AWS athena中的SQL下面的output如何计算

Question

I have the below result set.

with dataset AS (
    select 1 as total_users, ARRAY['google', 'meta', 'attentive', 'meta'] as path_list
    UNION ALL
    select 1, ARRAY['google', 'attentive', 'Direct Traffic', 'Direct Traffic', 'Direct Traffic', 'meta']
    UNION ALL
    select 4 , ARRAY ['google','meta', 'google']
    UNION ALL
    select 1, ARRAY['google', 'meta', 'meta', 'Direct Traffic' , 'meta']
    UNION ALL
    select 1, ARRAY['google', 'meta', 'meta']
    UNION ALL
    select 1, ARRAY['google', 'Direct Traffic', 'Direct Traffic','attentiva', 'attentiva', 'attentiva', 'Direct Traffic', 'meta']
)

SELECT path_list, total_users, path_list[2] as second_click, 
CASE WHEN CARDINALITY(path_list) > 2 THEN path_list[3] ELSE NULL END as third_click from dataset

The total_users column indicate number of users who traversed a particular path.

I want to calculate the following output.

• Second_click medium
• second_click_percentage
• third_click
• third_click_percentage

for each medium.

The result set looks as below

path_list total_users second_click third_click

[google, Direct Traffic, Direct Traffic, attentiva, attentiva, attentiva, Direct Traffic, meta]
1 Direct Traffic Direct Traffic

[google, meta, attentive, meta]
1 meta attentive

[google, attentive, Direct Traffic, Direct Traffic, Direct Traffic, meta]
1 attentive Direct Traffic

[google, meta, meta, Direct Traffic, meta]
1 meta meta

[google, meta, meta]
1 meta meta

[google, meta, google]
4 meta google

Now I need to calculate how users' percentage value that how many had meta as second click, attentive as second click.

Similarly need to identity third click percentage values for different mediums.

What can I do to write a SQL solution?

Answer 1

You can group data and calculate the percentages:

-- sample data
with dataset(total_users, path_list) AS (
    values (1, ARRAY['google', 'meta', 'attentive', 'meta']),
    (1, ARRAY['google', 'attentive', 'Direct Traffic', 'Direct Traffic', 'Direct Traffic', 'meta']),
    (4, ARRAY ['google','meta', 'google']),
    (1, ARRAY['google', 'meta', 'meta', 'Direct Traffic' , 'meta']),
    (1, ARRAY['google', 'meta', 'meta']),
    (1, ARRAY['google', 'Direct Traffic', 'Direct Traffic','attentiva', 'attentiva', 'attentiva', 'Direct Traffic', 'meta'])
),

-- query parts
clicks as (
SELECT total_users,
       path_list[2] as second_click,
       try(path_list[2]) as third_click,
       sum(total_users) over () as sum_total
from dataset)

select second_click, sum(total_users) * 1.0 / arbitrary(sum_total)
from clicks
group by second_click;

Output:

second_click	_col1
meta	0.8
Direct Traffic	0.1
attentive	0.1

If you want both groupings returned in one query you can look into grouping sets:

select second_click, third_click, sum(total_users) * 1.0 / arbitrary(sum_total)
from clicks
group by GROUPING SETS (
    (second_click),
    (third_click));

But I would argue that this will not be that convenient.