Question

Sorry about the complicated title.

I have two tables, customers and orders:

customers – names may be duplicated, ids are unique:

name | cid
 a   |  1
 a   |  2
 b   |  3
 b   |  4
 c   |  5

orders – pid is unique, join on cid:

pid | cid | date
 1  |  1  | 01/01/2012
 2  |  1  | 01/01/2012
 3  |  2  | 01/01/2012
 4  |  3  | 01/01/2012
 5  |  3  | 01/01/2012
 6  |  3  | 01/01/2012

So I used this code to get a count:

select customers.name, orders.date, count(*) as count
from customers
left JOIN orders ON customers.cid = orders.cid
where date between '01/01/2012' and '02/02/2012'
group by name,date

which worked fine but didnt give me null rows when the cid of customers didnt match a cid in orders, e.g. name-c, id-5

select customers.name, orders.date, count(*) as count
from customers
left JOIN orders ON customers.cid = orders.cid
AND date between '01/01/2012' and '02/02/2012'
group by name,date

So I changed the where to apply to the join instead, which works fine, it gives me the null rows.
So in this example I would get:

name | date       | count
 a   | 01/01/2012 | 3
 b   | null       | 1
 b   | 01/01/2012 | 3
 c   | null       | 1

But because names have different cid’s it is giving me a null row even if the name itself does have rows in orders, which I don’t want.

So I’m looking for a way for the null rows to only be returned when any other cid’s that share the same name also do not have any rows in orders.

Thanks for any help.

—EDIT—

I have edited the counts for null rows, count never returns null but 1.

The result of

select * from (select customers.name, orders.date, count(*) as count
from customers
left JOIN orders ON customers.cid = orders.cid
AND date between '01/01/2012' and '02/02/2012'
group by name,date) as t1 group by name

is

name | date       | count
 a   | 01/01/2012 | 3
 b   | null       | 1
 c   | null       | 1

Answer 1

First, select your date grouped by (name, date), excluding NULLs, then join with a set of distinct names:

SELECT names.name, grouped.date, grouped.count
FROM ( SELECT DISTINCT name FROM customers ) as names
LEFT JOIN (
  SELECT customers.name, orders.date, COUNT(*) as count
  FROM customers
  LEFT JOIN orders ON customers.cid = orders.cid
  WHERE date BETWEEN '01/01/2012' AND '02/02/2012'
  GROUP BY name,date
) grouped
ON names.name = grouped.name