Log file Viewer

[10:45:45] <ronny> dstufft: how does one install a wheel file with extras?

[10:46:14] <ronny> (pip tells me a version spec is missing when i add [foo] to a wheel file on the cli

[11:36:35] <ionelmc> ronny: you might have to use the --find-list + --no-index trick

[11:36:54] <ronny> ionelmc: this is for gumby elf ^^

[11:37:10] <ronny> ionelmc: i made it so that any install happesn via wheels, even develop installs

[11:37:38] <ronny> and suddently pip solves scripts for me

[11:37:39] <ionelmc> ronny: even develop? what would you put in the wheel there? a pth file?

[11:38:07] <ronny> ionelmc: still the bootstrap script, i had a quite fruitull discussion with a plone developer

[11:38:39] <ionelmc> i hope it was about getting rid of that bootstrap script :)

[11:44:04] <ronny> ionelmc: well the other way around scripts would either be underpowered or break

[11:44:33] <ionelmc> ronny: what do you mean?

[11:45:07] <ronny> ionelmc: cant explain in detail now, its a strange behaviour combination

[11:52:35] <ionelmc> ronny: humour me :-)

[11:57:23] <ronny> ionelmc: basically lots o bad experiences with changing pythonpath the last few days + simplistic develop install dont uninstall propperly

[12:00:48] <ionelmc> ronny: so what exactly is wrong with a pth (besides the fact it needs site.py to be activated)?

[12:01:52] <ronny> it changes sys.path

[12:12:07] <ionelmc> well, technically site.py does

[12:12:54] <ionelmc> i kinda forgot now what i didn't really like about that bootstrap script

[12:19:11] <ronny> well, yeah, i still need more tests, a dist command and a publish command

[12:39:15] <tomprince> ronny: How do you manage to have a 'develop' install without changing sys.path?

[12:39:44] <ronny> tomprince: by writing a python file that does evil?

[12:40:03] <ronny> bascially i generate a wheel with metadata for requirements/scripts/data and then tell pip to install that wheel

[12:43:43] <tomprince> Don't you still need to manipulate sys.path to use a package that isn't in site-packages?

[12:48:08] <ronny> tomprince: a python script is a pacakge once it has a __path__

[12:48:42] <ronny> tomprince: to just combine __file__ __path__ and a call to exec after compile and you get a python 2/3 compatible script that will get your code from soemwhere else

[15:23:13] <ionelmc> ronny: somehow i have a feeling that it will blow up on python3 or when nspackages are used

[15:25:11] <ronny> ionelmc: pep420 is broken as far as i can tell, normal namespaces will be fixed soon

[17:53:08] <tos9> Can I get entry_points/console_scripts to not use pkg_resources

[17:53:52] <ronny> tos9: as of now, nope, why?

[17:54:31] <tos9> ronny: Aright, thanks. I literally want `from mymodule import main; main()` in the file, so I can just go back to using scripts, but there's 30ms overhead for entry_points apparently, which is 3x the runtime of my script.

[17:54:53] <tos9> (This is the second time I've noticed, and I haven't profiled setuptools to figure out why, but.)

[17:55:09] <ronny> tos9: i suspect the reason is "details"

[17:55:28] <tos9> ronny: Well yeah it's always "details" :) but most of the time most of the details are bad ideas :P

[17:55:34] <tos9> So if I'm doing simpler things I'll survive.

[17:55:50] <ronny> dstufft: could pip perhaps ship a custom script writer that does the direct import, unlike setuptools installs pip installs wont need the extra guard

[17:56:26] <dstufft> ronny: tos9 install from a wheel, it'll use a simpler thing.

[17:56:28] <ronny> tos9: thing is, pip install can probably be thaught to skup setuptools scripts, and then use a custo mscript writer that does propper imports

[17:56:40] <ronny> dstufft: why only for the wheel?

[17:56:53] <dstufft> because pip writes the scripts for wheels, setuptools writes it for sdists

[17:57:03] <dstufft> maybe we can do it for sdists too, but we don't currently

[17:57:17] <ronny> i see

[17:57:19] <ronny> hmm

[17:57:26] <ronny> ya for gumby elf

[17:57:29] <tos9> dstufft: aha. OK, lemme try that first. Is there a shortcut for pip install -e . via wheel by the way?

[17:57:47] <tos9> well no that couldn't work?

[17:57:58] <ronny> dstufft: actually with the patch of liveless pip will use it for all things, wouldnt it?

[17:58:18] <ronny> tos9: develop install via wheel is impossible with setuptools

[17:58:31] <ronny> tos9: gmby elf can do, but its not at a 0.1 stage yet

[17:59:09] <dstufft> i don't know what "liveless pip" is

[17:59:21] <tos9> ronny: hrm how would that work?

[17:59:25] <tos9> can you put symlinks in wheels?

[17:59:34] <ronny> dstufft: the patches about building wheels for sdists

[18:00:13] <ronny> tos9: https://bitbucket.org/RonnyPfannschmidt/gumby_elf/src/76c93a6c84d31e52dd7bee1c8e46c1943d55c0bc/src/lowlevel.py?at=default#cl-13

[18:01:05] <ronny> dstufft: lifeless i meant (robert colins)

[18:01:09] <dstufft> ronny: oh

[18:01:56] <ronny> tos9: basically gumy elf builds a wheel with a evil python script and tells pip to install it

[18:02:08] <dstufft> ronny: currently that patch requires wheel to be installed, and you will be able to exclude packages from being installed as a wheel, and if building a wheel fails it'll install as sdist, but yes it means in more situations we'll use pip's script wrappers

[18:02:46] <tos9> ronny: hrm ok

[18:03:15] <ronny> dstufft: btw, cah pip write out a red complaining line for every package that has no wheel?

[18:05:03] <dstufft> ronny: no wheel on PyPI you mean?

[18:05:53] <ronny> dstufft: yeah

[18:08:31] <dstufft> ronny: I doubt that would be OK, it'd have to be yellow not red b/c it's not an error condition, but more importantly there are perfectly valid reasons not to have a wheel, and a lot of projects which don't currently. If we just spit out a ton of warnings then we'll introduce warning fatigue

[18:12:53] <tos9> I've already got warning fatigue from installation.

[18:13:07] <tos9> Considering how absurdly terrible MANIFEST thigns are

[18:13:44] <tos9> and even after my $PYTHONWARNINGS is set to default,ignore:Not importing directory:ImportWarning

[18:14:15] <dstufft> tos9: the latest pip shouldn't say anything about MANIFEST

[18:14:50] <tos9> How latest?

[18:14:56] <tos9> I have 6.0.8 I think

[18:16:26] <dstufft> 6.1.0 I think

[18:16:44] <tos9> Cool, trying.

[18:17:07] <tos9> Hrm no :/

[18:17:18] <dstufft> No?

[18:17:19] <tos9> pip install -U pip && virtualenv test && test/bin/pip install zope.interface && rm -r test

[18:17:25] <tos9> warning: no previously-included files matching '*.dll' found anywhere in distribution

[18:17:32] <tos9> (and 12 more of those)

[18:17:39] <dstufft> well a virtualenv installs it's own version of pip

[18:17:57] <tos9> Oh right

[18:18:36] <dstufft> https://bpaste.net/show/d8a376707419

[18:22:05] <tos9> dstufft: awesome yup, same now (modulo this internet connection)

[18:29:53] <tos9> dstufft: You're now my accomplice in an evil deed. Publishing a package named `l` PyPI.

[18:31:56] <jessamyn_> lol