Postgres function to turn regexp_matches SETOF result into an ARRAY - postgresql

Question

I am having a go at pgSQL for the first time.
Many thanks in advance indeed - Eugen
The problem I like to solve
I like to return an array of words with a minimum length from a text. Picked regexp_matches which returns SETOF text[] (not array). There may be better ways but this is what I picked.
regexp_matches ( string text, pattern text [, flags text ] ) → setof text[]
Postgres Reference for String functins
The Issue
When trying to use individual results from the resut set (seemingly arrays) they are in fact of type text but include the curly brackets of a Postgres array.
Example
Functionally the below example works because I added a workaround removing curly brackets with the trim function.
The output for the function described below
# select lower_words_arr('there is an answer','\w{3,}') AS words;
NOTICE: match = {there}
NOTICE: match = {answer}
words
----------------
{there,answer}
(1 row)
My Question
Why come, that individual regexp_matches entries match and return data as "{txt}" but behave as type text and not ARRAY?
Is there a way without my workaround?
Code
CREATE OR REPLACE FUNCTION lower_words_arr(input_str text, match_expr text) RETURNS text[] AS $$
DECLARE
words_arr text[];
one_match text;
BEGIN
FOR one_match IN SELECT regexp_matches(lower(input_str), match_expr, 'g')
AS match
LOOP
RAISE NOTICE 'match = %', one_match;
-- But if I write:
--
-- RAISE NOTICE 'match = %', one_match[0];
--
-- fails with:
---
-- ERROR: cannot subscript type text because it is not an array
-- CONTEXT: SQL statement "SELECT one_match[0]"
-- PL/pgSQL function lower_words_arr(text,text) line 10 at RAISE
--
-- even though this the result returned
---
-- NOTICE: match = {there}
-- NOTICE: match = {answewr}
words_arr := array_append(words_arr,trim(BOTH '{}' FROM one_match));
END LOOP;
RETURN words_arr;
END;
$$ LANGUAGE plpgsql;

Answer 1

I have now solved the issue without the trim workaround. The solution came right from the Postgres documentation. Look for SELECT (regexp_match('foobarbequebaz', 'bar.*que'))[1]; on that page. With that knowledge in hand I changed the function to
Improved Solution
CREATE OR REPLACE FUNCTION lower_words_arr(input_str text, match_expr text) RETURNS text[] AS $$
DECLARE
words_arr text[];
one_match text;
BEGIN
FOR one_match IN SELECT(regexp_matches(lower(input_str), match_expr, 'g'))[1] AS match
LOOP
RAISE NOTICE 'match = %', one_match;
words_arr := array_append(words_arr,one_match);
END LOOP;
RETURN words_arr;
END;
$$ LANGUAGE plpgsql STABLE;
Output
select lower_words_arr('there is an answer - you will find it in the Postgres documentation.','\w{4,}') AS words;
NOTICE: match = there
NOTICE: match = answer
NOTICE: match = will
NOTICE: match = find
NOTICE: match = postgres
NOTICE: match = documentation
words
-------------------------------------------------
{there,answer,will,find,postgres,documentation}
(1 row)
Conclusion
My original problem has been solved, but a puzzling piece understanding is still missing (not subject of my initial question):
Why does the one_match variable not behave like an array when [1] is removed from the SELECT?
Maybe I should open a question specific to that behaviour ...