vega_datasets source

[ChristopherDavisUCI]

Would there be a reliable way to transition from using https://github.com/altair-viz/vega_datasets as the default source of example datasets to instead https://github.com/vega/vega-datasets ? Or, if that's not realistic, would there be a way to synchronize the data in the altair-viz version with the data in the vega version? (As far as I can tell, they're not synchronized.)

I could imagine having no local datasets and having functionality like alt.vega_datasets.cars() which downloads the cars dataset from https://github.com/vega/vega-datasets/blob/main/data/cars.json

I don't know at all about the tradeoffs involved, it just feels like this approach might remove one step of maintenance.

Any thoughts on that?

[mattijn]

I like the idea. I remember it took me a while before realizing they are not the same repository.

The goal of vega_datasets was to make the datasets used by the Vega project easily accessible in Python in the form of a Pandas dataframe.

Since it is also used as teaching material, I can imagine we also like to provide a wider accessibility to choose which dataframe library a user prefers.

An interesting change with the past is that the Vega datasets are currently versioned by GitHub releases. Maybe that provides a new entrypoint for Altair.

Since this will change all examples in the documentation, we probably might think this through a bit how we can support users that are bound to older versions.

[dangotbanned]

I'm quite interested in this, so I opened #3631

I'd noticed a lot of examples using vega_datasets purely for the url, and that part of the package is really quite short.

[dangotbanned]

@ChristopherDavisUCI, @mattijn I feel like I've got this to good place now for feedback.

• feat(RFC): Adds altair.datasets #3631

I've got a lot of docs and tests already, which I'm hoping give a feel for the current API.
Any thoughts large or small would be super appreciated

[mattijn]

Will defining the dataset by URL goes through "Loader"?

[dangotbanned]

@ChristopherDavisUCI I hope it is okay with you, I've renamed this discussion to help with visibility.

I can revert if not, but I stumbled upon this only after searching for previous vega_datasets threads.
I was already working on #3631 locally, but thought it would be helpful to share after discovering this

[ChristopherDavisUCI]

Hi @dangotbanned Sure of course, thanks for looking into this! (I admit it is not very fresh in my memory what the different considerations were.)

[ChristopherDavisUCI]

Thank you for your work on this @dangotbanned ! I'm sorry that I haven't had a chance to look into this yet, but it is still on my radar. Please don't hesitate to send pings. -Chris

…

On Sat, Nov 16, 2024 at 3:16 PM Mattijn van Hoek ***@***.***> wrote: Will defining the dataset by URL goes through "Loader"? — Reply to this email directly, view it on GitHub <https://urldefense.com/v3/__https://github.com/vega/altair/discussions/3150*discussioncomment-11280516__;Iw!!CzAuKJ42GuquVTTmVmPViYEvSg!Kvdeataf7_6HtPnCbd_dFZsScA-usqaoR6qWRxHgeDR_fQSphDI7xC7kV43OBli1F56fwUTDwysqsJN2Ey8wFI0$>, or unsubscribe <https://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/ATXKWIBLWJF4HDZXNPNTJW32A7G5XAVCNFSM6AAAAABPMNYZRWVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTCMRYGA2TCNQ__;!!CzAuKJ42GuquVTTmVmPViYEvSg!Kvdeataf7_6HtPnCbd_dFZsScA-usqaoR6qWRxHgeDR_fQSphDI7xC7kV43OBli1F56fwUTDwysqsJN2cGSb1SE$> . You are receiving this because you were mentioned.Message ID: ***@***.***>