1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
use crate::{MySql, MySqlTypeInfo, MySqlValueRef};
use sqlx_core::decode::Decode;
use sqlx_core::encode::{Encode, IsNull};
use sqlx_core::error::BoxDynError;
use sqlx_core::types::{Text, Type};
use std::fmt::Display;
use std::str::FromStr;
impl<T> Type<MySql> for Text<T> {
fn type_info() -> MySqlTypeInfo {
<String as Type<MySql>>::type_info()
}
fn compatible(ty: &MySqlTypeInfo) -> bool {
<String as Type<MySql>>::compatible(ty)
}
}
impl<T> Encode<'_, MySql> for Text<T>
where
T: Display,
{
fn encode_by_ref(&self, buf: &mut Vec<u8>) -> Result<IsNull, BoxDynError> {
// We can't really do the trick like with Postgres where we reserve the space for the
// length up-front and then overwrite it later, because MySQL appears to enforce that
// length-encoded integers use the smallest encoding for the value:
// https://dev.mysql.com/doc/dev/mysql-server/latest/page_protocol_basic_dt_integers.html#sect_protocol_basic_dt_int_le
//
// So we'd have to reserve space for the max-width encoding, format into the buffer,
// then figure out how many bytes our length-encoded integer needs to be and move the
// value bytes down to use up the empty space.
//
// Copying from a completely separate buffer instead is easier. It may or may not be faster
// or slower depending on a ton of different variables, but I don't currently have the time
// to implement both approaches and compare their performance.
Encode::<MySql>::encode(self.0.to_string(), buf)
}
}
impl<'r, T> Decode<'r, MySql> for Text<T>
where
T: FromStr,
BoxDynError: From<<T as FromStr>::Err>,
{
fn decode(value: MySqlValueRef<'r>) -> Result<Self, BoxDynError> {
let s: &str = Decode::<MySql>::decode(value)?;
Ok(Self(s.parse()?))
}
}