Map/transform does not add a column but creates a new DataFrame with only the new column

[m4ttes4]

Map function do not modify the dataframe as expected ?

Minimal Reproducible Example:

using DataFrames, TableTransforms

A = DataFrame(a = 1)

A` is a DataFrame with one column: :a = 1

Using pipeline with Map/transform
B = A |> Map(:a => sum => :b)

Result:
B will be a new DataFrame with only one column :b, which equals 2 and A is the same.

[juliohm]

@m4ttes4 none of the transforms implemented in this package modify the input. This is expected behavior.

Notice that we provide parallel pipelines, which would be extremely difficult to implement with in-place changes.

[m4ttes4]

Hi @juliohm,

My question may sound trivial, but I used to have a pipeline where the code

B = A |> Map(....)

would create a new DataFrame containing the columns of A plus the new column(s) generated by the Map function.

This doesn’t seem to be the case anymore. Has this behavior changed in a recent release?

[juliohm]

Got it! I misunderstood your message.

Yes, we changed the behavior recently because it was an anti-pattern. If you need to attach the original columns you can use the \sqcup operator as described in the docs. For example:

t |> (Map(...) ⊔ Identity())

Most uses of Map we encounter in applications don't require carrying all other inputs forward.

[m4ttes4]

Thank you @juliohm !