iterative depth of tree - sql

Question

The problematic code:
CREATE OR REPLACE FUNCTION foo(searchid INTEGER)
RETURNS INTEGER AS
$$
DECLARE
level INTEGER := 0;
mid INTEGER := searchid;
BEGIN
WHILE EXISTS(SELECT id INTO mid FROM tbl1 WHERE parent_id=mid) LOOP
level := level + 1;
END LOOP;
RETURN level;
END;
$$
LANGUAGE 'plpgsql' IMMUTABLE;
I need to find the tree depth of an element with id searchid, I've written a slightly different function than above which uses mid NOTNULL as a condition of the while loop and it works.
However, when I try to use EXISTS directly in the WHILE condition as in the code posted above, postgresql says:
SQL error:
ERROR: syntax error at or near "$1"
LINE 1: SELECT EXISTS(SELECT id INTO $1 FROM tbl1 WHERE ...
So it does some strange transformations on my code which make it syntactically wrong.
How to fix it?
It runs on postgresql 8.3.17.

Answer 1

Just for the record:
If you were using an up-to-date version of Postgres, you could do this more efficiently with a single statement:
with recursive tree as (
select id, parent, 1 as level
from tbl1
where id = 1
union all
select c.id, c.parent, p.level + 1
from tbl1 c
join tree p on c.parent = p.id
)
select max(level)
from tree

Answer 2

The key mistake is that you cannot assign a variable with SELECT INTO inside an EXISTS construct. SELECT items inside the EXISTS construct are ignored.
I rewrote the function to simplify and make it more secure:
CREATE OR REPLACE FUNCTION foo(_searchid int, OUT _level int)
RETURNS int
LANGUAGE plpgsql STABLE AS
$func$
BEGIN
_level := 0;
LOOP
SELECT INTO _level, _searchid
_level + 1, t.id
FROM tbl1 t
WHERE t.parent_id = _searchid;
EXIT WHEN NOT FOUND;
END LOOP;
END
$func$;
Call:
SELECT foo(1);
Major points
It is your responsibility to prevent the loop from being infinite.
The _ prefix of parameters is to avoid naming conflicts with potential columns of the used table.
I use the special variable FOUND that is set TRUE after (and only after) certain SQL statements (like SELECT INTO) found a row.
Use the EXIT command to exit the loop when no row was found.
Increment _level inside the SELECT. (Or in the loop body, it's just a tiny simplification.)
Since PostgreSQL 9.1 you can assign to IN parameters, so I (ab)use _searchid and don't need to DECLARE any additional variables. Don't do this, if you need the original parameter value later in the function.
The function should not be declared IMMUTABLE, since it accesses a table. I made it STABLE instead. You can make the function IMMUTABLE to "cheat" and be able to use it in index creation (for instance) - but it's on you if such an index breaks after a change in the underlying table.
Recursive CTE
With modern PostgreSQL you could also use a recursive CTE for the job. That's what #a_horse hinted at in his comment - oh, and what he posted now as answer.
Another example (of many on SO) here.

Answer 3

Maybe instead of writing your own function you would like to use postgres extension with function that was created to present hierarchical data that is stored in a table?
It's called connectby and it is a part of tablefunc extension. How to use the function you can find here.
To install extension:
CREATE EXTENSION tablefunc;
You've got many possibilities like you can choose: key value of the row to start at, maximum depth to descend to, or zero for unlimited depth, or string to separate keys with in branch output.