Jimmy Lin

Captured discussion over Slack w/ @sahel-sh ... still unresolved points. This seems uncontroversial: ``` def rerank(self, query: Query, candidates: Candidates, k: int = 10) -> Candidates: ``` Note that `Candidates`...

I guess I'm hung up over naming of `Result`... it's really a `Tuple[Query, List[Candidates]]`: Let me propose S4 then? ``` def rerank(self, query: Query, candidates: Candidates, k: int = 10)...

S5: another strategy - use Python's ability to return multiple values: ``` def rerank(self, query: Query, candidates: List[Candidate], k: int = 10) -> (List[Candidate], ExecutionTraceObject) def rerank_batch(self, input: List[Tuple[Query, List[Candidate]]],...

Okay, after further thoughts, S7: ``` def rerank(self, query: Query, candidates: List[Candidate], k: int = 10) -> Result def rerank(self, request: Request, k: int = 10) -> Result def rerank_batch(self,...

> I also like s7. It follows many best practices like separation of the input param type and output param (even if they are very similar now (they might diverge...

Two more issues to sort out: 1. JSON format. 2. Anserini/Pyserini command-line options. --- For (2), in Anserini, `SearchCollection` currently takes `-output` to generate a standard TREC runfile. So I...

Instructions here! https://github.com/castorini/anserini/blob/master/docs/fatjar-regressions-v0.35.1.md

I've merged in #2014 - the merged in code produces exactly the same results as above, so whatever happened isn't due to that PR.

I've changed the 2CR values in #2143

All 2CRs passed on `tuna`.