+1 to ban them, considering we also do not have any regular performance testing as part of the project test suites. Jacek has a good point about the checkstyle - there is separate one for tests. Though I would not object if people want them banned from tests too.
I guess if we ban them this would mean - no new additions and fix only places where there is a clear problem? On Sun, 2 Jun 2024 at 14:03, Enrico Olivelli <eolive...@gmail.com> wrote: > +1 (from the peanuts gallery) > > Removing streams from anything that looks like an hot path is indeed a > good thing. > > Please balance with 'don't fix things that aren't broken'. > > While doing such changes seems a great idea, sometimes it may have side > effects that you don't see until you run on real datasets. > > Enrico > > > Il Dom 2 Giu 2024, 19:38 J. D. Jordan <jeremiah.jor...@gmail.com> ha > scritto: > >> +1 agree with all this. Also fine to just use in tests or ban completely. >> >> On Jun 2, 2024, at 11:58 AM, Jake Luciani <j...@apache.org> wrote: >> >> >> >> +1 Java streams cause perf issues in hot paths. Its fine for tests and >> slow paths. But for clairity its fine to ban it as well if the majority >> agrees. >> >> >> On Sun, Jun 2, 2024 at 12:34 PM Sam <samueldlightf...@gmail.com> wrote: >> >>> I completely agree with this. I suspect there is a lot of low-hanging >>> fruit with easy perf wins. >>> >>> On profiling a 90% write workload I found >>> StorageProxy::updateCoordinatorWriteLatencyTableMetric to be a hot-path, >>> consuming between 15-20% of ModificationStatement::executeWithoutCondition >>> cycles. >>> >>> https://github.com/apache/cassandra/pull/3344 >>> <image.png> >>> >>> >>> On Fri, 31 May 2024 at 21:19, Jacek Lewandowski < >>> lewandowski.ja...@gmail.com> wrote: >>> >>>> Usages of them in tests are ok I think. We have a separate checkstyle >>>> file for the test code. >>>> >>>> - - -- --- ----- -------- ------------- >>>> Jacek Lewandowski >>>> >>>> >>>> pt., 31 maj 2024 o 19:14 David Capwell <dcapw...@apple.com> napisał(a): >>>> >>>>> I am cool for forbidding with a callout that tests are ok. I am cool >>>>> with forbidding in tests as well, but thats just for consistency reasons >>>>> than anything. >>>>> >>>>> On May 31, 2024, at 8:12 AM, Brandon Williams <dri...@gmail.com> >>>>> wrote: >>>>> >>>>> >>>>> On Fri, May 31, 2024 at 9:35 AM Abe Ratnofsky <a...@aber.io> wrote: >>>>> >>>>>> +1 to forbidding Stream usage entirely; the convenience of using them >>>>>> outside of hot paths is less than the burden of figuring out whether or >>>>>> not >>>>>> a particular path is hot. >>>>>> >>>>> >>>>> I think I have most frequently appreciated them in tests, which I >>>>> think we could except, since these are categorically not in the hot path. >>>>> >>>>> Kind Regards, >>>>> Brandon >>>>> >>>>> >>>>> >>>>>
