#general
- l
limited-country-20626
10/13/2015, 6:05 PMAre there any documents or notes from the engine discussions at the previous summits? - w
witty-crayon-22786
10/13/2015, 8:51 PM@limited-country-20626: @salmon-summer-38098 has been working to polish something, and @enough-analyst-54434 may or may not be working from the ideas behind those notes. but i don't think there is anything comprehensive - s
salmon-summer-38098
10/13/2015, 8:52 PMyeah I shared that doc with @limited-country-20626 already - w
witty-crayon-22786
10/14/2015, 9:30 AMouch. that's not particular scalable. https://issues.scala-lang.org/browse/SI-4484 - w
witty-crayon-22786
10/14/2015, 3:36 PMok, so that ^ kept me up all night - w
witty-crayon-22786
10/14/2015, 3:36 PMbecause if there is one thing i've taken away from this whole bakeoff, it's that we should not be invalidating fully transitively for jvm compilation - w
witty-crayon-22786
10/14/2015, 3:37 PMso. i think a very easy thing we could add to start improving on this would be an option to include only direct deps at compile time - w
witty-crayon-22786
10/14/2015, 3:38 PMthis would be the enforced version of the jvm-dep-check "direct deps" check - w
witty-crayon-22786
10/14/2015, 3:38 PMthe upshots are many. - b
bored-art-40741
10/14/2015, 3:38 PMThe downside is that it's going to make buildgen very difficult to get right in a performant way - w
witty-crayon-22786
10/14/2015, 3:39 PMyou'd need to run it all the way to the typer - b
bored-art-40741
10/14/2015, 3:39 PMYeah, and that's infeasible - w
witty-crayon-22786
10/14/2015, 3:39 PMreally? - b
bored-art-40741
10/14/2015, 3:39 PMRight now we get away with doing it directly after parse, which is screaming fast - w
witty-crayon-22786
10/14/2015, 3:39 PMoh. because you need the classpath? - b
bored-art-40741
10/14/2015, 3:39 PMWe don't use any type information, it's entire lexical - w
witty-crayon-22786
10/14/2015, 3:39 PM@bored-art-40741: so you don't need the classpath to do this from parse - w
witty-crayon-22786
10/14/2015, 3:40 PMyea. - b
bored-art-40741
10/14/2015, 3:40 PMCorrect, but the issue is that scala creates lexically-invisible "direct" deps because of type inference - w
witty-crayon-22786
10/14/2015, 3:40 PMyep. see reference above. - b
bored-art-40741
10/14/2015, 3:40 PMRight now we're fine because those are always transitive deps - b
bored-art-40741
10/14/2015, 3:40 PMAh sorry, I didn't read the issue - w
witty-crayon-22786
10/14/2015, 3:41 PMthe impact on buildgens hadn't occurred to me. - w
witty-crayon-22786
10/14/2015, 3:41 PMsorry, i'm sure you guys all fully understood this before, but it's only now sinking in for me. - b
bored-art-40741
10/14/2015, 3:42 PMnp, scala is a beast to reason about - w
witty-crayon-22786
10/14/2015, 3:42 PMif there is anything at all worth throwing money at typesafe over, this might be it. - b
bored-art-40741
10/14/2015, 3:42 PMHa - b
bored-art-40741
10/14/2015, 3:43 PMIt's unclear if the semantics of the language even make this possible - w
witty-crayon-22786
10/14/2015, 3:43 PMdotty could be the opportunity to fix that. - w
witty-crayon-22786
10/14/2015, 3:43 PM...given sufficient motivation