Skip to main content

update

update(&self, values: T) where T derives #[derive(SqlxUpdate)] — updates one row by a key field, setting every other field.

#[derive(SqlxUpdate)]
pub struct UserStatsUpdate {
    #[key]
    pub user_id: i32,
    pub self_acceptance: i32,
}

async fn update(&self, values: UserStatsUpdate) -> sqlx::Result<UserStatsRow>;

Exactly one field must be marked #[key] — it becomes the WHERE target and is excluded from SET; every other field becomes a SET column. Zero or more than one #[key] field is a compile error, and so is a struct with no non-key fields (nothing to SET).

Always row-returning — there's no ()-returning form of update.

Per-dialect SQL​

Postgres/SQLite: UPDATE "table" SET col = $1, ... WHERE key = $n RETURNING *.

MySQL has no RETURNING. UPDATE `table` SET col = ?, ... WHERE key = ? followed by SELECT * FROM `table` WHERE key = ? on the same executor — the same Copy-executor requirement described in dsl, so update can't be combined with #[transactional] under the mysql feature.

What this is for vs. set_by​

update takes a whole values struct and updates exactly one row, keyed by #[key]. Use set_by instead for updating a fixed set of columns across many rows matched by an arbitrary filter, without a values struct.