GroupBy(multiple groupers) (#9372)

* GroupBy(multiple groupers)

* Add example to docs

fix docs

[pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix docs

* More docs

* fix doc

* fix doc again

* Fix bug.

* Add whats-new note

* edit

* Error on multi-variable groupby with MultiIndex

* Update doc/user-guide/groupby.rst

---------

Co-authored-by: Maximilian Roos <[email protected]>

doc/user-guide/groupby.rst

@@ -81,8 +81,7 @@ You can index out a particular group:
 
     ds.groupby("letters")["b"]
 
-Just like in pandas, creating a GroupBy object is cheap: it does not actually
-split the data until you access particular values.
+To group by multiple variables, see :ref:`this section <groupby.multiple>`.
 
 Binning
 ~~~~~~~
@@ -180,19 +179,6 @@ This last line is roughly equivalent to the following::
         results.append(group - alt.sel(letters=label))
     xr.concat(results, dim='x')
 
-Iterating and Squeezing
-~~~~~~~~~~~~~~~~~~~~~~~
-
-Previously, Xarray defaulted to squeezing out dimensions of size one when iterating over
-a GroupBy object. This behaviour is being removed.
-You can always squeeze explicitly later with the Dataset or DataArray
-:py:meth:`DataArray.squeeze` methods.
-
-.. ipython:: python
-
-    next(iter(arr.groupby("x", squeeze=False)))
-
-
 .. _groupby.multidim:
 
 Multidimensional Grouping
@@ -236,6 +222,8 @@ applying your function, and then unstacking the result:
     stacked = da.stack(gridcell=["ny", "nx"])
     stacked.groupby("gridcell").sum(...).unstack("gridcell")
 
+Alternatively, you can groupby both `lat` and `lon` at the :ref:`same time <groupby.multiple>`.
+
 .. _groupby.groupers:
 
 Grouper Objects
@@ -276,7 +264,8 @@ is identical to
 
     ds.groupby(x=UniqueGrouper())
 
-and
+
+Similarly,
 
 .. code-block:: python
 
@@ -303,3 +292,26 @@ is identical to
     from xarray.groupers import TimeResampler
 
     ds.resample(time=TimeResampler("ME"))
+
+
+.. _groupby.multiple:
+
+Grouping by multiple variables
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+Use grouper objects to group by multiple dimensions:
+
+.. ipython:: python
+
+    from xarray.groupers import UniqueGrouper
+
+    da.groupby(lat=UniqueGrouper(), lon=UniqueGrouper()).sum()
+
+
+Different groupers can be combined to construct sophisticated GroupBy operations.
+
+.. ipython:: python
+
+    from xarray.groupers import BinGrouper
+
+    ds.groupby(x=BinGrouper(bins=[5, 15, 25]), letters=UniqueGrouper()).sum()

doc/whats-new.rst

@@ -24,6 +24,11 @@ New Features
 ~~~~~~~~~~~~
 - Make chunk manager an option in ``set_options`` (:pull:`9362`).
   By `Tom White <https://github.com/tomwhite>`_.
+- Support for :ref:`grouping by multiple variables <groupby.multiple>`.
+  This is quite new, so please check your results and report bugs.
+  Binary operations after grouping by multiple arrays are not supported yet.
+  (:issue:`1056`, :issue:`9332`, :issue:`324`, :pull:`9372`).
+  By `Deepak Cherian <https://github.com/dcherian>`_.
 - Allow data variable specific ``constant_values`` in the dataset ``pad`` function (:pull:`9353``).
   By `Tiago Sanona <https://github.com/tsanona>`_.
 

xarray/core/dataarray.py

@@ -6801,27 +6801,22 @@ def groupby(
             groupers = either_dict_or_kwargs(group, groupers, "groupby")  # type: ignore
             group = None
 
-        grouper: Grouper
+        rgroupers: tuple[ResolvedGrouper, ...]
         if group is not None:
             if groupers:
                 raise ValueError(
                     "Providing a combination of `group` and **groupers is not supported."
                 )
-            grouper = UniqueGrouper()
+            rgroupers = (ResolvedGrouper(UniqueGrouper(), group, self),)
         else:
-            if len(groupers) > 1:
-                raise ValueError("grouping by multiple variables is not supported yet.")
             if not groupers:
                 raise ValueError("Either `group` or `**groupers` must be provided.")
-            group, grouper = next(iter(groupers.items()))
-
-        rgrouper = ResolvedGrouper(grouper, group, self)
+            rgroupers = tuple(
+                ResolvedGrouper(grouper, group, self)
+                for group, grouper in groupers.items()
+            )
 
-        return DataArrayGroupBy(
-            self,
-            (rgrouper,),
-            restore_coord_dims=restore_coord_dims,
-        )
+        return DataArrayGroupBy(self, rgroupers, restore_coord_dims=restore_coord_dims)
 
     @_deprecate_positional_args("v2024.07.0")
     def groupby_bins(

xarray/core/dataset.py

@@ -10397,25 +10397,22 @@ def groupby(
             groupers = either_dict_or_kwargs(group, groupers, "groupby")  # type: ignore
             group = None
 
+        rgroupers: tuple[ResolvedGrouper, ...]
         if group is not None:
             if groupers:
                 raise ValueError(
                     "Providing a combination of `group` and **groupers is not supported."
                 )
-            rgrouper = ResolvedGrouper(UniqueGrouper(), group, self)
+            rgroupers = (ResolvedGrouper(UniqueGrouper(), group, self),)
         else:
-            if len(groupers) > 1:
-                raise ValueError("Grouping by multiple variables is not supported yet.")
-            elif not groupers:
+            if not groupers:
                 raise ValueError("Either `group` or `**groupers` must be provided.")
-            for group, grouper in groupers.items():
-                rgrouper = ResolvedGrouper(grouper, group, self)
+            rgroupers = tuple(
+                ResolvedGrouper(grouper, group, self)
+                for group, grouper in groupers.items()
+            )
 
-        return DatasetGroupBy(
-            self,
-            (rgrouper,),
-            restore_coord_dims=restore_coord_dims,
-        )
+        return DatasetGroupBy(self, rgroupers, restore_coord_dims=restore_coord_dims)
 
     @_deprecate_positional_args("v2024.07.0")
     def groupby_bins(