Bigquery mysql to adjust dates for active weekdays - mysql

Question

In a previous question the answer was given to perform a query which would get revenue from the last year in a new column. This worked perfectly, but now I have a follow-up question. (please review this link to have a look at the previous question)
The query used to get this data (With thanks to Mikhail):
#standardSQL
SELECT
a.date, a.location, a.revenue,
DATE_SUB(a.date, INTERVAL 1 YEAR) date_last_year,
IFNULL(b.revenue, 0) revenue_last_year
FROM `project.dataset.table` a
LEFT JOIN `project.dataset.table` b
ON a.location = b.location
AND DATE_SUB(a.date, INTERVAL 1 YEAR) = b.date
The simplified outcome looks as follows (limited to 1 location):
date revenue revenue_last_year
2019-01-31 1471,2577 2185,406
2019-01-30 1291,1111 4723,7439
2019-01-29 2178,6532 2263,5283
2019-01-28 1531,8021 0
2019-01-26 1578,1247 2446,6234
2019-01-25 1299,644 1522,4537
2019-01-24 788,2669 1979,104
2019-01-23 787,441 4117,7927
2019-01-22 2437,2951 1876,2479
2019-01-21 1071,0476 0
2019-01-19 2291,0456 2289,8657
The follow up question relates to working with weekdays from last year. As you can see the revenue last year has '0' values. That's because it's a day that location A was closed. However, to make an accurate comparison on a day by day basis we need to locate the revenue for the day this revenue was open.
For more information, have a look at this table below to see what it looks like when we take the last 10 days of Jan this year and append two columns with the revenues of last year found manually:
date revenue revenue_last_year date revenue
2019-01-31 1471,2577 2185,406 2018-01-31 2185,406
2019-01-30 1291,1111 4723,7439 2018-01-30 4723,7439
2019-01-29 2178,6532 2263,5283 2018-01-29 2263,5283
2019-01-28 1531,8021 0 2018-01-27 2843,6616
2019-01-26 1578,1247 2446,6234 2018-01-26 2446,6234
2019-01-25 1299,644 1522,4537 2018-01-25 1522,4537
2019-01-24 788,2669 1979,104 2018-01-24 1979,104
2019-01-23 787,441 4117,7927 2018-01-23 4117,7927
2019-01-22 2437,2951 1876,2479 2018-01-22 1876,2479
2019-01-21 1071,0476 0 2018-01-20 2561,4086
2019-01-19 2291,0456 2289,8657 2018-01-19 2289,8657
Please note the differences in dates.
What would be a good way to solve this? Would it be necessary to adjust to a query for weekdays and how would you approach this?

Answer 1

Below is for BigQuery Standard SQL
#standardSQL
SELECT
a.date, a.location, ANY_VALUE(a.revenue) revenue,
ARRAY_AGG(
STRUCT(b.date AS date_last_year, b.revenue AS revenue_last_year)
ORDER BY b.date DESC LIMIT 1
)[OFFSET(0)].*
FROM `project.dataset.table` a
CROSS JOIN `project.dataset.table` b
WHERE a.location = b.location
AND b.date BETWEEN DATE_SUB(DATE_SUB(a.date, INTERVAL 1 YEAR), INTERVAL 7 DAY) AND DATE_SUB(a.date, INTERVAL 1 YEAR)
GROUP BY a.date, a.location
You can test, play with above using dummy/sample data (I used same as in my answer for your previous question) as in below example
#standardSQL
WITH `project.dataset.table` AS (
SELECT DATE '2018-02-20' `date`, 'A' location, 1 revenue UNION ALL
SELECT '2018-02-21', 'A', 3 UNION ALL
SELECT '2019-02-20', 'A', 5 UNION ALL
SELECT '2019-02-21', 'A', 7 UNION ALL
SELECT '2019-02-22', 'A', 9 UNION ALL
SELECT '2018-02-20', 'B', 2 UNION ALL
SELECT '2018-02-22', 'B', 4 UNION ALL
SELECT '2019-02-20', 'B', 6 UNION ALL
SELECT '2019-02-21', 'B', 8 UNION ALL
SELECT '2019-02-22', 'B', 10
)
SELECT
a.date, a.location, ANY_VALUE(a.revenue) revenue,
ARRAY_AGG(
STRUCT(b.date AS date_last_year, b.revenue AS revenue_last_year)
ORDER BY b.date DESC LIMIT 1
)[OFFSET(0)].*
FROM `project.dataset.table` a
CROSS JOIN `project.dataset.table` b
WHERE a.location = b.location
AND b.date BETWEEN DATE_SUB(DATE_SUB(a.date, INTERVAL 1 YEAR), INTERVAL 7 DAY) AND DATE_SUB(a.date, INTERVAL 1 YEAR)
GROUP BY a.date, a.location
-- ORDER BY a.date, a.location
with result
Row date location revenue date_last_year revenue_last_year
1 2019-02-20 A 5 2018-02-20 1
2 2019-02-20 B 6 2018-02-20 2
3 2019-02-21 A 7 2018-02-21 3
4 2019-02-21 B 8 2018-02-20 2
5 2019-02-22 A 9 2018-02-21 3
6 2019-02-22 B 10 2018-02-22 4
Note the differences in dates :o)