Skip to content

BUG: fix isin with nans and large arrays #36266

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 7 commits into from
Sep 19, 2020
Merged

BUG: fix isin with nans and large arrays #36266

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
656b6b4
fix isin with nans and large arrays
Hanspagh Sep 10, 2020
246cab5
use .any() instead of any() + whatsnew entry
Hanspagh Sep 11, 2020
25d48c0
test series.isin
Hanspagh Sep 14, 2020
859cbf6
update whats new
Hanspagh Sep 14, 2020
3679c14
docs
Hanspagh Sep 16, 2020
4e4359b
Update pandas/tests/test_algos.py
Hanspagh Sep 17, 2020
53ab240
Update pandas/tests/test_algos.py
Hanspagh Sep 17, 2020
File filter

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 6 additions & 1 deletion pandas/core/algorithms.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -440,7 +440,12 @@ def isin(comps: AnyArrayLike, values: AnyArrayLike) -> np.ndarray:
# GH16012
# Ensure np.in1d doesn't get object types or it *may* throw an exception
if len(comps) > 1_000_000 and not is_object_dtype(comps):
Copy link
Contributor

@realead realead Sep 21, 2020

Choose a reason for hiding this comment

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

Sorry a little bit too late. The whole point of doing this for len(comps) > 1_000_000, was that numpy was deemed to be faster (which is probably no loner the case btw, see #22205 (comment)), adding any, isnan, logical_or on top (with all the cache misses and temporary objects) will make this branch much slower. So probably it is best just to drop the whole branch and always keep f = htable.ismember_object (unless it is is_integer_dtype of cause).
Copy link
Contributor

@jreback jreback Sep 21, 2020

Choose a reason for hiding this comment

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

can u run the asvs and check here?
Copy link
Contributor

@realead realead Sep 24, 2020

Choose a reason for hiding this comment

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

@jreback I have opened RP #36611 with my suggestion and some benchmarks, which show that numpy's in1d is only faster when here are very few unique values.
f = np.in1d
# If the the values include nan we need to check for nan explicitly
# since np.nan it not equal to np.nan
if any(np.isnan(values)):
f = lambda c, v: np.logical_or(np.in1d(c, v), np.isnan(c))
else:
f = np.in1d
elif is_integer_dtype(comps):
try:
values = values.astype("int64", copy=False)
Expand Down
8 changes: 7 additions & 1 deletion pandas/tests/test_algos.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -801,7 +801,6 @@ def test_i8(self):
tm.assert_numpy_array_equal(result, expected)

def test_large(self):

s = pd.date_range("20000101", periods=2000000, freq="s").values
result = algos.isin(s, s[0:2])
expected = np.zeros(len(s), dtype=bool)
Expand Down Expand Up @@ -841,6 +840,13 @@ def test_same_nan_is_in(self):
result = algos.isin(comps, values)
tm.assert_numpy_array_equal(expected, result)

def test_same_nan_is_in_large(self):
s = np.tile(1.0, 1_000_001)
s[0] = np.nan
result = algos.isin(s, [np.nan, 1])
expected = np.ones(len(s), dtype=bool)
tm.assert_numpy_array_equal(result, expected)

def test_same_object_is_in(self):
# GH 22160
# there could be special treatment for nans
Expand Down