Jimmy Lin

I agree, this should be fixed, but it's a question of priorities... An additional consideration is that this fix will make a bunch of papers already published - both by...

Sorry, to clarify - "massive regression impact" means that all the regression numbers for _every_ collection will change (we now have 25 different collections that we have regressions for)... but...

Okay, here are the results, on Robust04: AP | Paper 1 | Paper 2 | :------------------|---------|---------| BM25+RM3 (default) | 0.2903 | 0.2903 | BM25+RM3 (default): fixed | 0.2920 | 0.2920...

Okay, I'll refactor and send a pull request... this will help me get my feet wet and learn the API, etc.

@t-k- Just to confirm - the current decision is _not_ to merge this to master for now. cc @MXueguang

I've converted it into a "draft" so it doesn't get merged accidentally. Thanks!

Agreed.

@MFajcik thanks for filing this! It seems like a bug, likely because `AnalyzerUtils.analyze` is not thread safe? We'll look into it, but can't guarantee when we'll get to it...

Hi @MFajcik - I'm working on a potential fix here: https://github.com/castorini/anserini/pull/1992

Merged into Anserini main trunk. Should work in Pyserini now - TODO: add test case.