Is it possible to run calculations while aggregating in Google BigQuery - mysql

Question

In StandardSQL, is it possible to run operations on each row during the grouping process? I'm not sure if I'm even asking the right question. Here's an example.
Let's say I have 3 rows like this:
| move_id | item_id | quantity | value |
|---------|---------|----------|-------|
| 1 | 1 | 10 | 100 |
| 1 | 2 | 20 | 150 |
| 1 | 3 | 30 | 200 |
I now want to group the table by move_id, summing values based on the proportion of each row's quantity to the minimum quantity.
For example the minimum quantity is 10, and row 2 has a quantity of 20, which means it's value should be cut in half before summing. Row 3 has a quantity of 30, which means it's value should be cut to a third before summing.
So my final value column should be 100 + (150 / 2) + (200 / 3) = 241.67.
My result should be:
| move_id | quantity | value |
|---------|----------|--------|
| 1 | 10 | 241.67 |
The query should be something like:
SELECT ANY_VALUE(move_id) AS move_id, MIN(quantity) AS quantity, SUM([THIS IS MY QUESTION, WHAT GOES HERE?]) as value FROM table GROUP BY move_id;
Is this possible?

Answer 1

Below is for BigQuery Standard SQL and does all in one shot
#standardSQL
SELECT move_id,
MIN(quantity) AS quantity,
SUM(value/quantity) * MIN(quantity) AS value
FROM `project.dataset.table`
GROUP BY move_id
If to apply to sample data from your question - result is
Row move_id quantity value
1 1 10 241.66666666666669
As you can see here - instead of splitting calculation/aggregation inside query, you can rather transform your formula such like below
100 + (150 / 2) + (200 / 3)
(100 * 10 / 10 + (150 * 10 / 20) + (200 * 10 / 30)
((100 / 10 + (150 / 20) + (200 / 30)) * 10
SUM(value / quantity) * MIN(quantity)
so, you ended up with just simple aggregation "in one shot"

Answer 2

The somewhat difficult part of your query is that you want to aggregate, but the sum you have in mind itself requires the result of an aggregation - the minimum quantity per each move_id group. One option here would be to first generate the min quantity in a CTE, then aggregate that CTE using your logic.
WITH cte AS (
SELECT *, MIN(quantity) OVER (PARTITION BY move_id) min_quantity
FROM yourTable
)
SELECT
move_id,
MIN(quantity) AS quantity,
SUM(value * min_quantity / quantity) AS value
FROM cte
GROUP BY
move_id;
Demo
Note: The above demo uses SQL Server, but the SQL used is ANSI compliant, and should also run on BigQuery without any issues.
Also, if your version of BigQuery does not support cte, then you may just inline the code contained inside the CTE as as a subquery.

Answer 3

In absence of CTEs, you can use Derived Table (subquery) to fetch minimum quantity for every move_id separately. And, then utilize them in the main query, to compute the sum:
SELECT t.move_id,
dt.min_quantity,
Sum(t.value / ( t.quantity / dt.min_quantity )) AS value
FROM your_table AS t
JOIN (SELECT move_id,
Min(quantity) AS min_quantity
FROM your_table
GROUP BY move_id) AS dt
ON dt.move_id = t.move_id
GROUP BY t.move_id
SQL Fiddle Demo