https://pantsbuild.org/ logo
Powered by
#development
  • a

    aloof-angle-91616

    03/10/2019, 2:15 AM
    i decided to shlex the arguments because i didn't want to parse them
  • a

    aloof-angle-91616

    03/10/2019, 2:15 AM
    going to make a PR maybe then look at pipelining query expressions
  • h

    hundreds-father-404

    03/10/2019, 2:19 AM
    Cool! If I understand correctly, this will result in less first class commands, e.g. 
    ./pants owner-of
    would be removed in favor of 
    ./pants --query='owner_of' list
    ? I remember being overwhelmed at 4S with discoverability / how many options there are in pants. So unifying around one single idea like query—and then making it be easy to find what are valid queries—seems like a win
  • a

    aloof-angle-91616

    03/10/2019, 2:36 AM
    yes!
  • a

    aloof-angle-91616

    03/10/2019, 2:36 AM
    see issue
  • a

    aloof-angle-91616

    03/10/2019, 2:36 AM
    i'm v pumped about it
  • a

    aloof-angle-91616

    03/10/2019, 2:36 AM
    it would also be able to supersede 
    dependees
    and 
    filter
    ❤️ 1
  • a

    aloof-angle-91616

    03/10/2019, 2:37 AM
    and probably several others or maybe at least some scripts we have internally
  • a

    aloof-angle-91616

    03/10/2019, 2:40 AM
    also the query pipeline is potentially going to be next level, i'm thinking something like:
    Copy code
    > ./pants --query='changes_since master' --query='filter type=junit_tests' --query='dependencies transitive=True' --query='filter type=jar_library' my-random-3rdparty-jar-task ::
    🔥 1
    h
    • 2
    • 16
  • a

    aloof-angle-91616

    03/10/2019, 2:40 AM
    this is a contrived example
  • a

    aloof-angle-91616

    03/10/2019, 2:40 AM
    but would take the target specs (in this case 
    ::
    ) as the initial input
  • a

    aloof-angle-91616

    03/10/2019, 2:41 AM
    then intersect that against the target with changes since master
  • a

    aloof-angle-91616

    03/10/2019, 2:41 AM
    then filter to junit test targets, then find the transitive dependencies of those targets which are jar_library()s
  • a

    aloof-angle-91616

    03/10/2019, 2:42 AM
    then run some task on those targets (in this case 
    my-random-3rdparty-jar-task
    ), because this is all still just target root selection
  • a

    aloof-angle-91616

    03/10/2019, 2:43 AM
    and these are also all v2 rules right now so they can be used by other rules extremely easily
  • a

    aloof-angle-91616

    03/10/2019, 2:43 AM
    everything came together extremely quickly and that was cool
  • a

    aloof-angle-91616

    03/10/2019, 2:43 AM
    going to push a PR with just an initial 
    --query
    option in a bit, no pipelining or anything
  • a

    aloof-angle-91616

    03/10/2019, 2:46 AM
    and since bazel has analogies to these in its query language it might make sense to support that too the only reason i didn't was because i didn't want to parse anything and like the pipelining idea is already different enough that i think we don't need to copy bazel here right now (maybe they can copy us, idk)
  • a

    aloof-angle-91616

    03/10/2019, 2:46 AM
    oh and this isn't magic, v2 rules have to be made for each query method to produce 
    BuildFileAddresses
    , but like that is fun and not hard
  • a

    aloof-angle-91616

    03/10/2019, 4:53 AM
  • a

    aloof-angle-91616

    03/10/2019, 7:51 AM
    i also realized that it’s weird we only have single dispatch in 
    yield Get(A, B, x)
    expressions
  • a

    aloof-angle-91616

    03/10/2019, 7:51 AM
    and like yes you can make a datatype containing the appropriate fields to simulate it
  • a

    aloof-angle-91616

    03/10/2019, 7:53 AM
    but it would also be cool to do like 
    yield Get(A, [Param(B, x), Param(C, y)])
    w
    • 2
    • 7
  • a

    aloof-angle-91616

    03/10/2019, 7:53 AM
    i don’t actually think that’s a good idea but it would be cool
  • a

    aloof-angle-91616

    03/10/2019, 8:03 AM
    the alternative is actually not bad and is what i am working on for 
    --query
    composition
    -.py
  • a

    aloof-angle-91616

    03/10/2019, 8:04 AM
    so this^ would work right now
  • a

    aloof-angle-91616

    03/10/2019, 8:06 AM
    and this is like what multiple dispatch in `Get()`s would look like if we didn't have to go through an intermediate object (which is why we need the 
    hydrate_with_input()
    method at all)
    -.py
  • a

    aloof-angle-91616

    03/10/2019, 8:08 AM
    i think i like the first better anyway
  • a

    aloof-angle-91616

    03/10/2019, 8:08 AM
    was just unable to figure it out for a bit
  • a

    aloof-angle-91616

    03/10/2019, 8:14 AM
    i get it now, each 
    yield Get(A, B, x)
    where 
    B
    is tagged with 
    @union
    is like dispatching and you get multiple dispatch by making successive `yield Get(...)`s
1...617618619...906Latest