It seems when parsing a literal beginning with digits, the sql parser simply ends this token when meets a non-numeric character see https://github.com/sqlparser-rs/sqlparser-rs/blob/main/src/tokenizer.rs#L547 For example, `123foobar` would be parsed into a number token `123` and a word token `foobar` instead of one word token `123foobar`. And `123foobar456` would be parsed into `123` and `foobar456`