Conversation
| let lowercase = lowercase.unwrap_or(true); | ||
|
|
||
| let language = language.unwrap_or_else(|| DEFAULT_LANGUAGE.to_string()); | ||
|
|
||
| let stemmer = match stemmer { | ||
| None => Stemmer::try_default_from_language(&language), | ||
| Some(stemmer_algorithm) => Some(Stemmer::from_algorithm(&stemmer_algorithm)), | ||
| }; | ||
|
|
||
| let stopwords_config = match stopwords { | ||
| None => { | ||
| // Try to create from the language | ||
| Language::from_str(&language) | ||
| .ok() | ||
| .map(StopwordsInterface::Language) | ||
| } | ||
| Some(stopwords_interface) => Some(stopwords_interface), | ||
| }; |
| fn hash<H: Hasher>(&self, state: &mut H) { | ||
| let options = match self { | ||
| DocumentOptions::Common(options) => Cow::Borrowed(options), | ||
| DocumentOptions::Bm25(bm25) => Cow::Owned(bm25.to_options()), |
There was a problem hiding this comment.
Don't expect Bm25 branch to ever be used, as we don't deserialize into Bm25
This comment was marked as resolved.
This comment was marked as resolved.
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
| }; | ||
|
|
||
| // Order of keys in the map should not affect the hash | ||
| let mut keys: Vec<_> = options.keys().collect(); |
There was a problem hiding this comment.
Can we maybe just collect the options into a BTreeMap instead? I think this custom hashing approach might cause hard to detect bugs in future.
Alternatively we could just replace the initial HashMap with a BTreeMap, the APIs should be the same.
There was a problem hiding this comment.
How about this?
// Order of keys in the map should not affect the hash
let mut keys: Vec<(_, _)> = options.iter().collect();
keys.sort_by_key(|(key, _)| *key);
keys.hash(state);I'm not too worried about it because it does not have to be portable.
There was a problem hiding this comment.
I'll merge this now so I can roll the release. We can handle this one separately if we like.
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
* make bm25 parameters compatible with default fastembed params * bm25 params in openapi schema * fmt * improve bm25 config deserialization * more explicit docstring for disabling english
languageparam, which is used to create default stopwords and stemmer