8000 ENH: random: Add the method `permuted` to Generator. by WarrenWeckesser · Pull Request #15121 · numpy/numpy · GitHub
[go: up one dir, main page]
Skip to content

ENH: random: Add the method permuted to Generator. #15121

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 14 commits into from
Sep 2, 2020
Merged

ENH: random: Add the method permuted to Generator. #15121

Changes from 1 commit
Commits
Show all changes
14 commits
Select commit Hold shift + click to select a range
a84cdf0
ENH: random: Make _shuffle_raw and _shuffle_int standalone functions.
WarrenWeckesser Dec 15, 2019
081298c
ENH: random: Add the method `permuted` to Generator.
WarrenWeckesser Dec 15, 2019
7fc1136
Use PyMem_Malloc/_Free instead of malloc/free
WarrenWeckesser Aug 13, 2020
bc49d4b
Raise TypeError instead ValueError when out is not an ndarray.
WarrenWeckesser Aug 13, 2020
3dbcbae
Use order='K' when copying x to out.
WarrenWeckesser Aug 13, 2020
50c7611
Require out.shape be the same as x.shape
WarrenWeckesser Aug 13, 2020
023114f
Handle axis=None more efficiently.
WarrenWeckesser Aug 14, 2020
91d208a
DOC: Overhaul section on shuffling methods.
WarrenWeckesser Aug 20, 2020
95eff00
Use np.copyto instead of out[...] = x
WarrenWeckesser Aug 20, 2020
677f862
Don't release the GIL when shuffling an object array.
WarrenWeckesser Aug 20, 2020
f5d6f0b
DOC: Fix examples and move method summary to the top.
WarrenWeckesser Aug 20, 2020
f016d3a
Fix C prototype declaration to fix compiler warning.
WarrenWeckesser Aug 20, 2020
ebed0e8
Refine the C prototype to eliminate another compiler warning.
WarrenWeckesser Aug 20, 2020
785e239
Remove cimport of unused malloc and free.
WarrenWeckesser Aug 21, 2020
8000
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
Next Next commit
Handle axis=None more efficiently.
  • Loading branch information
@WarrenWeckesser
WarrenWeckesser committed Aug 14, 2020
commit 023114fa656df68de15470261649bae695bb96b8
33 changes: 31 additions & 2 deletions numpy/random/_generator.pyx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,14 @@ from ._common cimport (POISSON_LAM_MAX, CONS_POSITIVE, CONS_NONE,
validate_output_shape
)

cdef extern from "numpy/arrayobject.h":
#int PyArray_SetWritebackIfCopyBase(np.ndarray, np.ndarray)
int PyArray_ResolveWritebackIfCopy(np.ndarray)
Copy link
Member

Choose a reason for hiding this comment

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

Suggested change
int PyArray_ResolveWritebackIfCopy(np.ndarray)
int PyArray_ResolveWritebackIfCopy(np.ndarray) except -1

object PyArray_FromArray(object, object, int)

enum:
NPY_ARRAY_WRITEBACKIFCOPY

np.import_array()

cdef int64_t _safe_sum_nonneg_int64(size_t num_colors, int64_t *colors):
Expand Down Expand Up @@ -4184,7 +4192,7 @@ cdef class Generator:

Parameters
----------
x : array_like
x : array_like, at least one-dimensional
Array to be shuffled.
axis : int, optional
Slices of `x` in this axis are shuffled. Each slice
Expand Down Expand Up @@ -4255,6 +4263,9 @@ cdef class Generator:
cdef np.npy_intp axlen, axstride, itemsize
cdef void *buf
cdef np.flatiter it
cdef np.ndarray to_shuffle
cdef int status
cdef int flags

x = np.asarray(x)

Expand All @@ -4268,7 +4279,25 @@ cdef class Generator:
out[...] = x
Copy link
Member

Choose a reason for hiding this comment

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

IIRC this uses force-casting, maybe we should use np.copyto(out, x, casting="safe"). This also supports a few types of broadcasting of input to output. That is fine as such, but I feel it may be confusing with respect to the axis argument if the number of dimensions do not match. Maybe we should just enforce identical shapes?


if axis is None:
self.shuffle(out.flat)
if x.ndim > 1:
if not (np.PyArray_FLAGS(out) & (np.NPY_ARRAY_C_CONTIGUOUS |
np.NPY_ARRAY_F_CONTIGUOUS)):
flags = (np.NPY_ARRAY_C_CONTIGUOUS |
NPY_ARRAY_WRITEBACKIFCOPY)
to_shuffle = PyArray_FromArray(out, <object>NULL, flags)
self.shuffle(to_shuffle.ravel(order='K'))
# Because we only execute this block if out is not
# contiguous, we know this call will always result in a
# copy of to_shuffle back to out. I.e. status will be 1.
status = PyArray_ResolveWritebackIfCopy(to_shuffle)
assert status == 1
else:
# out is n-d with n > 1, but is either C- or F-contiguous,
# so we know out.ravel(order='A') is a view.
self.shuffle(out.ravel(order='A'))
else:
# out is 1-d
self.shuffle(out)
return out

ax = normalize_axis_index(axis, np.ndim(out))
Copy link
Member

Choose a reason for hiding this comment

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

Suggested change
ax = normalize_axis_index(axis, np.ndim(out))
ax = normalize_axis_index(axis, out.ndim)

Expand Down
0