By using AWS re:Post, you agree to the Terms of Use

Importing json that has a nested array


My json looks like this:

"userid" : "10",
"actions" :
{"type":"flip", "timestamp":"5pm"},
{"type":"jump", "timestamp":"6pm"}

I want to bring that into a table in redshift that has columns "userid, type, timestamp"
Do I need to reformat the file into flat lines or is this possible with a standard copy command?

asked 3 years ago74 views
3 Answers
Accepted Answer

It is not currently possible to import such data using COPY. However, you can define the data as a Spectrum external table and use our nested data support to bring the data in.

CREATE EXTERNAL TABLE spectrum.nested_example
     ( userid     int
     , actions    array<struct<type:varchar(20), timestamp:varchar(20)>>
LOCATION 's3://benchmark-files/temp/nested_data/nested_example/'
SELECT u.userid
FROM spectrum.nested_example u
INNER JOIN c.actions a ON true
answered 3 years ago

I'm guessing this degrades query performance significantly over having it stored flat?

answered 3 years ago

Hi meyerovb,

You could always take the SELECT against the external table that Joe gave you and put it into a CTAS statement to materialize the data in local Redshift storage if you're concerned about the unnest the columns at query time. It's really an alternate load pattern for Redshift and one that only need be executed once each time the external table data changes.

However, don't be too surprised when you find case where Spectrum query performance on external data beats Redshift local storage performance, especially when not joining big tables. This happens because Redshift Spectrum parallelism is up to 10x wider than Redshift slice parallelism. So even though the data is read from S3 instead of local disk there are 10x more scan workers to get the heavy lifting at the access operator level of the query plan done faster, assuming there are enough splits in S3 to feed them all relatively evenly.

I hope this helps you at least a little,

answered 3 years ago

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions