I went through much of this a few weeks ago, and this article would have saved me quite a bit of time. Like others here, siglen is new to me, so I'll have to have a play around with that.

One thing that bothered me was how cumbersome the queries become once you start having to concatenate and coalesce a whole bunch of columns together. Postgres has a variadic function called "concat_ws" which ought to be useful here, especially as it implicitly coalesces NULL inputs to empty string. However, it's not declared to be immutable, so can't be used in an expression index.

The solution (from Erwin Brandstetter's answer on Stack Overflow) is to use an immutable variant of this, constrained to only handle data of type "text".

https://stackoverflow.com/questions/54372666/create-an-immut...

With this immutable concat function in place, and a trigram expression index making use of it, the WHERE clauses become a lot more concise, for example, the clause in the final query of the article reduces to just:

  WHERE immutable_concat_ws(' ', asin, reviewer_id, reviewer_name, summary) ilike '%' || input.q || '%'