Hi, just an idea: Sometimes it would be nice to be able to pass some function arguments directly when creating a kedro Node. This is especially useful for dynamic pipeline creation. For example, I want to create a preprocessing pipeline for each dataset. I have a function

create_data_pipeline(dataset)

, that downloads and processes the respective dataset via a function

make_data(dataset)

. Right now, I guess I could go for

partial(make_data, dataset)

, register dataset as param in the catalog, or create a separate download function for each dataset. All of these feel a bit silly/convoluted to me, since I already have the

dataset

value right there, in my pipeline creation function. Ideally, I would like to create something like

Node(make_data, python_inputs=dataset)

, i.e., inputs that are not looked up in the catalog. What do you think?

Agreed, it's rather frustrating!

Hi @Bjarne Hiller and @Laurens Vijnck, Thank you for the suggestions and letting us know about the not-so-great experience. This will help us improve the package. There has been an old discussion around this - https://github.com/kedro-org/kedro/discussions/1279 We will try to get this in our grooming sessions and spike on the possibility of having a way to pass node inputs. Thank you cc: @Dmitry Sorokin

As an FYI, this has been discussed several times before (e.g. in #213 and #526). Added a comment to the issue accordingly.