Tag autocomplete fuzzy matches on both sides of input

[hammeiam]

Problem: Tags currently only autocomplete based on a fuzzy match after the user's input, eg SQL("INPUT%"). This leads to users needing to do a lot of unnecessary typing in order to find their desired tag. This is further compounded by the fact that many users employ an information hierarchy in their tags eg Location: USA: Ohio: Cleveland. This means that a user wastes a lot of time typing because they can't just autocomplete "Cleveland" to the appropriate set of tags.

Solution: Simply fuzzy match on both sides of the user input, eg SQL("%INPUT%"). Per dstillman's feedback, this change is behind the user-set flag unboundedTagSearch, which defaults to false.

[dstillman]

I mean, it's not a bug — it's left-bound because that's an indexed search, and otherwise it's not.

We'll need to do some performance tests in large libraries to see how it performs non-left-bound on modern computers. We likely haven't tested this in years, but historically searches were slow in large libraries with many tags. And if that's still the case, we'd only do this either 1) as a pref, defaulting off or 2) if we added a separate word index on tags.

[quinn-p-mchugh]

as a pref, defaulting off

I would love to see this implemented as a preference, defaulting to off. Let the users decide whether the performance hit is acceptable.

[jankap]

Would love to see this search happening :) Can I test something? I have a real-life lib with ~400 items

[hammeiam]

@dstillman I put this change behind a user flag so people can turn it on if they choose.

[quinn-p-mchugh]

Thanks for your work @hammeiam. Is this able to be reviewed/merged?

[hammeiam]

@quinn-p-mchugh @dstillman rebased changes against latest main

[hammeiam]

@dstillman Congrats on the great looking beta launch! I think it is the perfect opportunity to add this quick enhancement :)

[rossant]

I agree this would be a great enhancement, I hope it will be merged soon :)

[hammeiam]

Hi @dstillman - just removed conflicts, ready to merge whenever you are :)