We were attempting to create a function with liquibase (using the redshift drivers), but I think this problem is more general to liqubase.
I submitted this bug back in May but haven’t seen even any acknowledgement of it? We had to hack around it by using CHR() instead of the unicode character itself.
This appears to definitely be a Liquibase issue because running updateSQL creates the bad result. I am using just plain old SQL files.
create or replace function test.fn_test () RETURNS VARCHAR IMMUTABLE AS $ SELECT '([^A-Za-zÀ-ÖØ-öø-ÿÂ-ẑĀ-ẕÃ-ᵶ''. -]+)'; $ LANGUAGE sql ;
create or replace function test.fn_test () RETURNS VARCHAR IMMUTABLE AS $ SELECT '([^A-Za-zÃ€-Ã–Ã˜-Ã¶Ã¸-Ã¿Ã‚-áº‘Ä€-áº•Ãƒ-áµ¶''. -]+)'; $ LANGUAGE sql;
Note that the file is definitely UTF-8.