Postgres pivot table with crosstab

Question

I have a table like the next one:

+------------+---------+---------+---------+
|    date    | value 1 | value 2 | value 3 |
+------------+---------+---------+---------+
| 01/01/2017 |     263 |       7 |     222 |
| 02/01/2017 |     275 |      -9 |     209 |
| 03/01/2017 |     331 |      -9 |     243 |
| .          |       . |       . |       . |
| .          |       . |       . |       . |
| .          |       . |       . |       . |
+------------+---------+---------+---------+

I want to create this other one in postgres:

+---------+---------------+------------+------------+
|         |    01/01/2017 | 02/01/2017 | 03/01/2017 |
+---------+---------------+------------+------------+
| value 1 |           263 |        275 |        331 |
| value 2 |             7 |         -9 |         -9 |
| value 3 |           222 |        209 |        243 |
+---------+---------------+------------+------------+

But my problem is that I dont know how many dates I will have, so I have to use something like this:

SELECT * FROM crosstab(
  $$ SELECT value1, date  FROM myTable ORDER BY 1 $$,
  $$ SELECT m FROM generate_series((select min(date) from myTable) ,(select max(date) from myTable), '1 month'::interval) m $$
) AS (
  ".." date, ".." date, ".." date, ".." date
);

Does someone can help me? Thanks.

Answer 1

Your basic issue is that PostgreSQL needs to know what the columns look like in order to plan the query. Consequently you need to return some sort of fixed-column structure. There are a number of ways you can do this:

1. Query dates first or allow them to be input, and then generate your query in the db client.
2. Wrap this in a stored procedure which returns a refcursor
3. Wrap in a stored procedure which returns a list of JSON representations of rows.

But either way you cannot do it in a query without dynamically generating the query somewhere.