Create pivot table using KornShell - shell

Question

I am trying to create a pivot table (csv) using KornShell (ksh).
This is the data (csv):
name a1 a2 amount
----------------------------
product1 | 1 | 1000 | 1.5
product1 | 2 | 2000 | 2.6
product1 | 3 | 3000 | 1.2
product1 | 1 | 3000 | 2.1
product1 | 2 | 3000 | 4.1
product1 | 3 | 2000 | 3.1
The result should be:
__| a2| 1000 | 2000 | 3000
a1 \----------------------
1 | 1.5 2.1
2 | 2.6 4.1
3 | 3.1 1.2
I want to "group" the data by the two attributes and create a table which contains the sums of the amount column for the respective attributes.
EDIT: The attributes a1 and a2 are dynamic. I do not know which one of them is going to exist and which not, or how many attributes there will be at all.

Answer 1

It sounds like you are not using the database as effectively as you might. Here is a thought to help you make some headway: You can generate SQL using the shell.
You know what the beginning of each query is going to look like:
select a1,
So you would want to start building some report SQL:
Report_SQL="select a1, "
You then would need to get the list of SQL statements for an arbitrarily sized set of columns for the pivot report(in MySQL - Other databases would require || concatenation):
select distinct concat('sum(case a2 when ', a2, ' then amount else null end) as "_', a2,'",')
from my_database_table
order by 1
;
Because this is in the shell, it is easy to pull this into a variable as follows:
SQL=" select distinct concat('sum(case a2 when ', a2, ' then amount else null end) as _', a2,',') "
SQL+=" from my_database_table "
SQL+=" order by 1 "
# You would have to define a runsql command for your database platform.
a2_columns=$(runsql "$SQL")
At this point you will have an extra comma at the end of the a2_columns variable. That is easily removed:
a2_columns=${a2_columns%,}
Now we can concatenate these variables to create the report SQL that you really seem to need:
Report_SQL+="${a2_columns}"
Report_SQL+=" from my_database_table "
Report_SQL+=" group by 1"
Report_SQL+=" order by 1"
The resulting report SQL would look something like this:
select a1,
sum(case a2 when 1000 then amount else null end) as _1000,
sum(case a2 when 2000 then amount else null end) as _2000,
sum(case a2 when 3000 then amount else null end) as _3000
from my_database_table
group by 1
order by 1
;
The formatting of the report header is left as an exercise to the reader. :)