UTF-8 null character in text causes "invalid byte sequence for encoding UTF8"

I detected this issue when user input (sanitized for SQL-injection with knex) started causing postgres errors.

Here is a sample that reproduces this behavior

client.query('SELECT $1::text as message', ['Hello world\u0000!']);

Output:

(node:94484) UnhandledPromiseRejectionWarning: error: invalid byte sequence for encoding "UTF8": 0x00

Is this something we can detect/prevent/sanitize on driver level or should I sanitize all strings that touch text type in postgreSQL on application level? (probably using some cumbersome REGEXP). The knex guys suggested that I should contact you.

https://github.com/knex/knex/issues/2284#issuecomment-575685625

[charmander]

@jbtobar Zero bytes should work fine with bytea. Are you passing the parameter as a Buffer or other typed array? That’s necessary. Also, make sure your query doesn’t try to decode the bytea as text at any point.

[charmander]

pg could detect it, but it would just throw the same kind of error upon doing so. Unless you’re getting accidental \0s in input that would warrant manual filtering, it’s probably safe to continue with default error handling.