New MultiIndex dimension behaviours

[ConnorSwales]

What is your issue?

As of the latest release (2022.6.0), the behaviour of MultiIndex coordinates and their constituents appears to have changed.

Creating a simple multi-indexed DataSet:

ds1 = xr.Dataset({'foo': (('x',), [1, 2, 3])}, {'x': [1, 2, 3], 'y': 'a'})
ds2 = xr.Dataset({'foo': (('x',), [4, 5, 6])}, {'x': [1, 2, 3], 'y': 'b'})

mult = xr.concat([ds1, ds2], dim='y').stack(yx=['y', 'x'])

Printing the DataSet and the coordinate indices on version 2022.3.0:

mult
<xarray.Dataset>
Dimensions:  (yx: 6)
Coordinates:
  * yx       (yx) MultiIndex
  - y        (yx) object 'a' 'a' 'a' 'b' 'b' 'b'
  - x        (yx) int64 1 2 3 1 2 3
Data variables:
    foo      (yx) int64 1 2 3 4 5 6

mult.coords.indexes
yx: MultiIndex([('a', 1),
                ('a', 2),
                ('a', 3),
                ('b', 1),
                ('b', 2),
                ('b', 3)],
               names=['y', 'x'])

Printing the DataSet and the coordinate indices on version 2022.6.0:

mult
<xarray.Dataset>
Dimensions:  (yx: 6)
Coordinates:
  * yx       (yx) object MultiIndex
  * y        (yx) <U1 'a' 'a' 'a' 'b' 'b' 'b'
  * x        (yx) int64 1 2 3 1 2 3
Data variables:
    foo      (yx) int64 1 2 3 4 5 6

mult.coords.indexes
Indexes:
yx: MultiIndex([('a', 1),
                ('a', 2),
                ('a', 3),
                ('b', 1),
                ('b', 2),
                ('b', 3)],
               name='yx')
y: MultiIndex([('a', 1),
               ('a', 2),
               ('a', 3),
               ('b', 1),
               ('b', 2),
               ('b', 3)],
              name='yx')
x: MultiIndex([('a', 1),
               ('a', 2),
               ('a', 3),
               ('b', 1),
               ('b', 2),
               ('b', 3)],
              name='yx')

On the latest version (2022.6.0), the constituent coordinates (x,y) of the multi-indexed coordinate (yx) have asterisks next to them, implying that they are dimensional (despite the them not being named equal to their sole dimension, as stated in the docs' coordinates section).

dimension coordinates are one dimensional coordinates with a name equal to their sole dimension (marked by * when printing a dataset or data array)

Further, looking into the .coords.indexes shows that each constituent coordinate is also indexed by an instance of the same MultiIndex.

I was wondering if this change been made design and whether the documentation just need updating, or I have misunderstood something along the way.

[benbovy]

Hi @ConnorSwales,

Yes this is an intentional change part of the ongoing explicit indexes refactor. There's still some work to do on the documentation. For more details, see:

• MultiIndex listed multiple times in Dataset.indexes property #6752
• Explicit indexes: next steps #6293
• https://docs.xarray.dev/en/stable/roadmap.html#flexible-indexes
• https://github.com/pydata/xarray/projects/1

[headtr1ck]

Closing for now.
Feel free to reopen or better to comment in the linked issues above.