Skip to content

Fix np.clip returning wrong shape when bounds broadcast larger than a #10767

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

Open
VenkateswarluNagineni wants to merge 5 commits into
base: main
Choose a base branch
from
Open

Fix np.clip returning wrong shape when bounds broadcast larger than a #10767

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
0675a65
Fix np.clip returning wrong shape when bounds broadcast larger than a
VenkateswarluNagineni Aug 8, 2026
d1d3aef
Add changelog entry for PR #10767
VenkateswarluNagineni Aug 8, 2026
e48ce6b
Fix RST title underline in changelog entry for #10767
VenkateswarluNagineni Sep 11, 2026
ffd7f75
Fix np.clip layout regression: preserve F-contiguous output
VenkateswarluNagineni Sep 12, 2026
8d5ea14
Address np.clip PR review feedback
VenkateswarluNagineni Sep 23, 2026
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
Prev Previous commit
Address np.clip PR review feedback
  • Loading branch information
@VenkateswarluNagineni
VenkateswarluNagineni committed Sep 23, 2026
commit 8d5ea140c5ceb56a60bd6b6e971c2aed6dde396e
19 changes: 5 additions & 14 deletions docs/upcoming_changes/10767.bug_fix.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,15 +1,6 @@
Fix ``np.clip`` shape and layout preservation
----------------------------------------------
Fix ``np.clip`` result shape for broadcast bounds
-------------------------------------------------

Two bugs are fixed:

1. **Wrong shape when bounds broadcast larger than** ``a``: the result was
previously allocated with ``np.empty_like(a)`` before the broadcast shape
was computed, so a bound that broadcasts to a shape larger than ``a``
silently produced the wrong shape and wrote outside the result buffer.

2. **F-contiguous layout not preserved**: switching from ``np.empty_like(a)``
to ``np.empty(shape, a.dtype)`` always produced a C-contiguous result even
when ``a`` was F-contiguous. Both issues are now fixed by dispatching the
allocation at the type level so that F-contiguous inputs produce
F-contiguous outputs.
``np.clip`` now returns the correct result shape when ``a_min`` or ``a_max``
broadcasts to a shape larger than ``a``. The output allocation preserves
NumPy-compatible layout behavior for F-contiguous inputs.
7 changes: 4 additions & 3 deletions numba/tests/test_array_methods.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2041,9 +2041,10 @@ def test_clip_fortran_order(self):
"one-None-bound clip lost F-contiguous layout")

# array bounds that broadcast to a larger shape (np_clip_impl path)
big = np.zeros((3, 4))
result = cfunc(a, big, 10.0)
expected = np.clip(a, big, 10.0)
narrow = np.asfortranarray(np.arange(6.0).reshape(2, 3, 1))
big = np.asfortranarray(np.zeros((2, 3, 4)))
result = cfunc(narrow, big, 10.0)
expected = np.clip(narrow, big, 10.0)
np.testing.assert_array_equal(result, expected)
self.assertTrue(result.flags.f_contiguous,
"broadcast clip lost F-contiguous layout")
Expand Down