Lucene's test framework makes heavy use of randomization in order to explore more of the vast space of possible states. You might be familiar with this as "fuzz testing"? There's a blog post about it here (from 2011!) https://blog.mikemccandless.com/2011/03/your-test-cases-should-sometimes-fail.html and an entire project spun out from Lucene (I think that's how it started anyway) in order to support this: https://github.com/randomizedtesting/randomizedtesting/wiki - you can even watch a video about it (linked from that github page)
On Thu, Sep 26, 2024 at 12:36 PM Seunghan Jung <ajtwlstmd...@gmail.com> wrote: > > Hello Lucene Developers, > > > I’m writing because I have a question regarding the Tokenizer-related test > code. > > > I was looking at the following code in the link below: > > https://github.com/apache/lucene/blob/7b4b0238d7048a0f8532ce55afb72f89dfd69b1c/lucene/test-framework/src/java/org/apache/lucene/tests/analysis/BaseTokenStreamTestCase.java#L1547-L1558 > > > I noticed that the `newAttributeFactory` method, which is used for Tokenizer > testing, contains a random element. I was wondering why randomness was > introduced here. > > From my understanding, random elements should not be included in tests > because they can produce different results on multiple test runs. Was there a > specific reason for this? > > If anyone is familiar with the history of this, I’d really appreciate your > insight. > > > Thank you. > > --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
