How to Dynamically Generate Columns with Count in SQL for Data Mining?

Use SQL to dynamically generate columns

This article discusses a common problem in the field of data mining: dynamically creating columns based on dynamic data. This challenge arises when data needs to be presented in a user-friendly format, especially when a count of values ​​is required in each dynamically generated column.

Problem Statement

We have three tables: Customers, CustomerRewards and Rewards. The goal is to generate a new table that shows each customer's name and the number of rewards they have in each reward type (e.g. Bronze, Silver, Gold, etc.). However, reward types are dynamic, meaning that new types can be added or removed over time.

Solution: Use the PIVOT function

Static PIVOT:

If the number of reward types is known in advance, we can use a hard-coded PIVOT function. For example:

select name, [Bronze], [Silver], [Gold], [Platinum], [AnotherOne]
from
(
  select c.name,
    cr.description,
    r.typeid
  from customers c
  left join rewards r
    on c.id = r.customerid
  left join customerrewards cr
    on r.typeid = cr.typeid
) x
pivot
(
  count(typeid)
  for description in ([Bronze], [Silver], [Gold], [Platinum], [AnotherOne])
) p;

Dynamic PIVOT:

If the number of reward types may vary, we can use dynamic SQL to perform PIVOT:

DECLARE @cols AS NVARCHAR(MAX),
    @query  AS NVARCHAR(MAX)

select @cols = STUFF((SELECT ',' + QUOTENAME(description)
                    from customerrewards
                    group by description, typeid
                    order by typeid
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)')
        ,1,1,'')

set @query = 'SELECT name,' + @cols + ' from
             (
                select c.name,
                  cr.description,
                  r.typeid
                from customers c
                left join rewards r
                  on c.id = r.customerid
                left join customerrewards cr
                  on r.typeid = cr.typeid
            ) x
            pivot
            (
                count(typeid)
                for description in (' + @cols + ')
            ) p '

execute(@query)

Contains total column

To include the total column we can use ROLLUP:

STATIC ROLLUP:

select name, sum([Bronze]) Bronze, sum([Silver]) Silver,
  sum([Gold]) Gold, sum([Platinum]) Platinum, sum([AnotherOne]) AnotherOne
from
(
  select name, [Bronze], [Silver], [Gold], [Platinum], [AnotherOne]
  from
  (
    select c.name,
      cr.description,
      r.typeid
    from customers c
    left join rewards r
      on c.id = r.customerid
    left join customerrewards cr
      on r.typeid = cr.typeid
  ) x
  pivot
  (
    count(typeid)
    for description in ([Bronze], [Silver], [Gold], [Platinum], [AnotherOne])
  ) p
) x
group by name with rollup

Dynamic ROLLUP:

DECLARE @cols AS NVARCHAR(MAX),
    @colsRollup AS NVARCHAR(MAX),
    @query  AS NVARCHAR(MAX)

select @cols = STUFF((SELECT ',' + QUOTENAME(description)
                    from customerrewards
                    group by description, typeid
                    order by typeid
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)')
        ,1,1,'')

select @colsRollup
      = STUFF((SELECT ', Sum(' + QUOTENAME(description) + ') as ' + QUOTENAME(description)
                    from customerrewards
                    group by description, typeid
                    order by typeid
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)')
        ,1,1,'')


set @query
          = 'SELECT name, ' + @colsRollup + '
             FROM
             (
                SELECT name,' + @cols + ' from
                 (
                    select c.name,
                      cr.description,
                      r.typeid
                    from customers c
                    left join rewards r
                      on c.id = r.customerid
                    left join customerrewards cr
                      on r.typeid = cr.typeid
                ) x
                pivot
                (
                    count(typeid)
                    for description in (' + @cols + ')
                ) p
              ) x1
              GROUP BY name with ROLLUP'

execute(@query)

The above is the detailed content of How to Dynamically Generate Columns with Count in SQL for Data Mining?. For more information, please follow other related articles on the PHP Chinese website!