Since ClickHouse doesn't support WITH clause for queries, I'm having troubles implementing a query for calculating MAU (Monthly active users calculated day by day and the 30 days range is moving day by day as well). In Postgres I would do this:
WITH days AS (
SELECT created_at::DATE AS day,
FROM events
WHERE created_at > '2019-04-01'
GROUP BY 1
)
SELECT day,
(SELECT count(distinct user_id)
FROM events
WHERE events.created_at::DATE BETWEEN days.day-29 AND days.day
AND created_at > '2019-04-01'
) AS mau
FROM days
Desired result looks like this
┌────────day─┬──mau─┐
│ 2019-04-04 │ 1278 │
│ 2019-04-05 │ 1375 │
│ 2019-04-06 │ 1162 │
│ 2019-04-07 │ 1237 │
│ 2019-04-08 │ 1272 │
│ 2019-04-09 │ 1263 │
│ 2019-04-10 │ 1336 │
│ 2019-04-11 │ 1457 │
│ 2019-04-12 │ 1286 │
│ 2019-04-13 │ 1210 │
│ 2019-04-14 │ 1253 │
│ 2019-04-15 │ 2342 │
│ 2019-04-16 │ 1464 │
│ 2019-04-17 │ 1513 │
│ 2019-04-18 │ 1158 │
│ 2019-04-19 │ 1207 │
│ 2019-04-20 │ 1222 │
│ 2019-04-21 │ 1054 │
│ 2019-04-22 │ 1505 │
│ 2019-04-23 │ 5287 │
│ 2019-04-24 │ 4367 │
│ 2019-04-25 │ 3624 │
│ 2019-04-26 │ 2415 │
│ 2019-04-27 │ 1962 │
│ 2019-04-28 │ 2032 │
│ 2019-04-29 │ 2547 │
│ 2019-04-30 │ 4059 │
└────────────┴──────┘
You could try something like this:
SELECT
created_at,
uniqExact(user_id) AS mau
FROM
(
SELECT
created_at + n AS created_at,
user_id
FROM
(
SELECT
today() - 14 AS created_at,
123 AS user_id
UNION ALL
SELECT
today() - 20 AS created_at,
456 AS user_id
)
ARRAY JOIN range(30) AS n
)
WHERE created_at <= today()
GROUP BY created_at
FORMAT TSV
2019-04-11 1
2019-04-12 1
2019-04-13 1
2019-04-14 1
2019-04-15 1
2019-04-16 1
2019-04-17 2
2019-04-18 2
2019-04-19 2
2019-04-20 2
2019-04-21 2
2019-04-22 2
2019-04-23 2
2019-04-24 2
2019-04-25 2
2019-04-26 2
2019-04-27 2
2019-04-28 2
2019-04-29 2
2019-04-30 2
2019-05-01 2
21 rows in set. Elapsed: 0.005 sec.