SELECT COUNT（*）GROUP BY零行

Question

Tables in my db:

CREATE TABLE items (
  id serial NOT NULL,
  user_id integer NOT NULL,
  status smallint
);

CREATE TABLE phones (
  phone character varying,
  users integer[]
);

My Query for find phone numbers where status = 1:

SELECT phones.phone, COUNT(*) AS "count" 
FROM phones,items 
WHERE phones.phone = ANY (Array['7924445544', '8985545444'])
AND   items.user_id = ALL (phones.users) AND items.status = 1
GROUP BY phones.phone;

Query out:

   phone    | count
------------+------
 7924445588 |     3

Need out with ZERO count:

   phone    | count 
------------+-------
 8985545444 |     0
 7924445544 |     3

How to get that?

Answer 1

You shouldn't do that in the query. However it is rather easy to do it if you want to:

WITH phone_seek AS(
   SELECT '8985545444' AS phone
   UNION ALL
   SELECT '7924445588 '
)
SELECT  phone_seek.phone, COUNT(items.id) AS "count"
FROM 
  phones_seek
  JOIN phones
    ON phones_seek.phone = phones.phones
  CROSS JOIN items
WHERE 
  items.user_id = ALL (phones.users) AND items.status = 1 
GROUP BY phones.phone;

Answer 2

It's a bit tricky to create non-existing rows. (There are billions of them, at least...) Do a UNION ALL with a select providing the result you want if that phone no doesn't exist.

<current query>
UNION ALL
select '8985545444',0
from one_row_table where not exists (select 1 from phones
                                     where phones.phone = '8985545444')

EDIT: If the phone numbers do exists, but not fulfill the WHERE clause conditions, use a correlated sub-select to do the count:

SELECT phones.phone,
       (select count(*) from items
        where items.status = 1
          and items.user_id = phones.users) as "Count"
FROM phones
WHERE phones.phone = ANY (Array['7924445544', '8985545444'])

Answer 3

Count is an aggregate function and generally db engines produces count 0 when no row matches the criteria.

The reason why your query produces no results is inclusion of your phones.phone field in db and the group by.

1) A cheap solution to this is select only count(*) as your application already be knowing the phone number:

SELECT  COUNT(*) AS "count"
FROM phones, items
WHERE phones.phone = '8985545444'
  AND items.user_id = ALL (phones.users) AND items.status = 1 ;

2) An ideal solution to this would be your application to handle 0 result returned by db.

3) And if you need db to do all job you may use something like:

SELECT  phones.phone, COUNT(*) AS "count"
FROM phones, items
WHERE phones.phone = '8985545444'
  AND items.user_id = ALL (phones.users) AND items.status = 1 
GROUP BY phones.phone;
UNION ALL
select '8985545444',0
from one_row_table where not exists (select 1 from phones
                                     where phones.phone = '8985545444')

Answer 4

I'm not well versed with PostreSQL's array syntax, but this seems to be a simple outer join:

SELECT phones.phone, COUNT(items.user_id) AS "count" 
FROM phones LEFT JOIN items 
  ON items.user_id = ALL (phones.users) 
 AND items.status = 1
WHERE phones.phone = ANY (Array['7924445544', '8985545444']) 
GROUP BY phones.phone;