How to search for username with hyphen

malle

Hello,
assume having two users
test-good
test-better
Seemingly the hyphen ist like a special char and won't work with asterisk for like-search.
You can try this in the web UI:
Search for username:test -> Users will be found
Search for username:test* -> Users will be found
Search for username:test-* -> No search result
Quoting the hyphen with backslash has no effect. Is there a way to achieve this?

Thank you and regards

dan

Hmmm.

Which search engine are you using, elasticsearch or the database search engine?

What version of FusionAuth are you using?

malle

Hey @dan
this seemingly occurs with V1.23.3 and also with V1.17.4.
And with both search engines, database and elasticsearch.

robotdan

@malle said in How to search for username with hyphen:

Search for username:test-* -> No search result

Does this work?

Search for username:"test-*"

dan

Hi @malle ,

Took some time to dig into this a bit more.

I have a system which has a user with a username test2-name.

If I use the database search for this user, all of these return the user:

test2
test2*
test2-
test2-*
test2-n
test2-n*

Note that you don't put the username: prefix on any database searches.

If on the other hand I use the elasticsearch search engine, and search for the same user, I see the following:

username:test2 -> found
username:test2* -> found
username:test2- -> found
username:test2-* -> not found
username:test2-n -> found
username:test2-n* -> not found

I'm not quite sure what's going on with the wildcard causing Elastic to not find the value, but don't think it is related to the -.

robotdan

@dan

Did you try quoting the value?

username:"test2-*"
username:"test2-n*"

dan

Running with quotes works in one case:

username:"test2-*" -> found
username:"test2-n*" -> not found

malle

@dan
username:"test2-n*" -> not found ...so still a problem.
And I do not really care about the FusionAuth Web UI, the problem actually is that I use the Java client via UserSearchCriteria ...so it is quoted anyways. Unfortunetally there are already a lot of 'hyphen-user' in the system and adjusting these by taking the hyphen out is not an option.

robotdan

I think the problem is that we are indexing username with the default text analyzer which uses a hyphen (-) as a separating character.

This means Elasticsearch has indexed test-good and found two tokens test and good.

Without changing the way this field is indexed, I think you can achieve your desired results by modifying your query.

For example, to search for test-g*, you would break this into two parts, and structure your search like this: username:"test" AND username:g*. This will find all usernames that include the token test and a have a second token token that begins with g.

Try this strategy and see if it works for you.

malle

In the meantime we could implemented it with the mentioned workaround.
Late thanks @robotdan !