Question

Select from great answer in How to find first free time in reservations table in PostgreSql

create table reservation (during tsrange,
 EXCLUDE USING gist (during WITH &&)
 );

is used to find gaps in schedule starting at given date and hour (2012-11-17 8: in sample below)
It finds saturday, sunday and public holidays also.
Public holidays are defined in table

create table pyha ( pyha date primary key)

How to exclude weekends and public holidays also?

Hard-coding free time as reserved to query like

with gaps as (
  select
    upper(during) as start,
    lead(lower(during),1,upper(during)) over (ORDER BY during) - upper(during) as gap
  from (
    select during
    from reservation
   union all values
     ('(,2012-11-17 8:)'::tsrange), -- given date and hour from which to find free work time
     ('[2012-11-17 0:,2012-11-18 24:)'::tsrange), -- exclude saturday
     ('[2012-11-18 0:,2012-11-19 8:)'::tsrange),  -- exclude sunday
     ('[2012-11-19 18:,2012-11-20 8:)'::tsrange),
     ('[2012-11-20 18:,2012-11-21 8:)'::tsrange),
     ('[2012-11-21 18:,2012-11-22 8:)'::tsrange),
     ('[2012-11-22 18:,2012-11-23 8:)'::tsrange),
     ('[2012-11-23 18:,2012-11-24 24:)'::tsrange),
     ('[2012-11-24 0:,2012-11-25 24:)'::tsrange), -- exclude saturday
     ('[2012-11-25 0:,2012-11-26 8:)'::tsrange)  -- exclude sunday
 ) as x
)
select *
  from gaps
where gap > '0'::interval
order by start

requires separate row in union for every free time range.

Which is best way to return free time in work days and work hours ( 8:00 .. 18:00 ) starting from given date and hour ?

Update

Select in answer returns free time at 8:00 always.
How to return free time not before specified start hour in specified start date, for example not before 2012-11-19 9:00 if start hour is 9 ?
Start hour may have only values 8,9,10,11,12,13,14,15,16 or 17

Even if 2012-11-19 8:00 if free it should return 2012-11-19 9:00.
It should return 8:00 only if there is no free time in 2012-11-19 at 9:00 and 8:00 is first free in succeeding work days.

I tried to fix this by adding 2012-11-19 9: to two places as shown in query below but this query still returns free time at 2012-11-19 8:00.
How to fix this so it returns free time at 2012-11-19 9:00 ?

create table reservation (during tsrange,
 EXCLUDE USING gist (during WITH &&)
 );
create table pyha ( pyha date primary key);
with gaps as (
    select
        upper(during) as start,
        lead(lower(during),1,upper(during)) over (ORDER BY during) - upper(during) as gap
    from (
        select during
          from reservation
             where upper(during)>= '2012-11-19 9:'
       union all values
         ('(,2012-11-19 9:)'::tsrange)
        union all
        select
            unnest(case
                when pyha is not null then array[tsrange(d, d + interval '1 day')]
                when date_part('dow', d) in (0, 6) then array[tsrange(d, d + interval '1 day')]
                else array[tsrange(d, d + interval '8 hours'),
                           tsrange(d + interval '18 hours', d + interval '1 day')]
            end)
        from generate_series(
            '2012-11-19'::timestamp without time zone,
            '2012-11-19'::timestamp without time zone+ interval '3 month',
            interval '1 day'
        ) as s(d)
        left join pyha on pyha = d::date
    ) as x
)

select start,
   date_part('epoch', gap) / (60*60) as hours
  from gaps
where gap > '0'::interval
order by start

Update2

I tried updated answer but it returns wrong data. Complete testcase is:

create temp table reservation  ( during tsrange ) on commit drop;
insert into reservation values(
'[2012-11-19 11:00:00,2012-11-19 11:30:00)'::tsrange );

with gaps as (
    select
        upper(during) as start,
        lead(lower(during),1,upper(during)) over (ORDER BY during) - upper(during) as gap
    from (
        select during
          from reservation
        union all
        select
            unnest(case
                when pyha is not null then array[tsrange(d, d + interval '1 day')]
                when date_part('dow', d) in (0, 6) then array[tsrange(d, d + interval '1 day')]
                when d::date =  DATE'2012-11-19' then array[
                            tsrange(d, '2012-11-19 12:'),  -- must return starting at 12:00
                            tsrange(d + interval '18 hours', d + interval '1 day')]
                else array[tsrange(d, d + interval '8 hours'), 
                           tsrange(d + interval '18 hours', d + interval '1 day')]
            end)
        from generate_series(
            DATE'2012-11-19'::timestamp without time zone,
            DATE'2012-11-19'::timestamp without time zone+ interval '3 month',
            interval '1 day'
        ) as s(d) 
        left join pyha on pyha = d::date
    ) as x 
)

select start,
   date_part('epoch', gap) / (60*60) as tunde
  from gaps 
where gap > '0'::interval
order by start

Observed first row:

"2012-11-19 11:30:00"

Expected :

"2012-11-19 12:00:00"

how to fix ?

Answer 1

You can use generate_series() function in order to mask-out non business hours:

with gaps as (
    select
        upper(during) as start,
        lead(lower(during),1,upper(during)) over (ORDER BY during) - upper(during) as gap
    from (
        select during
        from reservation
        union all
        select
            unnest(case
                when pyha is not null then array[tsrange(d, d + interval '1 day')]
                when date_part('dow', d) in (0, 6) then array[tsrange(d, d + interval '1 day')]
                when d::date = '2012-11-14' then array[tsrange(d, d + interval '9 hours'), tsrange(d + interval '18 hours', d + interval '1 day')]
                else array[tsrange(d, d + interval '8 hours'), tsrange(d + interval '18 hours', d + interval '1 day')]
            end)
        from generate_series(
            '2012-11-14'::timestamp without time zone, 
            '2012-11-14'::timestamp without time zone + interval '2 week', 
            interval '1 day'
        ) as s(d) 
        left join pyha on pyha = d::date
    ) as x 
)
select *
    from gaps
where gap > '0'::interval
order by start

Let me explain some tricky parts:

• you dont have to insert dates for sat/sun into pyha table because you can use date_part('dow', d) function. Use pyha table for public holidays only. ‘dow’ returns 0 or 6 for Sun or Sat respectively.
• public holidays and sat/sun can be represented as single interval (0..24). Weekdays have to be represented by two intervals (0..8) and (18..24) hence unnest() and array[]
• you can specify start date and length in generate_series() function

Based on your update to the question I added another when to case:

when d::date = '2012-11-14' then array[tsrange(d, d + interval '9 hours'), tsrange(d + interval '18 hours', d + interval '1 day')]

The idea is to produce different interval(s) for starting date (d::date = '2012-11-14'): (0..9) and (18..24)