SQL code for time and record specific selection? - sql

Question

I've been troubling my brain about the next problem. I work with a large dataset which contains records of species observations and location. Here's what I want to do:
For each period of time (5 years) I want to know the number of distinct species that were described in that particular period, but which were not described in an earlier period. So for each new period, the query needs to check if a certain species wasn't already described before. This needs to be done for 100 separate areas. The dataset already knows for each record in which 5 year period it was recorded.
My final result should be a table with areas on the x-axis, the periods on the y-axis, and in the cells the number of described species for each period, per area. It would be great if this is possible with one query. But I'm planning to do this in Excel, since I would be very happy to get for each area, the number of distinct species per time period.

Answer 1

PostgreSQL supports Windowed Aggregate Functions:
SELECT
area, period, SUM(x) AS newSpecies
FROM
(
SELECT area, period,
CASE -- check for the first description
WHEN date_col = MIN(date_col) OVER (PARTITION BY species) THEN 1
ELSE 0
END AS x
FROM au.trans
) AS dt
GROUP BY area, period
Depending on your data you might need to switch to ROW_NUMBER instead:
CASE -- check for the first description
WHEN ROW_NUMBER() OVER (PARTITION BY species ORDER BY date_col) = 1 THEN 1
ELSE 0
END AS x
Now you just have to pivot that data, don't know if there's a PIVOT function in PostgreSQL, otherwise you'll need to do the classical MAX(CASE). For each area you need to add a
SELECT period,
-- cut&paste&modify for each area
MAX(CASE WHEN area = 'area52' THEN newSpecies ELSE 0 END AS area52,
....
FROM (previous query) AS dt
GROUP BY period