Autocomplete is incomplete

[TranslatorIssueCreator]

Type: Bug Report

URL: https://ui.test.transltr.io/results?l=Sexual%20Deviancy&t=0&q=93d8fc12-ff27-4c44-a688-89708a9a9b30

ARS PK: 93d8fc12-ff27-4c44-a688-89708a9a9b30

Steps to reproduce:

Type "Sex" into the search bar - Using the what drugs may treat disease query. 5 options are returned, including one that starts with "sexual". However, if you type "sexual" into the search bar, many more options are returned. Seems like all of the "sexual *" options should be returned when just "sex" is entered.

Screenshots:

[dnsmith124]

This is a quirk of how we have to filter out non-relevant terms because there's no type selection available right now when we hit the NR or NN. Once we are able to request only specific biolink types from the NR this will no longer be an issue.

[codewarrior2000]

Taking a contrarian perspective, what kind of autocomplete do we expect when someone types "Se" into the search bar?

Edit:
My point is that fuzzy search (or fuzzy matching) in one form or another has been around for 20 years and users would be able to recognize this current search bar behavior as being something familiar and acceptable.

[gaurav]

Andy:

• Caused by lack of Biolink type information -- got 1000 results from NameRes, filtered out 999 drugs and only showed the one disease. As you get more specific, you start pushing out the drugs, and suddenly more results appear.
• Would like Biolink type filtering and CURIE type (e.g. MONDO/HP for disease) -- will not be as noticeable when that happens
• Simpler/quicker fix: max = 10,000
• Having these examples would be great for testing

Gaurav: we want to do improved testing one-character-at-a-time.

• Biolink type filtering: fixing one last bug
  • Andy: higher priority that CURIE filtering
  • David: it's going to solve 95% of the problem
• CURIE type (e.g. MONDO/HP for disease): @gaurav to add to NR and see if we can deploy this really quickly

Responding to @codewarrior2000:

• Andy: we considered starting autocomplete at 3-4 letters, but there is a 2-letter gene name that we need to worry about (AR?)

[sstemann]

this behavior described at the start of this ticket is still happening. is there something that should change?

[gprice1129]

I can't reproduce this behavior in the production environment @sstemann @dnsmith124

[dnsmith124]

I've retested as well and can't seem to either

[gprice1129]

@sierra-moxon do we need a third-party tester to decide if this is closeable? I would be more comfortable if we had someone outside of our team test.

[gaurav]

I think we've tested this enough that we're pretty sure it's fixed, so I'm going to close it. Please reopen if I missed something or if it fails testing at some point in the future.