Open Galileo-Galilei opened 10 months ago
Nice idea, I asked in the team and start tagging issues.
MatplotlibWriter
in various places. To retain backwards compatibility, they would probably have to do some try ... except
with imports. cc @rashidakanchwala @tynandebold _load
for certain datasets (including this one) because they're meant to be artifacts and not really full I/O components.Regardless, for consistency I agree we should rename it 👍🏽
An user ask why load
is not supported? Maybe this is the reason why it is called Writer but not Dataset. I don't use this dataset myself so I can't remember is it always like this.
Maybe we could have a MatplotlibFigureDataset
(that saves and loads figures objects as pickles) (although maybe we shouldn't and we should tell users to use PickleDataset
for this) and MatplotlibImageDataset
(that only saves images and has no _load
)
Actually it totally makes sense that some dataset don't have _save
method (API Dataset
didn't for a while, SQLQueryDataset
still doesn't...). Can I start a PR for this or the impact is too high on kedro-viz?
Yeah I don't think it's a problem that some datasets don't have _save
. What's problematic is the monkeypatching that kedro-viz does on some of them https://github.com/kedro-org/kedro-viz/issues/1352
About the impact on Kedro Viz, cc @rashidakanchwala @tynandebold
Yes, it will. We will be looking at making sure Kedro-viz is not so coupled with Kedro-datasets. And we are going to prioritise this work in the second half of November most likely. Until then, we could pin kedro-viz to kedro-datasets to the version it would work with.
I don't think this is a good first issue, there's some uncertainty still https://github.com/kedro-org/kedro-plugins/issues/353#issuecomment-1740974418 removing the label, when we have more clarity we'll update this issue
Description
MatplotlibWriter
has been the only dataset with inconsistent naming for a while. There are a couple of more subtle issues with it still to be fixed (see https://github.com/kedro-org/kedro-plugins/issues/529), but I think we should use the dataset renaming (which is a breaking change whatever) as an opportunity to make it consistent with the rest of the codebaseContext
Consistency is better :)
Possible Implementation
Rename
MatplotlibWriter
toMatplotlibDataset
.Possible Alternatives
Don't do it ;)
This is likely a great ticket for Hacktoberfest if the project is eligible?