Check for uniqueness within update statement - sql

Question

I have a table that links a class to the students that are in that class:
create table class_student
(class_id int,
student_id int,
constraint class_student_u unique nonclustered (class_id, student_id))
If I want to transfer all the classes from one student to another (remove one student from all the classes he/she is enrolled in and add another student to each of the classes the old student was enrolled in), I use the following query:
update class_student
set student_id = #newStudent
where student_id = #oldStudent
and class_id not in (select class_id
from class_student
where student_id = #newStudent)
delete from class_student
where student_id = #oldStudent
How can I transfer the classes from more than one student to the new student? I can't just put where student_id in (#oldStudent1, #oldStudent2) because if both old students are in the same class, after running the above query there will be a violation of the unique constraint. Also, I'd like to do the update in as few queries if possible (I could just run the above queries twice, but I'd like to do it in fewer).
I'm using SQL Server 2008 R2.
Edit: To clarify, here's an example:
class_id student_id
===================
1 1
1 2
2 3
3 1
3 3
4 2
4 3
This means that student 1 is in class 1 and 3, student 2 is in class 1 and 4, and student 3 is in class 2, 3, and 4. If I want to transfer all the classes from student 1 to student 3, I would run the following query:
update class_student
set student_id = 3
where student_id = 1
and class_id not in (select class_id
from class_student
where student_id = 3)
delete from class_student
where student_id = 1
Our data would look like this:
class_id student_id
===================
1 3
1 2
2 3
3 3
4 2
4 3
If, instead, I had run this query:
update class_student
set student_id = 3
where student_id in (1, 2)
and class_id not in (select class_id
from class_student
where student_id = 3)
delete from class_student
where student_id in (1, 2)
Ignoring the unique constraint on the table, the data would look like this:
class_id student_id
===================
1 3
1 3
2 3
3 3
4 3
The double (1, 3) record is what I'm trying to avoid, because it will cause a unique constraint violation in the table.

Answer 1

When setting up the original table you should always include a unique row id with which to reference any specific row (please see below the 'identity' column called row_id):
DROP TABLE class_student
create table class_student
(
row_id int identity(1,1),
class_id int,
student_id int,
constraint class_student_u unique nonclustered (class_id, student_id)
)
insert class_student (class_id,student_id) values (1,1)
insert class_student (class_id,student_id) values (1,2)
insert class_student (class_id,student_id) values (2,3)
insert class_student (class_id,student_id) values (3,1)
insert class_student (class_id,student_id) values (3,3)
insert class_student (class_id,student_id) values (4,2)
insert class_student (class_id,student_id) values (4,3)
In a situation where students 1 and 2 are leaving and you are passing any classes they were taking to student 3 (unless student 3 is already attending those classes), the code could
look something like this:
WITH CTE
AS
(
SELECT row_Id,class_id,student_id,RN = ROW_NUMBER()OVER(PARTITION BY
class_id ORDER BY class_id) FROM class_student WHERE student_id in (1,2,3)
)
DELETE FROM class_student where class_id in (select class_id from
class_student group by class_id having count(class_id) > 1) and student_id
<> 3 and row_id not in (select row_id from cte where student_id <> 3 and
rn >= 2)
Update class_student set student_id = 3
I am using a 'common table expression' with 'RANK' to number each class_id according to the number of rows bearing the same class_id. To see this you can run the code below after
creating the class_student table and inserting the data (see top) but before you run the CTE code above:
WITH CTE
AS
(
SELECT row_Id,class_id,student_id,RN = ROW_NUMBER()OVER(PARTITION BY
class_id ORDER BY class_id) FROM class_student WHERE student_id in (1,2,3)
)
SELECT * FROM CTE
Because class_id 1,3 and 4 are duplicated, they have a value of 2 in the RN (Row Number) column.
I'm using this result in the CTE to delete the rows we don't need from the class_student table and this is where the importance of always having a unique row_id can be seen.
The Delete query deletes rows in the class_student table which are Class ID duplicates. In the case of a class attended by both student 3 and one or both of the other students it
takes the rows where the Student ID is not 3 (because Student 3 is not leaving).
To do this successfully (without taking rows that we want to retain to be assigned to student 3), it requires (by comparing row_id's) that rows where RN = 2 (i.e. class_id is duplicated)
and student_id is not 3 are retained so that we keep one of the rows for Classes that both student 1 and 2 were doing but student 3 was not (i.e. where neither student_id was 3).
Finally, update all remaining rows in the table to a student_id of 3 so that Student 3 gets all the courses.
To see the result you can run:
select * from class_student

Answer 2

I think you'll need at least 2 DML statements to accomplish your goal. And if you really need it to happen in one go, then you can wrap the statements in a stored procedure.
insert into class_student (class_id, student_id)
select distinct class_id, #newStudent
from class_student
where student_id in (#oldStudent1, #oldStudent2)
and class_id not in (select class_id
from class_student
where student_id = #newStudent);
delete from class_student
where student_id in (#oldStudent1, #oldStudent2);
EDIT: Fixed insert to include the "not in" clause.