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

    witty-crayon-22786

    09/12/2016, 10:35 PM
    (that's essentially what inlining does, with the exception of the root values)
  • w

    witty-crayon-22786

    09/12/2016, 10:36 PM
    @fancy-queen-20734 : ^
  • w

    witty-crayon-22786

    09/12/2016, 10:37 PM
    in the presence of statically generated subgraphs, i think you could get rid of the 
    Select
    instances at the root too, and be 100% TaskNodes in the graph.
  • w

    witty-crayon-22786

    09/12/2016, 10:39 PM
    i agree about "I feel like this is a place where we could introduce a new type of declaration, either on the collection type, or as a rule that statically defines the ways a product can be merged".
  • w

    witty-crayon-22786

    09/12/2016, 10:39 PM
    ...and i think i see where you were going with items 1/2/3 above. they all depend on whether some type is mergeable.
  • w

    witty-crayon-22786

    09/12/2016, 10:40 PM
    sorry. jumped the gun there.
  • f

    fancy-queen-20734

    09/12/2016, 10:40 PM
    no problem.
  • w

    witty-crayon-22786

    09/12/2016, 10:41 PM
    one critical question then is: are types only mergeable in particular contexts, or is a mergeable type always mergeable, and therefore implicitly mergeable
  • f

    fancy-queen-20734

    09/12/2016, 10:41 PM
    that’s exactly the interesting question
  • w

    witty-crayon-22786

    09/12/2016, 10:42 PM
    because if things are implicitly mergeable, then 
    SelectDependencies(Collection.of(A), Collection.of(B))
    could be automatically available
  • w

    witty-crayon-22786

    09/12/2016, 10:43 PM
    ... ie, if 
    A -> B
    , then implictly also 
    Collection.of(A) -> Collection.of(B)
  • w

    witty-crayon-22786

    09/12/2016, 10:44 PM
    going to look for some counter examples.
  • f

    fancy-queen-20734

    09/12/2016, 10:49 PM
    let me know what you find
  • w

    witty-crayon-22786

    09/12/2016, 10:54 PM
    hm, there are no instances of 
    SelectDependencies(Collection.of(A), Collection.of(B))
    in master... only in the branch where i switch to a transitive walk
  • w

    witty-crayon-22786

    09/12/2016, 10:54 PM
    but for:
  • w

    witty-crayon-22786

    09/12/2016, 10:54 PM
    Copy code
    if (issubclass(product, Collection) or product.__name__.endswith('s')) and (issubclass(dep_product, Collection) or dep_product.__name__.endswith('s')):
      print('>>> {} implies a conversion?'.format(x))
  • w

    witty-crayon-22786

    09/12/2016, 10:54 PM
    i see
  • w

    witty-crayon-22786

    09/12/2016, 10:54 PM
    Copy code
    >>> SelectDependencies(Paths, PathGlobs, field_types=(PathWildcard, PathDirWildcard, PathRoot,)) implies a conversion?
>>> SelectDependencies(Dirs, Paths, u'link_stats', field_types=(Link,)) implies a conversion?
>>> SelectDependencies(Files, Paths, u'link_stats', field_types=(Link,)) implies a conversion?
  • w

    witty-crayon-22786

    09/12/2016, 10:57 PM
    have a meeting, but will look for counter examples afterward.
  • w

    witty-crayon-22786

    09/12/2016, 11:32 PM
    ...hm. counter examples are challenging, because 
    A->B
    is... the fundamental operation in the system.
  • w

    witty-crayon-22786

    09/12/2016, 11:33 PM
    any valid path through the graph, basically.
  • w

    witty-crayon-22786

    09/12/2016, 11:34 PM
    so instead, i guess it's about whether it would be safe to mark particular Collections mergeable.
  • w

    witty-crayon-22786

    09/12/2016, 11:57 PM
    ...and that definitely seems like a type-level decision.
  • f

    fancy-queen-20734

    09/13/2016, 12:02 AM
    yeah. I think we should start there and if we find places where making it context dependent, then do more thinking. Having the type know its mergeability is simpler.
  • w

    witty-crayon-22786

    09/13/2016, 12:55 AM
    @User uploaded a file:

    type mergability▾

    and commented: @fancy-queen-20734 :
  • w

    witty-crayon-22786

    09/13/2016, 1:07 AM
    the question of how to actually implement the generated Rule(s) remain, but.
  • w

    witty-crayon-22786

    09/13/2016, 1:08 AM
    (one declaration of mergeability will result in a lot more Rules, implicitly.)
  • w

    witty-crayon-22786

    09/13/2016, 5:57 PM
    @salmon-summer-38098 : 
    self.visualize_graph_to_file(roots, 'viz.0.dot')
  • w

    witty-crayon-22786

    09/13/2016, 5:57 PM
    export DOT=0; dot -Tsvg -oviz.$DOT.svg viz.$DOT.dot && open viz.$DOT.svg
  • f

    fancy-queen-20734

    09/13/2016, 6:00 PM
    @witty-crayon-22786 actually, I’m getting lunch first, so it might be a bit longer.
1...168169170...906Latest