534. Game Play Analysis III π
Description
Table: Activity
+--------------+---------+ | Column Name | Type | +--------------+---------+ | player_id | int | | device_id | int | | event_date | date | | games_played | int | +--------------+---------+ (player_id, event_date) is the primary key (column with unique values) of this table. This table shows the activity of players of some games. Each row is a record of a player who logged in and played a number of games (possibly 0) before logging out on someday using some device.
Write a solution to report for each player and date, how many games played so far by the player. That is, the total number of games played by the player until that date. Check the example for clarity.
Return the result table in any order.
The result format is in the following example.
Example 1:
Input: Activity table: +-----------+-----------+------------+--------------+ | player_id | device_id | event_date | games_played | +-----------+-----------+------------+--------------+ | 1 | 2 | 2016-03-01 | 5 | | 1 | 2 | 2016-05-02 | 6 | | 1 | 3 | 2017-06-25 | 1 | | 3 | 1 | 2016-03-02 | 0 | | 3 | 4 | 2018-07-03 | 5 | +-----------+-----------+------------+--------------+ Output: +-----------+------------+---------------------+ | player_id | event_date | games_played_so_far | +-----------+------------+---------------------+ | 1 | 2016-03-01 | 5 | | 1 | 2016-05-02 | 11 | | 1 | 2017-06-25 | 12 | | 3 | 2016-03-02 | 0 | | 3 | 2018-07-03 | 5 | +-----------+------------+---------------------+ Explanation: For the player with id 1, 5 + 6 = 11 games played by 2016-05-02, and 5 + 6 + 1 = 12 games played by 2017-06-25. For the player with id 3, 0 + 5 = 5 games played by 2018-07-03. Note that for each player we only care about the days when the player logged in.
Solutions
Solution 1: Window Function
Thinking
We need a running sum of games per player by date. A self-join works, but a window does it in one scan.
SUM(games_played) OVER (PARTITION BY player_id ORDER BY event_date) accumulates in date order inside each player. No extra group-and-join.
We can use the window function SUM() OVER() to group by player_id, sort by event_date, and calculate the total number of games played by each user up to the current date.
1 2 3 4 5 6 7 8 9 | |
Solution 2: Self-Join + Group By
Thinking
Without window functions, self-join each player's day to all of that player's days that are not later, then GROUP BY player and date.
The predicate t1.event_date >= t2.event_date keeps past and current rows. The sum matches the window prefix, at the cost of more paired rows.
We can also use a self-join to join the Activity table with itself on the condition of t1.player_id = t2.player_id AND t1.event_date >= t2.event_date, and then group by t1.player_id and t1.event_date, and calculate the cumulative sum of t2.games_played. This will give us the total number of games played by each user up to the current date.
1 2 3 4 5 6 7 8 9 10 | |
Solution 3
Thinking
Solution 2 writes the same predicate with a comma join. CROSS JOIN ... ON only moves the filter into the join clause.
The choice is stylistic; the grouped sums match Solution 2.
1 2 3 4 5 6 7 8 9 | |