Follow-up to 3c678df942b9322bfda5eb47d428b1c692ac0bf4
Can’t we just run the fancy regex when tokenising the search term, the fancy regex is never needed when indexing cooked
That would result in a change in behavior because the previous regexp will handle URLs within quotes as well.
I think that eating a=b for
“http://bob.com?a=b” is kind of unintentional, it is an oddity
It is intentional because we don’t index the query params in the search data so we need to strip it from the search term as well. Otherwise it doesn’t match.
Let’s merge this in then we can refine later