#development
- w
witty-crayon-22786
02/10/2017, 8:55 PM@happy-kitchen-89482 : we've been banging away on performance, and want to wait a little bit longer before enabling by default - w
witty-crayon-22786
02/10/2017, 8:55 PMcertainly this quarter - w
witty-crayon-22786
02/10/2017, 8:56 PMtwo outstanding performance tracks: one is executing all IO on the rust side to ensure that we get concurrency (should be reviewable next week), and the other is nick's work to eliminate noop subgraphs, which he's a few weeks away from we think - h
happy-kitchen-89482
02/10/2017, 8:57 PMcool - h
happy-kitchen-89482
02/10/2017, 8:57 PMI’m still a bit fuzzy on how the new engine interacts with existing tasks/targets - w
witty-crayon-22786
02/10/2017, 8:59 PM@happy-kitchen-89482 : as are we. - w
witty-crayon-22786
02/10/2017, 9:00 PMin particular, it is only being used for "private" APIs currently (build file parsing, target construction), and so what the integration with the Task API will look like is unknown - w
witty-crayon-22786
02/10/2017, 9:01 PMthe root of the API is "for these subjects, generate these products" - w
witty-crayon-22786
02/10/2017, 9:02 PMso how a Task would request that, and whether we integrate it into the product map (and whether there will still be a product map) is up in the air - h
happy-kitchen-89482
02/10/2017, 9:02 PMMaybe we don’t bother and just rewrite all the tasks? - h
happy-kitchen-89482
02/10/2017, 9:02 PMSince there would seem to be an impedance mismatch anyway - w
witty-crayon-22786
02/10/2017, 9:02 PMoh, we do. but we're just assuming a migration - h
happy-kitchen-89482
02/10/2017, 9:03 PMWell, clearly we rewrite all the tasks sooner or later - w
witty-crayon-22786
02/10/2017, 9:03 PMsortof. - w
witty-crayon-22786
02/10/2017, 9:03 PMin particular, "frontend" tasks never go away - w
witty-crayon-22786
02/10/2017, 9:03 PMthe repl will begin requesting a compile in a new way - w
witty-crayon-22786
02/10/2017, 9:04 PMbut then it will check out the results locally and do something synchronous in the foreground - w
witty-crayon-22786
02/10/2017, 9:04 PMditto bundle. - w
witty-crayon-22786
02/10/2017, 9:05 PMtests will be overhauled though. - h
happy-kitchen-89482
02/10/2017, 9:05 PMI have to assume that a JVM compile task under the new engine would be significantly simpler than the existing one - w
witty-crayon-22786
02/10/2017, 9:05 PM@happy-kitchen-89482 : the instant we're happy with the performance we're going to be thinking about UX though, and everyone should get involved in that - w
witty-crayon-22786
02/10/2017, 9:06 PM@happy-kitchen-89482 : yes - h
happy-kitchen-89482
02/10/2017, 9:06 PME.g., if we wanted to start with one that invoked javac/scalac directly without zinc, that would presumably be almost trivial - w
witty-crayon-22786
02/10/2017, 9:07 PMyea - w
witty-crayon-22786
02/10/2017, 9:07 PMexcept in that it needs to request codegen first - h
happy-kitchen-89482
02/10/2017, 9:07 PMright - h
happy-kitchen-89482
02/10/2017, 9:07 PMand remote sources - w
witty-crayon-22786
02/10/2017, 9:07 PMwhich is all modeled out, and we can definitely evolve toward it - h
happy-kitchen-89482
02/10/2017, 9:07 PMbut that’s just declarative, right? - h
happy-kitchen-89482
02/10/2017, 9:07 PMthe engine knows how to handle that