Log file Viewer

[11:57:45] <pombreda_> Hi https://github.com/nexB/fetchcode/issues/1#issuecomment-555475520 ... I would like to reuse the fine vcs code in pip... what's you take on the best way?

[16:51:41] <techalchemy> pombreda_, it's probably not possible yet

[16:52:06] <techalchemy> there is a long-ish term plan to refactor reusable elements of pip into separate libraries but pip itself is being massively refactored first

[18:08:10] <pombreda_> techalchemy: so best to fork the valuable bits for now then?

[18:09:12] <techalchemy> depends on how patient you are pombreda_

[18:09:16] <pombreda_> (I am interested in the vcs part that does exactly what I need: reliably fetch vcs repos using vcs strings)

[18:09:48] <pombreda_> techalchemy: I worked out trimming done the bit + their tests. And they know all pass.. :)

[18:10:08] <pombreda_> *the bit: the vcs module and its essential deps

[18:11:38] <techalchemy> pombreda_, if you want to resolve dependencies on them also you'll need slightly more but you can be slightly less patient because you can just wait for me to finish the updates i'm working on

[20:20:18] <altendky> is it possible to specify the version (not revision/hash/tag) for a git+ dependency similar to #egg= for the project name? specifically for `-e git+https://github.com/altendky/pip-tools@ac7ecec6a2961b1dd9589b7a12f853f0b5af07d6#egg=pip-tools`

[20:33:41] <techalchemy> altendky, no

[20:34:04] <altendky> techalchemy: thanks. i guess i've just got enough bugs coming together here that i'm in a corner then. :]

[20:34:18] <techalchemy> it's impossible to know a version before you check something out, was the rationale on that one I believe

[20:34:21] <altendky> checking to see if -m can save me

[20:34:26] <techalchemy> I am mostly in your position on this

[20:34:38] <altendky> techalchemy: mm, i mean let me specify it on the line. not have pip guess it

[20:34:47] <altendky> just like #egg=projectname

[20:34:52] <techalchemy> 'what if i wanted to validate after it gets downloaded'

[20:35:00] <techalchemy> yeah too bad

[20:35:02] <techalchemy> lol

[20:35:31] <techalchemy> tbh i dont know why we need egg fragments but

[20:35:35] <altendky> techalchemy: "it's impossible to know..." the same can be said for the projectname, right?

[20:35:49] <techalchemy> see previous statement

[20:36:20] <altendky> techalchemy: #egg= isn't always required but i know if had cases where it made things better. at least with the way things are. not so worried right now about how they could or should be

[20:37:20] <techalchemy> yeah there are edge cases and whatnot, the real reason egg fragments are needed but not versions is basically that egg fragments provide the name of the package and pip doesn't care about the version

[20:37:30] <techalchemy> if you supply a vcs dependency it trumps any version

[20:38:26] <techalchemy> altendky, what issue are you hitting?

[20:40:20] <altendky> techalchemy: https://bpaste.net/show/OIS7M line 186. i've got my own wrapper thingy around pip-tools and i think pip-tools is trying to update itself now (didn't before) because i'm trying switching to a git+ dep for it from a regular pypi one. the regular pypi one wouldn't trigger a reinstall but -e git+ does. so may just like you -m pip install --upgrade pip in windows (-m making it not fail) then maybe -m will

[20:40:20] <altendky> help here too

[20:40:31] <altendky> my first hope was to specify a version so that it didn't think it was changing

[20:40:42] <altendky> well, not changing, but also not triggering an update

[20:42:23] <techalchemy> glad to see pip-tools still has the same challenges

[20:43:00] <techalchemy> yeah i'm pretty sure it will always trigger an update

[20:43:31] <altendky> alrighty, well maybe -m will make it not fail

[20:45:57] <techalchemy> oh let me read the erro rmessage more carefully

[20:47:54] <techalchemy> oh because it's trying to update the script itself, got it... on windows i'm guessing no

[20:48:59] <techalchemy> i think piptools just shoots off subprocess calls to pip and if it's trying to call a subprocess that will replace the calling process's file it will need to do some weird hackery

[20:49:08] <techalchemy> if i had to guess randomly

[22:15:08] <altendky> techalchemy: yeah, i figured it was the same thing as you get with `pip install --upgrade pip` on windows. where `python -m pip install --upgrade pip` works. and yeah, `python -m piptools sync` rather than `pip-sync` did 'fix' my issue. i'd rather just not be reinstalling, but at least this works.

[22:16:15] <techalchemy> oh nice yeah if you're invoking piptools first you should be ok

[22:18:15] <altendky> yeah, first i pip install a minimal requirements file with pip-tools. then i use pip-sync to sync a bigger requirements file. also use azure to run pip-compile on multiple platforms. working on making it do it for multiple python versions. i mean, they are all entirely separate runs, but still handy for making locks for multiple scenarios