pmxbot IRC Log Viewer

[00:12:40] <wsanchez_> I wanna put mysql-connector-python==1.0.10 into a requirements file, but that results in:

[00:12:41] <wsanchez_> Some externally hosted files were ignored (use --allow-external mysql-connector-python to allow).

[00:12:41] <wsanchez_> Which is OK, but if I stick —allow-external into the file, I get an exception. Pilot error?

[00:26:37] <dstufft> wsanchez_: --allow-external mysql-connector-python

[00:27:21] <dstufft> https://gist.github.com/dstufft/f9e0b7a0183ac13ec77e

[00:27:34] <dstufft> pretend django-celery says mysql-connector-python

[00:36:25] <wsanchez_> That raises for me:

[00:36:25] <wsanchez_> Traceback (most recent call last):

[00:36:25] <wsanchez_> File "<stdin>", line 2, in <module>

[00:36:27] <wsanchez_> File "setup.py", line 99, in <module>

[00:36:29] <wsanchez_> install_requirements = read_requirements("py_base.txt")

[00:36:31] <wsanchez_> File "setup.py", line 93, in read_requirements

[00:36:33] <wsanchez_> parse_requirements(joinpath(requirements_dir, reqs_filename))

[00:36:35] <wsanchez_> File "/Users/wsanchez/Developer/IMS/Server/.develop/roots/py_modules/lib/python2.7/site-packages/pip/req.py", line 1605, in parse_requirements

[00:36:37] <wsanchez_> finder.allow_external |= set([normalize_name(line).lower()])

[00:36:39] <wsanchez_> AttributeError: 'NoneType' object has no attribute 'allow_external'

[00:36:56] <wsanchez_> lemme push my code up.

[00:37:51] <dstufft> are you using pip as an API

[00:38:20] <wsanchez_> …yes

[00:38:53] <dstufft> you probably need to construct a finder instancce and pass it into that function

[00:40:06] <wsanchez_> raise happens here: https://github.com/burningmantech/ranger-ims/blob/master/Server/setup.py#L93

[00:42:59] <wsanchez_> Why I'm doing this is less obvious in this project, but in others the idea is that there is a directory of requirements files. One for base (always needed) requirements, one for stuff you need while developing, and then optionally one each for optional requirements associated with optional features.

[00:45:40] <dstufft> any reason you don't just declare these in your setup.py with extras

[00:46:56] <wsanchez_> Because that's code, and it seems cleaner to just have some text files. Here's a better example: http://trac.calendarserver.org/browser/CalendarServer/trunk/requirements

[00:51:02] <dstufft> wsanchez_: some text files that you read by callng an undocmented private method of pip inside of code vs some static dictionaries in a .py file? ;)

[00:51:11] <dstufft> also https://caremad.io/blog/setup-vs-requirement/

[00:51:36] <wsanchez_> pip.req.parse_requirements is private?

[00:51:53] <wsanchez_> Or is all of pip's API private?

[00:52:50] <wsanchez_> Trying to remember how I found out about that function

[00:53:02] <dstufft> we techincally don't support being used as an API yea, I mean we try not to break it and what not

[00:53:12] <dstufft> but for instance, in 1.6 that function moves

[00:53:25] <dstufft> to pip.req.req_file.parse_requirements

[00:53:40] <wsanchez_> Ah, probably StackOverflow, eg: http://stackoverflow.com/questions/14399534/how-can-i-reference-requirements-txt-for-the-install-requires-kwarg-in-setuptool

[00:54:35] <dstufft> probably people are going to have some breakage

[00:54:38] <dstufft> if they were doing that

[00:54:47] <wsanchez_> ok so you're saying I'm off in the weeds so I should'a packed a machete.

[00:55:15] <dstufft> yea more or less

[00:55:25] <wsanchez_> fair enough

[00:58:39] <dstufft> wsanchez_: the correct answer to not duplicate stuff is to use "." or "-e ." in the requirements.txt

[00:58:41] <dstufft> fwiw

[00:58:47] <dstufft> you can also specify extras with that

[00:59:16] <wsanchez_> hrm.

[00:59:22] <wsanchez_> …reading that blog post

[01:00:05] <dstufft> I'm not super good at words

[01:00:13] <wsanchez_> :-)

[01:00:51] <dstufft> if you're familar with ruby and bundler at all yehuda katz has a similar post which inspired my post (linked at the bottom) which may be eaier to understand too

[01:01:36] <wsanchez_> I started to read a Ruby book a long time ago and saw something about the syntax for strings that made me close the book and move on.

[01:02:13] <dstufft> haha

[01:02:16] <dstufft> I know that feel

[01:03:12] <dstufft> the primary take away is that setup.py lists symbolic names that don't specify where to get stuff from, and requirements.txt has a "location to get stuff from" (defualting to PyPI) which when pared with that symbolic name gives you an "aboslute" name

[01:06:37] <Alex_Gaynor> dstufft: in practice isn't the distinction bullshit because setuptools will happilly make up a place to get some stuff from?

[01:08:47] <wsanchez_> It does seem like there's still unnecessary duplication. At a minimum it's too easy for information in setup.py and requirements files to get out of sync and for one not to notice.

[01:10:07] <dstufft> Alex_Gaynor: well setuptools is bad yes

[01:10:14] <dstufft> wsanchez_: you don't have to list things in your requirements file

[01:10:28] <dstufft> https://github.com/dstufft/warehouse/blob/master/requirements.txt#L1-L2

[01:10:35] <dstufft> ^ installs all of WArehouse's dependencies

[01:10:38] <dstufft> without duplicating them

[01:10:56] <wsanchez_> hrm

[01:11:08] <dstufft> you can also do like

[01:11:09] <dstufft> -e .[test]

[01:11:14] <dstufft> to install a "test" extra too

[01:11:32] <dstufft> less moving parts at install time this way

[01:11:43] <dstufft> afk for dinner

[01:33:26] <wsanchez_> dstufft: Thanks for the info. I posted on http://stackoverflow.com/questions/14399534/ that using the API may not be a great idea and will rethink my use of requirements.txt. Also off to dinner; laters.

[01:44:48] <Ivo> dstufft: do we actually mention anywhere that pip is explicitly not considered a library?

[01:44:57] <Ivo> (or should be)

[02:07:45] <dstufft> Ivo: dunno

[02:07:53] <dstufft> Ivo: we certainly don't document the API at all

[02:08:23] <Ivo> seems like a murphy's law

[02:08:41] <Ivo> could they have done so?

[02:08:46] <Ivo> yes, they have.

[11:04:06] <ronny> how do i teach pip self-signed ssl certs?

[11:09:31] <dstufft> ronny: --cert on the CLI

[11:09:36] <dstufft> or PIP_CERT env var

[11:09:43] <dstufft> or cert in the ~/.pip/pip.conf

[11:12:18] <ronny> ah, i had a old version

[11:12:42] <ronny> dstufft: btw, how do i stop pip install --download from asking for every single file?

[11:13:02] <dstufft> asking?

[11:13:08] <dstufft> what's it asking

[11:13:20] <ronny> if it should wipe/overwrite/keep

[11:14:20] <ronny> oO

[11:14:23] <ronny> wait a moment

[11:14:25] <ronny> fml

[11:15:02] <jezdez> --exists-action (s)witch, (i)gnore, (w)ipe, (b)ackup

[11:15:29] <ronny> i was on pip 1.1 again, forgot to update after a recreation of the virtualenv

[11:15:52] <Ivo> y u no update virtualenv

[11:16:05] <ronny> say hello debian

[11:16:22] <Ivo> sudo pip install pip setuptools virtualenv

[11:16:28] <Ivo> **** the system

[11:16:35] <Ivo> *add a -U in there

[11:16:35] <ronny> nope

[11:16:52] <dstufft> ronny: FWIW you don't have to install virtualenv to use it

[11:17:30] <dstufft> if you'd rather use a newer version without installing it :) or manually upgrade, it's up to you doesn't matter to me :)

[11:18:44] <ronny> just no system install

[11:19:12] <dstufft> you might also like pip install --user virtualenv

[11:19:42] <ronny> btw, does pip --cert completely replace the used certs, or does it add to the certs to check against?

[11:20:00] <dstufft> replaces

[11:20:29] <ronny> so id have to make a whole bundle for my company?

[11:20:35] <ronny> is there a howto for doing that?

[11:21:01] <ronny> (we have servers with self-signed certs)

[11:21:11] <Ivo> 'whole bundle'? you just use your company's CA

[11:22:37] <dstufft> ronny: CA bundles are just concated files

[11:23:12] <dstufft> https://github.com/kennethreitz/requests/blob/master/requests/cacert.pem

[11:41:25] <ronny> k

[11:41:32] <Ivo> dstufft: how do you find out whats wrong with pypi rst... :/

[11:42:16] <Ivo> python setup.py check -smr works just fine

[11:42:21] <Ivo> https://pypi.python.org/pypi/pipa

[11:42:30] <dstufft> Ivo: fuck around with https://bitbucket.org/pypa/pypi/src/ec1282bfe85c12c93005ca7271f56c9df5c9445e/description_utils.py?at=default#cl-102 until you figure it out

[11:46:20] <mgedmin> Ivo, pip install restview; restview --long-description [--strict]

[11:47:57] <Ivo> omfg cannot wait for cryptography 0.3

[11:48:49] <Ivo> "quickly install package"... uses pyopenssl... takes 13mb >_<

[11:51:22] <mgedmin> Ivo, actually restview --long-description is unhappy because pypa's setup.py refuses to run under python2, but restview -e 'python3 setup.py --long-description' --strict works

[11:51:27] <mgedmin> and emits this error message: <string>:65: (INFO/1) Hyperlink target "pip.conf" is not referenced.

[11:51:34] <mgedmin> HTH

[11:56:40] <Ivo> godamn i hate rst links

[11:56:59] <Ivo> they suck so much compared to markdown >_<

[11:58:18] <Ivo> mgedmin: thankyou

[11:59:01] <dstufft> someday pypi will support Md

[11:59:22] <Ivo> tektnoic submitted a patch for that

[11:59:35] <dstufft> yea

[12:00:09] <dstufft> It's not how it'll work long term, it's relies on the MD being unrenderable, which not all MD is, some of it just renders terribly wrong

[12:00:15] <Ivo> ronny: if you're interested, i made a simple local pypi server made to do pip --cert correctly, if you would like to 'see how it works' https://pypi.python.org/pypi/pipa

[12:00:28] <dstufft> plus it makes it pretty annoying if you want to add a third markup type

[12:00:45] <dstufft> long term the metadata will support different markup types

[12:01:03] <ronny> Ivo: we have a devpi behind https

[12:01:06] <Ivo> pfmoore: man, you gotta step away from the argument and just ask for actual working code branches :)

[12:01:16] <dstufft> oh hey pfmoore is here!?

[12:01:33] <Ivo> yah he has been running an irc client sometimes lately

[12:01:36] <Ivo> it seems

[12:01:45] <ronny> btw, is there a reason why pip refuses subsubdomains on a subdomain wildcard

[12:02:05] <ronny> wilcard is *.domain.com, servername is name.expossedvia.domain.com

[12:02:11] <mgedmin> browsers refuse them too, don't they?

[12:02:15] <dstufft> they should

[12:02:23] <dstufft> the RFC states that a * is only good for a single segment

[12:02:24] <ronny> firefox allows a exception

[12:03:02] <dstufft> probably we could add an option to not check, but it'd disable SSL alltogether

[12:03:08] <dstufft> Theuni2: ping ;)

[12:03:25] <ronny> hi Theuni2

[12:58:39] <Theuni2> yo!

[12:59:12] <Theuni2> dstufft: see our somewhat long issue last week when the fastly issue escalated o_O

[12:59:34] <dstufft> Theuni2: I see that! I'm glad y'all got it fixed :]

[12:59:45] <dstufft> apparently sending data around some tubes is kind of hard

[13:04:28] <Theuni2> especially when people f*$#& with the standards

[15:23:20] <pfmoore> dstufft: Ivo: I'm running an offline client (irccloud) so I just look like I'm here ;-) But yeah, Ivo, you're right re working code.

[15:38:20] <Ivo> Theuni2: life wouldn't be interesting if people didn't fuck with the standards

[15:39:13] <Theuni2> life would be much more interesting if people fucked people instead.

[15:40:40] <Ivo> I'm not sure we're ever satisfied with how many things we've fucked

[15:40:55] <Theuni2> no one ever is

[15:41:12] <Ivo> #justhumanthings

[17:24:28] <DanielHolth> What was that talk about cross-compiling using distutils? Is that a thing?

[17:25:06] <dstufft> kind of

[23:17:32] <nZac_> Question about local git repositories as Python packages. I am testing a Sphinx theme in directory X with project Y's documentation. I `pip install -e X` to install the theme in project Y. However if I make changes (git commit) to the theme how can I update project Y's copy? pip install --upgrade -e ../path/to/package/repo doesn't work, Would pip install --upgrade -e theme-name doesn't work.

[23:20:08] <pfmoore> nZac_: pip install -e should make project Y refer directly to X, not use a copy. So you shouldn't need to do anything, just make the changes in X and Y will see them immediately.

[23:20:36] <nZac_> pfmoore: thanks for the response. I am not seeing those changes.

[23:21:05] <nZac_> Can you think of anyplace I could have messed it up?

[23:21:28] <pfmoore> hmm, that's odd - but you refer to themes rather than Python files, so maybe there's something different going on. Sorry, I'm not sure what that could be

[23:21:54] <pfmoore> I don't know anything about how Sphinx organises themes, I'm afraid...

[23:22:43] <nZac_> Well, at least I know I am installing it correctly with -e for editable.

[23:23:16] <tomprince> Do you have sphinx installed in the virtualenv?

[23:25:30] <tomprince> ^--- nZac_

[23:25:36] <nZac_> tomprince: sure do

[23:26:29] <nZac_> I think I just figured it out... pfmoore was right it was something totally unrelated. I had to make clean and the make html to pull in the changes

[23:26:32] <tomprince> I'd be tempted to add some prints to conf.py to print out where sphinx is actully loading things from. So 'import sphinx_theme; print sphinx_theme.__file__' et al.

[23:27:21] <nZac_> -e appeared to be the correct action, I was just not completing all the proper steps. Thanks for the help tomprince ad pfmoore!

[23:56:14] <unstable> I have a package pyxmlsec in my requirements.txt, though it's failing on ...

[23:56:15] <unstable> Downloading/unpacking pyxmlsec Could not find any downloads that satisfy the requirement pyxmlsec Some externally hosted files were ignored (use --allow-external pyxmlsec to allow).

Log file Viewer

Help | Karma | Search:

#pypa logs for Wednesday the 26th of March, 2014