bug(expr): SUBSTR of unicode produces error: "byte index _ is not a char boundary"

[jon-chuang]

We should not calculate by bytes, but by unicode character:

To reproduce:
Risingwave:

=> select substr('Mér', 1, 2);
SSL SYSCALL error: EOF detected

PSQL:

=> select substr('Mér', 1, 2);
 substr 
--------
 Mé
(1 row)

Similar for ''Mér'::char(3)

---

resources:

src/expr/src/vector_op/substr.rs:48:23

risingwave/src/expr/src/vector_op/substr.rs

Line 37 in 95ab15c

pub fn substr_start_for(s: &str, start: i32, count: i32, writer: &mut dyn Write) -> Result<()> {

https://stackoverflow.com/questions/4249745/does-postgresql-varchar-count-using-unicode-character-length-or-ascii-character

[xiangjinwu]

Similar problem exists for other string functions:

• overlay
• ascii