Question

I have a PL/pgsql function like so

CREATE OR REPLACE FUNCTION foo(colname TEXT, col INT)
    RETURNS REAL AS $$
    BEGIN
        IF (colname = 'a') THEN
            RETURN (col * 1.5);
        ELSIF (colname = 'b') THEN
            RETURN (col * 2.5);
        ELSIF (colname = 'c') THEN
            RETURN (col * 3.5);

        .. and so on ..

        ELSE
            RAISE EXCEPTION 'Invalid column!';
        END IF;
        RETURN NULL;
    END;
    $$
    LANGUAGE plpgsql;

The function allows me to do stuff like

SELECT foo('a', a) FROM table WHERE

I would like to make this nicer, and not have to pass the column name, if I can help it. In other words, I would like to be able to do

SELECT foo(a) FROM table WHERE

and figure out the column name in the function based on the col passed to the function. Is that possible?

Answer 1

No. What’s passed to a function is just a value that is the result of an expression. It makes no difference the your expression happens to be a single column name. What if you were to write the query like this?

SELECT foo(a+2) FROM table WHERE ...

Or this?

SELECT foo(2+2) FROM table WHERE ...

What should the function infer is the column name in these cases?