Joiner - Make it possible to join on integer values of different size

A face this time to time. I join two tables where number columns a should join on are of different sizes. For instance, I could have Long value on the left and Int value on the right. I would believe I can always state the values equal or differ no matter how many bites each value occupies in memory.

But every time I was joining two tables using a pair of long and int column, I got an error: " Type mismatch found of Joining Column Pair “x” and “y”.This causes an empty output table."

Getting rid of it would be nice.

2 Likes

Definitely +1 on the issue. This happens to me, where I’ll receive a .csv file from another department where the customer ID information is stored as a string. When I join it with data that’s pulled from our table (stored as integer), I get the error you’re experiencing. It’s not a big deal, but it is annoying, especially if you’re working with a large workflow and end up having to convert several times throughout.

It would be great either to have a checkbox that converts all joining columns to strings, or to have a checkbox that converts formats on the “left” table to formats on the “right” table (or vice versa).

1 Like

Hello!

Just came across it today as well but with Joiner (Labs) which does not produce error but rather doesn’t match it :slight_smile:

Ticket created! (Internal reference: AP-14999)

Br,
Ivan

2 Likes

Hi Ivan,

did you created a ticket about Joiner(Labs) behaviour or a ticket about the problem I described?

Thanks

Hi Snowy,

Thanks for your contribution. I’d be careful with the case you replied with. What I suggested it should be possible to join tables on integer number columns with no respect to their size (integer, long, …). I don’t think we can apply this to joins on string and number columns each other safely too. At least it would take some performance costs if the node converted all values to string to perform the join.

Jan

1 Like

More a discussion point, but due to potential rounding issues, is it not usually best to convert to string then join?

3 Likes

Hello!

@jan_lender I created a ticket in general for joining based on different column types which covers problem you described as well :slight_smile:

Br,
Ivan

1 Like

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.

Hello there @Snowy, @jan_lender and others,

with KNIME version 4.4.0 Joiner node is now out of Labs and features a possibility to match different column types. Give it a try and feel free to comment.

Br,
Ivan

2 Likes