delta/python-packages/numpy.git/numpy/core/src/umath, branch v1.18.4 github.com: numpy/numpy.git BUG: core: use blas_ilp64 also for *_matmul, *_dot, and *_vdot 2019-12-15T20:43:54+00:00 Pauli Virtanen pav@iki.fi 2019-12-09T18:47:28+00:00 9466d31bd846e47f14ba0f22087cd7c420589a1e Changing these to support ILP64 blas was missed in gh-15012 
Changing these to support ILP64 blas was missed in gh-15012
Merge pull request #15036 from eric-wieser/fix-leak 2019-12-03T12:39:44+00:00 Matti Picus matti.picus@gmail.com 2019-12-03T12:39:44+00:00 e39aa70e47313e33afd80e3080b135144ed32be8 BUG: Fix refcounting in ufunc object loops 
BUG: Fix refcounting in ufunc object loops
 BUG: Fix refcounting in ufunc object loops 2019-12-03T11:42:17+00:00 Eric Wieser wieser.eric@gmail.com 2019-12-03T10:28:54+00:00 de6e644b1853f595683cfbedb3745b24c75ce63d Previously `func` was leaked on both success and failure paths. 
Previously `func` was leaked on both success and failure paths.
ENH: add isinf, isnan, fmin, fmax loops for datetime64, timedelta64 (#14841) 2019-11-22T00:07:38+00:00 Matti Picus matti.picus@gmail.com 2019-11-22T00:07:38+00:00 5923592155fa9301dfbdcc2b2b92972ea796dcc4 `np.datetime('NaT')` should behave more like `float('Nan')`. Add needed infrastructure so `np.isinf(a)` and `np.isnan(a)` will run on `datetime64` and `timedelta64` dtypes. Also added specific loops for `numpy.fmin` and `numpy.fmax` that mask `NaT`. 
`np.datetime('NaT')` should behave more like `float('Nan')`. 
Add needed infrastructure so `np.isinf(a)` and `np.isnan(a)`
will run on `datetime64` and `timedelta64` dtypes.
Also added specific loops for `numpy.fmin` and `numpy.fmax`
that mask `NaT`.
MAINT: Use templating to merge float loops 2019-11-01T00:43:03+00:00 Eric Wieser wieser.eric@gmail.com 2019-10-31T23:45:05+00:00 81f857d1010f3cf9701e44a2e937680473ca5f5e This deduplicates a bunch of code, making this file shorter and slightly more manageable. Note that this requires us to adjust the api parser to process templated files. 
This deduplicates a bunch of code, making this file shorter and slightly more manageable.

Note that this requires us to adjust the api parser to process templated files.
Merge pull request #14738 from eric-wieser/fix-intp-printf 2019-10-18T09:09:50+00:00 Matti Picus matti.picus@gmail.com 2019-10-18T09:09:50+00:00 d8a7c44b33221bb7b6fd926c93a9bca38ca45f69 BUG: Don't narrow intp to int when producing error messages 
BUG: Don't narrow intp to int when producing error messages
 Fix other incorrect narrowing for format strings 2019-10-17T21:58:58+00:00 Eric Wieser wieser.eric@gmail.com 2019-10-17T21:58:58+00:00 2d246b9fd647757543290537824438c979e6167e Note that zd is for Py_ssize_t, and `NPY_INTP_FMT` is for npy_intp. In all conceivable situations these are the same type, but we don't today check that at compile time, so for now we maintain the distinction. Found by searching for `Format\([^;]+\(int\)`. 
Note that zd is for Py_ssize_t, and `NPY_INTP_FMT` is for npy_intp. In all conceivable situations these are the same type, but we don't today check that at compile time, so for now we maintain the distinction.

Found by searching for `Format\([^;]+\(int\)`.
MAINT: umath: Change error message for unsupported bool subtraction. 2019-10-16T08:01:24+00:00 Warren Weckesser warren.weckesser@gmail.com 2019-10-16T07:59:24+00:00 d2aeb397c192c3b1a3efee8e031a1af6e95da18d The error message that results when attempting to subtract bool arrays is changed to say that the operation is "not supported", rather than "is deprecated". This is now consistent with a similar error message from the negative unary operator. 
The error message that results when attempting to subtract bool arrays
is changed to say that the operation is "not supported", rather than
"is deprecated".  This is now consistent with a similar error message
from the negative unary operator.
Merge pull request #14681 from seberg/remove-unused-boolean-loops 2019-10-16T07:46:59+00:00 Warren Weckesser warren.weckesser@gmail.com 2019-10-16T07:46:59+00:00 27c143c15da262fe2d233c0c310a184af21a8ff3 MAINT: Remove unused boolean negative/subtract loops 
MAINT: Remove unused boolean negative/subtract loops
 API: NaT (arg)min/max behavior (#14717) 2019-10-16T00:28:27+00:00 Zijie (ZJ) Poh 8103276+zjpoh@users.noreply.github.com 2019-10-16T00:28:27+00:00 67fc36206b7193fcac0d51568b94350165a038a9 As discussed, we want argmin/argmax/min/max to return NaT when it exists in the array. I modified some previous test cases that ignores NaT. I also combined the DATETIME_argmin function with more general argmin function. So that it is consistent with argmax. Closes gh-12797. 
As discussed, we want argmin/argmax/min/max to return NaT when it exists in the array.

I modified some previous test cases that ignores NaT. I also combined the DATETIME_argmin function with more general argmin function. So that it is consistent with argmax.

Closes gh-12797.