Summarise multiple columns at once in MySQL - mysql

Question

I have some data (~70,000 rows) that is in a similar format to the below.
+-----------+-----+-----+----+-----------+
| ID | A | B | C | Whatever |
+-----------+-----+-----+----+-----------+
| 1banana | 42 | 0 | 2 | Um |
| fhqwhgads | 514 | 6 | 9 | Nevermind |
| 2banana | 69 | 42 | 0 | NULL |
| pears | 18 | 96 | 2 | 8.8 |
| zubat2 | 96 | 2 | 14 | "NULL" |
+-----------+-----+-----+----+-----------+
I want to make an output table that counts how many times each number occurs in any of the three columns, such as:
+--------+---------+---------+---------+-----+
| Number | A count | B count | C count | sum |
+--------+---------+---------+---------+-----+
| 0 | 0 | 1 | 1 | 2 |
| 2 | 0 | 1 | 2 | 3 |
| 6 | 0 | 1 | 0 | 1 |
| 9 | 0 | 0 | 1 | 1 |
| 14 | 0 | 0 | 1 | 1 |
| 18 | 1 | 0 | 0 | 1 |
| 42 | 1 | 1 | 0 | 2 |
| 69 | 1 | 0 | 0 | 1 |
| 96 | 1 | 1 | 0 | 2 |
| 514 | 1 | 0 | 0 | 1 |
+--------+---------+---------+---------+-----+
(In my real-world use, there would be at least 10 times as many rows in the input table than in the query result)
Whether or not the query returns a row of zeros for numbers that aren't anywhere in those 3 columns isn't that important, as is a lack of a distinct sum column (though my preferences are that it does have the sum column and numbers not in any column are excluded).
Currently, I am using the following query to get ungrouped data:
SELECT * #Number, COUNT(DISTINCT A), COUNT(DISTINCT B), COUNT(DISTINCT C)
FROM
( # Generate a list of numbers to try
SELECT #ROW := #ROW + 1 AS `Number`
FROM DataTable t
join (SELECT #ROW := -9) t2
LIMIT 777 # None of the numbers I am interested in should be greater than this
) AS NumberList
INNER JOIN DataTable ON
Number = A
OR Number = B
OR Number = C
#WHERE <filters on DataTable columns to speed things up>
#WHERE NUMBER = 10 # speed things up
#GROUP BY Number
The above query with the commented-out parts of the code left as they are returns a table similar to the data table, but sorted by which number of the entry it matches. I would like to group together all rows starting with the same Number, and have the values in the "data" columns of the query result be the count of how many times the Number occured in the corresponding column of DataTable.
When I uncomment the grouping statements (and delete the * from the SELECT statement), I can get the count of how many rows each Number appeared in (useful for the sum column of the desired output). However, it does not give me the actual totals of how many times the Number matched each data column: I just get three copies of the number of rows where Number was found. How do I get the groupings to be by each actual column instead of the total number of matching rows?
Additionally, you may have noticed that I have some lines with comments regarding speeding things up. This query is slow, so I added a couple filters so testing it runs faster. I would very much like some way to make it run fast so that sending the results of the query from the complete set to a new table is not the only reasonable way to re-use this data, since I would like to have the ability to play around with the filters on DataTable for non-performance reasons. Is there a better way to structure the overall query so that it runs faster?

Answer 1

I think you want to unpivot using union all and then an aggregation:
select number, sum(a) as a, sum(b) as b, sum(c) as c, count(*) as `sum`
from ((select a as number, 1 as a, 0 as b, 0 as c from t
) union all
(select b, 0 as a, 1 as b, 0 as c from t
) union all
(select c, 0 as a, 0 as b, 1 as c from t
)
) abc
group by number
order by number;