Assigning values as new column based on percentages from another table

Question

I have two tables one is a row-level table another is a corresponding table with variables.

Table 1

Table 2

Output Required -

The variable should be populated in a new column based on the % of the variable.

Answer 1

First, don't use symbols as field names; user the word percentage rather than the symbol % .

Second, your mapping table ( table2 ) should probably have the lower and upper bounds to make things simpler later on... (You can accomplish that using window functions if you can't correct the mapping table.)

Then you can use window functions on your data to identify each row in terms of which row it is within just its own group.

Once done, it becomes a relatively simple join...

WITH
  map AS
(
  SELECT
    *,
    SUM(percentage) OVER (PARTITION BY State, Region ORDER BY Variable)  AS upper_bound
  FROM
    Table2  # lower_bound is just upper_bound - percentage
),
  data AS
(
  SELECT
    *,
    ROW_NUMBER() OVER (PARTITION BY State, Region) - 1  AS group_row_number,   # -1 to make the row number start from zero
    COUNT(*)     OVER (PARTITION BY State, Region)      AS group_size
  FROM
    Table1
)
SELECT
  *
FROM
  data
INNER JOIN
  map
    ON  data.Region = map.Region
    AND data.State  = map.State
    AND data.group_row_number >= data.group_size * (map.upper_bound - map.percentage)
    AND data.group_row_number <  data.group_size *  map.upper_bound

Answer 2

Below is for BigQuery Standard SQL

• Version 1

Non-orthodox version with use of RANGE_BUCKET function

#standardSQL
WITH buckets AS (
  SELECT state, region, 
      ARRAY_AGG(variable ORDER BY variable) variables, 
      ARRAY_AGG(percentage ORDER BY variable) bins
  FROM (
    SELECT state, region, variable, SUM(1. * percentage) OVER(win) percentage
    FROM table2  
    WINDOW win AS (PARTITION BY state, region ORDER BY variable)
  )  
  GROUP BY state, region
)
SELECT user, state, region, 
  variables[OFFSET(
    RANGE_BUCKET((ROW_NUMBER() OVER(win) - 1) / (COUNT(1) OVER(win)) * 100, bins)
  )] AS variable
FROM table1
JOIN buckets USING (state, region)
WINDOW win AS (PARTITION BY state, region)
-- ORDER BY user

If to apply to sample data from your question - output is

Row user    state   region  variable     
1   1       ORD     1       ABC  
2   2       ORD     1       ABC  
3   3       ORD     1       ABC  
4   4       ORD     1       XYZ  
5   5       ORD     1       XYZ  
6   6       ORD     1       XYZ  
7   7       IAD     2       ABC  
8   8       IAD     2       ABC  
9   9       IAD     2       ABC  
10  10      IAD     2       ABC  
11  11      IAD     2       AED  
12  12      IAD     2       AED  
13  13      IAD     2       XYZ  
14  14      IAD     2       XYZ    

• Version 2

Below is more traditional version (obviously with same output as in above/first version)

#standardSQL
WITH buckets AS (
  SELECT *, SUM(percentage) OVER(PARTITION BY state, region ORDER BY variable) AS bin
  FROM table2  
), table1_with_stats AS (
  SELECT *, 
    ROW_NUMBER() OVER(win) - 1 AS position, 
    COUNT(*) OVER(win) AS size
  FROM table1
  WINDOW win AS (PARTITION BY state, region)
)
SELECT user, state, region, variable
FROM table1_with_stats
INNER JOIN buckets
USING (state, region)
WHERE position BETWEEN size * (bin - percentage) / 100
  AND size * bin /100 - 1
-- ORDER BY user

• Test Data

You can test, play with above using below CTE

WITH table1 AS (
  SELECT 1 user, 'ORD' state, 1 region UNION ALL
  SELECT 2, 'ORD', 1 UNION ALL
  SELECT 3, 'ORD', 1 UNION ALL
  SELECT 4, 'ORD', 1 UNION ALL
  SELECT 5, 'ORD', 1 UNION ALL
  SELECT 6, 'ORD', 1 UNION ALL
  SELECT 7, 'IAD', 2 UNION ALL
  SELECT 8, 'IAD', 2 UNION ALL
  SELECT 9, 'IAD', 2 UNION ALL
  SELECT 10, 'IAD', 2 UNION ALL
  SELECT 11, 'IAD', 2 UNION ALL
  SELECT 12, 'IAD', 2 UNION ALL
  SELECT 13, 'IAD', 2 UNION ALL
  SELECT 14, 'IAD', 2 
), table2 AS (
  SELECT 'ORD' state, 1 region, 'ABC' variable, 50 percentage UNION ALL
  SELECT 'ORD', 1, 'XYZ', 50 UNION ALL
  SELECT 'IAD', 2, 'ABC', 50 UNION ALL
  SELECT 'IAD', 2, 'XYZ', 25 UNION ALL
  SELECT 'IAD', 2, 'AED', 25 
)