generate series using break time - postgresql

Question

I have a table that store opening hour and closing hour
CREATE TABLE public.open_hours
(
id bigint NOT NULL,
open_hour character varying(255),
end_hour character varying(255),
day character varying(255),
CONSTRAINT pk_open_hour_id PRIMARY KEY (id)
)
WITH (
OIDS=FALSE
);
ALTER TABLE public.open_hours
OWNER TO postgres;
I have another table that sotre
CREATE TABLE public.break_hours
(
id bigint ,
start_time character varying(255),
end_time character varying(255),
open_hour_id bigint ,
CONSTRAINT break_hours_pkey PRIMARY KEY (id),
CONSTRAINT fkinhl5x01pnn54nv15ol5ntxr5 FOREIGN KEY (open_hour_id )
REFERENCES public.open_hours(id) MATCH SIMPLE
ON UPDATE NO ACTION ON DELETE NO ACTION
)
WITH (
OIDS=FALSE
);
ALTER TABLE public.break_hours
OWNER TO postgres;
I need to generate time series of 30 minutest interval based on break times.
For eg: if my open hours is 08:00 AM and end hour is 06:00 PM and my break time is 11:00 AM to 11:30 and another break time is 03:00 PM to 03:15 PM then i need to generate series from 08:00 AM to 11:00 AM and 11:30 AM to 03:00 PM and 03:15 to 06:00 PM.
sample data
open_hours
-----------
id open_hours end_hour day
1 08:00 AM 06:00 PM Monday
break_hours
id start_time end_time open_hour_id
1 11:00 AM 11:30 AM 1
2 03:00 PM 03:15 PM 1
Sample out put
--------------
08:00 AM
08:30 AM
09:00 AM
09:30 AM
10:00 AM
10:30 AM
11:30 AM
12:00 PM
12:30 PM
01:00 PM
01:30 PM
02:PM PM
02:30 PM
03:15 PM
03:45 PM
04:15 PM
04:45 PM
05:15 PM
Query used for generating series between open hours is
SELECT DISTINCT gs AS start_time,gs + interval '30min' as end_time
FROM generate_series( timestamp '2018-11-09 08:00 AM', timestamp '2018-11-09 06:00 PM', interval '30min' )gs
ORDER BY start_time

Answer 1

It seems that your table modelling should be cleaned. E.g. you should not store times as text types but as time without time zone.
demo: db<>fiddle
WITH hours AS (
SELECT
oh.open_hour + '1970-01-01'::date as open_hour,
oh.end_hour + '1970-01-01'::date as end_hour,
bh.start_time + '1970-01-01'::date as break_start,
bh.end_time + '1970-01-01'::date as break_end,
lead(start_time + '1970-01-01'::date) OVER (ORDER BY start_time) as next_start_time
FROM open_hours oh
LEFT JOIN break_hours bh
ON oh.id = bh.start_date
)
SELECT generate_series(open_hour, break_start, interval '30 minutes')::time as time_slot
FROM (
SELECT
open_hour, break_start
FROM hours
ORDER BY break_start
LIMIT 1
)s
UNION
SELECT
generate_series(break_end, next_start_time, interval '30 minutes')::time
FROM (
SELECT
break_end, next_start_time
FROM
hours
WHERE next_start_time IS NOT NULL
) s
UNION
SELECT generate_series(break_end, end_hour, interval '30 minutes')::time
FROM (
SELECT
break_end, end_hour
FROM hours
ORDER BY break_start DESC
LIMIT 1
) s
Explanation:
WITH clause (CTE):
Merging both tables. I am adding a nonsense date because this results in a timestamp. The later used function generate_series only works for timestamps not for type time. The part is cut away later after the generation with the ::time cast.
The result of the CTE is:
open_hour end_hour break_start break_end next_start_time
1970-01-01 08:00:00 1970-01-01 18:00:00 1970-01-01 09:30:00 1970-01-01 09:45:00 1970-01-01 11:00:00
1970-01-01 08:00:00 1970-01-01 18:00:00 1970-01-01 11:00:00 1970-01-01 11:30:00 1970-01-01 15:00:00
1970-01-01 08:00:00 1970-01-01 18:00:00 1970-01-01 15:00:00 1970-01-01 15:15:00 (NULL)
UNION part:
This part contains three subparts. Because I have to merge the time series from both tables:
1. Taking the opening hour. Generate a time series to the first break beginning.
For this I only need the first row from the CTE above. That's why LIMIT 1 is used.
2. For all breaks: Generate a time series from current break ending to the next break beginning.
The CTE contains a window function lead() which shifts the start_time of the next row into the current one (have a look at the last column of the CTE result). So now I am able to get all break times, no matter how many there are. In my example I added a third break from 9:30 to 9:45 to demonstrate it. So the next time series can be generated from all these columns (current break_end to next_start_time). Only the last row does not contain a next_start_time because there is none.
3. Last step: Generate a time series from the last break ending to the closing hour.
This is quiet similar to (1). After iterating all break times I have to add the last time series from the last break time to the closing time. This could be achieved either by filtering the row without next_start_time or sorting DESC and using LIMIT 1 as I did.
More complex case with more day types:
demo: db<>fiddle
WITH hours AS (
SELECT
oh.id as day_id,
oh.open_hour + '1970-01-01'::date as open_hour,
oh.end_hour + '1970-01-01'::date as end_hour,
bh.start_time + '1970-01-01'::date as break_start,
bh.end_time + '1970-01-01'::date as break_end,
lead(start_time + '1970-01-01'::date) OVER (PARTITION BY oh.id ORDER BY start_time) as next_start_time
FROM open_hours oh
LEFT JOIN break_hours bh
ON oh.id = bh.start_date
)
SELECT day_id, generate_series(open_hour, break_start, interval '30 minutes')::time as time_slot
FROM (
SELECT DISTINCT ON (day_id)
day_id, open_hour, break_start
FROM hours
ORDER BY day_id, break_start
)s
UNION
SELECT
day_id, generate_series(break_end, next_start_time, interval '30 minutes')::time
FROM (
SELECT
day_id, break_end, next_start_time
FROM
hours
WHERE next_start_time IS NOT NULL
) s
UNION
SELECT day_id, generate_series(break_end, end_hour, interval '30 minutes')::time
FROM (
SELECT DISTINCT ON (day_id)
day_id, break_end, end_hour
FROM hours
ORDER BY day_id, break_start DESC
) s
ORDER BY day_id, time_slot
The main idea stays the same as in the example for only one day. The difference is that we have to consider the different day types. I expanded the example above and added a second day with different opening hours and break times.
Changes:
The window function in the CTE got a PARTITION BY part. This ensures that only the start_times are shifted that contains to the same day.
LIMIT 1 will not work anymore because it limits the whole table to one row. This has been changed to DISTINCT ON (day_id) which limits the table to the first row of each day.