Using CASE where need data from another record - mysql

Question

I'm using a CASE statement in a SELECT statement that arguably should be broken up into smaller parts but I'm one field off completing this task and have passed the point of no return.
Without dropping the whole query in here, the gist of the issue is this:
SELECT
Dogs,
Cats,
COUNT(DISTINCT CASE WHEN my.my_id = 2765 THEN my.data END) AS CountAccounts,
COUNT(DISTINCT CASE WHEN my.my_id = 3347 THEN my.data END) AS CountUpgradedAccounts
my refers to a table that is a lookup table of name value pairs. When my.my_id = 3347 that signifies an "upgraded" account and the data point is the date that the account was upgraded. When my.my_id = 2765 that signifies account creation and the corresponding data point is the accountID.
my It looks like this:
UserID | my_id | Data
374 | 2765 | 8826
487 | 3347 | 2013-09-01
662 | 2765 | 8826
321 | 2765 | 9213
722 | 3347 | 2014-10-14
852 | 2765 | 8826
487 | 2765 | 9213
When my_id = 2765, I'd like the distinct number of accountIDs that it relates to. In the table above that is 2: Accounts 8826 and 9213.
I know this would be really simple if I was pulling data from my only. But my is woven into my query in a way that complicates things.
In fact, here is the query, perhaps the problem will be easier to see. Note the last field being selected in the SELECT statement is the problem. I don;t want to count distinct dates, I want to count distinct accountIDs that have upgraded.:
SELECT
sub.name AS ARName,
sub.desc AS ARDescription,
m.name AS MessageName,
m.subj AS MessageDescription,
clk.type AS EventType,
COUNT(DISTINCT clk.eid) AS CountAdmins,
COUNT(DISTINCT CASE WHEN my.my_id = 3347 THEN clk.eid END) AS CountUpgradeAdmins,
COUNT(DISTINCT CASE WHEN my.my_id = 2765 THEN my.data END) AS CountAccounts,
COUNT(DISTINCT CASE WHEN my.my_id = 3347 THEN my.data END) AS CountUpgradedAccounts # <-- THIS LINE IS THE PROBLEM
FROM
bata.sseq seq
INNER JOIN bata.messages m
ON m.id = seq.mid
INNER JOIN bm_arc.clicks208 clk
ON clk.camp = seq.camp
INNER JOIN bemails.cid cid
ON cid.id = clk.eid
INNER JOIN bonfig.sub
ON sub.id = seq.sid
LEFT JOIN bemails.my208 my
ON cid.id = my.eid AND (my_id = 3347 OR my_id = 2765) # only return people who upgraded and accountIDs
WHERE
seq.cid = 208
AND
sub.desc REGEXP '^Home pg free trail (A|B)'
GROUP BY
ARName,
ARDescription,
MessageName,
MessageDescription,
EventType
I've found trying to word this question challenging so sorry if what I'm asking is not clear. If there's any more info I can add let me know.
Following discussion, what I'm asking for in other words:
"For each instance of 3347 get the corresponding instances of UserID
and with those UserIDs the count of distinct corresponding datapoints
in my.data WHERE my.my_id = 2765"

Answer 1

Maybe try changing the problem line to, select UserId, not data:
SELECT
sub.name AS ARName,
sub.desc AS ARDescription,
m.name AS MessageName,
m.subj AS MessageDescription,
clk.type AS EventType,
COUNT(DISTINCT clk.eid) AS CountAdmins,
COUNT(DISTINCT CASE WHEN my.my_id = 3347 THEN clk.eid END) AS CountUpgradeAdmins,
COUNT(DISTINCT CASE WHEN my.my_id = 2765 THEN my.data END) AS CountAccounts,
COUNT(DISTINCT my2.data ) AS CountUpgradedAccounts
FROM
bata.sseq seq
INNER JOIN bata.messages m
ON m.id = seq.mid
INNER JOIN bm_arc.clicks208 clk
ON clk.camp = seq.camp
INNER JOIN bemails.cid cid
ON cid.id = clk.eid
INNER JOIN bonfig.sub
ON sub.id = seq.sid
LEFT JOIN bemails.my208 my
ON cid.id = my.eid AND (my.my_id = 3347 OR my.my_id = 2765) # only return people who upgraded and
LEFT JOIN bemails.my208 my2
ON my2.my_id = 2765 and my2.userID = my.userID and my.my_id=3347 #get the accounts that the user belongs to
WHERE
seq.cid = 208
AND
sub.desc REGEXP '^Home pg free trail (A|B)'
GROUP BY
ARName,
ARDescription,
MessageName,
MessageDescription,
EventType