https://pantsbuild.org/ logo
Powered by
#general
  • w

    witty-crayon-22786

    08/26/2015, 5:06 AM
    i wonder whether all of the recursive options could eventually be deprecated in favor of default subsystems
  • w

    witty-crayon-22786

    08/26/2015, 6:14 AM
  • w

    witty-crayon-22786

    08/26/2015, 6:14 AM
    one: they use annotations/decorators to mark fields as part of a task's identity
  • w

    witty-crayon-22786

    08/26/2015, 6:14 AM
    i think we could do something similar to clean up payloads
  • h

    happy-kitchen-89482

    08/26/2015, 6:23 AM
    we only have two legitimate recursive options: --level and --color
  • j

    jolly-chef-92794

    08/26/2015, 2:33 PM
    @witty-crayon-22786 +1 using decorators to clean up payloads, I think that’d be much nicer and make inheritance cleaner also.
  • b

    bored-art-40741

    08/26/2015, 3:14 PM
    Not sure that I agree
  • b

    bored-art-40741

    08/26/2015, 3:14 PM
    That's going to involve metaclass magic to discover the payload fields
  • b

    bored-art-40741

    08/26/2015, 3:15 PM
    In my experience that's almost universally a bad idea, though I could see how it would work here. But more importantly, I think we should figure out serialization / caching / execution graph semantics before we mess around with the target interface
  • b

    bored-art-40741

    08/26/2015, 3:19 PM
    Before doing that I'd rather see Target get lifecycle methods that you override specifically to construct your payload. But that's basically what 
    __init__
    is doing for us now, and it's not clear what we'd gain from breaking it out
  • w

    witty-crayon-22786

    08/26/2015, 4:21 PM
    re: figuring out semantics: yea, probably.
  • w

    witty-crayon-22786

    08/26/2015, 4:22 PM
    but re: "override specifically to construct your payload", i'm not sure turing complete payload construction is beneficial
  • w

    witty-crayon-22786

    08/26/2015, 4:23 PM
    as opposed to whitelisting/blacklisting typed fields
  • b

    bored-art-40741

    08/26/2015, 4:23 PM
    I tend to agree
  • w

    witty-crayon-22786

    08/26/2015, 4:23 PM
    or parameters, i suppose
  • w

    witty-crayon-22786

    08/26/2015, 4:24 PM
    fields still allow for manipulation... parameters would not
  • w

    witty-crayon-22786

    08/26/2015, 4:24 PM
    but with option fingerprinting in place, i think parameter-whitelisting would be sufficient... no other inputs should affect behaviour
  • w

    witty-crayon-22786

    08/26/2015, 4:24 PM
    (right?)
  • w

    witty-crayon-22786

    08/26/2015, 4:25 PM
    (could reach out and inspect files, but)
  • w

    witty-crayon-22786

    08/26/2015, 4:25 PM
    ... and those wouldn't be recorded as parameters. hm.
  • w

    witty-crayon-22786

    08/26/2015, 4:25 PM
    so yea, fields.
  • w

    witty-crayon-22786

    08/26/2015, 4:29 PM
    another thought would be a sort of declarative "implicit" parameter for things like files.
  • w

    witty-crayon-22786

    08/26/2015, 4:30 PM
    oh... like field-based guice injection.
  • w

    witty-crayon-22786

    08/26/2015, 4:31 PM
    hm, yea... this is exactly DI, isn't it
  • a

    ancient-stone-50795

    08/26/2015, 8:57 PM
    @happy-kitchen-89482: have a nice burn...
  • h

    happy-kitchen-89482

    08/26/2015, 8:58 PM
    thanks!
  • h

    happy-kitchen-89482

    08/26/2015, 8:58 PM
    Don't convert pants to ruby while I'm gone
  • w

    witty-crayon-22786

    08/26/2015, 8:58 PM
    blugh
  • a

    ancient-stone-50795

    08/26/2015, 8:59 PM
    That won’t be an issue, OTOH, I’m convening a side channel conversation on converting it to Java!
  • a

    ancient-stone-50795

    08/26/2015, 9:00 PM
    can’t find the right emoji for that one
1...220221222...2420Latest