Romain Leroux

@maufl Yes good point, it might be better to use `StreamBuilder::with_state` instead. I guess I can give all this a try, at least until the last step I listed where...

@maufl FYI I implemented the steps I listed that bubble up `Automaton::State` in `AutomatonWeight::scorer` in [this branch](https://github.com/lerouxrgd/tantivy/tree/fst-state) following your advice. Now I guess the question is, how to score documents.

@maufl Yes, I was thinking of the same score. Actually the tricky part is still to get the automaton state because all we have in the API is `A: Automaton`...

Oh yes good idea, let's see what the maintainers think is best.

@fulmicoton I am not using it directly now, so as far as I am concerned it is fine to discard this PR. I see that correct scoring is indeed tricky,...

I usually find that [fake_clock](https://crates.io/crates/sn_fake_clock) is a nice solution if you can apply it to your use-case. This [blog post](https://klau.si/blog/mocking-in-rust-with-conditional-compilation/) is a good example of it.

I feel that having a `valid_paths` RegexSet (instead of an excluding one) that would be optional would be better ?

I see good points indeed. I was thinking that by excluding paths it might easier for someone on the web to spam your server with random paths and completely pollute...

Just FYI this should match your schema: ```rust #[derive(Debug, PartialEq, Clone, serde::Deserialize, serde::Serialize)] #[serde(default)] pub struct Updated { pub id: String, } impl Default for Updated { fn default() ->...

Thank you for porting those changes to `rsgen-avro` and for sharing your findings. As for the `serde` support on schemas it's a nice to have, but what I was really...