Is it possible to use pgx.RowToStructByName that silently ignore extra columns?

[katatrina]

I'm using pgx.RowToStructByName with "db" struct tag to avoid scanning manually. My problem is, for example, when the table has 4 fields but the struct only has 3 fields, an error is returned like "struct doesn't have corresponding row field ...".

type Province struct {
	Code      string    `db:"code"`
	FullName  string    `db:"full_name"`
	CreatedAt time.Time `db:"created_at"`
}

CREATE TABLE IF NOT EXISTS provinces
(
    code       TEXT PRIMARY KEY,
    full_name  TEXT        NOT NULL,
    is_active   TEXT NOT NULL DEFAULT true,    <-- struct does not map this column
    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
);

Like in the situation that we applied migration but the code was not synced yet. This problem leads to a decision: I must ensure the db schema and the struct must matched together at runtime, either I must specify explicitly columns insead of using * in the SELECT statement. Can the function somehow just ignore that column and scan result into existing struct fields?

[jackc]

You can use pgx.RowToStructByNameLax to not require every struct field be supplied by the query, but you cannot ignore columns from the database. Being too lax would make it too easy to not bind a value you were expecting to be bound. The application would use the zero value of the field thinking it was the actual value from the DB.

This problem leads to a decision: I must ensure the db schema and the struct must matched together at runtime, either I must specify explicitly columns insead of using * in the SELECT statement.

AFAIK, specifying the columns instead of * is a good practice for precisely this reason - it is more tolerant of DB changes (also it can have better performance by only selecting what is needed).

[katatrina]

Tks, it clear me now