PMXBOT Log file Viewer

Help | Karma | Search:

#pypa logs for Thursday the 2nd of October, 2014

(Back to #pypa overview) (Back to channel listing) (Animate logs)
[04:41:58] <sebleier> Anyone know if there is a service/tool to display the diff between two versions of a package on pypi?
[10:55:18] <pf_moore> dstufft: ping
[10:55:56] <dstufft> pf_moore: pong
[10:56:46] <pf_moore> dstufft: I'm looking at the PyPI apis for some scripts I'm writing and got hung up on PEP 438 (the rel=internal one)
[10:57:20] <pf_moore> If we go to phase 2, AIUI, tools like pip will only scrape rel=internal links. Is that right?
[10:58:26] <dstufft> No, we're already in Phase 2
[10:58:31] <dstufft> PEP 438 is compltely implemented
[10:59:28] <dstufft> it will only install rel=internal links by default
[10:59:37] <dstufft> but it adds options to re-enable the other classes
[10:59:39] <pf_moore> OK, so things that aren't rel=internal need --allow-external
[10:59:46] <pf_moore> Doh, I knew that...
[11:00:21] <pf_moore> So what about users putting up a local index? Dumping some files in a directory and exposing them via Apache or twisted or something
[11:00:43] <pf_moore> Those won't have rel=internal, so they should use --find-links rather than --extra-index-url?
[11:00:52] <dstufft> it requires a <meta> tag to trigger that behavior
[11:01:13] <dstufft> item #3 under the stuff talking about PyPI
[11:01:46] <pf_moore> Ah, got it.
[11:01:51] <dstufft> FWIW I have a PEP I've been working on on and off that removes PEP 438
[11:01:57] <dstufft> I think it was ultimately a bad idea
[11:02:32] <pf_moore> Yeah, the multi-index stuff looks like a better approach.
[11:02:43] <pf_moore> Is thatthe one you mean?
[11:02:51] <dstufft> yea
[11:03:03] <dstufft> I'm going over it again right now actually
[11:03:09] <dstufft> to update it, reword it, address comments people had
[11:03:49] <pf_moore> But it does mean that for little tools like the ones I'm forever writing, scraping the simple index (in a way that supports non-PyPI indexes) is an ugly beast.
[11:04:20] <pf_moore> distlib's locators cover it sort of, but I keep hitting issues with the API :-(
[11:04:48] <pf_moore> So I write my own and find all the reasons I shouldn't, but should just use existing code...
[11:05:17] <dstufft> You mean PEP 470 makes it harder?
[11:06:18] <pf_moore> sorry, no I meant that the existing stuff is hard, PEP 438 makes it easier if you're willing to ignore anything but rel=internal
[11:06:44] <pf_moore> but pep 438 doesn't help old-style indexes like local ones would be
[11:06:59] <dstufft> ah
[11:07:00] <dstufft> yea
[11:07:21] <pf_moore> PEP 470 might mean that more people end up with simple indexes that just throw up a load of links.
[11:07:44] <dstufft> PEP 470, assuming it gets accepted, should make things a lot easier, since it goes back to the older style of parsing links on PyPI
[11:07:49] <dstufft> no need to worry about rels
[11:07:53] <dstufft> just look for things to install
[11:08:00] <dstufft> except no more scraping indrectly linked things
[11:08:13] <dstufft> and PyPI itself removes everything but what is hosted on PyPI, so pip doesn't need that logic anymore
[11:08:21] <pf_moore> Yeah, that does help. Hadn't thought of that.
[11:08:47] <pf_moore> My code never bothers with going off PyPI anyway - too few cases where it makes a difference to me.
[11:09:09] <pf_moore> But rel=internal does skip some of the utter crap on there.
[11:09:17] <dstufft> yea
[11:09:27] <pf_moore> Have you seen the list of links on the setuptools simple page? Horrendous
[11:09:49] <pf_moore> Anyway, someone needs me to do something - gotta go. Thanks for the info
[11:10:36] <dstufft> see ya!
[13:34:48] <pf_moore> Sigh. Every time I write a program to query PyPI, it takes so long to set up the code to get the data that I forget why I wanted it in the first place
[13:34:53] <pf_moore> And every time I try to write a library to do the querying it takes so long I forget what my use cases were :-(
[14:46:47] <isomorphismes> thanks buck1
[14:47:25] <isomorphismes> buck1: do you mean I should make just a 3-line text file test.py: #!/usr/bin/python; import urllib; print urllib #and that's it?
[18:28:34] <nanonyme> pf_moore, how about next time writing documentation instead of code as the first thing you do? ;)
[19:22:31] <pf_moore> nanonyne: well yeah, but most of these things start as "hmm, I wonder how many packages on PyPI have wheels" or similar Not something I'd document
[19:23:02] <pf_moore> Then after hours fighting with xmlrpc and scraping webpages, I realise I've lost track of the original question
[19:23:26] <pf_moore> Essentially you're making the mistake of assuming I'm organised :-)
[19:25:03] <dstufft> pf_moore: I have an advantage, I answer most of those questions with SQL queries :V
[19:25:09] <dstufft> unless I care about things that aren't hosted on PyPI
[19:25:27] <pf_moore> I know. And as a DBA/SQL developer, I'm insanely jealous :-)
[19:27:58] <dstufft> I can probably arrange for database dumps with the user accounts stripped out fwiw
[19:28:06] <dstufft> if you ever have querstions like that and want one
[19:28:28] <dstufft> it uses postgresql though, I bet you're a MSSQL kind of guy ;P
[19:29:59] <pf_moore> dstufft: Actually Oracle... But I'm sort of trying to learn postgresql in my spare time, so if you ever did have the time, a dump would be really cool to play with.
[19:30:40] <dstufft> I actually jsut took a fresh pg_dump yesterday, so I can jsut clean out the personal data real quick
[19:31:38] <dstufft> pf_moore: oracle was going to be my second choice :D
[19:31:53] <pf_moore> lol - love being predictable ;-)
[20:02:01] <carlio> dstufft: oo, could i have a copy of the dump too?
[20:26:08] <xafer> any chance to also have it ? :p