initial repo commit

README.md

@@ -0,0 +1,12 @@
+# Cumulative Table Design Example
+
+This repo shows how to build a robust cumulative table design with an example of monthly active users. 
+
+In this repo, we'll be using **2022-01-01** as 'today' and **2021-12-31** as 'yesterday'
+
+
+The steps for this design are:
+
+- The Daily Table step
+  - In this step we aggregate just the events of today to see who is daily active
+  - This query is pretty simple and straight forward check 

queries/active_users_cumulated_populate.sql

@@ -0,0 +1,48 @@
+INSERT INTO active_users_cumulated
+
+-- First read in yesterday from the cumulated table
+WITH yesterday AS (
+    SELECT * FROM active_users_cumulated
+    WHERE snapshot_date = '2021-12-31'
+),
+-- Read in the daily active user numbers for just today from the daily table
+today AS (
+    SELECT * FROM active_users_daily
+    WHERE snapshot_date = '2022-01-01'
+),
+
+-- we FULL OUTER JOIN today and yesterday. We need to do some COALESCE both because
+-- activity_array may not exist yet for a given user (i.e. they are brand new)
+-- is_active_today may be null as well since it's null on days when a user didn't generate an event
+combined AS (
+SELECT
+ -- We need to COALESCE here since t.user_id and y.user_id may be
+ COALESCE(y.user_id, t.user_id) AS user_id,
+ COALESCE(
+        IF(CARDINALITY( y.activity_array) < 30,
+            ARRAY[COALESCE(t.is_active_today, 0)] || y.activity_array,
+            ARRAY[COALESCE(t.is_active_today, 0)] || SLICE(y.activity_array, -1, 29)
+         )
+       , ARRAY[t.is_active_today]
+ ) as activity_array,
+ t.snapshot_date
+ FROM yesterday y
+            FULL OUTER JOIN today t
+            ON y.user_id = t.user_id
+)
+
+SELECT
+      user_id,
+      activity_array[1] AS is_daily_active,
+      -- if any of the array values are 1, then the user was active in the last month
+      CASE WHEN ARRAY_SUM(activity_array) > 0 THEN 1 ELSE 0 END AS is_monthly_active,
+      -- if any of the first 7 array values are non-zero, then the user was active in the last week
+      CASE WHEN ARRAY_SUM(SLICE(activity_array, 1, 7)) > 0 THEN 1 ELSE 0 END AS is_weekly_active
+      activity_array,
+
+FROM combined
+
+
+
+
+

queries/active_users_daily_populate.sql

@@ -0,0 +1,9 @@
+INSERT INTO active_users_daily
+
+SELECT
+    user_id,
+    -- If the user_id has at least 1 event, they are daily active
+    IF(COUNT(user_id) > 0, 1, 0) as is_daily_active
+FROM events
+WHERE event_date = '2022-01-01'
+GROUP BY user_id

tables/active_users_cumulated.sql

@@ -0,0 +1,9 @@
+create table active_users_cumulated (
+    user_id: integer,
+    is_daily_active: integer,
+    is_weekly_active: integer,
+    is_monthly_active: integer,
+    activity_array: array<integer>,
+    snapshot_date: date
+)
+

tables/active_users_daily.sql

@@ -0,0 +1,5 @@
+create table active_users_cumulated (
+    user_id: integer,
+    is_active_today: integer,
+    snapshot≈_date: date
+)