Calculate mean, median and mode for each column in table and transpose so there is one row per column and the aggregates are columns (not using union) - sql

Question

Let's say I have a table with numerical columns x,y & z. For data quality purposes I want to calculate the mean, median and mode for each column and present as rows, like so:
https://ibb.co/HFz1qTw
The end goal is to create an up-to-date table like the one in sys.all_columns (I'm not admin)
Is there an elegant way to do this (preferably dynamically so I don't have to enter every column name, while not using UNION) ?

Answer 1

Feasible but complex.
Here are the steps :
1 : Generates dynamically the SQL to compute mean, median... for the columns you want
2 : Run the query as dynamic cursor.
If you need it to like really easy to use at the end, you can use a pipelined function :
3 : So basically create a function that takes as an argument the table name and will be used like :
select * from table(MyPipelinedFunction(MyTableName));
4 : And you will easily get the result
Here are some tips on each step :
Step 1 : generate SQL dynamically
SQL := 'Select average(#FIELD1#), mean or median(#FIELD2#)... from #TABLE#';
SQL := replace(SQL,'#FIELD1#','Whatever');
SQL := replace(SQL,'#TABLE#','Whatever');
Step 2 : Run the query as dynamic cursor.
Declare
RefCursor SYS_REFCURSOR;
Mean number;
Median nuber;
...
BEGIN
OPEN RefCursor FOR SQL;
FETCH RefCursor INTO Mean, Median...;
CLOSE RefCursor;
Step 3 : pipelined function
TYPE Statistics IS RECORD(Mean NUMBER, MEdian Number, ...);
TYPE StatisticsTable IS TABLE OF Statistics;
FUNCTION ComputeStatistics(TableName in varchar2) RETURN Statistics PIPELINED IS
S Statistics;
BEGIN
-- Get the table field names in case they are different using a static cursor
SELECT * FROM USER_TAB_COLUMNS where table_name = upper(TableName);
-- Use the results to build the SQL query and run the dynamic cursor as in Step 1 & 2
-- After this step, results are stored in the variables Mean, Median...
-- Output result
S := Statistics(Mean, Median, ...);
PIPE ROW (S);
RETURN;
END ComputeStatistics;
Step 4 : Get the result
select * from table(ComputeStatistics('WhateverTable');

Answer 2

I would just unpivot and aggregate:
select c.col,
min(c.val), max(c.val), avg(c.val), median(c.val), stats_mode(c.val)
from t cross join lateral
(select 'x' as col, x as val from dual union all
select 'y', y from dual union all
select 'z', z from dual
) c
group by c.col;
You can turn this into dynamic SQL using PL/SQL. In practice, though, I would just query for the column names and generate the SQL directly:
select replace('
select [col] as col,
min(t.[col]), max(t.[col]), avg(t.[col]), median(t.[col]), stats_mode(t.[col])
from t t union all
', '[col]', column_name)
from sys.all_tab_columns c join
sys.all_tables t
on c.owner = t.owner and
c.table_name = t.table_name;
You can then just remove the last union all and run the code.
Note: This doesn't use listagg() because of limits on string length. But you can combine this logic into a single string if you don't have too many columns.