Fuzzy queries

[curquiza]

Because of the GitHub bug regarding discussion transfer, I create this discussion so that the product team can take into account the suggestions.

meilisearch/meilisearch#1096

[gmourier]

@eikooc, sorry to ping you here but we have transferred the discussion here!

It may be possible to configure the fuzziness behavior to your need if we expose a way to configure the typo behavior since it seems to me that's two words (fuzziness and typo) for the same thing.

E.g Configure the minimum word size for 1 typo / Configure the minimum word size for 2 typos

I imagine that it will have an impact on the performance depending on the configuration. Meilisearch is limited to 2 typos now to favor speed and relevancy for most of the use-case. May not need to have more typos allowed if you can set the trigger to a configurable minimum size of words 🤔

[eikooc]

It's true that you can get some of the behavior by increasing the typo tolerance, but not exactly.

Fuzziness would match only one in this example:

// Documents: [{title: batman},{title: ant}]
// Query: "btm"

But increased typo tolerance would match them both.

[6543]

my usecase would be to specificly set fuzziness/typo to 0 so it only return exact (case insensitive) matches

(of course fuzzy search should still be possible on that index, so per query setting)

PS: my workaround is to post-process the result and iterate over all hits and manually filter again :/

[6543]

(go-gitea/gitea#29671)

[macraig]

Hi @6543 , I've opened a new discussion based on your needs here