I have been counting NULL and non-NULL columns with a subselect and some aggregate functions
CREATE TEMPORARY TABLE citizens(name text, country text,profession text,postalcode text);
INSERT INTO citizens VALUES
('Fred', 'USA', 'Professor', NULL),
('Amy', 'USA', 'Professor', NULL),
('Ted', 'USA', 'Professor', 90210),
('Barb', 'USA', 'Lawyer', 10248),
('Wally', 'USA', 'Lawyer', NULL),
('Fred', 'Canada', 'Professor', 'S0H'),
('Charles', 'Canada', 'Professor', 'S4L'),
('Nancy', 'Canada', 'Lawyer', NULL),
('Linda', 'Canada', 'Professor', NULL),
('Steph', 'France', 'Lawyer', 75008 ),
('Arnold', 'France', 'Lawyer', 75008 ),
('Penny', 'France', 'Lawyer', 75008 ),
('Harry', 'France', 'Lawyer', NULL);
SELECT country,
profession,
MAX(have_postalcode::int*num) AS num_have,
MAX((1-have_postalcode::int)*num) AS num_not_have
FROM
(
SELECT country, profession,
COUNT(*) AS num,
(postalcode IS NOT NULL) AS have_postalcode
FROM citizens
GROUP BY country, profession, have_postalcode
) AS d
GROUP BY country, profession
with the result
USA Professor 1 2
Canada Lawyer 0 1
USA Lawyer 1 1
France Lawyer 3 1
Canada Professor 2 1
but it seems like there ought to be a slicker way (for example it pains me that MAX is used just to grab the one nontrivial value). Does anybody have a cool idea?
http://sqlfiddle.com/#!1/17a9d/15