Major Description
Original description SPARK-37507:
to_binary(expr, fmt) is a common function available in many other systems to provide a unified entry for string to binary data conversion, where fmt can be utf8, base64, hex and base2 (or whatever the reverse operation to_char()supports).
https://docs.aws.amazon.com/redshift/latest/dg/r_TO_VARBYTE.html
https://docs.snowflake.com/en/sql-reference/functions/to_binary.html
https://docs.teradata.com/r/kmuOwjp1zEYg98JsB8fu_A/etRo5aTAY9n5fUPjxSEynw
Related Spark functions: unbase64, unhex
Expected improvement:
- `base64` behaves more strictly, i.e. does not allow symbols not included in base64 dictionary (A-Za-z0-9+/) and verifies correct padding and symbol groups (see RFC 4648 § 4). Whitespaces are ignored.
- Current implementation allows arbitrary strings and invalid symbols are skipped.
- `hex` converts only valid hexadecimal strings and throws errors otherwise. Whitespaces are not allowed.
- `utf-8` and `utf8` are interchangeable.
- Correct errors are thrown and classified for invalid input (CONVERSION_INVALID_INPUT) and invalid format (CONVERSION_INVALID_FORMAT)
