ENH: Add numba engine to groupby.aggregate

asv_bench/benchmarks/groupby.py

@@ -660,4 +660,62 @@ def function(values):
         self.grouper.transform(function, engine="cython")
 
 
+class AggEngine:
+    def setup(self):
+        N = 10 ** 3
+        data = DataFrame(
+            {0: [str(i) for i in range(100)] * N, 1: list(range(100)) * N},
+            columns=[0, 1],
+        )
+        self.grouper = data.groupby(0)
+
+    def time_series_numba(self):
+        def function(values, index):
+            total = 0
+            for i, value in enumerate(values):
+                if i % 2:
+                    total += value + 5
+                else:
+                    total += value * 2
+            return total
+
+        self.grouper[1].agg(function, engine="numba")
+
+    def time_series_cython(self):
+        def function(values):
+            total = 0
+            for i, value in enumerate(values):
+                if i % 2:
+                    total += value + 5
+                else:
+                    total += value * 2
+            return total
+
+        self.grouper[1].agg(function, engine="cython")
+
+    def time_dataframe_numba(self):
+        def function(values, index):
+            total = 0
+            for i, value in enumerate(values):
+                if i % 2:
+                    total += value + 5
+                else:
+                    total += value * 2
+            return total
+
+        self.grouper.agg(function, engine="numba")
+
+    def time_dataframe_cython(self):
+        def function(values):
+            total = 0
+            for i, value in enumerate(values):
+                if i % 2:
+                    total += value + 5
+                else:
+                    total += value * 2
+            return total
+
+        self.grouper.agg(function, engine="cython")
+
+
 from .pandas_vb_common import setup  # noqa: F401 isort:skip

doc/source/whatsnew/v1.1.0.rst

@@ -98,7 +98,7 @@ Other enhancements
   This can be used to set a custom compression level, e.g.,
   ``df.to_csv(path, compression={'method': 'gzip', 'compresslevel': 1}``
   (:issue:`33196`)
-- :meth:`~pandas.core.groupby.GroupBy.transform` has gained ``engine`` and ``engine_kwargs`` arguments that supports executing functions with ``Numba`` (:issue:`32854`)
+- :meth:`~pandas.core.groupby.GroupBy.transform` and :meth:`~pandas.core.groupby.GroupBy.aggregate` has gained ``engine`` and ``engine_kwargs`` arguments that supports executing functions with ``Numba`` (:issue:`32854`, :issue:`33388`)
 -
 
 .. ---------------------------------------------------------------------------

pandas/core/groupby/generic.py

@@ -79,6 +79,7 @@
     NUMBA_FUNC_CACHE,
     check_kwargs_and_nopython,
     get_jit_arguments,
+    is_numba_util_related_error,
     jit_user_function,
     split_for_numba,
     validate_udf,
@@ -244,7 +245,9 @@ def apply(self, func, *args, **kwargs):
         axis="",
     )
     @Appender(_shared_docs["aggregate"])
-    def aggregate(self, func=None, *args, **kwargs):
+    def aggregate(
+        self, func=None, *args, engine="cython", engine_kwargs=None, **kwargs
+    ):
 
         relabeling = func is None
         columns = None
@@ -272,11 +275,18 @@ def aggregate(self, func=None, *args, **kwargs):
                 return getattr(self, cyfunc)()
 
             if self.grouper.nkeys > 1:
-                return self._python_agg_general(func, *args, **kwargs)
+                return self._python_agg_general(
+                    func, *args, engine=engine, engine_kwargs=engine_kwargs, **kwargs
+                )
 
             try:
-                return self._python_agg_general(func, *args, **kwargs)
-            except (ValueError, KeyError):
+                return self._python_agg_general(
+                    func, *args, engine=engine, engine_kwargs=engine_kwargs, **kwargs
+                )
+            except (ValueError, KeyError) as err:
+                # Do not catch Numba errors here, we want to raise and not fall back.
+                if is_numba_util_related_error(str(err)):
+                    raise err
                 # TODO: KeyError is raised in _python_agg_general,
                 #  see see test_groupby.test_basic
                 result = self._aggregate_named(func, *args, **kwargs)
@@ -941,7 +951,9 @@ class DataFrameGroupBy(GroupBy[DataFrame]):
         axis="",
     )
     @Appender(_shared_docs["aggregate"])
-    def aggregate(self, func=None, *args, **kwargs):
+    def aggregate(
+        self, func=None, *args, engine="cython", engine_kwargs=None, **kwargs
+    ):
 
         relabeling = func is None and is_multi_agg_with_relabel(**kwargs)
         if relabeling:
@@ -962,6 +974,11 @@ def aggregate(self, func=None, *args, **kwargs):
 
         func = maybe_mangle_lambdas(func)
 
+        if engine == "numba":
+            return self._python_agg_general(
+                func, *args, engine=engine, engine_kwargs=engine_kwargs, **kwargs
+            )
+
         result, how = self._aggregate(func, *args, **kwargs)
         if how is None:
             return result

pandas/core/groupby/groupby.py

@@ -910,9 +910,12 @@ def _cython_agg_general(
 
         return self._wrap_aggregated_output(output)
 
-    def _python_agg_general(self, func, *args, **kwargs):
+    def _python_agg_general(
+        self, func, *args, engine="cython", engine_kwargs=None, **kwargs
+    ):
         func = self._is_builtin_func(func)
-        f = lambda x: func(x, *args, **kwargs)
+        if engine != "numba":
+            f = lambda x: func(x, *args, **kwargs)
 
         # iterate through "columns" ex exclusions to populate output dict
         output: Dict[base.OutputKey, np.ndarray] = {}
@@ -923,11 +926,21 @@ def _python_agg_general(self, func, *args, **kwargs):
                 # agg_series below assumes ngroups > 0
                 continue
 
-            try:
-                # if this function is invalid for this dtype, we will ignore it.
-                result, counts = self.grouper.agg_series(obj, f)
-            except TypeError:
-                continue
+            if engine == "numba":
+                result, counts = self.grouper.agg_series(
+                    obj,
+                    func,
+                    *args,
+                    engine=engine,
+                    engine_kwargs=engine_kwargs,
+                    **kwargs,
+                )
+            else:
+                try:
+                    # if this function is invalid for this dtype, we will ignore it.
+                    result, counts = self.grouper.agg_series(obj, f)
+                except TypeError:
+                    continue
 
             assert result is not None
             key = base.OutputKey(label=name, position=idx)

pandas/core/groupby/ops.py

@@ -54,6 +54,14 @@
     get_group_index_sorter,
     get_indexer_dict,
 )
+from pandas.core.util.numba_ import (
+    NUMBA_FUNC_CACHE,
+    check_kwargs_and_nopython,
+    get_jit_arguments,
+    jit_user_function,
+    split_for_numba,
+    validate_udf,
+)
 
 
 class BaseGrouper:
@@ -608,10 +616,16 @@ def _transform(
 
         return result
 
-    def agg_series(self, obj: Series, func):
+    def agg_series(
+        self, obj: Series, func, *args, engine="cython", engine_kwargs=None, **kwargs
+    ):
         # Caller is responsible for checking ngroups != 0
         assert self.ngroups != 0
 
+        if engine == "numba":
+            return self._aggregate_series_pure_python(
+                obj, func, *args, engine=engine, engine_kwargs=engine_kwargs, **kwargs
+            )
         if len(obj) == 0:
             # SeriesGrouper would raise if we were to call _aggregate_series_fast
             return self._aggregate_series_pure_python(obj, func)
@@ -656,7 +670,18 @@ def _aggregate_series_fast(self, obj: Series, func):
         result, counts = grouper.get_result()
         return result, counts
 
-    def _aggregate_series_pure_python(self, obj: Series, func):
+    def _aggregate_series_pure_python(
+        self, obj: Series, func, *args, engine="cython", engine_kwargs=None, **kwargs
+    ):
+
+        if engine == "numba":
+            nopython, nogil, parallel = get_jit_arguments(engine_kwargs)
+            check_kwargs_and_nopython(kwargs, nopython)
+            validate_udf(func)
+            cache_key = (func, "groupby_agg")
+            numba_func = NUMBA_FUNC_CACHE.get(
+                cache_key, jit_user_function(func, nopython, nogil, parallel)
+            )
 
         group_index, _, ngroups = self.group_info
 
@@ -666,7 +691,14 @@ def _aggregate_series_pure_python(self, obj: Series, func):
         splitter = get_splitter(obj, group_index, ngroups, axis=0)
 
         for label, group in splitter:
-            res = func(group)
+            if engine == "numba":
+                values, index = split_for_numba(group)
+                res = numba_func(values, index, *args)
+                if cache_key not in NUMBA_FUNC_CACHE:
+                    NUMBA_FUNC_CACHE[cache_key] = numba_func
+            else:
+                res = func(group, *args, **kwargs)
+
             if result is None:
                 if isinstance(res, (Series, Index, np.ndarray)):
                     if len(res) == 1:
@@ -842,7 +874,9 @@ def groupings(self) -> "List[grouper.Grouping]":
             for lvl, name in zip(self.levels, self.names)
         ]
 
-    def agg_series(self, obj: Series, func):
+    def agg_series(
+        self, obj: Series, func, *args, engine="cython", engine_kwargs=None, **kwargs
+    ):
         # Caller is responsible for checking ngroups != 0
         assert self.ngroups != 0
         assert len(self.bins) > 0  # otherwise we'd get IndexError in get_result

pandas/core/util/numba_.py

@@ -12,6 +12,25 @@
 NUMBA_FUNC_CACHE: Dict[Tuple[Callable, str], Callable] = dict()
 
 
+def is_numba_util_related_error(err_message: str) -> bool:
+    """
+    Check if an error was raised from one of the numba utility functions
+
+    For cases where a try/except block has mistakenly caught the error
+    and we want to re-raise
+
+    Parameters
+    ----------
+    err_message : str,
+        exception error message
+
+    Returns
+    -------
+    bool
+    """
+    return "The first" in err_message or "numba does not" in err_message
+
+
 def check_kwargs_and_nopython(
     kwargs: Optional[Dict] = None, nopython: Optional[bool] = None
 ) -> None:
@@ -76,7 +95,6 @@ def jit_user_function(
     ----------
     func : function
         user defined function
-
     nopython : bool
         nopython parameter for numba.JIT
     nogil : bool