How to replace nested cursors? - sql-server

Question

I'm in the middle of writing an achievement module for a website we run. This particular bit of code will run on SQL Server at the end of a given timeframe to award the achievements (which will be notified through the client the next time the site is visited).
So I need to look at all the teams and all the "end" achievements (which are independent of the teams). Each achievement has one or more rules that must be met. Once I've determined that the achievement is passed, I award the achievement to all of the participants of that team.
I was handling this through cursors, and I got an error, so when I went to google the problem, I got endless links on forums of "YOU DUMB $&#$ WHY ARE YOU USING CURSORS" (to paraphrase). I figured while I was solving my problem, I may as well replace what I have using a set-based approach (if possible).
The alternative examples I found were all basic update scenarios using single nested loops on tables that have keys back to each other, and honestly, I wouldn't have considered using cursors in those scenarios to begin with.
What I have below is not entirely syntactically correct, but I can figure that out on my own by playing around. This should give a clear idea of what I'm trying to accomplish, however:
declare #TimeframeID int;
declare #AchievementID int;
declare #q1 int;
declare #q2 int;
declare #TeamID int;
declare #TotalReg int;
declare #TotalMin int;
declare #AvgMin decimal(10, 2);
declare #AggType varchar(50);
declare #Pass bit = 1;
declare #Email varchar(50);
declare #ParticipantID int;
select #TimeframeID = MAX(TimeframeID) from Timeframes
where IsActive = 1;
declare team_cur CURSOR FOR
select
t.TeamID,
(select COUNT(1) from Registrations r
where r.TeamID = t.TeamID and r.TimeframeID = #TimeframeID) TotalReg,
(select SUM(Minutes) from Activities a
inner join Registrations r on r.RegistrationID = a.RegistrationID
where r.TeamID = t.TeamID and r.TimeframeID = #TimeframeID) TotalMin
from Teams t
where Active = 1
group by TeamID;
declare ach_cur CURSOR FOR
select AchievementID from luAchievements
where TimeframeID = #TimeframeID and AchievementType = 'End';
declare rule_cur CURSOR for
select Quantity1, Quantity2, AggregateType
from AchievementRule_Links arl
inner join luAchievementRules ar on ar.RuleID = arl.RuleID
where arl.AchievementID = #AchievementID;
open team_cur;
fetch next from team_cur
into #TeamID, #TotalReg, #TotalMin;
while ##FETCH_STATUS = 0
begin
open ach_cur;
fetch next from ach_cur
into #AchievementID;
while ##FETCH_STATUS = 0
begin
open rule_cur;
fetch next from rule_cur
into #q1, #q2, #AggType;
while ##FETCH_STATUS = 0
begin
if #AggType = 'Total'
begin
if #q1 > #TotalReg
begin
set #Pass = 0;
end
end
else if #AggType = 'Average'
begin
print 'do this later; need to get specs';
end
fetch next from rule_cur
into #q1, #q2, #AggType;
end
close rule_cur;
deallocate rule_cur;
-- if passed, award achievement to all team members
if #Pass = 1
begin
declare reg_cursor cursor for
select max(p.Email) from Participants p
inner join Registrations reg on reg.ParticipantID = p.ParticipantID
where reg.TeamID = #TeamID;
open reg_cursor;
fetch next from reg_cursor
into #Email;
while ##FETCH_STATUS = 0
begin
exec ProcessAchievement #AchievementID, #Email, 0;
fetch next from reg_cursor
into #Email;
end
close reg_cursor;
deallocate reg_cursor;
-- award achievement to team
exec ProcessTeamAchievement #AchievementID, #TeamID;
end
fetch next from ach_cur
into #AchievementID;
end
close ach_cur;
deallocate ach_cur;
fetch next from team_cur
into #TeamID, #TotalReg, #TotalMin;
end
close team_cur;
deallocate team_cur;
Is there a set-based alternative to what I'm doing here? I need to add that this is running against a small set of records, so performance isn't my concern; best practices for future gargantuan updates are.

Answer 1

To make this set-based you need to fix the procs you are calling so that they either have a table-valued parameter or the code in the proc joins to a table where you have the records you want to process. In the second case you would either mark them as processsed or delete them when you are finished.