Finding average value of three weekdays, broken down on hours - sql

Question

I have an origin-destination table like this in Bigquery with weekday, date, UTC time/hour and count of trips:
Origin Destination Day Date Time Count
NY Station Downtown Mon 02.09.2019 15 12
NY Station Downtown Mon 02.09.2019 16 10
City libry Eastside Mon 02.09.2019 17 10
NY Station Downtown Tue 03.09.2019 15 8
NY Station Downtown Tue 03.09.2019 16 5
City libry Eastside Tue 03.09.2019 17 5
NY Station Downtown Wed 04.09.2019 15 8
NY Station Downtown Wed 04.09.2019 16 10
City libry Eastside Wed 04.09.2019 17 11
I wish to get the average Count for
each origin-destination pair (NY Station-Downtown and City libry-Eastside)
the average of Monday-Wednesday at each given time
The output should then be something like
Origin Destination Avg_Day Period Time Avg_Count
NY Station Downtown Mon-Wed Week1 (02.09.19-04.09.19) 15 9,33
NY Station Downtown Mon-Wed Week1 (02.09.19-04.09.19) 16 8,33
City libry Eastside Mon-Wed Week1 (02.09.19-04.09.19) 17 8,67
Ignore the Avg_day and Period columns as its just for help/showing for which days and dates i wish to achieve the average for. In other words the aim is to have an idea of the average counts for each origin-destination pair on a normal weekday (in this case defined as mon-wed) on certain hours of the day. The average count of for example the time 15 for NY Station-Downtown pair is 9,33, calculated by taking the average of the count for 15 o'clock at Monday, at Tuesday and at Wednesday (that is the average of 12, 8 and 8).
I have tried variants of CASE and WHERE SQL queries, but not even close to grasping the logic on how to make the query for this so no point in posting any query. Possibly have to create a temporary table also. Can anyone help me? it is HUGELY appreciated

Answer 1

Below is for BigQuery Standard SQL
#standardSQL
select
Origin,
Destination,
'Mon-Wed' AS Avg_Day,
FORMAT('Week%i (%s-%s)', week, min_date, max_date) AS Period,
Time,
Avg_Count
from (
SELECT
Origin,
Destination,
'Mon-Wed' AS Avg_Day,
EXTRACT(WEEK FROM PARSE_DATE('%d.%m.%Y', date)) week,
MIN(date) AS min_date,
MAX(date) AS max_date,
Time,
ROUND(AVG(count), 2) AS Avg_Count
FROM `project.dataset.table`
WHERE day IN ('Mon', 'Tue', 'Wed')
GROUP BY Origin, Destination, Time, week
)
if to apply to sample data from your question - output is
Row Origin Destination Avg_Day Period Time Avg_Count
1 NY Station Downtown Mon-Wed Week35 (02.09.2019-04.09.2019) 15 9.33
2 NY Station Downtown Mon-Wed Week35 (02.09.2019-04.09.2019) 16 8.33
3 City libry Eastside Mon-Wed Week35 (02.09.2019-04.09.2019) 17 8.67