SQL Server performance issue with while loops - sql

Question

I just need advice on how I could speed up my code. I'm supposed to count on yearly base, how the grades of some students are improving and calculate in percentage. Also keep in mind that I have around 100k-150k records per year.
Basically end results look like this, so at end of 20150131, 2% of students had grade A finished with grade B and so on.
Grade Date B C
A 20150131 2% 3%
B 20150131 88% 85%
C 20150131 10% 12%
A 20140131 2% 3%
B 20140131 88% 85%
C 20140131 10% 12%
A 20130131 2% 3%
B 20130131 88% 85%
C 20130131 10% 12%
Input looks like this .. just info about student and his grade on certain date
Student Date Grade
1 20150131 A
2 20150131 C
3 20150131 A
1 20140131 B
2 20140131 B
3 20140131 A
My code looks like this:
WHILE #StartDateInt > #PeriodSpan
BEGIN
while #y <= #CategoriesCount
BEGIN
set #CurrentGr = (Select Grade from #Categories where RowID = #y)
set #CurrentGrCount = (Select COUNT(Students) from #TempTable where Period = #PeriodSpan and Grade = #CurrentGr)
set #DefaultCurrentGr = (Select Grade from #Categories where RowID = #y)
insert into Grade_MTRX (Student, Period, Grades_B, SessionID)
select temp1.Grade, #PeriodNextSpan as Period, COUNT(Grades_B)/#CurrentGrCount as 'Grades_B', #SessionID
from #TempTable temp1
join #TempTable temp2 on temp1.Student = temp2.Student and temp1.Period + 10000 = temp2.Period
where temp1.Grade = #CurrentGr and temp2.Grade = 'C' and temp1.Period = #PeriodSpan
group by temp1.Grade, temp1.Period
update Grade_MTRX set Grades_C = (
select COUNT(Grades_C)/#CurrentGrCount
from #TempTable
where Grade = 'C' and Period = #PeriodNextSpan)
where Category = #CurrentGr and Period = #PeriodNextSpan
end
end
I understand SQL Server doesn't like while loops, as I understand it kills it's performance... But I'm using while inside of while loop... going over years, for each grade and just counting them and... first I insert 1 row of current grade, and then I keep updating that row until its fully populated.
I do understand this is really bad, but at the end that's why I am here to learn better way to accomplish this.
Thank you in advance!

Answer 1

150,000 records per year is really nothing. Let's say you had this Grade table:
CREATE TABLE Grade(
student_id INT,
date INT,
grade CHAR);
With this info:
student_id date grade
1 2013 A
1 2014 A
1 2015 B
2 2013 B
2 2014 A
2 2015 C
3 2013 C
3 2014 A
3 2015 B
Then if you just run a query like:
SELECT this_year.date, last_year.grade AS last_year, this_year.grade AS this_year, COUNT(*) AS total,
(100.0 * COUNT(*)) / (SELECT COUNT(*) FROM Grade WHERE date = this_year.date) AS percent
FROM Grade AS this_year
INNER JOIN Grade AS last_year ON this_year.date = last_year.date + 1
AND this_year.student_id = last_year.student_id
GROUP BY this_year.date, this_year.grade, last_year.grade
ORDER BY 1, 2, 3;
you end up with these results:
date | last_year | this_year | total | percent
------+-----------+-----------+-------+---------------------
2014 | A | A | 1 | 33.3333333333333333
2014 | B | A | 1 | 33.3333333333333333
2014 | C | A | 1 | 33.3333333333333333
2015 | A | B | 2 | 66.6666666666666667
2015 | A | C | 1 | 33.3333333333333333
(5 rows)
Having a few million rows of data with this kind of query shouldn't be any real trouble. Even tens of millions of rows. But if you need things to be faster still then check out windowing functions that you can do with Postgres, Oracle, and MSSQL server.