SQL Find latest date where condition doesn't exist - sql

Question

I'm quite new to SQL and working on a query that has been thoroughly defeating me for a while now. I come to this site often - it's a terrific resource thanks to all of your expertise, and generally I find what I need, but this time I think my query is a bit too specific and I've not found something applicable. Could someone give me a hand, please?
I have two tables: one Client table and one Contact (aka appointment) table. What I need to find are all of the clients' most recent appointment days (before a certain date, in this case '11/08/2015') where the Outcome for any appointment on that day is NOT a '2'. Each client may have more than one appointment on a single day, and an Outcome of '2' for any of those appointments means that we have to ignore the whole day and move back to the next most recent day..
For example, Client '3' should have a returned appointment date of '01/07/2015' (and just one row) and not the two rows for '16/07/2015', because one of the appointments on '16/07/2015' had an Outcome of '2'. All other values for Outcome are acceptable (including NULL), just not '2'.
The multiple appointment on the same day bit is the part that I'm finding tricky - I can find the latest appointment day using a Select MAX (or TOP 1) statement, but when I add on a "<> '2'" it still continues to return the same days that may have an Outcome '2' because other appointments on that same day have another Outcome. I've been trying to play around with my tables and GROUP BY and NOT EXIST, but I don't seem to be making any headway.
Contact
ClientID AppDate Outcome
1 30/07/2015 17:00 2
1 01/07/2015 17:00 3
2 03/03/2015 16:00 NULL
2 01/03/2015 16:00 NULL
3 16/07/2015 15:40 6
3 16/07/2015 15:40 2
3 01/07/2015 15:40 3
4 05/08/2015 12:30 6
4 05/08/2015 12:30 2
4 01/08/2015 12:30 3
5 23/07/2015 15:30 2
5 23/07/2015 15:30 NULL
5 01/07/2015 15:30 4
6 20/07/2015 10:10 NULL
6 20/07/2015 10:10 2
6 01/07/2015 10:10 6
7 23/07/2015 15:40 2
7 01/07/2015 15:40 1
7 23/06/2015 15:40 8
8 13/07/2015 11:30 2
8 13/07/2015 11:30 6
8 01/07/2015 11:30 2
8 01/06/2015 11:30 3
9 29/07/2015 17:00 3
9 29/07/2015 17:00 6
10 14/07/2015 11:00 NULL
10 01/07/2015 11:00 5
Client
ClientID Forename Surname
1 I B
2 J B
3 S C
4 S T
5 P C
6 K D
7 P E
8 P H
9 S F
10 A G
Apologies if I'm missing something glaringly obvious! Thanks for reading and for any responses. I attach my truncated query for your general amusement...
SELECT
cli.ClientID ,
cli.Forename ,
cli.Surname ,
con.AppDate ,
con.Outcome
FROM
Client AS cli
INNER JOIN
Contact AS con
ON cli.ClientID = con.ClientID
AND con.AppDate =
(SELECT MAX(con1.AppDate)
FROM Contact AS con1
WHERE con.ClientID = con1.ClientID
AND con1.AppDate < '11/08/2015 00:00:00'
AND con1.Outcome <> '2')
ORDER BY
cli.ClientID
EDIT:
Thank you to Mr Linoff for the Cross Apply query, it worked perfectly.
Sorry that I didn't include the expected output earlier. For reference (for anyone else working with a similar problem in future) I was looking to obtain:
Appointments
Client ID Act Date and Time Outcome
1 01/07/2015 17:00 3
2 03/03/2015 16:00 NULL
3 01/07/2015 15:40 3
4 01/08/2015 12:30 3
5 01/07/2015 15:30 4
6 01/07/2015 10:10 6
7 01/07/2015 15:40 1
8 01/06/2015 11:30 3
9 29/07/2015 17:00 3
9 29/07/2015 17:00 6
10 14/07/2015 11:00 NULL

Answer 1

I think cross apply is the best approach to this:
select c.*, con.*
from client c cross apply
(select top 1 con.*
from (select con.*,
sum(case when Outcome = 2 then 1 else 0 end) over (partition by ClientId, AppDate) as num2s
from contact con
where con.ClientId = c.ClientId and
con.AppDate < '2015-11-08'
) con
where num2s = 0
order by AppDate desc
) con;
In this case, cross apply works a lot like a correlated subquery, but you can return multiple values. The subquery uses window functions to count the number of "2" on a given day and the rest of the logic should be pretty obvious.
This returns one row from the most recent date with appropriate appointments. If you want multiple such rows, use with ties.

Answer 2

What you need to do is to have a condition which picks out all the Dates where Outcome is two and filter them out.
Something like this:
WITH ClientCTE AS
(
SELECT MAX(con1.AppDate) AS AppDate ,ClientID
FROM Contact AS con1
WHERE con.ClientID = con1.ClientID
AND con1.AppDate < '11/08/2015 00:00:00'
AND con1.Outcome <> '2'
AND con1.AppDate NOT IN (SELECT AppDate FROM Contact WHERE Outcome = '2'))
SELECT
* FROM Client C
INNER JOIN
Contact AS con
ON cli.ClientID = con.ClientID
INNER JOIN ClientCTE CTE
ON cli.ClientID = CTE.ClientCTE
AND CTE.AppDate = con.AppDate
Let me know if it works

Answer 3

Please try the below query.
It uses the MAX(case...) OVER( partition BY CAST(AppDate as DATE) windowed function to flag all appointments which have a 2 outcome on same day along with WHERE clause to filter out dates.
This flag is then used in outer WHERE clause to remove unwanted data and joined on contacts and client tables
SELECT
cli.ClientID ,
cli.Forename ,
cli.Surname ,
con.AppDate ,
con.Outcome
from
client cli right join contact con
on con.ClientID=cli.ClientID
inner join
(
select ClientID,MAX(AppDate)as lastDate from
( select *,
MAX(CASE when ISNULL(Outcome,1) =2 then 1 else 0 end) OVER(PARTITION BY CAST(AppDate as DATE),ClientID ORDER BY AppDate) as flag
from Contact
where AppDate< '2015-11-08'
) p
where flag =0 group by ClientID
) s
on s.ClientId=con.ClientID and s.lastDate=con.AppDate
Also here's the link to sql fiddle for demo:
http://sqlfiddle.com/#!6/c519d/2