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

    witty-crayon-22786

    01/27/2016, 12:24 AM
    the thing is: the report file has to exist anyway
  • b

    bored-art-40741

    01/27/2016, 12:24 AM
    It's very useful when you differentiate between .pants.d and .local_artifact_cache
  • b

    bored-art-40741

    01/27/2016, 12:24 AM
    But yeah, I think the PR I've been working on is a good PoC and road toward separating ivy resolution from fetching
  • b

    bored-art-40741

    01/27/2016, 12:25 AM
    It very nicely isolates the directory where ivy stores the xmls and such that it generates as part of resolution
  • b

    bored-art-40741

    01/27/2016, 12:25 AM
    Which can then be rebased and cached properly
  • w

    witty-crayon-22786

    01/27/2016, 12:25 AM
    oh. neat. um, please coordinate with @fancy-queen-20734 ... i think he may or may not have started that.
  • w

    witty-crayon-22786

    01/27/2016, 12:25 AM
    he mentioned storing an intermediate format which was not the report itself
  • b

    bored-art-40741

    01/27/2016, 12:25 AM
  • b

    bored-art-40741

    01/27/2016, 12:25 AM
    Kind of messy, sorry
  • b

    bored-art-40741

    01/27/2016, 12:25 AM
    I'll clean it up sometime this week
  • w

    witty-crayon-22786

    01/27/2016, 12:26 AM
    ok, thanks.
  • b

    bored-art-40741

    01/27/2016, 12:26 AM
    Yeah, we can also just rebase the report itself
  • b

    bored-art-40741

    01/27/2016, 12:26 AM
    In the same fashion we rebase zinc analysis
  • b

    bored-art-40741

    01/27/2016, 12:26 AM
    Just replace a known realpath prefix with a sentinel token, and then swap it back out when ingesting artifacts
  • b

    bored-art-40741

    01/27/2016, 12:27 AM
    Fetching on a cache hit for a resolve is a slightly different story, but splitting that doesn't seem hard to me
  • b

    bored-art-40741

    01/27/2016, 12:27 AM
    We could either implement our own fetcher (we have all of the info on how to fetch the artifact and where to fetch it to), or we could probably rig up ivy to do the fetch
  • b

    bored-art-40741

    01/27/2016, 12:27 AM
    Or a bunch of parallel fetches of individual artifacts
  • b

    bored-art-40741

    01/27/2016, 12:28 AM
    Or, as you mentioned before, we could couple the ivy pom/jar/etc artifact with the report artifact
  • b

    bored-art-40741

    01/27/2016, 12:28 AM
    So it's all just self-contained
  • b

    bored-art-40741

    01/27/2016, 12:28 AM
    Mildly wasteful of space, but eh
  • b

    bored-art-40741

    01/27/2016, 12:28 AM
    If only there was a content addressed blob store to help normalize out that kind of thing...
  • w

    witty-crayon-22786

    01/27/2016, 12:29 AM
    @bored-art-40741: the thing about the report is that it isn't super useful to cache prolly;
  • b

    bored-art-40741

    01/27/2016, 12:29 AM
    I buy that too
  • w

    witty-crayon-22786

    01/27/2016, 12:29 AM
    it doesn't tell you where to find things
  • w

    witty-crayon-22786

    01/27/2016, 12:29 AM
    that is where nick has been exploring
  • b

    bored-art-40741

    01/27/2016, 12:29 AM
    If we're already going to normalize it into our own internal graph representation
  • w

    witty-crayon-22786

    01/27/2016, 12:29 AM
    yea
  • b

    bored-art-40741

    01/27/2016, 12:29 AM
    Just pickle it up (or whatever) and cache it. Sure
  • b

    bored-art-40741

    01/27/2016, 12:30 AM
    I'd be really happy to see this sort of incremental work on ivy precede the actual overhauls in our models
  • b

    bored-art-40741

    01/27/2016, 12:30 AM
    The enemy of better is perfect, and right now I'd describe our current state as surprisingly broken
1...918919920...2420Latest