Log file Viewer

[07:08:00] <leomark> Im using pip 1.5.4, running in ubuntu 14.04.. I got this error while installing "Connection to pypi.python.org timed out" does pip supports http proxy?

[07:11:47] <dstufft> leomark: it supports CONNECT proxies

[09:23:16] <straycat> Is there a specification for the package json I find on pypi.python.org e.g. https://pypi.python.org/pypi/acpi/json ?

[10:27:47] <mgedmin> straycat, pypi front page -> left sidebar -> PyPI Developer Info

[10:28:22] <mgedmin> then near the top: PyPI APIs: JSON

[10:28:35] <mgedmin> links to https://wiki.python.org/moin/PyPIJSON

[10:28:51] <mgedmin> which is not very informative, I'm afraid

[10:29:08] <mgedmin> (also, why is pypi so slow at the moment?)

[10:29:59] <straycat> mgedmin, Indeed :) I'm using the xmlrpc now anyway.

[10:31:45] <mgedmin> source code is available

[10:32:00] <mgedmin> and I think dstufft is very interested in documenting the existing pypi APIs so he can reimplement them for warehouse

[10:33:12] <mgedmin> https://warehouse.pypa.io/en/latest/api-reference/ currently describes XML-RPC and the "legacy" simple API

[10:33:21] <mgedmin> I'm guessing JSON is in the todo list somewhere

[10:35:31] <dstufft> The JSON api is already re-reimplemented in WArehouse if I recall, there might be some documentation. Generally my plans are "add newer better APIs and leave the old ones for backwards compat" because the old ones kinda blow

[10:37:41] <mgedmin> how do you do a 'git grep' on bitbucket?

[10:38:46] <mgedmin> straycat, https://bitbucket.org/pypa/pypi/src/bb928c418c8bee2762f65640da077f336e7cb67b/webui.py?at=default#cl-1439 is the source code of the /json method

[10:44:03] <straycat> mgedmin, thanks :)

[11:26:49] <mgedmin> so... why is pip install --download /tmp python-nation executing the setup.py?

[11:27:20] <mgedmin> oh, to discover install_requires?

[11:27:54] <mgedmin> in that case why is pip install --download /tmp --no-deps python-nation still executing the setup.py?

[11:32:28] <dstufft> mgedmin: because the --no-deps conditional happens later in the code

[11:32:37] <dstufft> I don't think there is a specific actual reason other than that

[12:26:40] <doismellburning> package_data, data_files, both/neither?

[12:30:48] <DanielHolth> package_data

[12:33:03] <doismellburning> DanielHolth: thanks - got a link to any deprecation notice?

[12:34:35] <doismellburning> I just read http://stackoverflow.com/a/14159430/928098 and if it's right I am sad

[12:41:11] <doismellburning> oops, should have gone straight to https://packaging.python.org/en/latest/distributing.html#package-data

[12:48:21] <dstufft> pacakge_data and data_files do different things

[12:48:47] <dstufft> that SO post is also wrong

[12:49:25] <dstufft> The packaging.p.o link is correct

[12:52:34] <doismellburning> yeah it looked wrong

[12:52:45] <doismellburning> cheers

[12:53:10] <dstufft> the difference is that the package_data flag says "install these files" while the MANIFEST.in says "include these files in the sdist"

[12:53:20] <doismellburning> 'Data files are files to be placed `"outside of your packages". We do not do that.' huzzah

[12:53:20] <dstufft> or why that SO post is wrong

[12:53:50] <dstufft> if you specify package_data, but your tarball doesn't contain those files, then they can't very well be added

[12:53:52] <doismellburning> I am not convinced I grok the difference between "install these files" and "include these files in the sdist"

[12:54:01] <doismellburning> oh right

[12:54:11] <ionelmc> doismellburning: the SO answer assumes you're going to always use include_package_data flag

[12:54:13] <dstufft> doismellburning: the sdist is the tarball itself

[12:54:26] <ionelmc> in that context is correct, but that flag is false by default :|

[12:54:29] <dstufft> there are files you want to add to the tarball but might not want to install yea?

[12:54:38] <dstufft> like a README or a LICENSE file, or the setup.py file itself

[12:55:03] <ionelmc> doismellburning: see http://pythonhosted.org/setuptools/setuptools.html?highlight=include_package_data#including-data-files

[12:56:08] <doismellburning> I wonder if the simpler solution isn't to define a dir where "stuff to be packaged" lives

[12:56:15] <doismellburning> c.f. Maven's `src/main/resources`

[12:57:12] <ionelmc> doismellburning: i like that idea

[12:57:36] <doismellburning> that's not totally dissimilar to current convention

[12:57:57] <ionelmc> doismellburning: most people just lump everything in the root dir

[12:58:11] <ionelmc> code, packaging configuration, tests ... everything in the same places

[12:58:23] <doismellburning> where if / is my repo root, /LICENSE and /README etc. I don't want in some install, and then my actual code and other files that I do live in /packagename

[12:58:34] <ionelmc> doismellburning: i wrote about this actually: http://blog.ionelmc.ro/2014/05/25/python-packaging/

[12:58:37] <doismellburning> (so granted, that mixes data with code, but still)

[13:28:24] <doismellburning> ionelmc: can I clarify re http://blog.ionelmc.ro/2014/05/25/python-packaging/ - do you encourage including tests inside the package then?

[13:29:00] <ionelmc> doismellburning: can you point me to the exact paragraph ?

[13:30:09] <doismellburning> ionelmc: nope it's a general question I can't tell if you address

[13:30:15] <ionelmc> i'm a proponent of tests outside, roughly because tests have all sorts dependencies and pure unittests (no deps) are not always practical/useful

[13:30:25] <doismellburning> ionelmc: but afaict with tox it's either "install in dev mode" or "package tests"

[13:30:54] <ionelmc> with tox, you install the package and run tests from whereever you want

[13:31:01] <doismellburning> ionelmc: and if I'm right, your proposal means "tox testing develop install" doesn't work

[13:31:23] <ionelmc> not necessarily the tests that are in the installed package

[13:31:38] <doismellburning> sorry, I guess I should clarify; with tox running Django tests, because of the way discovery works

[13:32:17] <ionelmc> doismellburning: did you read this part http://blog.ionelmc.ro/2014/05/25/python-packaging/#id4 ?

[13:32:41] <ionelmc> the paragraphs from there till the next title

[13:33:35] <doismellburning> ionelmc: sorry, I managed to completely miss that

[13:36:00] <doismellburning> I wish Octopress auto-added anchors at all

[13:36:30] <ionelmc> doismellburning: you think pelican provides those heading anchors ?

[13:36:35] <ionelmc> :)

[13:36:50] <doismellburning> whatcha use?

[13:37:00] <ionelmc> doismellburning:

[13:37:02] <ionelmc> https://www.irccloud.com/pastebin/kckeaTFC

[13:37:56] <doismellburning> iä

[13:37:58] <doismellburning> ok, thanks

[13:41:24] <ionelmc> doismellburning: also, medium has comments on paragraphs

[13:41:39] <ionelmc> that would be very useful for long articles

[13:41:59] <ionelmc> but it's medium :|

[13:42:57] <doismellburning> indeed

[13:43:03] <doismellburning> I disable comments

[13:43:09] <doismellburning> I just want to be able to go "read this section"

[13:44:20] <ionelmc> doismellburning: disable comments?

[13:44:50] <doismellburning> well

[13:44:54] <doismellburning> s/disable/don't provide/

[13:46:21] <ionelmc> how can you even disable comments?

[13:47:54] <ionelmc> aah

[13:47:57] <ionelmc> they call it notes

[15:33:11] <ghickman> is there a recommended place I can read up on the rationale behind removing dependency links from pip?

[15:33:41] <ghickman> would love to know more and alternative methods of using git checkouts in setup.pys

[16:01:51] <dstufft> ghickman: the rationale is basically "end users should know which URLs pip is going to hit and dependency links invert that control"

[16:02:07] <dstufft> this matters a lot more for public projects instead of company specific ones

[16:02:15] <ghickman> ok cool, thanks

[16:02:30] <ghickman> I found the mailing list post on it which was really imformative

[16:02:58] <ghickman> thanks for keeping the opt-in version for private packaging =)

[16:03:09] <dstufft> ghickman: you can use requirements files, though we're not going to kill --process-dependency-links (or w/e it's called, I forget) until the replacement is there

[16:03:27] <ghickman> brilliant, thank you

[16:03:32] <ghickman> that's really good to know

[16:03:40] <dstufft> the replacement will likely be doing something like... install_requires = ["mycompanyproject @ http://wahtever/..."]

[16:03:44] <dstufft> or something like that

[16:03:56] <dstufft> and PyPI will just reject uploads of projects that have it

[16:04:07] <ghickman> that makes a lot of sense

[20:41:07] <josePhoenix> Hi all

[20:41:25] <josePhoenix> I'm trying to untangle some setup customizations, and was wondering if anyone recognized the format of this line

[20:41:33] <josePhoenix> "Installed /private/var/folders/2t/m92rpkyn2ps064ffwv0qxrbm0001dr/T/easy_install-Ox9Wwe/stsci.distutils-0.3.2/d2to1-0.2.11-py2.7.egg"

[20:42:25] <josePhoenix> The stsci.distutils-0.3.2 is several versions behind, and I can't figure out what in setup.py is causing it to be pulled in. I have grepped without success for that version number

[20:43:12] <josePhoenix> It is the second line after running egg_info for pysynphot when attempting a pip install pysynphot

[20:43:44] <josePhoenix> full pip log here https://gist.github.com/josePhoenix/0f5e2c6fa129ae96f6cb

[21:01:08] <_jesse_> hi all, I'm having trouble registering a new version of my package. I see the latest version listed at https://pypi.python.org/pypi/tory_client, but https://pypi.python.org/simple/tory-client/ does not have the newest version (we used to have the package registered as 'tory-client', but this was changed to 'tory_client' at one point for consistency)

[21:02:31] <_jesse_> https://pypi.python.org/simple/tory_client/ redirects to https://pypi.python.org/simple/tory-client/

[21:02:41] <ionelmc> _jesse_: log it to https://bitbucket.org/pypa/pypi/issues

[21:03:04] <_jesse_> ionelmc: will do, thank you

[21:06:00] <aclark> _jesse_: 0.9.2 has no distributions so what would you expect to see under /simple/?

[21:06:30] <aclark> _jesse_: i.e. it's registered not uploaded

[21:08:36] <_jesse_> aaaah

[21:09:02] <_jesse_> sorry, I'm a bit new to python packaging, let me read and try again

[21:10:17] <aclark> np, just looks like you need to upload a file (AKA distribution)

[21:26:57] <_jesse_> aclark: yep, that was it, sorry for the confusion