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

    happy-kitchen-89482

    05/04/2016, 10:37 PM
    buy why all the changes from .format to %s?
  • r

    rough-minister-58256

    05/04/2016, 10:38 PM
    purely for debug logging efficiency
  • r

    rough-minister-58256

    05/04/2016, 10:38 PM
    or the start of it, at least
  • r

    rough-minister-58256

    05/04/2016, 10:43 PM
    that particular instance you flagged was generating massive strings unnecessarily
  • h

    happy-kitchen-89482

    05/04/2016, 10:46 PM
    It’s worth a comment at least, because we supposedly standardized on format
  • h

    happy-kitchen-89482

    05/04/2016, 10:46 PM
    also, I’m surprised that there’s a noticeable perf difference between the two
  • w

    witty-crayon-22786

    05/04/2016, 10:46 PM
    @happy-kitchen-89482: one is lazy, the other isn't
  • w

    witty-crayon-22786

    05/04/2016, 10:47 PM
    it's definitely best practice on the JVM at least to use loggers that will lazily construct strings only if debug is enabled
  • h

    happy-kitchen-89482

    05/04/2016, 10:47 PM
    Definitely
  • w

    witty-crayon-22786

    05/04/2016, 10:47 PM
    i've always been surprised that we don't
  • w

    witty-crayon-22786

    05/04/2016, 10:47 PM
    ...in pants
  • h

    happy-kitchen-89482

    05/04/2016, 10:49 PM
    So why did we standardize on format?
  • w

    witty-crayon-22786

    05/04/2016, 10:50 PM
    well, i don't think this is a matter of format vs 
    %
    interpolation
  • w

    witty-crayon-22786

    05/04/2016, 10:50 PM
    it's about how it is possible to achieve lazy debug logging
  • w

    witty-crayon-22786

    05/04/2016, 10:50 PM
    if there is a 
    {}
    based debug-logging approach, then that would be great
  • r

    rough-minister-58256

    05/04/2016, 10:53 PM
    yeah, we’d have to write one - probably worthwhile to circle back on that at some point and then generally apply it to all 
    logger.debug
    usages and measure the global impact against various tasks & workloads. this is really just testing the waters a bit - seems at least a couple of folks are finding it controversial.
  • m

    mammoth-pharmacist-54215

    05/04/2016, 10:53 PM
  • m

    mammoth-pharmacist-54215

    05/04/2016, 10:54 PM
    DeferredValue specifically
  • r

    rough-minister-58256

    05/04/2016, 10:55 PM
    seems fairly trivial to write our own helper unless there are other benefits to pulling in an extra dep - but likely modeled in kind
  • m

    mammoth-pharmacist-54215

    05/04/2016, 10:56 PM
    yup, writing something similar wouldn’t be hard.
  • e

    enough-analyst-54434

    05/04/2016, 10:57 PM
    It seems to me learning to read for 
    ', arg
    vs 
    ' % arg
    in the context of log statements is not a big burden
  • e

    enough-analyst-54434

    05/04/2016, 10:57 PM
    Its the stdlib
  • e

    enough-analyst-54434

    05/04/2016, 10:57 PM
    and not exotic like NotImplemented - the type, not the Exception weird 
    __eq__
    thing
  • w

    witty-crayon-22786

    05/04/2016, 10:58 PM
    yea, i really don't care either way. that's the api of the stdlib logger, so that seems to make it clearcut.
  • e

    enough-analyst-54434

    05/04/2016, 11:02 PM
    I appreciate the position of trying to swim instead of sink, but has anyone been spammed by Wiliam Morgan / Bouyant / linkerd?
  • e

    enough-analyst-54434

    05/04/2016, 11:02 PM
    Just got blasts on aurora and mesos dev lists with ~ads
  • e

    enough-analyst-54434

    05/04/2016, 11:02 PM
    We should never do this imo!
  • r

    rough-minister-58256

    05/04/2016, 11:02 PM
    completely fine w/ that and agree its not a big burden other than perhaps tickling folks OCD tendencies - so as a happy medium, a wrapper may be helpful to smooth controversial opinions if nothing else
  • e

    enough-analyst-54434

    05/04/2016, 11:03 PM
    OMG no
  • e

    enough-analyst-54434

    05/04/2016, 11:03 PM
    We waste time on stupid stuff
1...124812491250...2420Latest