Generate series of week intervals for given month

Question

In a Postgres 9.1 database, I am trying to generate a series of weeks for a given month but with some constraints. I need all weeks to start on Monday and get cut when they start or end in another month.

Example:

For February, 2013 I want to generate a series like this:

         start
------------------------
2013-02-01 00:00:00+00
2013-02-04 00:00:00+00
2013-02-11 00:00:00+00
2013-02-18 00:00:00+00
2013-02-25 00:00:00+00

The query that I have now looks like this:

SELECT GREATEST(date_trunc('week', dates.d),
                date_trunc('month',dates.d)) as start
FROM generate_series(to_timestamp(1359676800),to_timestamp(1362095999), '1 week') as dates(d)

This query gets me the first 4 weeks but it's missing the week from the 25th. Is it possible to get the last week?

Answer 1

SELECT generate_series(date_trunc('week', date '2013-02-01' + interval '6 days')
                     , date_trunc('week', date '2013-02-01' + interval '1 month - 1 day')
                     , interval '1 week')::date AS day
UNION  SELECT date '2013-02-01'
ORDER  BY 1;

This variant does not need a subselect, GREATEST or GROUP BY and only generates the required rows. Simpler, faster. It's cheaper to UNION one row.

• Add 6 days to the first day of the month before date_trunc('week', ...) to compute the first Monday of the month.

• Add 1 month and subtract 1 day before date_trunc('week', ...) to get the last Monday of the month.
  This can conveniently be stuffed into a single interval expression: '1 month - 1 day'

• UNION (not UNION ALL) the first day of the month to add it unless it's already included as Monday.

• Note that date + interval results in timestamp, which is the optimum here. Detailed explanation:

  • Generating time series between two dates in PostgreSQL

Automation

You can provide the start of the date series in a CTE:

WITH t(d) AS (SELECT date '2013-02-01')  -- enter 1st of month once
SELECT generate_series(date_trunc('week', d + interval '6 days')
                     , date_trunc('week', d + interval '1 month - 1 day')
                     , interval '1 week')::date AS day
FROM   t
UNION  SELECT d FROM t
ORDER  BY 1;

Or wrap it into a simple SQL function for convenience with repeated calls:

CREATE OR REPLACE FUNCTION f_week_starts_this_month(date)
  RETURNS SETOF date AS
$func$
SELECT generate_series(date_trunc('week', $1 + interval '6 days')
                     , date_trunc('week', $1 + interval '1 month - 1 day')
                     , interval '1 week')::date AS day
UNION
SELECT $1
ORDER  BY 1
$func$  LANGUAGE sql IMMUTABLE;

Call:

SELECT * FROM f_week_starts_this_month('2013-02-01');

You would pass the date for the first day of the month, but it works for any date. You the first day and all Mondays for the following month.