Finding nth occurrence of a pattern within a string in SQL (Presto) - regex

Question

I am writing a query in Presto SQL using the function regexp_extract
I have a string that may look like the following examples:
'1A2B2C3D3E'
'1A1B2C2D3E'
'1A2B1C2D2E'
What I'm trying to do is find for example the second occurrence of 1[A-E].
If I try
regexp_extract(col, '(1[A-E])(1[A-E])', 2)
This will work for the second example (and the first since it returns nothing since there is no second occurence). However, this will fail for the third example. It returns nothing. I know that is because my regex is searching for a 1[A-E] followed directly by another 1[A-E].
So then I tried
regexp_extract(col, '(1[A-E])(.*)(1[A-E])', 3)
But this does not work either. I am not sure how I can account for the fact that I may have 1A1B2C or 1A2B1C to find that second 1. Any help?

Answer 1

Your second pattern does work in the latest version of Trino (formerly known as Presto SQL):
WITH t(col) AS (
VALUES
'1A2B2C3D3E',
'1A1B2C2D3E',
'1A2B1C2D2E')
SELECT regexp_extract(col, '(1[A-E])(.*)(1[A-E])', 3)
FROM t
_col0
-------
NULL
1B
1C
(3 rows)
As others have commented, you don't need the capture groups for the first match or for the .*, and you should use the lazy quantifier to avoid .* eagerly matching all characters between the first and last occurrence:
WITH t(col) AS (
VALUES
'1A2B2C3D3E',
'1A1B2C2D3E',
'1A2B1C2D2E',
'1A2B1C2D1E')
SELECT regexp_extract(col, '1[A-E].*?(1[A-E])', 1)
FROM t
_col0
-------
NULL
1B
1C
1C
(4 rows)

Answer 2

You don't need the second capture group (.*) to keep the 2 capture groups in the result, and you can optionally match the allowed characters in between.
From what I read on this page you might also consider using regexp_extract_all to get all the matches, as regexp_extract returns the first match.
As the example data consists of a digit followed by a char A-E, you could exclude matching the 1 from the character class to prevent overmatching and backtracking.
(1[A-E])[02-9A-E]*(1[A-E])
Regex demo
If using a single capture group to get the second value is also ok, you can use
1[A-E][02-9A-E]*(1[A-E])
Regex demo