Combine Output of Queries in Pivot Table with PostgreSQL

Question

Suppose I have a table called orders that looks like this:

id	order date	Orders_Wanted	Orders_Given
1	2020-11-29 19:12:44.417	2	6
1	2020-11-29 20:12:44.417	2	6
1	2020-11-30 23:37:28.692	8	2
1	2020-11-30 23:37:28.692	2	6

How do I write a query that shows the count of orders_wanted - orders_given by hour broken down into two columns, one that counts positive results and one that counts negative results (a note that orders_wanted and orders_given are times, so that's why I am calculating orders_wanted - orders_given). I would also like to add a final column that calculates the percentage of total orders per hour that are positive (count_orders_positive/ (count_orders_negative + Count_orders_positive)).

The output of the query would look something like this:

week	day	hour	count_orders_positive	count_orders_negative	Percentage_orders_positive
48	7	19	0	1	100%
48	7	20	0	1	100%
49	1	23	1	1	50%

So far I am able to get the bottom two results using these queries, but I don't know how to combine them.

SELECT  
extract (week from (order_date at time zone 'MST' at time zone 'UTC') ) as "week", 
extract (isodow from (order_date at time zone 'MST' at time zone 'UTC') ) as "day",
extract (hour from (order_date at time zone 'MST' at time zone 'UTC') ) as "hour",
Count (extract (hour from (order_date at time zone 'MST' at time zone 'UTC') )) as 
"count_orders_positive" 
from orders  
WHERE orders_wanted - orders_given >= 0
group by week, day, hour
order by week, day, hour;

week	day	hour	count_orders_positive
49	1	23	1

SELECT
extract (week from (order_date at time zone 'MST' at time zone 'UTC') ) as "week", 
extract (isodow from (order_date at time zone 'MST' at time zone 'UTC') ) as "day",
extract (hour from (order_date at time zone 'MST' at time zone 'UTC') ) as "hour",
Count (extract (hour from (order_date at time zone 'MST' at time zone 'UTC') )) as 
"count_orders_negative" 
from orders  
WHERE orders_wanted - orders_given < 0
group by week, day, hour
order by week, day, hour;

week	day	hour	count_orders_negative
48	7	19	1
48	7	20	1
49	1	23	1

Answer 1

You can do conditional aggregation. avg() comes handy to compute the percentage:

select  
    extract (week   from (order_date at time zone 'MST' at time zone 'UTC') ) as "week", 
    extract (isodow from (order_date at time zone 'MST' at time zone 'UTC') ) as "day",
    extract (hour   from (order_date at time zone 'MST' at time zone 'UTC') ) as "hour",
    count(*) filter (where orders_wanted - orders_given >= 0) as count_orders_positive,
    count(*) filter (where orders_wanted - orders_given <  0) as count_orders_negative,
    100 * avg((orders_wanted - orders_given >= 0)::int) as percent_orders_positive
from orders
group by week, day, hour
order by week, day, hour;