pmxbot IRC Log Viewer

[06:29:34] <kennethreitz> is there a particular reason pip is unable to find a package version i just released?

[06:29:48] <kennethreitz> when it is plainly listed on pypi

[06:30:33] <apollo13> kennethreitz: which package & version exactly?

[06:30:45] <kennethreitz> apollo13 em-keyboard==0.0.2

[06:31:14] <apollo13> works for me

[06:31:23] <apollo13> https://dpaste.de/f2qs/raw

[06:31:28] <kennethreitz> Could not find a version that satisfies the requirement em-keyboard==0.0.2 (from versions: 0.0.1)

[06:31:29] <apollo13> maybe you ran into the old cached version?

[06:31:29] <kennethreitz> over and over

[06:32:08] <kennethreitz> is pip caching simple responses now or something?

[06:32:23] <apollo13> I ment the fastly cache

[06:33:02] <apollo13> I don't think that pip caches search/simple pages, only the actual downloads

[06:33:08] <kennethreitz> this has not been a problem before

[06:33:18] <kennethreitz> went through now

[06:33:26] <kennethreitz> 7 minutes later

[06:33:28] <apollo13> did you upload to pypi or warehouse?

[06:33:34] <kennethreitz> pypi

[06:34:11] <apollo13> anyways, I can imagine that you request it way to fast after upload that fastly would cache the wrong result for a short time

[06:38:04] <kennethreitz> i'm trying to debug how to ship a non-python file with a single-file script via sdist

[06:38:20] <kennethreitz> got it in the bundle, now just need to find out how to reference it. i've done this a million times, but it's been a while

[06:39:19] <kennethreitz> package_data prob only works when using 'packages'

[06:42:07] <kennethreitz> can't do 0.0.3 now, of course

[06:42:34] <kennethreitz> the whole point of using fastly is that you can invalidate caches easily lol

[06:43:38] <apollo13> well, in general it works, worst case you can issue the -XPURGE yourself I think

[06:44:07] <kennethreitz> well pypi should invalidate old caches for pages when new releases are made

[06:44:20] <apollo13> I think it does

[06:44:25] <kennethreitz> clearly not :)

[06:44:32] <apollo13> well, patches welcome…

[06:44:39] <apollo13> either way, I'd upload via warehouse

[06:45:10] <kennethreitz> i stick to what the masses do, as they are my userbase

[06:45:43] <apollo13> aha

[06:46:03] <apollo13> that reminds me of: "if everyone else jumps I am gonna jump too"

[06:46:23] <kennethreitz> haha sort of the opposite

[06:46:48] <kennethreitz> "you should jump! it's better!" "no, i'm going to stay with the heard over here until you get that all sorted out"

[06:47:06] <apollo13> we are not sorting it out on pypi anymore most likely

[06:47:24] <kennethreitz> i meant warehouse getting sorted out :)

[06:48:01] <apollo13> mhm, but it wouldn't hurt if people would test on warehouse…

[06:48:23] <apollo13> then again, the warehouse code is nicer, so I would not be surprised if uploads work better there

[07:01:33] <kennethreitz> this is the most annoying thing in the world

[07:01:53] <apollo13> lol

[07:02:02] <apollo13> can we switch problems? :D

[07:02:37] <kennethreitz> for a fee :D

[07:03:05] <apollo13> well, if that is the most annoying thing you have, I'll happily pay to get rid of mine

[07:03:39] <kennethreitz> little known to most, i was the original volunteer to "add s3" to the cheeseshop, along with someone else, a long while ago

[07:03:52] <kennethreitz> took one look at the codebase and was like nope

[07:03:57] <apollo13> so you are in a perfect position to fix bugs :þ

[07:04:01] <kennethreitz> hahahahaha

[07:04:12] <apollo13> but yeah, I know that feeling, looked at it once too

[07:04:18] <apollo13> warehouse code is really nice though!

[07:04:27] <apollo13> donald is doing a fantastic job there

[07:04:40] <apollo13> (and everyone else helping)

[07:06:14] <kennethreitz> indeed, i'm excited

[07:06:17] <kennethreitz> long way to go though :)

[07:06:34] <kennethreitz> i'm strangely not pumped about the new interface. it'll be fine, but i'm just so used to the old one

[07:07:27] <apollo13> I do not care much about ui, so dunno.

[07:07:38] <kennethreitz> that's like all i care about: ux

[07:07:40] <kennethreitz> lol

[07:07:51] <apollo13> ui != ux, but yeah

[07:08:08] <apollo13> then again most of my interactions are with pip itself not pypi, so I usually never see the webpage?

[07:08:17] <kennethreitz> i think ui is just a (big) part of ux

[07:08:45] <kennethreitz> pip and setuptools behavior is another

[07:08:58] <kennethreitz> which i ran into a huge annoyance tonight as well

[07:09:19] <kennethreitz> when you change your credentials and try to do anything it's like "you need to log in"

[07:09:22] <kennethreitz> doesn't tell you how

[07:09:34] <kennethreitz> i resorted to just deleting my pypirc

[07:09:45] <kennethreitz> something a user should never have to resort to doing, imo

[07:11:03] <apollo13> to be honest, I couldn't really care less. Of the problems pypi/pip has/had that is pretty much low on any list imo

[07:11:52] <kennethreitz> and a few weeks ago i had an emergency when i needed to reset my password immediately, and it wasn't possible

[07:11:56] <kennethreitz> the form was broken lol

[07:12:04] <kennethreitz> luckily donald fixed it asap

[07:12:06] <kennethreitz> but still

[07:12:11] <kennethreitz> horrible ux :)

[07:12:23] <kennethreitz> always been that way though, and it's gotten a lot better

[07:12:34] <kennethreitz> but the better it gets, the higher the expectations get

[07:12:43] <apollo13> yeah, but first things first

[07:15:06] <kennethreitz> agreed. my first thing's mental health, so i ain't going near any of that :P

[07:45:42] <ionelmc> kennethreitz: for packaging arbitrary files see https://github.com/pytest-dev/pytest-cov for an example (it add that special ".pth" file)

[07:46:18] <ionelmc> it handles more than just the sdist of course

[16:13:59] <dstufft> kennethreitz: pip itself caches /simple/*/ for ~10 minutes

[16:14:51] <dstufft> We do instantly purge on the Fastly side, though if you're uploading to PyPI instead of Warehouse that sometimes fails and doesn't get retried

[16:15:12] <dstufft> and if you're using the latest twine to upload, you're uploading to Warehouse by default ;)

[16:22:49] <xafer> hello dstufft, regarding https://github.com/pypa/setuptools/pull/631 , I don't see why we wouldn't want to implement Requires-External ? I agree it won't be generally useful but it has some usecases.

[16:24:24] <dstufft> xafer: it's the same level of attractive nuissance as ``requires`` is. It's something that sort of sounds like the thing you want, but because it's completely advisory it will likely just end up confusing people

[16:27:22] <xafer> Well I did not push External-Requires because at the time, I thought I could easily piggyback on requires instead but requires performs some data validation

[16:27:51] <xafer> I would not push for it on public packages

[16:28:29] <xafer> But in the context of private packages where you can control the metadata, it would be the perfect place to define system dependencies

[21:02:00] <[Tritium]> how are packages counted for the from page of legacy pypi?

[21:02:24] <[Tritium]> is that individual releases? each file in a release?

[21:03:15] <dstufft> [Tritium]: what is the "from page"

[21:04:28] <[Tritium]> https://pypi.python.org/pypi ...the front page

[21:04:49] <[Tritium]> "The Python Package Index is a repository of software for the Python programming language. There are currently 85194 packages here."

[21:05:15] <dstufft> [Tritium]: oh, that's "projects" in Warehouse terms

[21:05:28] <dstufft> e.g. top level names like "Django", "requests", etc

[21:05:46] <[Tritium]> ok, that is the final nail in the coffin of a bad idea

[21:06:19] <dstufft> there are 525,563 releases (e.g., name + version tuples, ("Django", "1.0"), and 640,927 files.

[21:06:27] <[Tritium]> it was suggested in -offtopic that pypi should... provide a torrent tracker for packages.

[21:07:42] <[Tritium]> average file size, as far as i can tell then, is 524K

[21:07:49] <[Tritium]> in that area at least

[21:08:05] <dstufft> average as in, mean?

[21:08:14] <[Tritium]> arithmatic mean

[21:08:33] <[Tritium]> IIUC pypi is roughly 320gb

[21:08:58] <dstufft> 525489.630900671510

[21:09:11] <dstufft> I just asked PostgreSQL :)

[21:09:13] <dstufft> that's in bytes

[21:09:37] <[Tritium]> i was within 10k

[21:09:49] <dstufft> pretty good guess :]

[21:10:42] <[Tritium]> Is that information in the public data?

[21:11:38] <dstufft> [Tritium]: um, well you could compute it by making 85194 HTTP requests

[21:11:55] <[Tritium]> I am not going to subject my pipe to that heh

[21:15:19] <[Tritium]> A robust statistical api is something i might suggest/request/...pr ...in a year or two

[21:24:26] <dstufft> [Tritium]: sure! :) As far as stats go, a lot of that largely comes down to me not being very good at extracting what's reallly useful out of some data so not really knowing what to expose

[21:24:35] <dstufft> that's why the BigQuery stuff just exposes more or less raw data

[21:25:33] <[Tritium]> aye. that makes sense. The fire-hose method is usually the right way to go unless you have a reason not to

Log file Viewer

Help | Karma | Search:

#pypa logs for Monday the 25th of July, 2016