Assume in Postgresql, I have a table T and one of its column is C1.
I want to trigger a function when a new record is adding to the table T. The function should check the value of column C1 in the new record and if it is null/empty then set its value to 'X'.
Is this possible?
You are correct that you need a trigger, because setting a default value for the column won't work for you - default values only work for null values and don't help you in preventing blank values.
In postgres there are a couple of steps to creating a trigger:
Step 1: Create a function that returns type trigger:
CREATE FUNCTION my_trigger_function()
RETURNS trigger AS $$
BEGIN
IF NEW.C1 IS NULL OR NEW.C1 = '' THEN
NEW.C1 := 'X';
END IF;
RETURN NEW;
END;
$$ LANGUAGE plpgsql
Step 2: Create a trigger that invokes the above function and fires BEFORE INSERT, which allows you to change the incoming values before they hit the table:
CREATE TRIGGER my_trigger
BEFORE INSERT ON T
FOR EACH ROW
EXECUTE PROCEDURE my_trigger_function()
And you're done.
See the above code executing on SQLFIddle.
You mention in a comment that the value 'X' is retrieved from a subquery. If so, change the relevant line so something like:
NEW.C1 := (select some_column from some_table where some_condition);
$$ but instead '?
$$. I have been deprived of using functions all this time until you pointed that out!
IF, you could use a single line with coalesce() , something like : new.c1 := coalesce (new.c1,X);It is possible but you are likely to be better off setting a default constraint on the column instead. When creating the table that would look like:
create table mytable as (
C1 thetype not null default X
);
This says that if you add a row to the table and don't specify the value for C1 then X will be used instead. The not null is not necessary, but prevents updates from nulling that column assuming that's what you want.
EDIT: This only works for constant X, from your comments it seems there are two possible solutions.
Using a trigger would look something like this:
create function update_row_trigger() returns trigger as $$
begin
if new.C1 is NULL then
new.C1 := X;
end if;
return new;
end
$$ language plpgsql;
create trigger mytrigger before insert on mytable for each row execute procedure update_row_trigger();
The new variable in a trigger function is special, representing the row being inserted. Specifying the trigger as a before insert trigger means you can modify the row before it is written to the table.
The second solution would be to use a computed column which Postgres defines in an unusual way:
create or replace function C1(row mytable) returns columntype immutable as $$
begin
return X; -- where X is an expression using values from `row`
end
$$ language plpgsql;
This creates a function that takes a row of your table and returns a value, you can call it using . notation though, meaning you can do:
select
*,
t.C1
from
mytable t;
The declaration of the function being immutable is optional, but it's needed if you want to index the "column". You would be able to index this column like this:
create index on mytable (C1(mytable));
