issue_comments

I'm sorry, but just to reiterate, "x" is referred to as a dimension several times in the document. It is then, one time, mistakenly referred to as a coordinate. This pull request, which should be accepted, is to make the document consistent in how it refers to "x".

Yeah, as a total new comer, this explanation seems not only unintuitive but also like we're fighting with common terminology used in math and everywhere else.

When you define the coordinates,

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

It looks like you are defining a specific coordinate system [10,20] on a specific dimension "x". The key of the object-value pair set in coords (ie. "x") MUST refer to a dimension, because "x" is a dimension that we are defining a coordinate system for. As we don't define any coordinate system on "y", asking xarray for data.coords["y"] gives us a array([0,1,2]) which we'd consider to be an intuitive default coordinate system.

Intuitively, we think: An xarray contains dimensions along which specific coordinates can be defined: xarray.dimension.coordinate

In the attribute definition, we say:

data.x.attrs["units"] = "x units" which feels like we are defining an attribute on the entire dimension. As opposed to data.x.10 (I realize the problem here) which would be definiting an attribute on a single coordinate within a dimension.

I'm done, thank you for your time. I will continue reading and learning about xarray.

Well, it's not consistent within this particular document, which is reasonably the first anyone reads. The dimension, as its refered to, in this example is called "x", we are assigning an attribute to "x", the coordinates of "x" are [10,20], we are not assigning an attribute to a coordinate within the dimension "x".