Possible to do a fuzzy-match LIKE query? about activestash HOT 2 CLOSED

Hi @timburgan

ActiveStash supports fuzzy string matching. It all comes down to how the indexing is configured in your model.

So to answer your question: yes that is what the example you provided does. For the gory details on how it works, keep reading :)

By default, using stash_index on a string field in your model, e.g. stash_index :email, will allow you to perform exact, range and match queries on that field. The index type "match" is what provides the fuzzy matching.

The text filtering and tokenisation that happens behind the scenes looks like this:

"[email protected]" -> downcase -> tokenise (split on whitespace and special chars *) -> ngram (with token length of 3)

(edit: It's just splitting on whitespace right now)

When saving the record for "[email protected]" you'd end up with something like the following terms in the CipherStash index for the email field on your Users collection:

• jen
• exa
• xam
• amp
• mpl
• net

When querying, the same text tokenisation pipeline is applied to your query terms (i.e. the right hand side of the =~ operator), and if any tokens generated from the query term match any of the tokens from the email index it will return the record.

Text tokenisation pipelines are not yet configurable in ActiveStash but that is on our roadmap.

All terms stored in the CipherStash index and terms produced by a query are of course encrypted and CipherStash cannot see them.

Note that matching using a regular expression is not supported. We take your point that (ab)using the =~ operator may be misleading. I think that the least we can do is provide a sensible error message if someone tries to use it with a regular expression.

Edit: spelling

from activestash.

🥇 love it, thank you @freshtonic

from activestash.