Add string representation to rio acccessor

[remi-braun]

To help to quickly grasp a dataset information, a bit like printing ds.profile in rasterio or running gdalinfo, it could be great to add a string representation to the accessor.

Something like :

import rioxarray
>>> xda = rioxarray.open_rasterio(r"my_raster.tif")
>>> print(xda.rio)

rioxarray accessor (.rio) | RasterDataset
Variables:
        Name: 'Slope'
        'Slope' attributes:
                count: 1
                crs: 4326
                rasterio_dtype: float64
                nodata: -9999.0
                transform: | 0.00, 0.00,-94.30|
| 0.00,-0.00, 42.35|
| 0.00, 0.00, 1.00|
                height: 50
                width: 35
                blockxsize: 35
                blockysize: 1
                bounds: (-94.30306000000002, 42.350229999999996, -94.30271000000002, 42.350730000000006)
        Name: 'Red'
        'Red' attributes:
                count: 1
                crs: 4326
                rasterio_dtype: float64
                nodata: -9999.0
                transform: | 0.00, 0.00,-94.30|
| 0.00,-0.00, 42.35|
| 0.00, 0.00, 1.00|
                height: 50
                width: 35
                blockxsize: 35
                blockysize: 1
                bounds: (-94.30306000000002, 42.350229999999996, -94.30271000000002, 42.350730000000006)

Issue based on #911

[remi-braun]

@snowman2 Any thoughts on that? 😄

[snowman2]

Thank you for your patience. I am hoping for a view that looks similar to the current xarray dataset that is more compact with further details as you view each dataarray.

Also, have it say RasterDataset instead of Dataset and RasterArray instead of array.

The other component to figure out is which information to show.

[remi-braun]

Sadly, I'm really not fluent with nice HTML repr 😢
I think what you also mean is that you wish to have rio information displayed at the xda/xds level (without asking specifically for the rio accessor), right?
So, it would need to extend xarray's repr (which is very complex I must say, with both str and html reprs)

Maybe the PR can be an intermediate milestone in order to at least have a repr, even if very simple?
Another issue could be opened to improve it.

On the topic on what to display, we can base ourself on what is available in rasterio's profile or what is printed with gdalinfo.

[snowman2]

HTML is not required. Just similar formatting.

Here is an example of what I am thinking:

rioxarray.RasterDataArray: 'Slope' (y[latitude]: <height>, x[longitude]: <width>, (Optional: z[time]: <count>) 
Profile:
  crs: ...
  dtype: float64
  nodata: ...
  transform: ...
  blockxsize: 35
  blockysize: 1
  bounds: (-94.30306000000002, 42.350229999999996, -94.30271000000002, 42.350730000000006)

rioxarray.RasterDataset
Dimensions:  (y[latitude]: <height>, x[longitude]: <width>, (Optional: z[time]: <count>) )
Data variables:
  Slope     (y[latitude], x[longitude], z[time])
  Red     (y[latitude], x[longitude], z[time])
Profile:
  crs: ...
  dtype: float64
  nodata: ...
  transform: ...
  blockxsize: 35
  blockysize: 1
  bounds: (-94.30306000000002, 42.350229999999996, -94.30271000000002, 42.350730000000006)

[remi-braun]

How should we manage Datasets with different profile per variable?

[snowman2]

How should we manage Datasets with different profile per variable?

I would give instructions to see data variables for details if they are all different.