pmxbot IRC Log Viewer

[11:11:17] <ionelmc> how can i avoid this sort of warning: /usr/lib/python2.7/pkgutil.py:186: ImportWarning: Not importing directory '/home/ionel/foobar/.ve/lib/python2.7/site-packages/sphinxcontrib': missing __init__.py ?

[11:11:28] <ionelmc> seems it's a namespace package

[12:55:11] <savior> hi guys

[12:55:16] <savior> I have question about pip

[12:55:40] <savior> how can I make pip pick up requirements.txt from a tarball?

[12:55:53] <savior> so if I want to do this: pip install whatever.tar.gz

[12:56:11] <savior> setup.py is run, but apparently requirements.txt is not taken into account

[12:56:15] <savior> pip 1.4.1

[12:56:43] <mgedmin> savior, use install_requires from setup.py instead of requirements.txt

[12:59:04] <david_cliqz> mgedmin: It's savior again

[12:59:14] <david_cliqz> just the connection was lost

[12:59:24] <david_cliqz> what I wanted to ask is

[12:59:27] <david_cliqz> mgedmin: thanks, but really my requirements.txt is just one line: --index-url XXX

[12:59:34] <david_cliqz> can I put that into install_requires as well?

[12:59:49] <apollo13> no

[13:00:22] <david_cliqz> so how can I make it take effect?

[13:00:32] <apollo13> you can't

[13:09:33] <david_cliqz> thanks

[15:08:26] <Bob_George33> Hey

[15:09:25] <Bob_George33> I've tried searching around and can't find anything that will help me

[15:10:21] <Bob_George33> I'm using setuptools to install a package and it makes the one file I need executable, non executable, despite the face that it is marked as an executable

[15:53:16] <malinoff> hi, #python seems to ignore my question, so i'm trying to ask here

[15:53:18] <malinoff> if i have a namespaced package X and a couple of subpackages X.A, X.B, X.C, can i still have a dedicated package X ?

[15:53:44] <malinoff> maybe with some clever tricks using importlib machinery?

[16:05:07] <pjdelport> malinoff: I'm not sure, but are you sure you want the excitement of maintaining clever tricks when it comes to packaging?

[16:06:33] <malinoff> pjdelport, well, i can live with just one package applying such tricks

[16:07:01] <malinoff> i just want to avoid creating mylib.core/mylib.common/whatever package

[16:07:19] <malinoff> simply because it is not *core*

[16:07:39] <malinoff> it can (and maybe should) be used as standalone library

[16:09:31] <malinoff> i should probably just use a submodule as namespace package, like mylib.ext

[16:09:48] <malinoff> and instead of having mylib.A, mylib.B i should have mylib.ext.A, mylib.ext.B etc

[16:21:10] <malinoff> oh, i can't do that

[17:23:48] <ionelmc> malinoff: you could implement some custom namespacing

[17:23:56] <ionelmc> then you can do whatever you want

[17:24:09] <ionelmc> take a look at how pylib/pytest do it

[17:24:24] <malinoff> ionelmc, you mean py. namespace?

[17:24:31] <ionelmc> malinoff: yeah

[17:24:47] <malinoff> ionelmc, yes, i know about it, same for zc, zope etc

[17:25:24] <ionelmc> tho I'd really hate to be the user reporting terrible bugs with your namespacing thing

[17:25:44] <ionelmc> cause I'd be a sad and angry user :-)

[17:25:46] <malinoff> the question is: when i have something. namespace, can i have a dedicated "something" package, so import something will import this package and import something.subpackage will import namespace package

[17:26:49] <malinoff> but i figured already this is not possible

[17:27:27] <ionelmc> malinoff: it is? https://pypi.python.org/pypi/py

[17:28:00] <ionelmc> i briefly looked once, didn't look simple at all

[17:28:24] <malinoff> oh

[17:28:25] <malinoff> right

[17:28:33] <malinoff> i completely forgot that py is such dedicated package

[17:28:35] <malinoff> thank you

[17:30:32] <malinoff> so there are "clever tricks" i was talking about :)

[17:44:50] <ionelmc> malinoff: and for the next trick: angry mob of users with pitchforks waiting in the bugtracker

[17:44:53] <ionelmc> hehe

[17:46:47] <malinoff> ionelmc, yeah, i'll keep that tricks in mind, but i definitely not gonna use it for now

[17:48:23] <malinoff> this reminds me these Armin's talks playing with sys._getframe and importlib machinery, allowing to import json files as modules

[18:33:37] <ronny> ionelmc, malinoff: pytest is not in the py namespace ever since it got split out

[18:34:44] <ronny> ionelmc, malinoff: bascially package namespaces are broken on py2, and afair a mess on py3

[18:35:02] <ionelmc> ronny: why does it have that weird _pytest package then?

[18:35:28] <ronny> ionelmc: python -m pytest broke strangely on some python versions unless it was a module

[18:35:52] <ionelmc> aaah ... 2.6 support

[18:35:54] <ionelmc> nasty

[18:36:36] <ionelmc> but didn't you drop support for 2.6?

[18:36:50] <mathieui> 2.6 is a PITA

[18:36:56] <ronny> no, only 2.5

[18:37:14] <ronny> 2.6 has another few years left to support i think

[18:37:31] <malinoff> yeah, 2.6 is pretty easy to support with additional dependencies if 2.7 is supported too

[18:38:31] <ionelmc> yeah, if your dependency is pain_tolerance

[18:38:32] <ionelmc> hehe

[18:38:57] <malinoff> nope, just importlib in my case :)

[19:31:40] <muchasgracias> hi, I have an outdated python program in my "pip list --outdated", what's the proper command to update it?

[19:34:06] <tos9> muchasgracias: pip install -U

[19:34:39] <muchasgracias> tos9: with the package name?

[19:34:44] <tos9> muchasgracias: yep

[19:35:13] <muchasgracias> tos9: is there a command to update all outdated package at once?

[19:35:31] <tos9> muchasgracias: pip list --outdated | parallel -X pip install -U

[19:35:32] <tos9> :)

[19:35:41] <muchasgracias> cool

[19:35:42] <muchasgracias> danke!

[19:40:13] <Wooble> (it's probably usually a better idea to do 1 at a time and run your test suite ;) )

[19:44:51] <pf_moore> tos9: won't that choke on the "(Current: X.Y Latest: V.W)" bit of the pip list output?

[19:45:04] <tos9> pf_moore: uh, yeah, I forgot that that was in there

[19:45:21] <tos9> so you need a sed in there as well then :)

[19:45:39] <pf_moore> np, I was just working on pip list and it occurred to me that the output is for human consumption, but not that scriptable

[19:45:47] <tos9> Yeah. Agreed.

[19:47:25] <tos9> Heh. pip list is pretty unhappy with its pipe closing.

[19:48:26] <tos9> oh ew

[19:48:32] <tos9> also location errors go to stdout

[19:50:27] <pf_moore> tos9: What do you mean by location errors?

[19:50:50] <tos9> pf_moore: Could not find a download satisfying ...

[19:50:55] <tos9> Some external sources were ...

[19:51:50] <pf_moore> Got a precise error message? I'm too lazy to look, and grepping the source is easy :-)

[19:51:58] <tos9> Sorry :)

[19:51:59] <tos9> yeah

[19:51:59] <tos9> sec

[19:52:10] <pf_moore> no problem

[19:52:24] <tos9> pf_moore: Could not find any downloads that satisfy the requirement settings

[19:52:41] <tos9> Some externally hosted files were ignored as access to them may be unreliable (use --allow-external settings to allow).

[19:54:56] <pf_moore> OK, pip just uses the stdlib logger. That one's logger.critical(...) and the normal output is logger.info(...).

[19:55:14] <pf_moore> Looks like we don't split the log levels by stream

[19:55:20] <pf_moore> Blech.

[19:58:16] <tos9> Yeah :/

[20:01:47] <pf_moore> Seems to me that using logging for actual user output (as opposed to, well, logging) is pretty icky.

[20:02:03] <pf_moore> But I can't see it changing soon, for hysterical raisins

[20:02:23] <tos9> Yeah. I think probably a new machine readable command might need to be the thing

[20:02:37] <tos9> unless pip wants to handle it entirely

[20:06:45] <pf_moore> Sigh. This feels similar to the old "pip doesn't have a Python API" issue - pip just isn't written to be used as a component.

[20:17:30] <tos9> well, pip doesn't have an API has an "answer" which is "the CLI is the API"

[20:17:56] <tos9> and then if anyone says "yeah but that stinks it's written in Python" you can say "hey mercurial does the same, and so does ..." :P

[20:18:34] <tos9> although I think every time that one comes up the conversation seems to steer towards "What does the api need to have and who wants to do the work for it"

[20:19:05] <dstufft> yes that's basically the question

[20:19:40] <dstufft> figuring out what an API would look like, and whether or not we're happy with the current APIs exposed to do particular things

[20:19:43] <tos9> :) and since the answer doesn't have "tos9 does" in it (at least not right now) I get to not complain about that one :P

[20:20:00] <dstufft> and whether the API should belong inside pip, or if packaging or some other lib is where it belongs

[20:20:10] <dstufft> (like version parsing sin't pip specific, so it doesn't belong there)

[20:20:16] <tos9> ah.

[21:28:34] <pf_moore> dstufft: I do think that the pip CLI should be a bit more scripting-friendly, if only in the sense of data on stdout, errors on stderr for things like pip list

[21:29:05] <pf_moore> But as I also don't have any plans to do anything about it (mainly because I can't figure out how) it's still just idle chatter...

[21:29:31] <dstufft> pf_moore: agreed

[21:46:46] <tdsmith> oh, i've kvetched about trying to pipe list outdated into upgrade before

[21:47:07] <tdsmith> got as far as observing how messages were logged and gave up :p

[21:55:04] <tos9> I mean. I can be horrible :) Here's pip's API for that: pip list --outdated | grep -v ' ' | parallel -X pip install -U

[21:55:13] <tos9> now don't have one word logging messages and we're all good and can go home

[21:56:15] <tdsmith> ha

[21:57:10] <tdsmith> package names aren't one word either though :

[21:57:11] <tdsmith> p

[21:57:46] <tdsmith> aren't printed as*, anyway

[22:22:09] <tos9> are there people who've uplaoded packages with spaces in them

[22:22:19] <tos9> is hurting them allowed

[22:24:41] <doismellburning> Registering Can you do the fandango? to https://pypi.python.org/pypi

[22:24:41] <doismellburning> Server response (400): Invalid package name. Names must contain only ASCII letters, digits, underscores, hyphens, and periods

[22:24:55] <doismellburning> (cannot speak for legacy packages)

[22:25:27] <tdsmith> i just meant that pip list --outdated output looks like "astroid (Current: 1.2.1 Latest: 1.3.4)"

[22:26:00] <tdsmith> anyway, point taken, it can be done but it isn't fun :p you could deploy a regex against that

[22:28:09] <doismellburning> otoh I now have https://pypi.python.org/pypi/0-._.-._.-._.-._.-._.-._.-0

[22:28:41] <tdsmith> ow

[22:30:09] <doismellburning> interestingly I just figured that probably conflicts with 0_._._._._._._._._._._._._0

[22:35:19] <tos9> tdsmith: oh yeah sorry I posted the dumb sed before which you'd still need. So hooray it's possible end of story we can go home :P

[23:40:09] <paulproteus> BTW, dstufft, please meet plathrop, a friend of mine who is excited about all the work you've been doing in packaging-land, and surely is excited about the work of others but doesn't know their names as well yet.

[23:40:35] <dstufft> hey :)

[23:40:53] <plathrop> dstufft: Your post about setup.py vs. requirements.pip is one I paste links to about once per week :-)

[23:41:51] <dstufft> heh, it ended up being a lot more useful than I originally anticipated

[23:42:21] <dstufft> I was frustrated about trying to repeatidly explain to people why reading a requirements.txt inside of a setup.py to populate install_requires was wrong

[23:43:32] <plathrop> I understand the drive - nobody likes maintaining two separate lists that feel like the same thing.

[23:43:55] <plathrop> I'm working on a post trying to explain my experiences with related problems. Specifically around repeatable deploys of python apps.

Log file Viewer

Help | Karma | Search:

#pypa logs for Wednesday the 4th of March, 2015