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

    witty-crayon-22786

    02/22/2017, 12:49 AM
    yea
  • w

    witty-crayon-22786

    02/22/2017, 12:49 AM
    right... not used in any rules
  • f

    fancy-queen-20734

    02/22/2017, 12:49 AM
    I think TypeConstraint(Vec<TypeId>) makes a lot of sense. It should be a smaller change too.
  • w

    witty-crayon-22786

    02/22/2017, 12:50 AM
    but it should wait for this Key thing to go out.
  • w

    witty-crayon-22786

    02/22/2017, 12:51 AM
    @fancy-queen-20734 : i'm not certain it is a small change. but i think it's an important one: https://github.com/pantsbuild/pants/issues/4005
  • f

    fancy-queen-20734

    02/22/2017, 12:52 AM
    Ah.
  • w

    witty-crayon-22786

    02/22/2017, 12:52 AM
    we currently use equality matches on TypeConstraints sometimes, which is... odd
  • f

    fancy-queen-20734

    02/22/2017, 12:52 AM
    yeah. I was thinking of the smaller, change the representation from Exactly to tuple
  • w

    witty-crayon-22786

    02/22/2017, 12:52 AM
    i guess you could do only that and not clean up the underlying bit.
  • f

    fancy-queen-20734

    02/22/2017, 12:53 AM
    it’s a mechanical change that removes noise, so I think it’s worth doing.
  • w

    witty-crayon-22786

    02/22/2017, 12:54 AM
    sure
  • f

    fancy-queen-20734

    02/22/2017, 12:54 AM
    I think the python type constraint classes add a bit too much friction right now
  • w

    witty-crayon-22786

    02/22/2017, 12:54 AM
    yea.
  • f

    fancy-queen-20734

    02/22/2017, 12:55 AM
    the tricky one is the symbol table, which is annoying to have as just a literal
  • f

    fancy-queen-20734

    02/22/2017, 12:55 AM
    and now I remember why I adopted the constraints
  • w

    witty-crayon-22786

    02/22/2017, 12:57 AM
    @fancy-queen-20734 : could 
    ...
    it.
  • w

    witty-crayon-22786

    02/22/2017, 12:57 AM
    a description is fine, but doing something like 
    (Target, Struct, PythonTarget,... 32 more)
  • f

    fancy-queen-20734

    02/22/2017, 12:57 AM
    yeah—that makes sense. Then in the future if it’s confusing for users, we could add a name table back in as a separate feature.
  • w

    witty-crayon-22786

    02/22/2017, 12:58 AM
    would be fine too
  • w

    witty-crayon-22786

    02/22/2017, 1:08 AM
    @fancy-queen-20734 : hm... don't wait on the Key change... looks like introducing millions of calls to python in order to clone the Values held in the Key is... slow.
  • w

    witty-crayon-22786

    02/22/2017, 1:08 AM
    going to need to split that back out and actually store a mapping.
  • w

    witty-crayon-22786

    02/22/2017, 1:14 AM
    or 
    Arc
    them
  • w

    witty-crayon-22786

    02/22/2017, 1:29 AM
    ...hm. Arcing value would probably be quite helpful as well.
  • w

    witty-crayon-22786

    02/22/2017, 1:51 AM
    ...Arcing value is likely to be the only helpful thing.
  • w

    witty-crayon-22786

    02/22/2017, 1:51 AM
    @fancy-queen-20734 : you don't have anything outstanding that touches externs?
  • w

    witty-crayon-22786

    02/22/2017, 6:50 PM
    @rough-minister-58256 : in case you want to take up the cffi tack, i posted some partial work on https://github.com/pantsbuild/pants/issues/4036
  • r

    rough-minister-58256

    02/22/2017, 6:50 PM
    sweet
  • r

    rough-minister-58256

    02/22/2017, 6:50 PM
    thx
  • f

    fancy-queen-20734

    02/22/2017, 7:30 PM
    @witty-crayon-22786 not modifying externs, though I am using some in the string representations
  • w

    witty-crayon-22786

    02/22/2017, 7:30 PM
    @fancy-queen-20734 : i think the key stuff will still be valuable, but it's really, really invasive.
1...245246247...906Latest