Regarding Pivoting

Question

Folks,
I have a table 'a' as follows:
VAL1     VAL2     VAL3
1     2     100
1     3     200
2     2     300
2     3     400Now I need the output as follows after writing a query:
1 100 200
2 300 400I have accomplished this by pivoting as follows after reading asktom articles:
SELECT   val1,
         MAX (DECODE (val2, 2, val3, NULL)) a,
         MAX (DECODE (val2, 3, val3, NULL)) b
    FROM a
GROUP BY val1My question is I am not still clear on how this internallly works. If someone can explain what happens stepwise.....
Specially while using max, Lets say I ignore Max function (though it will throw an error), at a given point how would the data look like? 1 100 200 and then 2 300 400 and then max would give me the same values.... or is it that at a given point data would look like 1 100 null, 1,null,200, 2,200,null , 2,null,100 ???
I mean would really appreciate how each step happens.
Thanks very much
saff

Answer 1

Hi,
You want to understand how you got this result set:
val1 a    b
==== =    =
1    100  200
2    300  400Let's look at any one cell of this output, say the column a for the row where val1=1. The number in that cell is 100. How did that number get there? First of all, you're GROUPing BY val1, so everything on that row has to have come from a row where val1=1.
But there were two rows in the table with val1=1. How did 100 get picked? Look at the expression that produces the a column:
        MAX (DECODE (val2, 2, val3, NULL)) a,The DECODE expression says, "If val2=2, then return val3; otherwise, return NULL." That value (either val3 or NULL) will thenbe processed by MAX. Like all aggregate functions, MAX, ignores NULLs, so only the rows with val2=2 will do anything. So this cell will contains the greatest val3 value from all the rows that have val1=1 and val2=2. There is only one such row, so it will necessarily be the MAX. (In fact, if your table is unique for val1 and val2, then all such cells will have a most one value that contributes to it, and it won't matter if the aggregate function you use for the pivot is MAX, MIN, SUM or AVG: when there is only one non-NULL number in the group, they all return that number.)
Using DECODE like this is sort of like having a WHERE-clause that only applies to a single column. The only difference is that a WHERE-clause quits looking at the row, but DECODE turns it into NULL. (MAX, then, quits looking at the row, so the net result is the same.) To undestand better how DECODE works like WHERE, try re-writing any query that has a WHERE-clause to use DECODE (or CASE) instead.
The important thing about pivots is that the data you start with has to identify to which row and column of output each row of input go. In your example, val1 tells to which row it will go, and val2 tells which column (val2=2 goes to column a, val2=3 to column b, anything else isn't displayed.)
If you want to know what the data looks lie at any point in time, try it and see! Try this query
SELECT    val1
,         val2
,         DECODE (val2, 2, val3, NULL) a
,         val3
,         DECODE (val2, 3, val3, NULL) b
FROM      a
ORDER BY  val1
,         val2;The query above will show you what the data looks like before the GROUP BY.