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

    rough-minister-58256

    03/22/2016, 9:30 PM
    er sorry, immutable
  • r

    rough-minister-58256

    03/22/2016, 9:30 PM
    words
  • w

    witty-crayon-22786

    03/22/2016, 9:30 PM
    yea, @rough-minister-58256 : i added memoization for identity, and it helped a bit
  • w

    witty-crayon-22786

    03/22/2016, 9:31 PM
    but memoizing in a few places wasn't quite enough. fundamentally, we need to minimize CA in-process
  • w

    witty-crayon-22786

    03/22/2016, 9:32 PM
    or have a fingerprint implementation as thorough as CA, but without actually pickling.
  • f

    full-kilobyte-93429

    03/22/2016, 9:32 PM
    CA is content address?
  • w

    witty-crayon-22786

    03/22/2016, 9:32 PM
    yes
  • r

    rough-minister-58256

    03/22/2016, 9:33 PM
    yeah, an object literal style key lookup
  • f

    full-kilobyte-93429

    03/22/2016, 9:34 PM
    seems worth pursuing, possible we still have CA everywhere but more efficient
  • w

    witty-crayon-22786

    03/22/2016, 9:34 PM
    er, what is worth pursuing
  • f

    full-kilobyte-93429

    03/22/2016, 9:35 PM
    fingerprint CA or object literal?
  • w

    witty-crayon-22786

    03/22/2016, 9:36 PM
    i don't know what that means
  • f

    full-kilobyte-93429

    03/22/2016, 9:36 PM
    not pickling the entire content
  • f

    full-kilobyte-93429

    03/22/2016, 9:36 PM
    but its fingerprint which is done once
  • w

    witty-crayon-22786

    03/22/2016, 9:37 PM
    sure... that's what we're already doing for the dependency values, right?
  • f

    full-kilobyte-93429

    03/22/2016, 9:40 PM
    the difference is this can be at different granularity, currently is done at subject level, it can be at the node level?
  • w

    witty-crayon-22786

    03/22/2016, 9:41 PM
    ah. yea, probably.,
  • w

    witty-crayon-22786

    03/22/2016, 9:41 PM
    ...well
  • w

    witty-crayon-22786

    03/22/2016, 9:41 PM
    yea.
  • w

    witty-crayon-22786

    03/22/2016, 9:42 PM
    i think that's true. because somewhere there is a Select node that will inspect the Subject and actually return it
  • w

    witty-crayon-22786

    03/22/2016, 9:42 PM
    and that's where it will be content addressed.
  • w

    witty-crayon-22786

    03/22/2016, 9:42 PM
    as 
    Return(subject)
  • f

    full-kilobyte-93429

    03/22/2016, 9:45 PM
    i don't quite understand this 'because...'
  • w

    witty-crayon-22786

    03/22/2016, 9:47 PM
    @full-kilobyte-93429: the only way for a task to view a subject is by `Select`ing it
  • w

    witty-crayon-22786

    03/22/2016, 9:48 PM
    and in that case it will be content-addressed as 
    Return(subject)
  • w

    witty-crayon-22786

    03/22/2016, 9:48 PM
    so... it probably doesn't make sense to content-address the subject field of a Node.
  • f

    full-kilobyte-93429

    03/22/2016, 9:49 PM
    i see
  • w

    witty-crayon-22786

    03/22/2016, 9:50 PM
    so yes, i think i'm agreeing with you there
  • f

    full-kilobyte-93429

    03/22/2016, 9:51 PM
    like a 
    node_key
    ?
  • w

    witty-crayon-22786

    03/22/2016, 9:51 PM
    ?
1...585960...906Latest