Looking for a special way to join on SQL server - sql-server

Question

I would like to join two tables but couldn't find any existing joins(I tried left, right, full, cross) to do that.
I want to combine table 1 and table 2 into the table 3.
The data order was based on chronological order, and I would like to see the same order in desired table.
TABLE 1:
Student---- Score1
A------------ 90
A------------ 80
B------------ 85
B------------ 60
C------------ 50
C------------ 40
Table2:
Student---- Score2
A------------ 66
A------------ 70
A------------ 85
B------------ 60
C------------ 40
Table 3: Desired Table
Student---- Score1-----Score2
A------------ 90 ----------- 66
A------------ 80 ----------- 70
A------------null -----------85
B------------ 85 ----------- 60
B------------ 60 ----------- null
C------------ 50 ----------- 40
C------------ 40 ----------- null
Thank you!

Answer 1

Ok, we need as first thing try to find a way to add a positional column to your table at runtime. This can be done with ROW_NUMBER() function:
SELECT *, ROW_NUMBER() OVER (PARTITION BY Student ORDER BY Student) Position FROM Table1
SELECT *, ROW_NUMBER() OVER (PARTITION BY Student ORDER BY Student) Position FROM Table2
This creates a nice Position column in our result:
Student Score1 Position
---------- ----------- --------------------
A 90 1
A 80 2
B 85 1
B 60 2
C 50 1
C 40 2
(6 rows affected)
Student Score2 Position
---------- ----------- --------------------
A 66 1
A 70 2
A 85 3
B 60 1
C 40 1
(5 rows affected)
Now we need to join these two temporary results. Since you want to include all the rows from each table, leaving empty (NULL) the spaces left from non-matching rows. FULL OUTER JOIN comes to the rescue, in all its beauty:
SELECT *
FROM (SELECT *, ROW_NUMBER() OVER (PARTITION BY Student ORDER BY Student) Position FROM Table1) T1
FULL OUTER JOIN
(SELECT *, ROW_NUMBER() OVER (PARTITION BY Student ORDER BY Student) Position FROM Table2) T2
ON T1.Student = T2.Student AND T1.Position = T2.Position
We get this:
Student Score1 Position Student Score2 Position
---------- ----------- -------------------- ---------- ----------- --------------------
A 90 1 A 66 1
A 80 2 A 70 2
NULL NULL NULL A 85 3
B 85 1 B 60 1
B 60 2 NULL NULL NULL
C 50 1 C 40 1
C 40 2 NULL NULL NULL
(7 rows affected)
Now just select what you are interested in:
SELECT COALESCE(T1.student, T2.student) Student,
T1.score1,
T2.score2
FROM (SELECT *, ROW_NUMBER() OVER (PARTITION BY Student ORDER BY Student) Position FROM Table1) T1
FULL OUTER JOIN
(SELECT *, ROW_NUMBER() OVER (PARTITION BY Student ORDER BY Student) Position FROM Table2) T2
ON T1.Student = T2.Student AND T1.Position = T2.Position
And voilà:
Student score1 score2
---------- ----------- -----------
A 90 66
A 80 70
A NULL 85
B 85 60
B 60 NULL
C 50 40
C 40 NULL
(7 rows affected)
Be aware though: with many records, this could not be the most efficient way of storing and retrieving your data...
Edit: what follows has been added after answer acceptance
Really important: since a small diatribe is born in comments, let's state the obvious.
The database design proposed by OP has many defect, for first it's based on the assumption that the order of the records in the table will always be the one in which the records have been inserted.
This could not be true and my solution can not work as expected until some more robust way of sorting records is implemented.
Would it be better to add a CreatedAt column to both tables, of type datetime, in which to store record insert date:
ALTER TABLE dbo.Table1 ADD
CreatedAt datetime NOT NULL CONSTRAINT DF_Table1_CreatedAt DEFAULT getdate()
ALTER TABLE dbo.Table2 ADD
CreatedAt datetime NOT NULL CONSTRAINT DF_Table2_CreatedAt DEFAULT getdate()
This could allow to more safely order the records.
The solution would change as follows:
SELECT COALESCE(T1.student, T2.student) Student,
T1.score1,
T2.score2
FROM (SELECT *, ROW_NUMBER() OVER (PARTITION BY Student ORDER BY CreatedAt) Position FROM Table1) T1
FULL OUTER JOIN
(SELECT *, ROW_NUMBER() OVER (PARTITION BY Student ORDER BY CreatedAt) Position FROM Table2) T2
ON T1.Student = T2.Student AND T1.Position = T2.Position

Answer 2

Well, technically:
with t1 as (
select *, ord = row_number() over(partition by student order by score1 desc)
from table1
),
t2 as (
select *, ord = row_number() over(partition by student order by score2)
from table1
)
select student = isnull(t1.student, t2.student),
t1.score1,
t2.score2
from t1
full join t2 on t1.student = t2.student and t1.ord = t2.ord;
But I doubt your desire is to order scores in table1 in descending order and scores in table2 in ascending order. So you're going to have to pin that down. Is there a column for time the test was administered? Probably best to order by that.
Better still, do this in your front end software, such as with SSRS, crystal reports, or the like. I say this because I assume this is for a reporting need where the rows don't really represent 'records' anymore.