Question

Evidentally, I can’t do this:

SELECT 
  SUM( t.MyField IS NULL ) AS totalNulls, 
  SUM( t.MyField LIKE '[0-9]') AS totalNumbers
FROM MyTable AS t;

I don’t know why these don’t work, since booleans in SQL are just numbers (0 and 1). But the errors I’m getting suggest it is not legal to have ‘is null’ or ‘like’ anywhere in the select clause. Why aren’t they legal there? How do I achieve the intended effect, as suggested by the (pseudo) SQL above?

Answer 1

If your data isn’t indexed on the queried column you should use the CASE-based solution Aaron recommended because the rows will only be queried once.

select  totalNulls = sum(case when MyField is null then 1 else 0 end),
        totalNumbers = sum(case when MyField like '[0-9]' then 1 else 0 end)
from    MyTable;

You should know that you only accept e.g. ‘7’ as a number not ’12’ in this case.
If you want to accept any numbers you’d have to ask

... when MyField not like '%[^0-9]%' ...

In this case you could do a charm using isnumeric:

select  totalNonNumbers = count(*) - sum(isnumeric(MyField)),
        totalNumbers = sum(isnumeric(MyField))
from    MyTable;

If the column is indexed a variation of Michaels solution could be fastest:

select  totalNulls = (select count(*) from MyTable where MyField is null),
        totalNumbers = 
          (select count(*) from MyTable where MyField between '0' and '9');

There are so many ways …