| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |\
| |
| | |
ENH: Do not emit compiler warning if forcing old API
|
| | | |
|
| |\ \
| | |
| | | |
BUG: Fix regression on np.dtype(ctypes.c_void_p)
|
| | | |
| | |
| | |
| | |
| | |
| | | |
Historically, this always returned np.uintp - we should continue to do so, even though other pointer types have never been supported
Fixes gh-12416
|
| | | | |
|
| |\ \ \
| | | |
| | | | |
MAINT: Update cversion hash.
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
A new function was added to the ufunc_funcs_api.
Closes #12412.
|
| |\ \ \ \
| |_|/ /
|/| | | |
BUG: Do not double-quote arguments to the command line
|
| |/ / /
| | |
| | |
| | |
| | |
| | | |
After the recent patch to CCompiler.spawn, the file-paths no longer need manual quoting - that's handled as needed within subprocess.
Fixes #12411
|
| |/ /
| |
| |
| | |
Fixes #12273
|
| |\ \
| | |
| | | |
BUG: Fall back to 'ascii' locale in build (if needed)
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | | |
If locale.getpreferredencoding(False) returns Null,
the build fails, since commit 5652b0785.
This seems to be not really necessary; here we
provide a fallback to 'ascii' locale if needed.
|
| |\ \ \
| |/ /
|/| | |
MAINT: correct the dtype.descr docstring
|
| | | |
| | |
| | |
| | | |
[ci skip]
|
| | | | |
|
| |\ \ \
| | | |
| | | | |
DOC: Fix typo in polyint. Fixes #12386.
|
| | | | |
| | | |
| | | |
| | | |
| | | | |
Change differentiate to integrate in parameter `p` description.
Fixes #12386.
|
| |\ \ \ \
| | | | |
| | | | | |
BUG: fix possible overlap issues with avx enabled
|
| | |/ / /
| | | |
| | | |
| | | |
| | | | |
Fix the overlap check for vectors larger than 16 bytes. Only an issue
when avx is enabled at compile time.
|
| |\ \ \ \
| |/ / /
|/| | | |
DEP: raise on a call to deprecated numpy.lib.function_base.unique
|
| | | | | |
|
| |\ \ \ \
| | | | |
| | | | | |
BUG: graceful DataSource __del__ when __init__ fails
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* DataSource __del__ could raise an AttributeError
if __init__ failed for any reason; __del__ now gracefully
handles the case where __init__ fails, as happens in the
internals of refguide_check when handling DataSource with
an unexpected kwarg
|
| |\ \ \ \ \
| | | | | |
| | | | | | |
DEV: remove shim added in 1.4
|
| | | |/ / /
| |/| | | |
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* BUG: Fix for #10533 np.dtype(ctype) does not respect endianness
See: #10533
Added some code that uses the __name__ of the ctypes type. Also
added some new test to make sure we'll know if the ctypes data
that allows this workaround breaks in the future.
* Fix for #10533 np.dtype(ctype) does not respect endianness
Removed the TODO for this bug.
* Fix for #10533 np.dtype(ctype) does not respect endianness
Solved the recursiveness issue that failed the automated test.
Also did some of Eric's requested changes.
* Small mistake in the new function fixed.
* Changed 'if t.__name__.endswith('_be') ' with
'if t.__ctypes_be__ is t' on Eric's suggestion.
* Code cleanup on Eric's recommendations.
* Cleaned up the tests on Eric's suggestions.
|
| |\ \ \ \ \
| | | | | |
| | | | | | |
DOC: Add release notes for ctypes improvements
|
| |/ / / / /
| | | | |
| | | | |
| | | | | |
Fixes gh-12272
|
| |\ \ \ \ \
| |/ / / /
|/| | | | |
MAINT: remove exec_command usage from ccompiler.py
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* replaced the usage of exec_command() in
ccompiler distutils module with subprocess
equivalent
|
| |\ \ \ \ \
| | | | | |
| | | | | | |
ENH: set correct __module__ for objects in numpy's public API
|
| | | | | | | |
|
| | | | | | | |
|
| | | | | | | |
|
| | | | | | | |
|
| | | | | | | |
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Fixes GH-12271
Tests verify that everything in ``dir(numpy)`` either has ``__module__`` set to
``'numpy'``, or appears in an explicit whitelist of undocumented functions and
exported bulitins. These should eventually be documented or removed.
I also identified a handful of functions for which I had accidentally not setup
dispatch for with ``__array_function__`` before, because they were listed under
"ndarray methods" in ``_add_newdocs.py``. I guess that should be a lesson in
trusting code comments :).
|
| |\ \ \ \ \ \
| | | | | | |
| | | | | | | |
BUG: test, fix loading structured dtypes with padding
|
| | | | | | | | |
|
| | | | | | | | |
|
| | |/ / / / / |
|
| |\ \ \ \ \ \
| |_|_|_|_|/
|/| | | | | |
BUG: do not override exception on import failure
|
| | | | | | | |
|
| |\ \ \ \ \ \
| |_|_|_|/ /
|/| | | | | |
TST: Update complex long double precision tests.
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
It seems that special casing the long double precision tests in order to
deal with PPC is no longer needed now that finfo has been fixed for the
IBM double double and the blacklisting of failing functions.
Consequently the tests can be simplified and no longer needed functions
removed. See #8566 for discussion.
|
| |\ \ \ \ \ \
| | | | | | |
| | | | | | | |
MAINT: move ufunc override code to umath and multiarray as much as possible.
|
| | | | | | | | |
|
| | | | | | | | |
|
| |\ \ \ \ \ \ \
| | | | | | | |
| | | | | | | | |
DOC: add a docstring for the function 'compare_chararrays' (See #10106)
|
| | | | | | | | |
| | | | | | | |
| | | | | | | | |
[ci skip]
|
