sql_fb, Sql Server

sql_fb

--- FB Spam friend request---

CREATE TABLE users
    ([user_id] int, [country] varchar(2), [facebook_age] int)
;
    
INSERT INTO users
    ([user_id], [country], [facebook_age])
VALUES
    (1001, 'US', 120),
    (1002, 'CA', 400),
    (1003, 'MX', 6),
    (5555, 'US', 8)
;

CREATE TABLE friend_requests
    ([date] datetime, [sender_id] int, [receiver_id] int, [result] int)
;
    
INSERT INTO friend_requests
    ([date], [sender_id], [receiver_id], [result])
VALUES
    ('2020-01-01 00:00:00', 1001, 9991, 0),
    ('2020-01-01 00:00:00', 1001, 1032, 1),
    ('2020-01-01 00:00:00', 1002, 5555, 2),
    ('2020-01-01 00:00:00', 1002, 1234, 1),
    ('2020-01-02 00:00:00', 1003, 1002, 1)
;

-- Q1: For each month in 2020, what is the friend request acceptance rate?
/*
SELECT month(date), SUM(CASE WHEN result=1 THEN 1.00 ELSE 0 END)/COUNT(*) AS accept_rate
FROM friend_requests
GROUP BY month(date)
*/
----

/*
WITH month_table AS (
    SELECT '2020-01-01' AS month
    UNION
    SELECT '2020-02-01' AS month
    UNION
    SELECT '2020-03-01' AS month
    UNION
    SELECT '2020-04-01' AS month
    UNION
    SELECT '2020-05-01' AS month
    UNION
    SELECT '2020-06-01' AS month
    UNION
    SELECT '2020-07-01' AS month
    UNION
    SELECT '2020-08-01' AS month
    UNION
    SELECT '2020-09-01' AS month
    UNION
    SELECT '2020-10-01' AS month
    UNION
    SELECT '2020-11-01' AS month
    UNION
    SELECT '2020-12-01' AS month),
    monthly_request AS (
    SELECT a.month, SUM(CASE WHEN b.result=1 THEN 1.00 ELSE 0 END) AS accepts, COUNT(b.result) AS requests
    FROM month_table AS a LEFT JOIN friend_requests AS b ON month(a.month)=month(b.date)
    GROUP BY a.month)
    SELECT month, CASE WHEN requests = 0 then 0.00 ELSE accepts/requests END AS accept_rate
    FROM monthly_request;
    
-- Q2: What percent of users in the United States accepted a friend request last week?
WITH R7users AS (
    SELECT *
    FROM users
    Where facebook_age>=7 AND country='US')

SELECT CAST(COUNT(DISTINCT receiver_id)/COUNT(DISTINCT a.user_id) AS DECIMAL(5,2)) AS US_acceptedRateUser
FROM R7users AS a LEFT JOIN 
(SELECT DISTINCT receiver_id FROM friend_requests WHERE date >= '2020-01-01' AND result=1) AS b ON a.user_id = b.receiver_id

*/
--- Dwell time ---

CREATE TABLE user_sessions
    ([date] datetime, [user_id] int, [start_time] int, [end_time] int, [end_reason] varchar(9))
;
    
INSERT INTO user_sessions
    ([date], [user_id], [start_time], [end_time], [end_reason])
VALUES
    ('2020-09-01 00:00:00', 1, 1964783746, 1964783924, 'close_app'),
    ('2020-09-01 00:00:00', 1, 1964783528, 1964783809, 'crash'),
    ('2020-09-02 00:00:00', 2, 1964783123, 1964783345, 'close_app'),
    ('2020-09-02 00:00:00', 3, 1964783252, 1964783658, 'crash')
;

CREATE TABLE dim_all_users
    ([user_id] int, [country] varchar(10), [is_active] varchar(5))
;
    
INSERT INTO dim_all_users
    ([user_id], [country], [is_active])
VALUES
    (1, 'US', 'true'),
    (2, 'US', 'false'),
    (3, 'CANADA', 'true')
;

/*
WITH date_tbl AS (
    SELECT date
    FROM user_sessions
    GROUP BY date)
    ,
    user_date AS (
        SELECT user_id, country, is_active, date
        FROM dim_all_users AS a CROSS JOIN date_tbl AS b)

SELECT a.user_id, a.date, SUM(CASE WHEN b.start_time IS NULL THEN 0.00 ELSE b.end_time-start_time END)/1000 AS dwell_time
FROM user_date AS a LEFT JOIN user_sessions AS b ON a.user_id = b.user_id AND a.date = b.date
GROUP BY a.user_id, a.date;
*/
-- 2. What is the percentage of active users who crashed when using Facebook from the US
-- area on a specific date?

/*
WITH user_sessions1 AS (
    SELECT a.* FROM user_sessions AS a INNER JOIN dim_all_users AS b ON a.user_id = b.user_id WHERE date = '2020-09-01' AND country = 'US')
    
SELECT CAST(COUNT(DISTINCT b.user_id)/COUNT(DISTINCT a.user_id) AS decimal(4,2)) AS crash_rate_us
FROM user_sessions1 AS a LEFT JOIN
(SELECT * FROM user_sessions1 WHERE end_reason='crash') AS b 
on a.user_id = b.user_id
*/
-- 3. Calculate the average dwell time in seconds across all sessions (i.e. return one number)? Dwell time is the length of time between opening and closing the menu.
-- SELECT (SUM(end_time)-SUM(start_time))/COUNT(*) AS avg_dwell_time
-- FROM user_sessions

------------ Portal -----------

CREATE TABLE video_calls
    ([caller] int, [recepient] int, [date] datetime, [call_id] int, [duration] int)
;
    
INSERT INTO video_calls
    ([caller], [recepient], [date], [call_id], [duration])
VALUES
    (1111, 1112, '2020-01-01 00:00:00', 1, 10),
    (1111, 1003, '2020-01-01 00:00:00', 2, 15),
    (1112, 1145, '2020-01-01 00:00:00', 3, 5)
;

CREATE TABLE fb_dau
    ([user_id] int, [dau_fg] int, [date] datetime, [country] varchar(2))
;
    
INSERT INTO fb_dau
    ([user_id], [dau_fg], [date], [country])
VALUES
    (1111, 1, '2020-01-01 00:00:00', 'US'),
    (1112, 0, '2020-01-01 00:00:00', 'CA'),
    (1003, 0, '2020-01-01 00:00:00', 'US')
;
-- Q1: On 2020-01-01 how many people initiated multiple calls?

SELECT COUNT(*)
FROM(
SELECT caller
FROM video_calls
WHERE date = '2020-01-01' 
GROUP BY caller
HAVING COUNT(DISTINCT call_id)>1) AS t;

-- Q2: % of DAU used the video calls function by each country on 2020-01-01?

WITH dau AS (
    SELECT country, user_id
    FROM fb_dau
    WHERE date='2020-01-01'),
    video_user AS (
        SELECT caller AS video_user
        FROM video_calls
        GROUP BY caller
        UNION 
        SELECT recepient AS video_user
        FROM video_calls
        GROUP BY recepient)
SELECT country, CAST(COUNT(DISTINCT video_user)/COUNT(DISTINCT user_id) AS decimal(4,2)) AS percent_video_users
FROM dau AS a LEFT JOIN video_user AS b ON a.user_id = b.video_user
GROUP BY country;

-- CONVERT(date, xxx) AS date

----------- posts ---------

CREATE TABLE user_actions
    ([date] datetime, [actor_id] int, [post_id] int, [relationship] varchar(7), [interaction] varchar(7))
;
    
INSERT INTO user_actions
    ([date], [actor_id], [post_id], [relationship], [interaction])
VALUES
    ('2021-03-20 00:00:00', 1001, 2057, 'friend', 'like'),
    ('2021-03-21 00:00:00', 2347, 1245, 'folower', 'dislike'),
    ('2021-03-21 00:00:00', 2348, 2057, 'friend', 'love'),
    ('2021-03-21 00:00:00', 3333, 1245, 'friend', 'like')
;

CREATE TABLE user_posts
    ([date] datetime, [poster_id] int, [post_id] int)
;
    
INSERT INTO user_posts
    ([date], [poster_id], [post_id])
VALUES
    ('2021-03-20 00:00:00', 1111, 2057),
    ('2021-03-20 00:00:00', 2222, 1245),
        ('2021-03-21 00:00:00', 1111, 1250)
;
WITH t AS (
    SELECT post_id
    FROM user_posts
    WHERE poster_id=1111 AND date BETWEEN GETDATE()-7 AND GETDATE())

SELECT SUM(CASE WHEN interaction = 'like' THEN 1.00 ELSE 0 END)/COUNT(DISTINCT t.post_id) AS avg_like
FROM t LEFT JOIN user_actions AS t1 ON t.post_id = t1.post_id;

-------------------------------------------------------

CREATE TABLE user_network_requests
    ([user_id] int, [timestamp] int, [data_center] varchar(1), [success] int)
;
    
INSERT INTO user_network_requests
    ([user_id], [timestamp], [data_center], [success])
VALUES
    (10032, 15009, 'A', 1),
    (10032, 15097, 'C', 0),
    (10033, 15000, 'A', 1),
    (10033, 16000, 'A', 0),
    (10033, 16700, 'C', 1),
    (15009, 16800, 'C', 1)
;

CREATE TABLE user_country
    ([user_id] int, [country] varchar(2))
;
    
INSERT INTO user_country
    ([user_id], [country])
VALUES
    (10032, 'US'),
    (15009, 'US')
;

-- 1. request failure rate for each data center
/*
SELECT data_center, 1-SUM(success)*1.00/COUNT(*) AS failure_rate
FROM user_network_requests
GROUP BY data_center

-- 2. request failure rate per country
SELECT country, 1-SUM(success)*1.00/COUNT(*) AS failure_rate
*/
-- 3. for each country, how many uers never had a failed request
/*
WITH t AS (
SELECT country, a.user_id, SUM(success) AS success_cnt, COUNT(*) AS requests
FROM user_network_requests AS a INNER JOIN user_country AS b ON a.user_id = b.user_id
GROUP BY country, a.user_id)

SELECT country, COUNT(DISTINCT user_id) AS users
FROM t 
WHERE success_cnt = requests
GROUP BY country
*/
----------------------- ads ----------------------------

CREATE TABLE ads
    ([ad_id] int, [user_id] int, [status] varchar(5), [timestamp] datetime)
;
    
INSERT INTO ads
    ([ad_id], [user_id], [status], [timestamp])
VALUES
    (1111, 4324, 'click', '2020-01-01 00:00:00'),
    (1112, 4324, 'click', '2020-01-01 00:00:00'),
    (1112, 32646, 'click', '2020-01-03 00:00:00'),
    (1111, 4542, 'hide', '2020-01-02 00:00:00'),
    (1112, 4542, 'click', '2020-01-01 00:00:00'),
    (1112, 2349, 'hide', '2020-01-02 00:00:00')
;
/*
-- 1. how to measure ad performance?
SELECT ad_id, SUM(CASE WHEN status NOT IN ('hide') THEN 1.00 ELSE 0 END)/COUNT(*) AS engagement_rate
FROM ads
GROUP BY ad_id;

-- 2. 如果有个user click其中一个ad, 写SQL推荐下一个show给这个user的ad.

WITH t AS (
SELECT *
FROM ads
WHERE status = 'click' ),
t1 AS (
    SELECT * 
    FROM ads
    WHERE status !='hide')
    
SELECT a.user_id, t1.ad_id
FROM (
    SELECT t.user_id, t.ad_id, t1.user_id AS rec_user
    FROM t INNER JOIN t1 ON t.ad_id = t1.ad_id
    WHERE t.user_id != t1.user_id) AS a INNER JOIN t1 ON a.rec_user = t1.user_id
WHERE a.ad_id != t1.ad_id
*/
------------- spam ------------------

DROP TABLE user_actions;
CREATE TABLE user_actions
    ([date] datetime, [user_id] int, [post_id] int, [action] varchar(6))
;
    
INSERT INTO user_actions
    ([date], [user_id], [post_id], [action])
VALUES
    ('2020-01-01 00:00:00', 1234, 1111, 'view'),
    ('2020-01-01 00:00:00', 2345, 1111, 'report'),
    ('2020-01-02 00:00:00', 2314, 2111, 'view'),
    ('2020-01-02 00:00:00', 4235, 2111, 'view'),
    ('2020-01-02 00:00:00', 543, 1324, 'view'),
    ('2020-01-02 00:00:00', 2346, 1324, 'view'),
    ('2020-01-02 00:00:00', 2314, 1324, 'report')
;

CREATE TABLE review
    ([date] datetime, [reviewer_id] int, [post_id] int)
;
    
INSERT INTO review
    ([date], [reviewer_id], [post_id])
VALUES
    ('2020-01-03 00:00:00', 567990, 1111),
    ('2020-01-03 00:00:00', 567990, 2111)
;

-- Q1: What percent of the content views was reported for spam each day?

SELECT a.date, SUM(CASE WHEN a.action = 'view' AND b.post_id IS NOT NULL THEN 1.00 ELSE 0 END)/SUM(CASE WHEN a.action='view' THEN 1 ELSE 0 END) AS percent_spam
FROM user_actions AS a LEFT JOIN review AS b ON a.post_id = b.post_id
GROUP BY a.date

-- Q2: What percentage of the posts viewedby the users was actually spam each day?

---------------------- comment on post ----------------------------------

CREATE TABLE comments
    ([user_id] int, [date] datetime, [comments] int)
;
    
INSERT INTO comments
    ([user_id], [date], [comments])
VALUES
    (1111, '2021-03-01 00:00:00', 3),
    (1111, '2021-03-02 00:00:00', 2),
    (1112, '2021-03-01 00:00:00', 5),
    (1112, '2021-03-03 00:00:00', 3),
    (1112, '2021-03-05 00:00:00', 7),
    (1113, '2021-03-05 00:00:00', 2)
;

CREATE TABLE ActiveUsers
    ([user_id] int, [country] varchar(2))
;
    
INSERT INTO ActiveUsers
    ([user_id], [country])
VALUES
    (1111, 'US'),
    (1112, 'CA'),
    (1113, 'MX'),
    (1115, 'US')
;

-- total number of comments each user in past 30 days,
--  percent of active users who comment,

SELECT user_id, SUM(comments) AS total_comments
FROM comments
WHERE date >= GETDATE()-30
--DATEDIFF(day, date, CONVERT(date, GETDATE()))<=30
GROUP BY user_id;

SELECT COUNT(b.user_id)*1.00/COUNT(a.user_id) AS percent_comment_active
FROM ActiveUsers AS a 
LEFT JOIN
(SELECT user_id FROM comments WHERE date>=GETDATE()-30 GROUP BY user_id) AS b ON a.user_id = b.user_id

----------- search ----------------

-- sql：search/search_result
-- 1）#of users have made more than 1/10 searches in  last 7  days
-- 2) % of uders clicked the search result of event

CREATE TABLE searches
    ([search_id] int, [user_id] int, [date] datetime)
;
    
INSERT INTO searches
    ([search_id], [user_id], [date])
VALUES
    (1, 213, '2021-01-01 00:00:00'),
    (2, 34567, '2021-01-01 00:00:00'),
    (3, 21312, '2021-01-01 00:00:00'),
    (4, 213, '2021-01-02 00:00:00'),
    (5, 345, '2021-01-02 00:00:00')
;

/*
SELECT COUNT(*) AS cnt
FROM(
SELECT user_id
FROM searches
    GROUP BY user_id
HAVING COUNT(*)>(SELECT 0.1*COUNT(*) FROM searches WHERE date >=GETDATE()-7)) AS t
*/

------------------ friend request -----------------------

CREATE TABLE friendRequests
    ([user_id_send] int, [user_id_receive] int, [date] datetime, [action] varchar(6))
;
    
INSERT INTO friendRequests
    ([user_id_send], [user_id_receive], [date], [action])
VALUES
    (123, 234, '2021-01-01 00:00:00', 'send'),
    (234, 345, '2021-01-01 00:00:00', 'send'),
    (123, 345, '2021-01-01 00:00:00', 'send'),
    (234, 123, '2021-01-02 00:00:00', 'accept'),
    (345, 234, '2021-01-03 00:00:00', 'accept')
;

WITH send_table AS(
SELECT *
FROM friendRequests 
WHERE action = 'send'),
accept_table AS (
    SELECT *
    FROM friendRequests
    WHERE action = 'accept')

SELECT COUNT(b.user_id_send)*1.00/COUNT(a.user_id_send) AS request_accept_rate
FROM send_table AS a LEFT JOIN accept_table AS b ON a.user_id_send = b.user_id_receive AND a.user_id_receive = b.user_id_send

------------------ interactions ---------

CREATE TABLE Interactions
    ([user_a] int, [user_b] int, [date] datetime)
;
    
INSERT INTO Interactions
    ([user_a], [user_b], [date])
VALUES
    (123, 234, '2020-01-01 00:00:00'),
    (233, 345, '2020-01-01 00:00:00'),
    (234, 3245, '2020-01-01 00:00:00'),
    (123, 546, '2020-01-01 00:00:00')
;

-- given a table of interaction between users (user_a | user_b | day), find number of users who had more than 5 interactions yesterday (assume there is only one unique interaction between a pair of users per day).

SELECT user_id, SUM(user_cnt) AS interactions
FROM(
SELECT user_a AS user_id, COUNT(DISTINCT user_b) AS user_cnt
FROM Interactions
    GROUP BY user_a
UNION ALL 
SELECT user_b AS user_id, COUNT(DISTINCT user_a) AS user_cnt
FROM Interactions
    GROUP BY user_b) AS t
GROUP BY user_id;

-------------------
SELECT 1+NULL
-----------------------------------
-- We have a table called ad_accounts(account_id, date, status). Status can be active/closed/fraud.
-- A) what percent of active accounts are fraud?
-- B) How many accounts became fraud today for the first time?
-- C) What would be the financial impact of letting fraud accounts become active (how would you approach this question)?

CREATE TABLE ad_accounts
    ([account_id] int, [status] varchar(6), [date] datetime)
;
    
INSERT INTO ad_accounts
    ([account_id], [status], [date])
VALUES
    (1111, 'fraud', '2021-01-01 00:00:00'),
    (1111, 'active', '2021-01-02 00:00:00'),
    (1101, 'active', '2021-01-03 00:00:00'),
     (1102, 'fraud', '2021-01-02 00:00:00'),
    (1102, 'active', '2021-01-03 00:00:00'),
    (1111, 'fraud', '2021-01-03 00:00:00'),
     (1112, 'fraud', '2021-01-03 00:00:00')
;
WITH t AS (
SELECT account_id, status, ROW_NUMBER() OVER(PARTITION BY account_id ORDER BY date DESC) AS account_rank
    FROM ad_accounts),
    latest AS (
        SELECT account_id
        FROM t 
        WHERE account_rank=1 AND status = 'active')
 SELECT COUNT(b.account_id)*1.00/COUNT(a.account_id) AS fraud_perc
 FROM latest AS a LEFT JOIN
 (SELECT * FROM ad_accounts WHERE status='fraud') AS b ON a.account_id = b.account_id;

SELECT account_id
FROM ad_accounts
WHERE status='fraud' AND date = '2021-01-03' 
AND account_id IN (SELECT account_id FROM ad_accounts WHERE status='fraud' GROUP BY account_id HAVING COUNT(*)=1)


run \| edit \| history \| help	0

SELECT account_id FROM ad_accounts WHERE status='fraud' AND date = '2021-01-03' AND account_id IN (SELECT account_id FROM ad_accounts WHERE status='fraud' GROUP BY account_id HAVING COUNT(*)=1)

by Userabc, 1 years ago

Please log in to post a comment.