mappings: add ngram analyzer

[jrcastro2]

• Allows partial matches on search
• closes Some users cannot be found when inviting or setting restrictions CERNDocumentServer/cds-rdm#114
• closes Some users cannot be found when inviting or setting restrictions invenio-app-rdm#2752

[carlinmack]

does this need rebasing? it seems like v2.0.0 was edited by Alex a few days after this PR https://github.com/inveniosoftware/invenio-users-resources/blob/master/invenio_users_resources/records/mappings/os-v1/users/user-v2.0.0.json

[carlinmack]

Cases which are failing:

query	matching usernames	my hypothesis
preferences.visibility:restricted	['pubres']	res
[email protected]	['pub', 'pubres']	inveniosoftware (from user email) and res
[email protected]	['pubres', 'pub']	pub
email:[email protected]	['pub']	inveniosoftware (from user email)
email:[email protected]	['pub']	inveniosoftware (from user email)
preferences.email_visibility:restricted	['pubres']	res
invalid:test	['pub']	inv

[carlinmack]

The tests that were failing were searches which should produce no results. They were failing because the n-gram analyser was taking n-grams of the word (pubres -> pub, ... res) and matching on the query. We have swapped to an edge_ngram tokeniser with min 2 max 20.

This means it will match from 2 characters up to 20 from the beginning of the token. This won't have a huge impact on the number of tokens (compared with n-gram) as it is dependant on the length of the token. If we had 2-5 on edge_ngram then it would not produce tokens for 6+ letter words

We also had to change two of the test cases to be specific matches

All the tests now pass as they do not match on the search queries

[carlinmack]

to be discussed https://opensearch.org/docs/latest/analyzers/#built-in-analyzers

[jrcastro2]

LGTM! 🚀

[carlinmack]

New test case fails on current master

but passes with PR