ENH: make vega_datasets accessible from alt.datasets

[jakevdp]

Addresses #796

This makes the contents of vega_datasets available from Altair via alt.datasets. So, for example, you can do:

import altair as alt
cars = alt.datasets.cars()
alt.Chart(cars).mark_point()

If the vega_datasets package is not installed, an error will be raised when you try to access a dataset, but not when you import the alt.datasets object (which is present in the main namespace)

In [1]: from altair import datasets
In [2]: datasets.cars()
---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
<ipython-input-2-a4a02b6877f0> in <module>()
----> 1 datasets.cars()

~/github/altair-viz/altair/altair/_datasets.py in __getattr__(self, attr)
     21 
     22     def __getattr__(self, attr):
---> 23         raise ImportError(ERR_MESSAGE)
     24 
     25 

ImportError: 
The vega_datasets package must be installed in order to use altair.datasets.
See http://github.com/altair-viz/vega_datasets/ for information.

[jakevdp]

I'm still about 50/50 on this change... or maybe 60/40 in favor.

• Pros

  • fewer imports leads to simpler examples and an easier learning curve for beginners
  • there is still no hard dependency on vega_datasets, so Altair can be used without it
  • this will open the way for more streamlined tools for things like choropleth maps, which require access to the data

• Cons

  • this obscures the existence of the vega_datasets package and might make it less clear how to debug issues that arise, particularly for beginners

[arokem]

You might want to also mention somewhere in the documentation (in case that's not already mentioned) that for these data to be available, users should install vega_datasets as well.

On the other hand If you don't, users will get that message on the first time they refer to alt.datasets so meh.

[jakevdp]

All the installation instructions list vega_datasets, so I think we have that covered: https://altair-viz.github.io/getting_started/installation.html

[arokem]

You're probably right, but to be on the safe side, in case someone shows up in the examples, instead of this page, I would add something like "Some of these examples also require an installation of vega_datasets " on this page: https://altair-viz.github.io/gallery/index.html

[palewire]

@arokem, we've already done that in #829

[arokem]

Whoops. Sorry about the noise!

[ellisonbg]

Jake, I am fine either way on this one

…

-- Brian E. Granger Associate Professor of Physics and Data Science Cal Poly State University, San Luis Obispo @ellisonbg on Twitter and GitHub bgranger@calpoly.edu and ellisonbg@gmail.com

[jakevdp]

Having slept on this a bit, I think I'd prefer not to wrap datasets into altair.