Skip to content

REF: Make numba function cache globally accessible #33621

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Apr 19, 2020
Merged

REF: Make numba function cache globally accessible #33621

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
11c2432
REF: Make numba function cache globally accessible
Apr 18, 2020
a475102
dont make variable private
Apr 18, 2020
e319ac6
Explicity pass the function to construct the cache key
Apr 18, 2020
5db6c6c
pop original func for rolling groupby
Apr 18, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 11 additions & 12 deletions pandas/core/groupby/generic.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -76,6 +76,7 @@
from pandas.core.internals import BlockManager, make_block
from pandas.core.series import Series
from pandas.core.util.numba_ import (
_numba_func_cache,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If it’s cross-module, can it be non-private?

check_kwargs_and_nopython,
get_jit_arguments,
jit_user_function,
Expand Down Expand Up @@ -161,8 +162,6 @@ def pinner(cls):
class SeriesGroupBy(GroupBy[Series]):
_apply_whitelist = base.series_apply_whitelist

_numba_func_cache: Dict[Callable, Callable] = {}

def _iterate_slices(self) -> Iterable[Series]:
yield self._selected_obj

Expand Down Expand Up @@ -504,8 +503,9 @@ def _transform_general(
nopython, nogil, parallel = get_jit_arguments(engine_kwargs)
check_kwargs_and_nopython(kwargs, nopython)
validate_udf(func)
numba_func = self._numba_func_cache.get(
func, jit_user_function(func, nopython, nogil, parallel)
cache_key = (func, "groupby_transform")
numba_func = _numba_func_cache.get(
cache_key, jit_user_function(func, nopython, nogil, parallel)
)

klass = type(self._selected_obj)
Expand All @@ -516,8 +516,8 @@ def _transform_general(
if engine == "numba":
values, index = split_for_numba(group)
res = numba_func(values, index, *args)
if func not in self._numba_func_cache:
self._numba_func_cache[func] = numba_func
if cache_key not in _numba_func_cache:
_numba_func_cache[cache_key] = numba_func
else:
res = func(group, *args, **kwargs)

Expand Down Expand Up @@ -847,8 +847,6 @@ class DataFrameGroupBy(GroupBy[DataFrame]):

_apply_whitelist = base.dataframe_apply_whitelist

_numba_func_cache: Dict[Callable, Callable] = {}

_agg_see_also_doc = dedent(
"""
See Also
Expand Down Expand Up @@ -1397,8 +1395,9 @@ def _transform_general(
nopython, nogil, parallel = get_jit_arguments(engine_kwargs)
check_kwargs_and_nopython(kwargs, nopython)
validate_udf(func)
numba_func = self._numba_func_cache.get(
func, jit_user_function(func, nopython, nogil, parallel)
cache_key = (func, "groupby_transform")
numba_func = _numba_func_cache.get(
cache_key, jit_user_function(func, nopython, nogil, parallel)
)
else:
fast_path, slow_path = self._define_paths(func, *args, **kwargs)
Expand All @@ -1409,8 +1408,8 @@ def _transform_general(
if engine == "numba":
values, index = split_for_numba(group)
res = numba_func(values, index, *args)
if func not in self._numba_func_cache:
self._numba_func_cache[func] = numba_func
if cache_key not in _numba_func_cache:
_numba_func_cache[cache_key] = numba_func
# Return the result as a DataFrame for concatenation later
res = DataFrame(res, index=group.index, columns=group.columns)
else:
Expand Down
2 changes: 2 additions & 0 deletions pandas/core/util/numba_.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,8 @@
from pandas._typing import FrameOrSeries
from pandas.compat._optional import import_optional_dependency

_numba_func_cache: Dict[Tuple[Callable, str], Callable] = dict()


def check_kwargs_and_nopython(
kwargs: Optional[Dict] = None, nopython: Optional[bool] = None
Expand Down
9 changes: 5 additions & 4 deletions pandas/core/window/rolling.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,6 +38,7 @@
from pandas.core.base import DataError, PandasObject, SelectionMixin, ShallowMixin
import pandas.core.common as com
from pandas.core.indexes.api import Index, ensure_index
from pandas.core.util.numba_ import _numba_func_cache
from pandas.core.window.common import (
WindowGroupByMixin,
_doc_template,
Expand Down Expand Up @@ -93,7 +94,6 @@ def __init__(
self.win_freq = None
self.axis = obj._get_axis_number(axis) if axis is not None else None
self.validate()
self._numba_func_cache: Dict[Optional[str], Callable] = dict()

@property
def _constructor(self):
Expand Down Expand Up @@ -505,7 +505,7 @@ def calc(x):
result = np.asarray(result)

if use_numba_cache:
self._numba_func_cache[name] = func
_numba_func_cache[(name, "rolling_apply")] = func

if center:
result = self._center_window(result, window)
Expand Down Expand Up @@ -1278,9 +1278,10 @@ def apply(
elif engine == "numba":
if raw is False:
raise ValueError("raw must be `True` when using the numba engine")
if func in self._numba_func_cache:
cache_key = (func, "rolling_apply")
if cache_key in _numba_func_cache:
# Return an already compiled version of roll_apply if available
apply_func = self._numba_func_cache[func]
apply_func = _numba_func_cache[cache_key]
else:
apply_func = generate_numba_apply_func(
args, kwargs, func, engine_kwargs
Expand Down
5 changes: 3 additions & 2 deletions pandas/tests/groupby/transform/test_numba.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@

from pandas import DataFrame
import pandas._testing as tm
from pandas.core.util.numba_ import _numba_func_cache


@td.skip_if_no("numba", "0.46.0")
Expand Down Expand Up @@ -98,13 +99,13 @@ def func_2(values, index):
expected = grouped.transform(lambda x: x + 1, engine="cython")
tm.assert_equal(result, expected)
# func_1 should be in the cache now
assert func_1 in grouped._numba_func_cache
assert (func_1, "groupby_transform") in _numba_func_cache

# Add func_2 to the cache
result = grouped.transform(func_2, engine="numba", engine_kwargs=engine_kwargs)
expected = grouped.transform(lambda x: x * 5, engine="cython")
tm.assert_equal(result, expected)
assert func_2 in grouped._numba_func_cache
assert (func_2, "groupby_transform") in _numba_func_cache

# Retest func_1 which should use the cache
result = grouped.transform(func_1, engine="numba", engine_kwargs=engine_kwargs)
Expand Down
3 changes: 2 additions & 1 deletion pandas/tests/window/test_numba.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@

from pandas import Series
import pandas._testing as tm
from pandas.core.util.numba_ import _numba_func_cache


@td.skip_if_no("numba", "0.46.0")
Expand Down Expand Up @@ -59,7 +60,7 @@ def func_2(x):
tm.assert_series_equal(result, expected)

# func_1 should be in the cache now
assert func_1 in roll._numba_func_cache
assert (func_1, "rolling_apply") in _numba_func_cache

result = roll.apply(
func_2, engine="numba", engine_kwargs=engine_kwargs, raw=True
Expand Down