Creating a somewhat complicated cartesian result - sql

Question

I have a table called "where_clauses" which contains a bunch of conditions I would like to use for building dynamic queries. I would like to know all possible queries I could perform using this data. Here is my "where_clauses" data...
INSERT INTO where_clauses (id,col_name,clause) VALUES (1,'x','x < 1');
INSERT INTO where_clauses (id,col_name,clause) VALUES (2,'x','x < 2');
INSERT INTO where_clauses (id,col_name,clause) VALUES (3,'x','x < 3');
INSERT INTO where_clauses (id,col_name,clause) VALUES (4,'y','y < 1');
INSERT INTO where_clauses (id,col_name,clause) VALUES (5,'y','y < 2');
INSERT INTO where_clauses (id,col_name,clause) VALUES (6,'y','y < 3');
INSERT INTO where_clauses (id,col_name,clause) VALUES (7,'z','z < 1');
Ideally I would like the "all possible queries" in the form of an array of ids. For example, the "all possible queries" result would be...
{1}
{1,4}
{1,4,7}
{1,5}
{1,5,7}
{1,6}
{1,6,7}
{2}
{2,4}
{2,4,7}
{2,5}
{2,5,7}
{2,6}
{2,6,7}
{3}
{3,4}
{3,4,7}
{3,5}
{3,5,7}
{3,6}
{3,6,7}
{4}
{4,7}
{5}
{5,7}
{6}
{6,7}
{7}
Note that im throwing out joining on equal columns. What is a query that would give all possible where_clauses?

Answer 1

This is the sort of problem that the new WITH RECURSIVE is intended to solve. The following generalizes to any number of column names (not just x, y, z).
WITH RECURSIVE subq(a, x) AS
( VALUES (ARRAY[]::int[], NULL) /* initial */
UNION ALL
SELECT subq.a || id, col_name FROM subq JOIN where_clauses
ON x IS NULL OR x < col_name )
SELECT a FROM subq
WHERE x IS NOT NULL; /* discard the initial empty array */

Answer 2

SELECT string_to_array(TRIM(x || ',' || y || ',' || z, ','), ',')
FROM (
WITH sq AS (
SELECT a.id x, b.id y, c.id z
FROM where_clauses a, where_clauses b, where_clauses c
WHERE a.col_name != b.col_name AND
a.col_name != c.col_name AND
b.col_name != c.col_name AND
a.id < b.id AND
b.id < c.id
)
SELECT x, y, z FROM sq
UNION ALL
SELECT distinct x, y, null::int FROM sq
UNION ALL
SELECT distinct y, z, null::int FROM sq
UNION ALL
SELECT distinct x, null::int, null::int FROM sq
UNION ALL
SELECT distinct y, null::int, null::int FROM sq
UNION ALL
SELECT distinct z, null::int, null::int FROM sq
) ORDER BY 1;
Does above query helps you out ?

Answer 3

Try this code, it selects three columns, those that are not used for a clause are left NULL, you could concatenate or manipulate that result further:
--all possibilities with only one clause
SELECT
id AS ID1, NULL ID2, NULL AS ID3
FROM where_clauses
--all possibilities with two clauses (xy,xz,yz)
UNION
SELECT
WC1.id AS ID1, WC2.id AS ID2, NULL AS ID3
FROM where_clauses WC1
CROSS JOIN where_clauses WC2
WHERE
WC1.col_name != WC2.col_name
AND WC1.id > WC2.id
--all possibilities with an x and a y and a z clause
UNION
SELECT
WC1.id AS ID1, WC2.id AS ID2, WC3.id AS ID3
FROM where_clauses WC1
CROSS JOIN where_clauses WC2
CROSS JOIN where_clauses WC3
WHERE
WC1.col_name != WC2.col_name
AND WC1.id > WC2.id
AND WC1.col_name != WC3.col_name
AND WC1.id > WC3.id
AND WC2.col_name != WC3.col_name
AND WC2.id > WC3.id
Here is a fiddle.
EDIT: modified fiddle slightly