postgres LIKE % operator vs similarity - pg_trgm.similarity_threshold . different query plans

If I use:

SET pg_trgm.similarity_threshold = 0.9;
... where column % 'some string s';

PostgreSQL does an index scan on: gin (column gin_trgm_ops)

In contrast, which should be the same:

... where similarity(column, 'some string s' ) >= 0.9

That does a seqential scan instead of using the index.

From the documentation: https://www.postgresql.org/docs/11/pgtrgm.html

text % text boolean
Returns true if its arguments have a similarity that is greater than the current similarity threshold set by pg_trgm.similarity_threshold.

Why?

Fundamentally, a WHERE condition must look like this to allow an index scan:

<indexed expression> <operator> <constant>

Here, <indexed expression> is what you created the index on <operator> is an operator supported by the operator family of the index's operator class, and <constant> is an expression that is constant for the duration of the index scan (in particular, it can only contain STABLE functions).

The one exception in PostgreSQL is if you use a function returning boolean in the WHERE condition, and that function has an optimizer support function that allows PostgreSQL to substitute an index scan. But similarity is no such function.