These are chat archives for astropy/astropy

21st
May 2014
Thomas Robitaille
@astrofrog
May 21 2014 10:21
I think someone should write a song, 'Waiting for Travis'
Francesco Biscani
@bluescarni
May 21 2014 10:42
maybe they should rename Travis -> Godot :)
Brigitta Sipocz
@bsipocz
May 21 2014 10:44
@astrofrog - I'm thinking about getting the "hide prompt and output" js feature. E.g. numpy uses in its docs for the code blocks and I found it very useful from time to time.
My questions is whether I should put it into astropy/sphinx/themes or into astropy_helpers?
Thomas Robitaille
@astrofrog
May 21 2014 11:40
@bluescarni - haha, true :)
@bsipocz - I think it would need to go into the astropy_helpers/sphinx because astropy/sphinx will be deprecated in 1.0 (since it's in astropy_helpers)
Brigitta Sipocz
@bsipocz
May 21 2014 11:41
OK
Thomas Robitaille
@astrofrog
May 21 2014 11:43
@bsipocz - you should in principle be able to install astropy_helpers like a normal package to try it out, but you will likely run into this bug: astropy/astropy#2521
I haven't had time to look into it yet
Brigitta Sipocz
@bsipocz
May 21 2014 11:46
OK. Well I made it work in astropy, so I guess it should be straightforward to do it in the helpers, too.
Thomas Robitaille
@astrofrog
May 21 2014 11:46
Ok, great :)
Brigitta Sipocz
@bsipocz
May 21 2014 12:26
This message was deleted
Ghost
@ghost~53638049048862e761fa0380
May 21 2014 17:20

More astropy_helpers fun.. I just checked out astropy master, made a few changes, and then did git diff. At the end there is something related to the astropy helpers subrepo:

diff --git a/astropy_helpers b/astropy_helpers
index 5f3659b..ecb2905 160000
--- a/astropy_helpers
+++ b/astropy_helpers
@@ -1 +1 @@
-Subproject commit 5f3659b2395ff799b27972c54545d90b9871e163
+Subproject commit ecb29056e257bbb0984d71e16d00abc697e581e2

Should I be worried?

@astrofrog ??
Thomas Robitaille
@astrofrog
May 21 2014 18:09
@taldcroft - huh, weird
did you figure it out?
for some reason it looks like it thinks you want to use the older version
ecb29056e257bbb0984d71e16d00abc697e581e2 was the commit before today
@taldcroft - let me know if you are still seeing this issue
Thomas Robitaille
@astrofrog
May 21 2014 18:17
(I'm afk for 30 min, but just leave a reply and I'll check when I get back)
Erik Tollerud
@eteq
May 21 2014 18:57
@taldcroft @astrofrog - I think I understand what @taldcroft was seeing: if you do git pull and there’s an update to astropy_helpers (or just checkout an older commit), and then immediately do git diff, you’ll see a message just like what @taldcroft saw. This is because you have to do git submodule update to update the submodule (or, for us python setup.py build)
Also, @taldcroft, FYI, your note above about python setup.py build_sphinx always re-building was a bug - it’s fixed in #2351 / #2535 . So your safest bet is to do it that way (as that should also keep astropy_helpers in sync)
Erik Tollerud
@eteq
May 21 2014 19:12
@astrofrog or @embray , are you around?
Thomas Robitaille
@astrofrog
May 21 2014 19:23
I'm around
ok, I see what you mean
@eteq - I'm around
Erik Tollerud
@eteq
May 21 2014 19:25
Ok, cool: can you clarify for me what exactly is supposed to end up in astropy_helpers vs. the core package?
I see all the sphinx stuff is in the helpers, but also in the core
(note also #2545 where I ask that this go in the actual docs, but I would also just like to know)
Thomas Robitaille
@astrofrog
May 21 2014 19:26
actually everything in the helpers is in the core right now for backward-compatibility
Erik Tollerud
@eteq
May 21 2014 19:27
But it’s all going through helpers if they’re present, is that the idea?
Thomas Robitaille
@astrofrog
May 21 2014 19:27
so anything required for building astropy or affiliated packages is in astropy-helpers
yes
and in the long term we'll remove them from the core
Erik Tollerud
@eteq
May 21 2014 19:27
And the plan is to keep it in for 0.4, or take it out just before release?
ah, ok
Thomas Robitaille
@astrofrog
May 21 2014 19:27
I think we are going to probably leave it until 1.0?
so anything required for building
strictly, sphinx isn't
Erik Tollerud
@eteq
May 21 2014 19:28
ah right, that’s true
Thomas Robitaille
@astrofrog
May 21 2014 19:28
but we had a discussion about this and it's silly to have to install astropy as a whole to build docs
so by moving sphinx to the helpers we make it then possible to have RTD not need to install astropy by default
Erik Tollerud
@eteq
May 21 2014 19:28
yeah, and it also has the advantage of allowing us to finally completely stop 2to3ing once we take it out of astropy (I’m skeptical that it’ll ever be un-2to3ed)
Thomas Robitaille
@astrofrog
May 21 2014 19:29
indeed
Erik Tollerud
@eteq
May 21 2014 19:29
oh, but RTD will still need astropy to be able to auto-generate the docs...
Thomas Robitaille
@astrofrog
May 21 2014 19:29
well actually I have a PR for the numpydoc stuff that makes it 2/3 compatible
@eteq - for astropy of course, but not affiliated packages
Erik Tollerud
@eteq
May 21 2014 19:29
ahhh, gotcha
My main concern on the 2to3ing is the stuff that subclasses sphinx components: that has to wait for sphinx to update, which is not guaranteed to ever happen
Thomas Robitaille
@astrofrog
May 21 2014 19:30
yep
well eventually it probably will
Erik Tollerud
@eteq
May 21 2014 19:30
well “in the future” ;)
Thomas Robitaille
@astrofrog
May 21 2014 19:30
but yeah without the sphinx stuff in core, we can then go 2to3 free
hehe yes
sorry for the hiccups with astropy-helpers in the last day or so
hopefully most issues are squashed
Erik Tollerud
@eteq
May 21 2014 19:31
oh, that’s fine - totally understandable with a rollout like that - I’m surprised how smooth it was, actually
My confusion/concern now is that having it in both the core and in astropy_helpers may lead to confusion and trouble among 0.4 users
Thomas Robitaille
@astrofrog
May 21 2014 19:32
I guess the question is whether users ever directly use these
Erik Tollerud
@eteq
May 21 2014 19:32
It depends on what “users” mean - probably not the average novice, but the advanced user who wants to try their hand at contributing, or anyone who encounters problems will indeed need that
Thomas Robitaille
@astrofrog
May 21 2014 19:33
right I agre
Erik Tollerud
@eteq
May 21 2014 19:33
I guess I’m not clear what what the down side is to removing it right away in 0.4 ?
Thomas Robitaille
@astrofrog
May 21 2014 19:33
hmm
well if affiliated packages rely on it
and someone has astropy 0.3
and updates astropy
hmm
Erik Tollerud
@eteq
May 21 2014 19:33
ahh, gotcha
Thomas Robitaille
@astrofrog
May 21 2014 19:33
but then it's only used for setup.py
Erik Tollerud
@eteq
May 21 2014 19:33
hmm indeed
Thomas Robitaille
@astrofrog
May 21 2014 19:33
so once something is installed it doens't matte
ah but basically the thing is that then all affilated packages have to have a release that is compatible with the new framework
so say astroquery 0.1 will not install with astropy 0.4
Erik Tollerud
@eteq
May 21 2014 19:34
hmm… yeah, that’s tricky
can we replace the relevant packages in astropy with shell modules that just import from helpers? Then anyone who actually looks at the code will know that’s the wrong place...
Thomas Robitaille
@astrofrog
May 21 2014 19:35
no because astroquery 0.1 doesn't know how to install the helpers
so the helpers won't be present during the install
but we could put big warnings at the top of the deprecated files
Erik Tollerud
@eteq
May 21 2014 19:36
Ah, right, because they aren’t actually installed in the system at all
Thomas Robitaille
@astrofrog
May 21 2014 19:36
right
Erik Tollerud
@eteq
May 21 2014 19:36
yeah, I think big, prominent warnings at the top of all relevant source code would help. I’m not sure it needs actual code deprecation, but just a note warning anyone who looks at the source would be fine
Thomas Robitaille
@astrofrog
May 21 2014 19:36
ok cool
Erik Tollerud
@eteq
May 21 2014 19:37
And #2545 will help, too
Thomas Robitaille
@astrofrog
May 21 2014 19:37
Yep
I also started working on astropy-helpers/docs
no reason it can't have it's own RTD docs
when was the original scheduled RC
Erik Tollerud
@eteq
May 21 2014 19:40
https://github.com/astropy/astropy/wiki/Release-Calendar says May 27. Or you mean the origin milestone? I think it was the 19th?
Thomas Robitaille
@astrofrog
May 21 2014 19:40
Not sure which I meant, but I guess what the current plan is for the RC
I guess we can aim for May 27 for the RC, and see how it goes
Erik Tollerud
@eteq
May 21 2014 19:40
MAy 27
yeah, and if we have to delay the release a week or two due to helper-related issues, that’s fine: that’s what RCs are for, after all!
Thomas Robitaille
@astrofrog
May 21 2014 19:43
hopefully it should be ok
I'm going to be at a conference for the next two weeks (two conferences)
Erik Tollerud
@eteq
May 21 2014 19:44
Hmm.. noticed something weird that may or may not be #2531 / #2535 related. If I build astropy once, and then do it again, the second time there’s still some compiling. But subsequent builds don’t require any re-building. Are you seeing the same @astrofrog ?
Thomas Robitaille
@astrofrog
May 21 2014 19:44
but I'll do my best to keep up
I'll check
Erik Tollerud
@eteq
May 21 2014 19:44
hmm, I’m at a workshop the next two weeks too. Should be more in contact than at a full conference, but still… Anyway, we’ll just do what we can
Thomas Robitaille
@astrofrog
May 21 2014 19:45
yep
Erik Tollerud
@eteq
May 21 2014 19:45
I guess my subtle planning of “June is conference season, so we want to release at the start of June” sort of backfired :wink:
Thomas Robitaille
@astrofrog
May 21 2014 19:45
let me check about the compiling thing
@eteq - is there as much compiling the second time?
if you do it three times does it still build?
Erik Tollerud
@eteq
May 21 2014 19:48
the second time is definitely less, and the third time no compiling happens, but it does show messages about copying over wcslib headers and skipping of Cython stuff
Thomas Robitaille
@astrofrog
May 21 2014 19:48
yep, I get the same
I don't quite understand why it still does a little the second time
Erik Tollerud
@eteq
May 21 2014 19:50
it looks like it’s only the convolution stuff
Thomas Robitaille
@astrofrog
May 21 2014 19:50
what about erfa_time
is it all the cython stuff?
we should check if it did that before though
can you check if pre-astropy-helpers it did the same?
Erik Tollerud
@eteq
May 21 2014 19:51
I’m 98% sure it finished the first time pre-helpers
but I’ll check
nope, just one build was sufficient pre-helpers
Thomas Robitaille
@astrofrog
May 21 2014 19:54
ok
can you open an issue on astropy-helpers?
Erik Tollerud
@eteq
May 21 2014 19:55
will do
Thomas Robitaille
@astrofrog
May 21 2014 19:55
will try and look later tonight
Erik Tollerud
@eteq
May 21 2014 20:03
@astrofrog : astropy/astropy-helpers#19
Thomas Robitaille
@astrofrog
May 21 2014 20:07
thanks!
Erik Tollerud
@eteq
May 21 2014 20:09
Another thing I noticed: I’m getting a ConfigurationDefaultMissingError when I try to do sphinx builds (and I think other times), post-helpers. I’m not sure if this is a problem with the config stuff vs. helpers, though. Where should I report it?
Thomas Robitaille
@astrofrog
May 21 2014 20:12
I'd put it in astropy core if you aren't sure
Erik Tollerud
@eteq
May 21 2014 20:13
k
sorry to lean on you like this - I’d normally just try to figure some of this out myself, but I haven’t really been able to wrap my head around the helpers too much yet
Thomas Robitaille
@astrofrog
May 21 2014 20:14
it's ok :)
actually it was mostly @embray until yesterday - I just learned fast ;)
Erik Tollerud
@eteq
May 21 2014 20:14
heh, well, that’s a promising sign