CHAR_LENGTH for postgres

[pruzko]

I've added support for CHAR_LEN[GTH] for postgres. I've also reused the concept of exp.Length(this='x', binary=True) from bigquery/snowflake (though I don't think is ever used in those dialects).

[VaggelisD]

Thanks for the PR!

Could you share concrete examples in which CHAR_LENGTH and LENGTH differ in Postgres? The psql docs specify that LENGTH returns the number of characters as does CHAR_LENGTH, in contrast to e.g MySQL:

mysql> SELECT LENGTH('€'), CHAR_LENGTH('€');
3 1

psql> SELECT LENGTH('€'), LENGTH('€', 'UTF8'), CHAR_LENGTH('€');
1 1 1

[pruzko]

OOPS! You're right, I messed something up in my notes. It is now fixed.

Also, parse_one('SELECT LENGTH(x)', 'mysql').sq('mysql') == 'SELECT LENGTH(x)' as it should and parse_one('SELECT LENGTH(x)', 'sqlite').sql('mysql') == 'SELECT CHAR_LENGTH(x)' as it also should.

[VaggelisD]

1. Do note that Postgres supports an encoding arg in LENGTH, currently this will be dropped afaict. If you want, you can extend exp.Length to also store that
2. Could we add a couple Postgres tests as well? They'd look like:

self.validate_identity("LENGTH(x)")
self.validate_identity("LENGTH(x, y)")
self.validate_identity("CHAR_LENGTH(x)", "LENGTH(x)")
self.validate_identity("CHAR_LEN(x)", "LENGTH(x)")

With these we should cover all Postgres roundtrips (the 2nd test requires bullet 1)

[VaggelisD]

Should this also be removed? Can't find LEN on Postgres's docs, and even if it exists it's probably an alias of LENGTH which is not binary (?)