[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [rdf4j-dev] package naming issue in MemoryBenchmark?
- From: "HÃvard M. Ottestad" <hmottestad@xxxxxxxxx>
- Date: Sun, 6 May 2018 11:05:54 +0200
- Delivered-to: firstname.lastname@example.org
I can push a fix.
Junit only runs tests that are in classes named something with test and methods annotated with @test. The benchmarks arenât run as part of the tests, and I donât see much benefit in them doing so either.
I would love a system that would do regression testing with benchmarks, would be awesome to get continuous feedback on how our code affects performance. Havenât found anything that does this though.
> On 6 May 2018, at 03:29, Jeen Broekstra <jeen.broekstra@xxxxxxxxx> wrote:
>> On 01/05/18 20:20, "HÃvard M. Ottestad" wrote:
>> Hi Jeen,
>> I wrote that code. The package should be fixed. Copy paste mistake.
> No worries, could you perhaps push a quick fix for that package name?
> What I still don't understand is why neither Jenkins nor Travis complains about this. Is this class somehow ignored when it runs the unit tests?
>> Iâve found that putting the benchmarks in the test directory makes it
>> easier to run the benchmarks.
> The problem I see is that they really _shouldn't_ be run as part of every unit test run. Is there a switch or option we can use to turn them on or off?
>> Iâve also had trouble with having
>> multiple repos and getting changes to become visible across modules,
>> e.g. I tried to fix a performance issue in the query parser and mvn
>> install wouldnât propagate it to the storage repo where my benchmark
> To be honest I've had some issues along those lines as well. I haven't quite gotten the multiple-repo setup working smoothly for me; there's a lot of forced refresh/rebuild going on. I'll try and figure if that's a a config issue on my end or if there's something we can tweak in the poms or the snapshot repo to make this easier to handle.
>> This benchmark here I developed when I discovered that every insert
>> triggers a get from the underlying sail. This is meant to be used for
>> the notifying sail code. However, even when there were no attached
>> sails to notify, it would still do the get. This made inserts slow,
>> especially for serialisable.
> Good find. Don't get me wrong I think having the benchmarks available is incredibly valuable. I'm just a bit wary of mixing them in with unit tests.
> rdf4j-dev mailing list
> To change your delivery options, retrieve your password, or unsubscribe from this list, visit