc)
Cadair my comment on the PR was the git tag
Cadair I mean I have done much worse than 3 hours with sunpy
bsipocz yeah, git tag happeded about the same time as the last backport as we needed that to build the wheels, but those failed at the first attempt and needed some more hacking, etc.
matrixbot
@matrixbot
Jul 12 2017 23:01
equant on Freenode When I use WCS(hdu.header) on ALMA data, I get 4 axes because it's reading in STOKES and FREQ data. Is there a way to have that information ignored so that it's 2-axis just like the Horse Head example in the docs?
equant on Freenode Nevermind... I think this is what I want.... wcs.dropaxis(3).dropaxis(2)