Typo about "len(arr.dims) <= len(arr.coords) in general"

[OrionSmedley]

What happened?

Its on the website https://docs.xarray.dev/en/stable/user-guide/terminology.html, but isn't true in general.

What did you expect to happen?

the following code should return true, if the website is correct.

Minimal Complete Verifiable Example

import xarray as xr
import numpy as np
data = xr.DataArray(np.random.randn(2, 3), dims=("x", "y"), coords={"x": [10, 20]})

print(len(data.dims ))
print(len(data.coords))

print( len(data.dims) <= len(data.coords) )

MVCE confirmation

• Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.
• Complete example — the example is self-contained, including all data and the text of any traceback.
• Verifiable example — the example copy & pastes into an IPython prompt or Binder notebook, returning the result.
• New issue — a search of GitHub Issues suggests this is not a duplicate.

Relevant log output

No response

Anything else we need to know?

No response

Environment

Details

[welcome]

Thanks for opening your first issue here at xarray! Be sure to follow the issue template!
If you have an idea for a solution, we would really welcome a Pull Request with proposed changes.
See the Contributing Guide for more.
It may take us a while to respond here, but we really value your contribution. Contributors like you help make xarray better.
Thank you!

[headtr1ck]

Thanks for the issue, always happy to improve the documentation.

However, I don't really see how this implies that you cannot have more coordinates as dimensions. It is written that a coordinate labels a dimension or set of dimensions, so clearly you can have multiple coordinates per dimension.

If you consider dimension-coordinates only, your example is correct.

[dcherian]

@ionsme can you suggest an improvement to the text please? It really helps when someone who's not a maintainer edits these things.

[OrionSmedley]

@dcherian Ok, I'll think about a good wording a bit later and do that.

[OrionSmedley]

@headtr1ck I agree that you can have multiple coordinates per dimension.

However, the documentation explicitly says that you can't have it the other way around.
That is the documentation currently says len(data.dims) <= len(data.coords) is always true.
This code snip bit I provided is a counter example; it explicitly returns false.

I also agree that "if you consider dimension-coordinates only" is correct. In that case, the code given in the documentation should be changed to len(data.dims) <= len(data.dimension-coords) (where data.dimension-coords is a stand in, because I don't know the correct expression to put)

[headtr1ck]

Oh wow, I completely missed the part where it says:

As a consequence, len(arr.dims) <= len(arr.coords) in general.

This is simply wrong, you are right.
You can have any number of coordinates, from none to many.

I don't think there is a direct way to retrieve dimension coordinates only.

Maybe we should retire the concept of dimension coordinate and purely talk about indexes. At the end it is just a 1D index with the same name as the dimension. I think there are still special use cases for them, like automatic tics in plotting.
@dcherian what do you think? Maybe this should be discussed in the call.

[dcherian]

concept of dimension coordinate and purely talk about indexes. At the end it is just a 1D index with the same name as the dimension.

Possibly. I think we'll have to reword this section significantly to talk about indexes (xref #6975)

cc @benbovy

[OrionSmedley]

Maybe we should retire the concept of dimension coordinate and purely talk about indexes. At the end it is just a 1D index with the same name as the dimension. I

Would it be possible to just make all coordinates dimension coordinates?
Because it seems to me that "non dimension coordinates" are just coordinates that have been unnecessarily demoted.

At least, from the user interface point of view.

By demoted I mean that you can't use it for things like selection, unless you use the workaround
ds.sel(MyDimCoord = ds.MyNonDimCoord == 'something')
rather than simply using
ds.sel(MyNonDimCoord =="something")

Also In terms of the graphing you mentioned, could just have the user select which coordinates to plot.

Under the hood,

I was thinking maybe there was some reason for this demoting. Naively though, I feel it should be possible to keep all coordinates as dimension coordinates. Maybe not the most efficient, but just to say it's possible, one could implement it as follows:
Make a pandas table that had the currently existing "dimension coordinate" indexed with the "non-dimension coordinates"
Then, if the user tries to do ds.sel(MyNonDimCoord =="something"), you find the "dimension" coordinate in the pandas table. And then run the ordinary ds.sel(MyDimCoord = "thingCorrespondingToSomething").