Get 0 value from a count with no rows

Question

I have SELECT:

SELECT c FROM (
    SELECT "candidate_id" as id, count("candidate_id") as c
    FROM "Applicaions"
    GROUP BY "candidate_id"
) as s WHERE id= _SOME_ID_;

But this only returns a value if count > 0. If count = 0 it returns nothing. How can I get 0 for a "Candidate" that doesn't have any application?

There is table "Candidates".
I need to get 0 if candidate has no applications or does not exist.

EDIT

I have now:

SELECT COALESCE ((SELECT count("candidate_id") as c
FROM "Applicaions" WHERE "candidate_id"=_SOME_ID_
GROUP BY "candidate_id"), 0);

It works perfectly. But is it possible to write it simpler or is this the best solution? Should I create any indexes?

Solution 1

You need to use the COALESCE function in PostgreSQL http://developer.postgresql.org/pgdocs/postgres/functions-conditional.html

Essentially you need to tell SQL how to handle NULLs. i.e. When NULL return 0.

Solution 2

You can't.

If your candidate has no applications, then you have no way to read their candidate_id value

How would you have knowledge that a candidate exists without them being in the Applications table?

In your example code, there is no candidate and therfore you couldn't possible say that a specific candidate had zero applications. By that logic there are an infinite number of candidates having zero applications.

You will need a table of candidates in order to derive this information... unless your intention is to presume a candidate exists because you're asking for it by ID?

EDIT

Now that you have a Candidates table you can do this:

SELECT c.ID, (SELECT COUNT(a.*) FROM Applications a WHERE a.candidate_id = c.ID) 
FROM Candidate c
WHERE ....

Solution 3

^{I was made aware of this question by the author of this related one who was led to believe his problem could not be solved after reading here. Well, it can.}

This answer is almost two years old, but the final questions were still pending.

Query

Is it possible to write it simpler or is this the best solution?

To test for a single ID, the query you found is good. You can simplify:

SELECT coalesce((SELECT count(candidate_id)
FROM   "Applications" WHERE candidate_id = _SOME_ID_), 0) AS c;

• The WHERE condition limits to a single candidate_id and there is a single aggregate function in the SELECT list. GROUP BY candidate_id was redundant.

• The column alias was swallowed by COALESCE(). If you want to name the resulting column move the alias to the end.

• You don't need double quotes for a regular lower case identifier.

Another, cleaner (IMHO) form would be to use LEFT JOIN:

SELECT count(a.candidate_id) AS c
FROM  (SELECT _SOME_ID_ AS candidate_id) x
LEFT   JOIN "Applicaions" a USING (candidate_id)

This works nicely for multiple IDs, too:

WITH x(candidate_id) AS (
   VALUES
     (123::bigint)
    ,(345)
    ,(789)
   )
SELECT x.candidate_id, count(a.candidate_id) AS c
FROM   x
LEFT   JOIN "Applicaions" a USING (candidate_id)
GROUP  BY x.candidate_id;

• LEFT JOIN is typically faster for a long list than multiple WHERE clauses or an IN expression.

Or, for all rows in your table "Candidates":

SELECT x.candidate_id, count(a.candidate_id) AS c
FROM   "Candidates" x
LEFT   JOIN "Applications" a USING (candidate_id)
GROUP  BY x.candidate_id;

Indexes

Should I create any indexes?

If read performance is important and the table holds more than just a couple of rows you definitely need an index of the form:

CREATE INDEX foo_idx ON "Applications" (candidate_id);

Since this seems to be a foreign key column referencing "Candidates".candidate_id, you should most probably have that to begin with.

Solution 4

Perhaps:

SELECT CASE c WHEN NULL THEN 0 ELSE c END
    FROM (
    SELECT "candidate_id" as id, count("candidate_id") as c
    FROM "Applicaions"
    GROUP BY "candidate_id"
) as s WHERE id= _SOME_ID_;

assuming that the 'nothing' is really NULL

Solution 5

You could try something like the following (assuming you have a candidate table and my assumption of table/column names are correct).

SELECT c FROM (
 SELECT "Candidate"."candidate_id" as id, 
 count("Applications"."candidate_id") as c
 FROM "Candidate" LEFT JOIN "Applications" 
 ON "Applications"."candidate_id"="Candidate"."id"     
GROUP BY "Candidate"."candidate_id" ) as s WHERE id= _SOME_ID_;

View more solutions

Comments

[Matthew]

You need a Candidate table unless you just want to return your queried ID with a "zero found" results or something.

[Miko Kronn]

There is "Candidates" table. How can I get 0 for those that doesn't have applications?

[Matthew]

Without using the Candidates table there are an infinite number of candidates without applications... 10001,10002,10003,...99999 ad infinitum

[Miko Kronn]

Yeah. 'nothing' is NULL. Your answer is good, but now I like COALESCE :)

[Miko Kronn]

I know. I removed them. Just wanted to tell that I need 0 value, not '0' string for displaying. But removing 's is something I can figure out on my own as opposed to joins, triggers etc :P. I don't know SQL well :(