I've been trying to wrap my head around creating aggregates in PostgreSQL (either 8.4 or 9.1) that accept one or more option parameters.
An example would be creating a PL/R extension to compute the p-th quantile, with 0 <= p <= 1. This would look something like quantile(x,p), and as part of a query:
select category,quantile(x,0.25)
from TABLE
group by category
order by category;
Where TABLE (category:text, x:float).
Suggestions?
Hopefully this example will help. You need a function that takes (accumulator, aggregate-arguments) and returns the new accumulator value. Play around with the code below and that should give you a feel for how it all fits together.
BEGIN;
CREATE FUNCTION sum_product_fn(int,int,int) RETURNS int AS $$
SELECT $1 + ($2 * $3);
$$ LANGUAGE SQL;
CREATE AGGREGATE sum_product(int, int) (
sfunc = sum_product_fn,
stype = int,
initcond = 0
);
SELECT
sum(i) AS one,
sum_product(i, 2) AS double,
sum_product(i,3) AS triple
FROM generate_series(1,3) i;
ROLLBACK;
That should give you something like:
one | double | triple
-----+--------+--------
6 | 12 | 18
This can be achieved with the ntile windowing function
-- To calculate flexible quantile ranges in postgresql, for example to calculate n equal
-- frequency buckets for your data for use in a visualisation (such as binning for a
-- choropleth map), you can use the following SQL:
-- this functions returns 6 equal frequency bucket ranges for my_column.
SELECT ntile, avg(my_column) AS avgAmount, max(my_column) AS maxAmount, min(my_column) AS minAmount
FROM (SELECT my_column, ntile(6) OVER (ORDER BY my_column) AS ntile FROM my_table) x
GROUP BY ntile ORDER BY ntile
You can find more on the ntile() function and windowing at http://database-programmer.blogspot.com/2010/11/really-cool-ntile-window-function.html
