https://kedro.org/ logo
#questions
Title
# questions
f

Filip Wójcik

04/13/2023, 7:51 AM
Hi guys, our team is wondering is it possible to have collapsible pipelines in Kedro viz, without applying namespace to every dataset corresponding to that pipeline? Sometimes it seems to over-complicate things, if your only use case is to have a nice, collapsible look 🙂 Thanks in advance Filip
d

datajoely

04/13/2023, 7:55 AM
so the short answer is no - but would love to hear your feedback on how it compliments things, it’s something I’m keen to improve the UX of
m

marrrcin

04/13/2023, 7:58 AM
collapsible pipelines in Kedro viz,
What do you mean by that?
d

datajoely

04/13/2023, 7:59 AM

https://docs.kedro.org/en/0.18.1/_images/modular_ds.gif

m

marrrcin

04/13/2023, 7:59 AM
Isn’t the use of
pipeline()
within
pipeline()
enough for that?
d

datajoely

04/13/2023, 8:00 AM
the
namespace
argument is required IIRC
👍 1
f

Filip Wójcik

04/13/2023, 8:00 AM
Hi! Thanks for the quick replay. The story is short. We often have pipelines that are composed of several nodes. When doing presentations, hiding such multiple steps and showing only one general block to focus on a high-level view is often desirable. Unfortunately, in creating a namespace, we need to include a whole section of mapping inputs or parameters in the pipeline definition, which is tedious, if you need only a collapsible tile in the UI. But after all - the functionality is excellent, and this is only a minor inconvenience 🙂
👍 1
d

datajoely

04/13/2023, 8:12 AM
I think that’s valid, I find the experience of doing so too fiddly myself
the best solution I can offer today is to namespace from the start
👍 1
but I appreciate that’s not helpful in this situation
f

Filip Wójcik

04/13/2023, 8:13 AM
No worries. After all, Kedro & Kedro-Viz are absolutely cool even at this point.Thanks a lot!
a

Antony Milne

04/13/2023, 8:17 AM
This is a very valid problem I think. Basically I think using namespaces to do this collapsible grouping is the right thing to do - the problem is that it’s awkward to do that if all you want the namespace for is to make something collapsible in kedro viz. There’s two things we could do that might make this easier in the future: • some special value
ALL
that “freezes” all dataset names in the call to pipeline, e.g.
pipeline(inputs=ALL, outputs=ALL, namespace="blah")
• making the catalog hierarchical in the same way parameters are, so that to add a namespace
blah
to a group of dataset names you don’t need to manually prepend
blah.
to each of them but instead make a top-level key
blah
and then indent all the datasets with that namespace underneath it
This way you’d still need to namespace something to make it collapsible, but it would be much easier to apply that namespace quickly without needing to rename all your datasets.