Do people still care about dummy inputs / outputs ...
# user-research
n
Do people still care about dummy inputs / outputs when there is need to force Kedro execute node in a specific order? https://github.com/kedro-org/kedro/discussions/3758
j
yes I think it's a crucial workaround until there's a better solution, I know @Alex Ferrero and team use this
n
Was planning to do a half knowledge sharing+ tech design but no hurry for this since most people accept that Kedro don’t have a nice way to deal with this.
d
it doesn’t by design
but i think it’s clear we see the user value now
b
Why not let people influence the topological sort in some way? Not so that it violates any constraints, but there are sub-parts of the DAG where some kind of
__gt__
/
__lt__
comparison might be possible on nodes?
n
@Ben Horsburgh can you elaborate on the last sentence? The discussion is indeed proposing to open up some ways of changing the order when there are multiplies solutions available, as toposort give one valid solve only.
b
I have a good idea - let me write up an example in issue over the weekend 🙂
🚀 1
n
did you end up writing it?
a
I'm extremely late to this and I don't know if it is still in the spotlight but yes, we use dummy variables to set the order of the nodes in a pipeline. To be honest, I have been trying to find a few days to play with the current logic but it's been impossible... I wanted to test an approach based on my knowledge on AWS lambdas definition string, which have a parameter 'next' to indicate which is the next node to run, use a custom method to create the order and use the current logic used by Kedro to allow or not allow the order set by the user. Again, this is an just an idea and I'm not a pure software developer so I'm not sure if it's the best approach. PS: I'm turning my notifications on not to miss these threads again 😀
❤️ 1