Louis Dureuil

hello @camgaertner :wave: can you retry with [`--experimental-reduce-indexing-memory-usage`](https://github.com/orgs/meilisearch/discussions/652)? Does that change the result?

Sure, thank you for the initial test. If you can confirm with more data that `--experimental-reduce-indexing-memory-usage` with improves the situation, it will point us in the right direction

I'm surprised that the new indexer would use more RAM as the previous one. Do you happen to have many threads? If using `--max-indexing-memory`, Meilisearch uses **10%** of the provided...

Hello @justinseesdragons 👋 We have known that the situation regarding dump import is less than ideal, we haven't been dedicating much time to fixing the issue as it is, as...

Thanks for reporting, both you and @zperzendetta I reproduced the bug with `movies.json` in the default settings and the two queries: ```jsonc { "q": "spring", "showRankingScore": true, "rankingScoreThreshold": 0.96 }...

hey @teodorstupnicki , the logic to compute the ranking score and decide whether to keep the document is in https://github.com/meilisearch/meilisearch/blob/main/milli/src/search/new/bucket_sort.rs#L240. Surely there's an off-by-one in the number of ranking score...

Hello, our [BENCHMARKS.md](https://github.com/meilisearch/meilisearch/blob/main/BENCHMARKS.md) file might help with that, but to troubleshoot this issue you mostly need the movies.json *dataset* rather than *workload* (dataset is the data, here the movie descriptions...

hello and thanks for the investigation @Gnosnay, indeed that one is a bit on the harder side of "good first issue" 😳 as you might have found, the logic happens...

Hello, thank you again for the time you put into investigating 👍 I'll try to answer your questions: - `bucket_sort` is about applying the ranking rules in the order they...

Closing this as it was fixed in v1.14 👍