Survey banner
Switching to Dataiku - a new area to help users who are transitioning from other tools and diving into Dataiku! CHECK IT OUT

Multi join returns with null columns while single join works.

Eulogitos
Level 1
Multi join returns with null columns while single join works.

Hello,

I have a dataset that I am breaking it down to 5 different ones.

Each dataset has a common column, column_id.

When I am done preparing each dataset, I multi left join each prepared dataset back to the original based on column_id. I am doing that to bring back the prepared columns.

My problem is that a few of the prepared columns come back as completely null; however, if I perform a single join instead of the a multi join, the prepared columns come back with results.

I face this issue only when I am using a multi join.

Has anyone experienced something like this or might be of help?

Thanks!

P.S.

What I am doing is basically folding columns together, in order to create single collumns with the desired values. I used to be an Alteryx user and recently started using Dataiku. So, when I was using Alteryx to achieve the same result I would transpose my data and then join back the transposed columns to the original dataset.


Operating system used: Windows 11

0 Kudos
1 Reply
ChrisH
Dataiker

Hello!  Happy to help with this. When I hear blank columns my first thought would be that for some reason the column id field isn't matching for some of the datasets. To figure out why it works with joining one dataset and not multiple would it be possible to provide some additional info? It would be great to understand:

 

  • What does the configuration of your multi-join look like? Is it all 5 of the new datasets joined back to the original or are they joined to each other in some  more complex way? Could you post a screenshot of the configuration of your Join recipe?

  • Is it possible to share any sample data of what you're trying to join and the result you're trying to end up with? Definitely understand the goal of folding columns but if I can better understand the final result you're aiming for it would be very helpful.

 

Thanks!

0 Kudos