postgreSQL- Count for value between previous month start date and end date - sql

Question

I have a table as follows
user_id date month year visiting_id
123 11-04-2017 APRIL 2017 4500
123 12-05-2017 MAY 2017 4567
123 13-05-2017 MAY 2017 4568
123 17-05-2017 MAY 2017 4569
123 22-05-2017 MAY 2017 4570
123 11-06-2017 JUNE 2017 4571
123 12-06-2017 JUNE 2017 4572
I want to calculate the visiting count for the current month and last month at the monthly level as follows:
user_id month year visit_count_this_month visit_count_last_month
123 APRIL 2017 1 0
123 MAY 2017 4 1
123 JUNE 2017 2 4
I was able to calculate visit_count_this_month using the following query
SELECT v.user_id, v.month, v.year,
SUM(is_visit_this_month) as visit_count_this_month
FROM
(SELECT user_id, date, month, year,
CASE WHEN TO_CHAR(date, 'MM/YYYY') = TO_CHAR(date, 'MM/YYYY')
THEN 1 ELSE 0
END as is_visit_this_month
FROM visits
GROUP BY user_id, date, month, year
HAVING user_id = 123) v
GROUP BY v.user_id, v.month, v.year
However, I'm stuck with calculating visit_count_last_month. Similar to this, I also want to calculate visit_count_last_2months.
Can somebody help?

Answer 1

You can use a LATERAL JOIN like this:
SELECT user_id, month, year, COUNT(*) as visit_count_this_month, visit_count_last_month
FROM visits v
CROSS JOIN LATERAL (
SELECT COUNT(*) as visit_count_last_month
FROM visits
WHERE user_id = v.user_id
AND date = (CAST(v.date AS date) - interval '1 month')
) l
GROUP BY user_id, month, year, visit_count_last_month;
SQLFiddle - http://sqlfiddle.com/#!15/393c8/2

Answer 2

Assuming there are values for every month, you can get the counts per month first and use lag to get the previous month's values per user.
SELECT T.*
,COALESCE(LAG(visits,1) OVER(PARTITION BY USER_ID ORDER BY year,mth),0) as last_month_visits
,COALESCE(LAG(visits,2) OVER(PARTITION BY USER_ID ORDER BY year,mth),0) as last_2_month_visits
FROM (
SELECT user_id, extract(month from date) as mth, year, COUNT(*) as visits
FROM visits
GROUP BY user_id, extract(month from date), year
) T
If there can be missing months, it is best to generate all months within a specified timeframe and left join ing the table on to that. (This example shows it for all the months in 2017).
select user_id,yr,mth,visits
,coalesce(lag(visits,1) over(PARTITION BY USER_ID ORDER BY yr,mth),0) as last_month_visits
,coalesce(lag(visits,2) OVER(PARTITION BY USER_ID ORDER BY yr,mth),0) as last_2_month_visits
from (select u.user_id,extract(year from d.dt) as yr, extract(month from d.dt) as mth,count(v.visiting_id) as visits
from generate_series(date '2017-01-01', date '2017-12-31',interval '1 month') d(dt)
cross join (select distinct user_id from visits) u
left join visits v on extract(month from v.dt)=extract(month from d.dt) and extract(year from v.dt)=extract(year from d.dt) and u.user_id=v.user_id
group by u.user_id,extract(year from d.dt), extract(month from d.dt)
) t