In the database, I have various alpha-numeric strings in the following format:
10_asdaasda
100_inkskabsjd
11_kancaascjas
45_aksndsialcn
22_dsdaskjca
100_skdn
The ideal way would be to normalize your data and split the two components of the column into two individual columns. One of type integer, one text.
With the current table, you can do something like demonstrated here:
WITH x(t) AS (
VALUES
('10_asdaasda')
,('100_inkskabsjd')
,('11_kancaascjas')
,('45_aksndsialcn')
,('22_dsdaskjca')
,('100_skdnascbka')
)
SELECT t
FROM x
ORDER BY (substring(t, '^[0-9]+'))::int -- cast to integer
,substring(t, '[^0-9_].*$') -- works as text
The same substring() expressions can be used to split the column.
The regular expressions are somewhat fault tolerant:
The first regex picks the longest numeric string from the left, NULL if no digits are found, so the cast to integer can't go wrong.
The second regex picks the rest of the string from the first character that is not a digit or '_'.
If the underscore is unambiguous as separator anyway, split_part() is faster:
ORDER BY (split_part(t, '_', 1)::int
,split_part(t, '_', 2)
SELECT name
FROM nametable
ORDER BY (split_part(name, '_', 1)::int
,split_part(name, '_', 2)