Pivoting columns from STRUCT type - google-bigquery

Question

I have a SQL table with information about email campaigns that my company has created. Each line of the table is an action that a user has taken on a specific campaign:
UserID
properties.CampaignName
properties.Status
01
Campaign#01
opened
01
Campaign#01
clicked
01
Campaign#02
opened
02
Campaign#01
opened
02
Campaign#02
opened
I wanted to Pivot this on SQL, in a way that would render the unique number of people who have opened and clicked on each campaign:
CampaignName
Opened
Clicked
Campaign#01
2149
122
Campaign#02
4223
141
My initial thought was to use the PIVOT query:
select * from table
pivot (count(distinct UserID) for properties.Status in ('opened', 'clicked'))
But I didn't realize something: the CampaignName and Status columns are nested under the "properties" column - so basically I have properties.status, properties.country, properties.campaignname, etc. Therefore the error I am getting is:
Error running query Column raw_properties of type STRUCT cannot be used as an implicit grouping column of a PIVOT clause at [2:1]

Answer 1

Consider below
select * from (select UserID, properties.* from your_table)
pivot (count(distinct UserID) for Status in ('opened', 'clicked'))
if applied to sample data in your question
with your_table as (
select '01' as UserID, struct('Campaign#01' as CampaignName, 'opened' as Status) as properties union all
select '01', struct('Campaign#01', 'clicked') union all
select '01', struct('Campaign#02', 'opened') union all
select '02', struct('Campaign#01', 'opened') union all
select '02', struct('Campaign#02', 'opened')
)
output is
So, technically this is exactly the same answer that I already provided you with - https://stackoverflow.com/a/70115346/5221944 - you just needed to flatten struct into separate columns first

Answer 2

You can still use your PIVOT clause, the deal here is with the struct values. So my approach was to unnest that data and set is as same level data and then proceed to reuse your query. Below are the steps I follow:
working table
create table `project.dataset.table`(
userid INT64,
properties STRUCT<CampaignName STRING,Status STRING>
);
dummy data
INSERT INTO `project.dataset.table`(userid,properties) VALUES(01,STRUCT("Campaign#01","opened"));
INSERT INTO `project.dataset.table`(userid,properties) VALUES(01,STRUCT("Campaign#01","clicked"));
INSERT INTO `project.dataset.table`(userid,properties) VALUES(01,STRUCT("Campaign#02","opened"));
INSERT INTO `project.dataset.table`(userid,properties) VALUES(02,STRUCT("Campaign#01","opened"));
INSERT INTO `project.dataset.table`(userid,properties) VALUES(02,STRUCT("Campaign#02","opened"));
query
with campaign_unnest as (
select o.userid,
prp.Status,prp.CampaignName
from `project.dataset.table` o, unnest([o.properties]) as prp
)
select * from campaign_unnest pivot (count(distinct userid) for Status in ('opened', 'clicked'))
output
CampaignName
Opened
Clicked
Campaign#01
2
1
Campaign#02
2
0
Let me know if this answer fits what you are trying to achieve. If not, please let me know so I can update my answer.
For reference I use the google bigquery documentation about arrays & structs, unnest and with.