Question

Assume a range of values inserted in a schema table and in the end of the month i want to apply for these records (i.e. 2500 rows = numeric values) the algorithm: sort the values descending (from the smallest to highest value) and then find the 80% value of the sorted column.

In my example, if each row increases by one starting from 1, the 80% value will be the 2000 row=value (=2500-2500*20/100). This algorithm needs to be implemented in a procedure where the number of rows is not constant, for example it can varries from 2500 to 1,000,000 per month

Answer 1

Write a query using the percentile_disc function to solve your problem. Sounds like it does what you want.

An example would be

select percentile_disc(0.8) within group (order by the_value)
from my_table