How to shift a year-week field in bigquery - sql

Question

This question is about shifting values of a year-week field in bigquery.
run_id year_week value
0001 201451 13
0001 201452 6
0001 201503 3
0003 201351 8
0003 201352 5
0003 201403 1
Here for each year the week can range from 01 to 53. For example year 2014 has last week which is 201452 but year 2015 has last week which is 201553.
Now I want to shift the values for each year_week in each run_id by 5 weeks. For the weeks there is no value it is assumed that they have a value of 0. For example the output from the example table above should look like this:
run_id year_week value
0001 201504 13
0001 201505 6
0001 201506 0
0001 201507 0
0001 201508 3
0003 201404 8
0003 201405 5
0003 201406 0
0003 201407 0
0003 201408 1
Explanation of the output: In the table above for run_id 0001 the year_week 201504 has a value of 13 because in the input table we had a value of 13 for year_week 201451 which is 5 weeks before 201504.
I could create a table programmatically by creating a mapping from a year_week to a shifted year_week and then doing a join to get the output, but I was wondering if there is any other way to do it by just using sql.

Answer 1

#standardSQL
WITH `project.dataset.table` AS (
SELECT '001' run_id, 201451 year_week, 13 value UNION ALL
SELECT '001', 201452, 6 UNION ALL
SELECT '001', 201503, 3
), weeks AS (
SELECT 100 * year + week year_week
FROM UNNEST([2013, 2014, 2015, 2016, 2017]) year,
UNNEST(GENERATE_ARRAY(1, IF(EXTRACT(ISOWEEK FROM DATE(1+year,1,1)) = 1, 52, 53))) week
), temp AS (
SELECT i.run_id, w.year_week, d.year_week week2, value
FROM weeks w
CROSS JOIN (SELECT DISTINCT run_id FROM `project.dataset.table`) i
LEFT JOIN `project.dataset.table` d
USING(year_week, run_id)
)
SELECT * FROM (
SELECT run_id, year_week,
SUM(value) OVER(win) value
FROM temp
WINDOW win AS (
PARTITION BY run_id ORDER BY year_week ROWS BETWEEN 5 PRECEDING AND 5 PRECEDING
)
)
WHERE NOT value IS NULL
ORDER BY run_id, year_week
with result as
Row run_id year_week value
1 001 201504 13
2 001 201505 6
3 001 201508 3
if you need to "preserve" zero rows - just change below portion
SELECT i.run_id, w.year_week, d.year_week week2, value
FROM weeks w
to
SELECT i.run_id, w.year_week, d.year_week week2, IFNULL(value, 0) value
FROM weeks w
or
SUM(value) OVER(win) value
FROM temp
to
SUM(IFNULL(value, 0)) OVER(win) value
FROM temp

Answer 2

If you have data in the table for all year-weeks, then you can do:
with yw as (
select year_week, row_number() over (order by year_week) as seqnum
from t
group by year_week
)
select t.*, yw5, year_week as new_year_week
from t join
yw
on t.year_week = yw.year_week left join
yw yw5
on yw5.seqnum = yw.seqnum + 5;
If you don't have a table of year weeks, then I would advise you to create such a table, so you can do such manipulations -- or a more general calendar table.