Combining an INNER JOIN query with Count of different values - mysql

Question

I am trying to link two tables with similar column. I need to find out how many values differ from table1.column1 and table 2.column1:
My current query:
SELECT i10_descr.i10_code, gems_pcsi9.i10_code
FROM i10_descr INNER JOIN gems_pcsi9 ON i10_descr.i10_code = gems_pcsi9.i10_code
ORDER BY i10_descr.i10_code;
I know this query shows the matching codes of each table: I cannot figure out how to COUNT the missing/different codes in the tables.
Also, I have to compute the ratio of codes.
Any help, tips, or direction is much appreciated.
Thanks

Answer 1

You could use an anti-join pattern to get a list of i10_code that exist in one table, but not the other. For example:
SELECT i.i10_code
FROM i10_descr i
LEFT
JOIN gems_pcsi9 g
ON g.i10_code = i.i10_code
WHERE g.i10_code IS NULL
ORDER BY i.i10_code
If you just want a count, you could use COUNT(i.i10_code) and/or COUNT(DISINCT i.i10_code) in the SELECT list and remove the ORDER BY clause.
To get the i10_code in the gems table that aren't in the i10 table, you'd do the same thing but invert the query so that gems is the "driving" table. e.g.
SELECT COUNT(DISTINCT g.i10_code) AS cnt_diff
FROM gems_pcsi9 g
LEFT
JOIN i10_descr i
ON i.i10_code = g.i10_code
WHERE i.i10_code IS NULL
If you want to combine the number of differences, you can combine the two queries by making them inline views:
SELECT d.cnt_diff + e.cnt_diff AS total_diff
FROM (
SELECT COUNT(DISTINCT g.i10_code) AS cnt_diff
FROM gems_pcsi9 g
LEFT
JOIN i10_descr i
ON i.i10_code = g.i10_code
WHERE i.i10_code IS NULL
) d
CROSS
JOIN (
SELECT COUNT(DISTINCT i.i10_code) AS cnt_diff
FROM i10_descr i
LEFT
JOIN gems_pcsi9 g
ON g.i10_code = i.i10_code
WHERE g.i10_code IS NULL
) e
NOTE: the COUNT aggregate will omit NULL values. The query would need to be tweaked if you also wanted to "count" rows that had NULL values for i10_code. You'd use COUNT(DISTINCT ) if you want just a number of distinct values that are different. A COUNT() would give a number of rows. These two results would be different if you had multiple rows with the same i10_code value.
To get a "ratio" of codes, assuming that at this point, the "differences" don't matter, you get a count of codes from each table. The queries to do that could be used inline views:
SELECT d.cnt / e.cnt AS ratio_cnt_g_over_cnt_i
, d.cnt AS cnt_g
, e.cnt AS cnt_i
FROM (
SELECT COUNT(DISTINCT g.i10_code) AS cnt
FROM gems_pcsi9 g
) d
CROSS
JOIN (
SELECT COUNT(DISTINCT i.i10_code) AS cnt
FROM i10_descr i
) e

Answer 2

An alternative method is to use union all with aggregation:
select in_i10descr, in_gems_pcsi9, count(*) as numcodes
from (select code, max(in_i10descr) as in_i10descr, max(in_gems_pcsi9) as in_gems_pcsi9
from ((select i10_descr.i10_code as code, 1 as in_i10descr, 0 as in_gems_pcsi9
from i10_descr
) union all
(select gems_pcsi9.i10_code, 0, 1
gems_pcsi9.i10_code
)
) t
group by code
) c
group by in_i10descr, in_gems_pcsi9;
This will calculate counts of things in each table separately and in both tables.