opening hours table design postgresql: how to query this? - postgresql

Question

I want to store opening hours of a store in a PostgreSQL database. I have a few question about this. First of all, my table design:
Table: opening_hours
--------------------
* id
* store_id
* start_day (0->6)
* start_time (HH:mm:ss)
* duration (HH:mm:ss)
* timezone (default: Europe/Brussels)
Example:
--------
1 1 0 08:00:00 12:00:00 Europe/Brussels
// store ID 1 is open on sundays from 08AM to 08PM (12 hours in total) local time
2 2 6 20:00:00 04:00:00 Europe/Brussels
// store ID 2 is open on saturday from 08PM till 04AM (on sunday) (this is like a pub or something with opening hours at night) local time (so 20:00:00 in Belgium open = 19:00:00 in London open)
I think this is a good design, because now I can make opening hours spanning more than 1 day (like when a pub is open at night). Before that, I was storing opening hours in each day separately so I had to enter 'saturday 20:00:00 -> 23:59:59' + 'sunday 00:00:00 -> 04:00:00' for the opening hours for store ID 2.
How can I query against these rows?
I want to check if a time already exists in the database before adding a new one, and I want to check if a row exists based on the current time of the user (2 different queries).
CURRENT ROWS:
-------------
1 1 1 08:00:00 08:00:00 Europe/Brussels
// open on monday from 08AM until 04PM
2 1 6 20:00:00 12:00:00 Europe/Brussels
// open on saturday from 08PM until sunday 08AM
NEW ROWS:
---------
3 1 1 14:00:00 04:00:00 Europe/Brussels (should not insert this row because it interferes with row ID 1)
// open on monday from 02PM until 06PM
4 1 1 17:00:00 04:00:00 Europe/Brussels (can insert because not interference with row 1)
// open on monday from 05PM until 09PM
5 1 0 07:00:00 05:00:00 Europe/Brussels (can't insert this row because it interferes with row 2).
// open on sunday from 07AM until 12PM
I hope my question is clear. If not, please correct me, I'll try to ask it differently then.

Answer 1

With your new table design actually, it is now a bit tricky, because of durations that can possibly span over the end of a week cycle. Consider for example:
start_day | start_time | duration
-----------+------------+----------
6 | 14:00:00 | 18:00:00
It actually extends into day=0, so it would need to match day=0, t=07:00:00.
When looking for matches against a specific time, you need to check two possibilities (either day, time or day + 7, time fall into one of your intervals). Same thing for overlaps (3 possibilities).
You can define some helper functions:
-- oh: opening hours helper functions
-- oh_tst: convert dow, time into a timestamp after the Epoch
-- (note: Epoch was not a Monday (dow 0), but it doesn't matter,
-- we could use any arbitrary date)
create or replace function oh_tst(dow int, t time) returns timestamp as $$
select '1970-01-01'::timestamp + $1 * interval'1 day' + $2;
$$
language sql immutable;
-- oh_single_matches: internal function (no handling of wrap around)
create or replace function oh_single_matches(start_day int, start_time time, d interval, day int, t time) returns boolean as $$
select oh_tst($4, $5) between oh_tst($1, $2) and oh_tst($1, $2) + $3 - interval'1 millisecond';
$$
language sql immutable;
-- oh_matches: tests if a (day, time) is within an oh interval
-- handle wrap around at end of week
create or replace function oh_matches(start_day int, start_time time, d interval, day int, t time) returns boolean as $$
select oh_single_matches($1, $2, $3, $4, $5)
or oh_single_matches($1, $2, $3, $4, $5 + 7);
$$
language sql immutable;
-- oh_overlaps: test for oh defs overlap (incl wrap-around)
create or replace function oh_overlaps(adow int, astart_time time, aduration interval,
bdow int, bstart_time time, bduration interval) returns boolean as $$
select (oh_tst($1, $2), $3) overlaps (oh_tst($4, $5), $6)
or (oh_tst($1, $2), $3) overlaps (oh_tst($4 + 7, $5), $6)
or (oh_tst($1, $2), $3) overlaps (oh_tst($4 - 7, $5), $6);
$$
language sql immutable;
Examples:
Single match (one day,time against an opening hours definition):
-- intervals are left-close:
select oh_matches(6, '14:00:00'::time, interval'2 hours', 6, '14:00:00'::time);
oh_matches
------------
t
-- ...and right-open (as Nature intended):
select oh_matches(6, '14:00:00'::time, interval'2 hours', 6, '15:59:59'::time);
t
select oh_matches(6, '14:00:00'::time, interval'2 hours', 6, '16:00:00'::time);
f
-- wrap around the end of week
select oh_matches(6, '14:00:00'::time, interval'18 hours', 0, '02:00:00'::time);
t
Interval overlaps:
select oh_overlaps(2, '14:00:00'::time, interval'8 hours',
2, '04:00:00'::time, interval'8 hours');
f
select oh_overlaps(2, '14:00:00'::time, interval'8 hours',
2, '08:00:00'::time, interval'8 hours');
t
-- wraparoud ok
select oh_overlaps(0, '01:00:00'::time, interval'8 hours',
6, '22:00:00'::time, interval'18 hours');
t
Test against a table:
Single timestamp (now()):
select * from mytable
where store_id=5
and oh_matches(start_day, start_time, duration, now()::timestamp);
Check if prospective rows in a candidates table would overlap existing definitions in mytable:
select * from candidates a inner join mytable b using (store_id)
where oh_overlaps(b.start_day, b.start_time, b.duration,
a.start_day, a.start_time, a.duration);