Assign to NEW by key in a Postgres trigger

Question

In the trigger body, how can I assign a value to NEW by it's field name?
That's what I want to do:

some_key = "some_column";
NEW[some_key] = 5;

• Ben: I know that, thats why I raised this question. There is a hacky way to access variable from NEW dynamically like this EXECUTE 'SELECT (1ドル).'||column_name||' ;' INTO some_var USING NEW; but I cant assign to NEW like that. @dezso
• Ben: And @dezso , whenever someone calls map an array - my heart bleeds
• Ben: Yes, but can you access in postgres array by key? Im just saying that some scripts don't show difference between array and map, and I know people that programm such scripts and don't know the difference between those two.
• Ben: Thanks for your comments @dezso, I guess you just can't achieve this in postgres.

Answer 1

First of all, there is no "trigger body" (unlike Oracle). In Postgres you have a trigger function (also, misleadingly, called "procedure") with a function body and 0-n triggers (without body) calling this function.

The special variable NEW in plpgsql trigger functions is neither a map nor an array; it's a record holding the new row:

NEW

Data type RECORD; variable holding the new database row for INSERT/UPDATE operations in row-level triggers. This variable is unassigned in statement-level triggers and for DELETE operations.

Assigning to a field (or column) of NEW is simple. The documented assignment operator is :=. (Since Postgres 9.4 also =.)

NEW.some_key := 5;

What you seem to be looking for is to parameterize the column name, which isn't quite as simple.
The additional module hstore provides the #= operator. (It's included in pretty much all standard distributions.) Install the module once per database with:

CREATE EXTENSION hstore;

Then you can:

NEW := NEW #= '"some_key"=>"5"'::hstore;

Sets the column some_key to '5' - if the column exists.

• An explicit cast to hstore is optional. The operator #= coerces a string literal to the right data type automatically.
• hstore only stores text strings, so a given literal for the value may have to be cast twice - a very minor drawback compared to alternative solutions.
• The given string literal has to fit the data type of the column, or an exception is raised.
• If no column with the given name exists, nothing is changed, no exception raised.

Related answer with details and an alternative solution:

• How to set value of composite variable field using dynamic SQL

Code example

CREATE OR REPLACE FUNCTION trg_tbl_ins_bef()
 RETURNS trigger AS
$func$
BEGIN
 NEW := NEW #= '"some_key"=>"5"';
 RETURN NEW;
END
$func$ LANGUAGE plpgsql;
CREATE TRIGGER ins_bef
BEFORE INSERT ON tbl
FOR EACH ROW EXECUTE PROCEDURE trg_tbl_ins_bef();

Postgres 11 or later allows to replace the misleding term PROCEDURE (the old syntax will keep working for the foreseeable future):

...
FOR EACH ROW EXECUTE FUNCTION trg_tbl_ins_bef();

• András Váczi: I haven't thought about hstore, it's much simpler than my approach.
• Erwin Brandstetter: Check out the linked answer. Pavel and I worked hard on a solution before the #= operator was introduced - and found one, too.
• Ben: I didn't said NEW is a map or array, writing this NEW[some_key] = 5; I was just giving a hint what I want to achieve. The thing with hstore is genius, would never think of it by myself.

Answer 2

I must admit that this is no easy way of solving it, but at least it's a way. I created the below example as a standalone one, to avoid all clutter with trigger creation and such. If you use it in a trigger, you can remove the declaration and initialization of p and replace the remaining use with NEW.

DO $$
DECLARE p members_test; 
BEGIN
 p := (1,2,3);
 CREATE TEMP TABLE t ON COMMIT DROP AS SELECT p.*; -- a one row table holding 
 -- the values of the variable
 EXECUTE format($e$UPDATE t SET %s = 43$e,ドル 'b'); -- this way you can access 
 -- the columns dynamically
 SELECT * INTO p FROM t; -- assign the new values back to the variable
 RAISE INFO 'p: %', p;
END;
$$;
INFO: p: (1,43,3)