Hey there! I have a field that hold "references/codes" for the items in my collection. Such as: • "2011_PL24-0015" • "AS-19/2024 - AJT/65177/2022" Those should be treated as whole words, including spaces and symbols. I'm performing the search through many fields at the same time, including this and a "description" one. I can't find this items by reference, since Typesense

q

won't treat them as single words (as expected). Is there a way to workaround this for only for one field of the

query_by

? Or any other solution?

By default

_

and

-

are ignored, so

2011_PL24-0015

is actually treated as

2011PL240015

-- with space it's not possible to ignore it as a word separator.

And doind something through filters? Or another symbol?

You can replace space with some other character then everything will be treated as a single word.

Is there a way not to ignore symbols?

Yes, check

symbols_to_index

parameter in the schema.

Thanks! That was it, I'm blind.

Yes same format. And it's a global parameter, not possible to apply it per field.

Damn, I only need it for one field 😅 My use case is, I only want to apply it to the "reference" but not to the "description" as the later can also contain that symbols but they should be treated as spaces rather as characters.

Haha I know, it's annoying. One of those many things we want to fix.

It's fine! Do you think that a left join would work without a lot of performance impact?

For small collections won't be a problem. Say upto few hundred k docs

😅 4M+ and growing. So I guess joining a 4M+ with another 4M+ won't be a good idea

The thing is it might still work but case to case basis. With joins we are splitting the docs to make them smaller to it might help to reduce IO.

Moving both fields from schema to the particular field

You mean the symbols configuration?

and the token_separator

I've to check the complexity. I'll get back to you..

No hurry, if it's not even on your roadmap, we'll look for workarounds

The work around will be to do a multi search with the modified version of query and then to merge results.

Naah, that would impact the ranking and sorting of stuff

Yeah