How can I get the maximum amount of the total amounts for different products in a month in Postgresql? - postgresql

Question

I've just begun using Postgresql recently. I have a table named 'sales'.
create table sales
(
cust varchar(20),
prod varchar(20),
day integer,
month integer,
year integer,
state char(2),
quant integer
)
insert into sales values ('Bloom', 'Pepsi', 2, 12, 2001, 'NY', 4232);
insert into sales values ('Knuth', 'Bread', 23, 5, 2005, 'PA', 4167);
insert into sales values ('Emily', 'Pepsi', 22, 1, 2006, 'CT', 4404);
insert into sales values ('Emily', 'Fruits', 11, 1, 2000, 'NJ', 4369);
insert into sales values ('Helen', 'Milk', 7, 11, 2006, 'CT', 210);
...
There are 500 rows, 10 distinct products and 5 distinct customers in total.
It looks like this:
Now I need to , find the most “popular” and least “popular” products (those products with most and least total sales quantities) and the corresponding total sales quantities (i.e., SUMs) for each of the 12 months (regardless of the year).
The result should be like this:
Now I can only write query like this:
select month,
prod,
sum(quant)
from sales
group by month,prod
order by month,prod;
And it gives me the result like this:
Now I need to pick up the maximum value for each month. For example, the biggest value in the first 10 sums of month 1, and so on...
I also need to get the minimum value of the sums (regardless of the year). And combine them horizontally... I have no idea about this...

Answer 1

Note: for a TLDR, skip to the end.
Your problem is a very interesting textbook case as it involves multiple facets of Postgres.
I often find it very helpful to decompose the problem into multiple subproblems before joining them together for the final result set.
In your case, I see two subproblems: finding the most popular product for each month, and finding the least popular product for each month.
Let's start with the most popular products:
WITH months AS (
SELECT generate_series AS month
FROM generate_series(1, 12)
)
SELECT DISTINCT ON (month)
month,
prod,
SUM(quant)
FROM months
LEFT JOIN sales USING (month)
GROUP BY month, prod
ORDER BY month, sum DESC;
Explanations:
WITH is a common table
expression,
which acts as a temporary table (for the duration of the query) and
helps clarify the query. If you find it confusing, you could also opt
for a subquery.
generate_series(1, 12) is a Postgres function which generate a series of integers, in this case from 1 to 12.
the LEFT JOIN allows us to associate each sale to the corresponding month. If no sale can be found for a given month, a row is returned with the month and the joined columns with NULL values. More information on joins can be found here. In your case, using LEFT JOIN is important, as using INNER JOIN would exclude products that have never been sold (which in that case should be the least popular product).
GROUP BY is used to sum over the quantities.
at this stage, you should -potentially- have multiple products for any given month. We only want to keep those with the most quantities for each month. DISTINCT ON is especially useful for that purpose. Given a column, it allows us to keep the first iteration of each value. It is therefore important to ORDER the sales by sum first, as only the first one will be selected. We want the bigger numbers first, so DESC (for descending order) should be used.
We can now repeat the process for the least popular products:
WITH months AS (
SELECT generate_series AS month
FROM generate_series(1, 12)
)
SELECT DISTINCT ON (month)
month,
prod,
SUM(quant)
FROM months
LEFT JOIN sales USING (month)
GROUP BY month, prod
ORDER BY month, sum;
Conclusion (and TLDR):
Now we need to merge the two queries into one final query.
WITH months AS (
SELECT generate_series AS month
FROM generate_series(1, 12)
), agg_sales AS (
SELECT
month,
prod,
SUM(quant)
FROM months
LEFT JOIN sales USING (month)
GROUP BY month, prod
), most_popular AS (
SELECT DISTINCT ON (month)
month,
prod,
sum
FROM agg_sales
ORDER BY month, sum DESC
), least_popular AS (
SELECT DISTINCT ON (month)
month,
prod,
sum
FROM agg_sales
ORDER BY month, sum
)
SELECT
most_popular.month,
most_popular.prod AS most_popular_prod,
most_popular.sum AS most_pop_total_q,
least_popular.prod AS least_popular_prod,
least_popular.sum AS least_pop_total_q
FROM most_popular
JOIN least_popular USING (month);
Note that I used an intermediate agg_sales CTE to try and make the query a bit clearer and avoid repeating the same operation twice, although it shouldn't be a problem for Postgres' optimizer.
I hope you find my answer satisfactory. Do not hesitate to comment otherwise!
EDIT: although this solution should work as is, I would suggest storing your dates as a single column of type TIMESTAMPTZ. It is often much easier to manipulate dates using that type and it is always good practice in case you need to analyze and audit your database further down the line.
You can get the month of any date by simply using EXTRACT(MONTH FROM date).