#general
- a
alert-monitor-62116
10/25/2017, 4:49 PMhowever this 'test/jvm/com/disney/studio/ds/testutil/kafka:kafka' guy is also pulling Jackson 2.8, though its seem if I run this target individually it works every time - a
alert-monitor-62116
10/25/2017, 4:50 PMbut if I run 2 junit targets the order of running this tests matters - a
alert-monitor-62116
10/25/2017, 4:51 PMother junit_tests target doesn't have jackson2.6 dependency and its pulling 2.8 - h
high-advantage-47550
10/25/2017, 7:19 PMi don't suppose there's a way to output the
/dependencies
graph as a dot file? 🙂dependees - w
witty-crayon-22786
10/25/2017, 7:26 PMyes../pants depmap --graph - h
high-advantage-47550
10/25/2017, 7:27 PMah sorry, missed it in
, thank you!help - f
fancy-queen-20734
10/26/2017, 4:28 PMgood morning - w
witty-crayon-22786
10/26/2017, 4:28 PM@fancy-queen-20734: hey! so re: https://github.com/pantsbuild/pants/pull/4976 - w
witty-crayon-22786
10/26/2017, 4:28 PMthe model change i'm suggesting is that we begin injecting targets bottom up - w
witty-crayon-22786
10/26/2017, 4:29 PMwhich means we're never injecting a target while it has dependents - w
witty-crayon-22786
10/26/2017, 4:29 PMi don't understand why we didn't do this before, but given how much we've had to do to avoid performance problems, it seems like low hanging fruit - w
witty-crayon-22786
10/26/2017, 4:30 PMwe know that the graph doesn't have cycles - w
witty-crayon-22786
10/26/2017, 4:30 PM(or, we will, because of the toposort) - w
witty-crayon-22786
10/26/2017, 4:31 PM...oh. hm. they would still have dependents, huh. i take that part back. - w
witty-crayon-22786
10/26/2017, 4:31 PMbut the complete list of dependencies would already exist. - f
fancy-queen-20734
10/26/2017, 4:32 PMI’d put together a patch to do something similar a while ago. I wish I’d wrote up a doc for it so I could remember the approach. - w
witty-crayon-22786
10/26/2017, 4:34 PMgiven that they would still have dependents, it isn't as much of a slam dunk as i had thought. - f
fancy-queen-20734
10/26/2017, 4:35 PMyeah - w
witty-crayon-22786
10/26/2017, 4:36 PM...but. - w
witty-crayon-22786
10/26/2017, 4:36 PMit does ease the export checking, right? - f
fancy-queen-20734
10/26/2017, 4:37 PMmy idea was that if you could assume that a given codegen target doesn’t depend on the results of previous targets, you could generate all of the targets with no dependencies / exports, and then do all the injection in a separate step that walked them in a topo ordering - w
witty-crayon-22786
10/26/2017, 4:37 PMbecause you would have the complete list of exports/dependencies guaranteed to have been generated already - f
fancy-queen-20734
10/26/2017, 4:37 PMif the assumption held, you could parallelize codegen - w
witty-crayon-22786
10/26/2017, 4:37 PMyea. but that's not a goal at the moment - f
fancy-queen-20734
10/26/2017, 4:38 PMyeah, but it’s related, and I’m happy that I remembered what it was I was trying to do - w
witty-crayon-22786
10/26/2017, 4:38 PMthe toposort is a flip of a switch to enable - f
fancy-queen-20734
10/26/2017, 4:38 PMyep - w
witty-crayon-22786
10/26/2017, 4:38 PMyea, sorry... didn't mean to dismiss it - f
fancy-queen-20734
10/26/2017, 4:38 PMnp. I wasn’t suggesting I’d do it as part of this change. I don’t think that would make sense - f
fancy-queen-20734
10/26/2017, 4:39 PMSo, right now, the patch is toposorted, so dependencies/exports will exist before the dependency/export edges are injected