On Tue, Feb 16, 2016 at 11:44 AM, James K. Lowden <jklowden at schemamania.org> wrote:
> On Mon, 15 Feb 2016 14:19:12 -0700 > Scott Robison <scott at casaderobison.com> wrote: > > > Each job will take some amount of time to process. The order doesn't > > matter as long as all jobs are eventually processed and you have a > > single process running the jobs. Limit 1 is a reasonable way to grab > > a single job. > > Reasonable, perhaps, but not logical. The logical approach is to use a > feature in the data to select the "single job". One obvious way in > your example would be to use min(jobid) or somesuch. > I have a hard time seeing into the future and understanding the implications of every problem statement in the world to know what the best solution would be to each and every one of them, so you could be right. :) Certainly there are algorithms in this world that depend on randomness as a feature. That being said, my self diagnosed OCD (my particular form is spelled COD, because it should be as symmetrical as possible) would probably require that I pick a very specific deterministic row. Not necessarily because it was right, but so that I could sleep at night. -- Scott Robison
