Postgres pg_trgm how to compare similarity for array of strings - sql

Question

I'm attempting to use pg_trgm for string fuzzy matching and I know it may be used like this:
SELECT * FROM artists WHERE SIMILARITY(name, 'Claud Monay') > 0.4;
where a scalar value may be used to compare against the similarity. However, I've seen this way of using SIMILARITY with an array of strings:
SELECT * FROM artists WHERE 'Cadinsky' % ANY(STRING_TO_ARRAY(name, ' '));
which uses the % operator which is a shorthand for comparing against the default value of 0.3. I'm trying to find the proper syntax to use ANY(STRING_TO_ARRAY(...)) but with the first form where an arbitrary scalar value may be given to compare the similarity against.
This is, most likely, just a simple question of properly using the syntax for ANY, but I'm failing at understanding what the correct form is.

Answer 1

There is no syntax to use ANY with 3 arguments (the string, the array of strings, and the similarity threshold). The way to do it is to set pg_trgm.similarity_threshold to the value you want rather than the default of 0.3, and then use % ANY.
If you want to use different thresholds in different parts of the query, you are out of luck with the ANY construct.
You can always define your own function, but you will probably not be able to get it to use an index.
create or replace function most_similar(text, text[]) returns double precision
language sql as $$
select max(similarity($1,x)) from unnest($2) f(x)
$$;
SELECT * FROM artists WHERE most_similar('Cadinsky', STRING_TO_ARRAY(name, ' '))>0.4;

Answer 2

I am not a DB expert nor good at SQL but here is my solution.
I basically use a function called unnest(). Thus, I can iterate over the array and check the similarity value for each item then compare it to similarity input, which is a float.
Using something like set pg_trgm.similarity_threshold=0.6; is a global setting as far as I know. The question is specifically asking for an explicit threshold.
Also, if you create a function to do the job and the function is not VOLATILE but is STABLE, you cannot use set pg_trgm.similarity_threshold. (At least that was what happened to me).
Caution: I didn't compare my approach to (ANY) approach in terms of performance.
Example Code:
CREATE OR REPLACE FUNCTION your_function_name (input text, similarity float) RETURNS
SELECT * FROM your_table_name
WHERE EXISTS
(SELECT
FROM unnest(ARRAY['item','anotherItem', 'third-ish']) element
WHERE SIMILARITY (input, element) > similarity
);
$ function $