Question

I’m having trouble with a mySQL query.

We have a table that tracks the amount of times a user clicks. It is a single integer field that increases by 1 each click.

Schema:

+------------+--------------+------+-----+
| Field      | Type         | Null | Key | 
+------------+--------------+------+-----+
| id         | int(11)      | NO   | PRI | 
| arrival_id | int(11)      | NO   |     |
| clickouts  | int(11)      | NO   |     | 
+------------+--------------+------+-----+

Example Data:

id        arrival_id        clickouts
 5            22                0
 6            23                1
 7            24                7

If I want to determine the percentage of arrivals that generated at least one click, I can write the following query. (Note – not every arrival has a clickout, hence the LEFT JOIN):

SELECT SUM(click_tracker.clickouts > 0)/COUNT(DISTINCT arrivals.id) 
FROM arrivals 
LEFT JOIN click_tracker ON arrivals.id = click_tracker.arrival_id;

The issue comes in when I also LEFT JOIN arrivals to another table (not click_tracker) during the same query.

I have another table called advertisements, that tracks various types of ads a user will see. The ads are tied to an arrival, and an arrival can have many ads. The schema is as follows:

+------------+--------------+------+-----+
| Field      | Type         | Null | Key | 
+------------+--------------+------+-----+
| id         | int(11)      | NO   | PRI | 
| arrival_id | int(11)      | NO   |     |
| ad_type    | varchar(255) | NO   |     | 
+------------+--------------+------+-----+

Now, I want to find the percentage of arrivals with ad_type = “ZXY” along with the percentage of arrivals that generated at least one click (regardless of the ad_type). Doing the following won’t work:

SELECT COUNT(DISTINCT advertisements.id)/COUNT(DISTINCT arrivals.id) AS ad_pct,
SUM(click_tracker.clickouts > 0)/COUNT(DISTINCT arrivals.id) AS click_pct
FROM arrivals
LEFT JOIN advertisements ON arrivals.id = advertisements.arrival_id
LEFT JOIN click_tracker ON arrivals.id = click_tracker.arrival_id

The LEFT JOIN of the advertisements table creates a result set duplicate arrival id’s. Thus, when I subsequently LEFT JOIN onto the click_tracker table, I get duplicate click_tracker id’s as well.

I need a way to do SUM(click_tracker.clickouts > 0) for distinct click_tracker rows only. I tried SUM(DISTINCT click_tracker.clickouts > 0) but that did not give me the correct results.

Answer 1

I hate to say it, but those two calculations are completely independent, and I think you’ll need to execute two queries for them. Here they are (and an sqlfiddle with the data I used):

Percentage of arrivals by ad_type:

SELECT
  ad_type,
  (COUNT(arrival_id) / x.total) as ad_pct
  FROM advertisements, 
       (SELECT COUNT(id) AS total FROM arrivals) x
  GROUP BY ad_type;

(Or you could change the GROUP BY to a WHERE clause for a single ad_type.)

Percentage of arrivals with at least one click:

SELECT
  (COUNT(id) / x.total) AS click_pct
  FROM click_tracker,
       (SELECT COUNT(id) AS total FROM arrivals) x
  WHERE click_tracker.clickouts > 0;