pmxbot IRC Log Viewer

[02:31:39] <glyph> dstufft: I still haven't figured out an even halfway clean monkey patch ;-)

[07:53:37] <marius__> how do I install https://pypi.python.org/pypi/progressbar/2.3-dev

[07:54:02] <marius__> Could not find a version that satisfies the requirement progressbar==2.3-dev (from versions: 2.1, 2.2)

[07:54:14] <marius__> pip install progressbar==2.3-dev

[07:54:27] <marius__> what is the correct cmd to install?

[09:46:03] <agronholm> marius__: I don't see any downloads for that project

[09:46:49] <agronholm> marius__: pip install https://python-progressbar.googlecode.com/files/progressbar-2.3-dev.tar.gz

[09:47:21] <agronholm> but do you really want to install that? it is dated aug 2010

[09:47:29] <agronholm> and the released 2.3 is dated may 2011

[10:15:39] <arch_> Hey. Is python3.4 supposed to include pip?

[10:16:52] <arch_> because after installing python3.4, there's no pip

[10:17:38] <jezdez> arch_: nope, see https://docs.python.org/3/library/ensurepip.html

[10:18:43] <arch_> running python3.4 -m ensurepip complains that /usr/bin/python3.4: No module named ensurepip

[10:26:16] <jezdez> hm, not sure why that is

[10:26:25] <jezdez> are you on debian or something?

[10:26:41] <jezdez> I recall dstufft saying something about ensurepip being different there

[10:28:19] <jezdez> see https://bugs.launchpad.net/ubuntu/+source/python3.4/+bug/1290847

[10:28:24] <jezdez> in case you use ubuntu

[10:41:51] <arch_> hmm. Ok, seems to work.

[11:12:20] <arch_> jezdez: do the official docs explain the builtin pip and venv usage somewhere? Seems there's a passing reference in the installation and packaging tutorial but nothing specific. Not sure if there should be

[11:30:09] <nanonyme> Uargh. Is Debian/Ubuntu intending to break virtualenv or am I misinterpreting that Launchpad bug?

[11:46:00] <mitsuhiko> nanonyme: nothing new

[11:46:14] <mitsuhiko> there is a reason why i recommend people against using system libraries

[11:52:59] <nanonyme> Hrm

[11:56:42] <nanonyme> mitsuhiko, I really think we should have an easily creatable relocatable runnable package format

[11:56:59] <mitsuhiko> you mean like … eggs?

[11:57:07] <nanonyme> Can they be?

[11:57:12] <mitsuhiko> they could

[11:57:18] <mitsuhiko> that was exactly what they were

[11:57:18] <nanonyme> Oh, eggs

[11:57:23] <nanonyme> No, I don't actually want those

[11:57:32] <nanonyme> I was rather thinking of the executable zips

[11:57:37] <mitsuhiko> like eggs?

[11:58:06] <nanonyme> Is it the same thing? Ie an arbitrary zip file with __main__

[11:58:20] <nanonyme> Erm, __main__.py even

[11:59:46] <nanonyme> I've never taken a look at what eggs have eaten to be honest

[12:00:14] <nanonyme> mitsuhiko, did they use to be able to bundle their deps?

[12:00:42] <mitsuhiko> not within the egg but they could just load other eggs

[12:01:06] <nanonyme> Yeah, not what I'm thinking of then

[12:01:31] <nanonyme> What I'm thinking of is you'd self-host all your deps inside the zip file, then give that to someone and they only have to have system Python to run it

[12:02:08] <mitsuhiko> well. people were against eggs

[12:02:12] <mitsuhiko> maybe the time is right now

[12:02:16] <mitsuhiko> but i doubt there will be support

[12:02:26] <nanonyme> I thought it was because they were full of black magic and undocumented bits

[12:02:36] <nanonyme> Or that's the impresion I got from what people were talking

[12:03:36] <mitsuhiko> that's the same with pip, setuptools and wheel

[12:03:45] <mitsuhiko> everything of eggs besides the format was kept anyways

[12:04:03] <nanonyme> Weeeell, I actually have been told the opposite of wheel

[12:04:17] <nanonyme> Ie that it was well documented while it was designed

[12:05:11] <nanonyme> Of course all this could have been objective :)

[12:06:37] <nanonyme> Also I think the task of wheel and what I was thinking of are distinct anyway. If egg was trying to do both, it already had unnecessarily complexity. Just do one thing and do it well

[12:07:02] <mitsuhiko> what was the unnecessary complexity?

[12:07:19] <mitsuhiko> all the complexity in egg is and was necessary to achieve the task of eggs

[12:07:25] <mitsuhiko> and to be honest, there is very little complexity in eggs

[12:07:48] <nanonyme> Dependency-based package installation is one thing, executable self-hosted package is another

[12:07:49] <mitsuhiko> eggs had two problems: unstable python bytecode, so they were version dependent and lack of stable python abi on linux

[12:08:04] <mitsuhiko> nanonyme: eggs had nothing to do with dependency based package installation

[12:08:13] <mitsuhiko> that was setuptools

[12:08:14] <mitsuhiko> and still is

[12:08:33] <mitsuhiko> nanonyme: the vast majority of criticism against eggs was from people that did not even try to use it

[12:08:50] <mitsuhiko> and just complained that people dared to have binary distributions

[12:09:20] <nanonyme> Well, *I* do think binary distributions are an unavoidable thing on certain platforms

[12:09:56] <nanonyme> Basically the *only* thing I did not like about eggs was that they did not work with pip

[12:10:22] <nanonyme> When I started complaining loudly, I was told some superior format called wheel was about to replace them anyway so I stopped caring

[12:11:26] <mitsuhiko> nanonyme: i wonder why they did not work with pi

[12:11:33] <mitsuhiko> oh well, because pip intentionally broke support

[12:11:40] <mitsuhiko> and then when it took off, it was reinvented as wheels

[12:11:43] <mitsuhiko> history repeating itself

[12:12:01] <mitsuhiko> (yes i am still bitter about eggs being killed off and recreated with nearly exactly the same design but more restrictions)

[12:12:18] <mitsuhiko> we could have had a honest discussion about eggs 4 years ago and be further today

[12:22:19] <nanonyme> That sounds not much before I moved from Perl to Python

[12:25:21] <nanonyme> Erm, rather roughly at same time

[12:25:22] <nanonyme> Anyway, I guess my point is that I personally haven't probably been competent for a real packaging conversation until about two years

[12:40:52] <nanonyme> mitsuhiko, the only thing I'm bitter about is that pip has been unusable for a lot of Windows users until very recently (1.5.x) so when I've tried to advocate it instead of easy_install, I've repeatedly had to go with "oh, it doesn't work. oh, I guess just use easy_install like before then)

[12:41:42] <mitsuhiko> nanonyme: i am very well aware of that, and that is been mostly the removal of eggs

[12:41:48] <nanonyme> And the fact that easy_install is bundled with Windows Python2 and pip is not hasn't helped much

[12:41:49] <mitsuhiko> eggs worked really well on windows

[12:42:03] <mitsuhiko> pip intentionally broke it "because we do not need them"

[12:42:41] <nanonyme> Might as well done it properly and killed all support for C extension modules on all platforms

[12:43:02] <nanonyme> With *that* pip with source packages only would have worked fine

[12:43:22] <jezdez> arch_: tbh, I don't know

[12:43:35] <jezdez> dstufft may know more as he was actively involved in that discussion

[12:43:42] <nanonyme> (nothing to compile, everything Just Works)

[12:43:53] <nanonyme> (of course, that would also have killed pip :p)

[12:45:01] <nanonyme> mitsuhiko, btw, came into other thoughts during this train trip. Maybe we *should* have platform-specific hooks so OS could recommend installing some distro package instead in interactive mode?

[12:45:27] <nanonyme> And have it done in a standard way

[12:46:38] <nanonyme> Ie so that if this integration is missing, pip would work normally. It would possibly prevent Linux distros from mutilating setuptools/pip/virtualenv from arbitrary locations in the future

[12:48:05] <nanonyme> (and packages would still be installable even with the integration, it would just notify end-user

[12:48:06] <nanonyme> )

[12:49:19] <nanonyme> Okay, fine, this isn't #pypa-dev

[13:43:39] <nanonyme> Urf, I can't do pip install --user --upgrade virtualenv if machine has OS-package virtualenv installed, right?

[13:47:35] <apollo13> most likely not

[14:58:32] <mitsuhiko> nanonyme: that hook thing was discussed before :)

[15:19:59] <dstufft> mitsuhiko: eggs, like much of what came out of setuptools, are reasonable ideas and terrible execution

[15:22:45] <dstufft> (The restrictions that Wheel placed that Eggs didn't have were important restrictions)

[15:33:33] <Alex_Gaynor> mitsuhiko: in what ways do wheels not work on windows? We've shipped 10s of thousands of 'no compiler needed' installs with them

[15:34:05] <mitsuhiko> Alex_Gaynor: where did i say that wheels do not work on windows?

[15:34:26] <Alex_Gaynor> Perhaps I misread, sounded like you were saying eggs worked better than wheels on windows

[15:34:48] <mitsuhiko> easy_install + eggs worked better than pip without wheels

[15:34:53] <mitsuhiko> wheels are very, very recent

[15:34:58] <Alex_Gaynor> Ah, yes, sure.

[16:14:16] <nanonyme> Alex_Gaynor, *pip* did not basically work on Windows before wheels for packages with C extensions. easy_install (with eggs) did

[16:15:17] <nanonyme> Ie primarily: good that the problem got solved now with pip 1.5.x, bad that it took this long to be solved

[16:34:31] <DanielHolth> wheel 2012-?

[16:51:42] <DanielHolth> weekend project idea: figure out how to publish a version of tinycc as a wheel and compile Python extensions with it

[16:52:00] <nanonyme> mitsuhiko, what was the resolution?

[16:52:29] <nanonyme> DanielHolth, what 2012-?

[16:53:03] <DanielHolth> that's when the project started

[16:53:14] <DanielHolth> (wheel)

[16:53:44] <DanielHolth> 🎡

[16:53:55] <nanonyme> Yeah but without pip 1.5.x they're not used by default == not proper solution on Windows

[16:54:00] <DanielHolth> sweet

[16:54:49] <nanonyme> I can't find release notes but that means pretty new

[16:55:03] <DanielHolth> yes

[16:57:02] <nanonyme> IOW it's not that uncommon that you get eggs installed with easy_install into virtualenvs on Windows still. Sometimes it works, sometimes it doesn't

[16:57:24] <DanielHolth> meh

[17:02:55] <DanielHolth> dstufft how did the 'can't install extras of already-installed packages' bug get into pip?

[17:03:10] <DanielHolth> was it just a simple dist-info incompatibility?

[17:03:38] <dstufft> I dunno

[17:04:38] <nanonyme> dstufft, basically I'm just ranting about easy_install still and how things were not that long ago

[17:05:05] <dstufft> it wasn't that long ago that the only way to install stuff from PyPI was to acess half the internet over HTTP

[17:05:14] <dstufft> we've come a long way

[17:08:31] <nanonyme> Well, yes

[17:09:11] <nanonyme> Will just have to remember to tell colleagues at work to update their pip and then start using that only when I get back to from Summer holiday

Log file Viewer

Help | Karma | Search:

#pypa logs for Thursday the 10th of July, 2014