On Mon, Jun 9, 2014 at 6:10 PM, Charles R Harris
On Mon, Jun 9, 2014 at 5:21 PM, Christoph Gohlke
wrote: On 6/8/2014 1:34 PM, Julian Taylor wrote:
Hello,
I'm happy to announce the fist beta release of Numpy 1.9.0. 1.9.0 will be a new feature release supporting Python 2.6 - 2.7 and 3.2 - 3.4. Due to low demand windows binaries for the beta are only available for Python 2.7, 3.3 and 3.4. Please try it and report any issues to the numpy-discussion mailing list or on github.
The 1.9 release will consists of mainly of many small improvements and bugfixes. The highlights are:
* Addition of __numpy_ufunc__ to allow overriding ufuncs in ndarray subclasses. Please note that there are still some known issues with this mechanism which we hope to resolve before the final release (e.g. #4753) * Numerous performance improvements in various areas, most notably indexing and operations on small arrays are significantly faster. Indexing operations now also release the GIL. * Addition of nanmedian and nanpercentile rounds out the nanfunction set.
The changes involve a lot of small changes that might affect some applications, please read the release notes for the full details on all changes: https://github.com/numpy/numpy/blob/maintenance/1.9.x/ doc/release/1.9.0-notes.rst Please also take special note of the future changes section which will apply to the following release 1.10.0 and make sure to check if your applications would be affected by them.
Source tarballs, windows installers and release notes can be found at https://sourceforge.net/projects/numpy/files/NumPy/1.9.0b1
Cheers, Julian Taylor
Hello,
I tested numpy-MKL-1.9.0b1 (msvc9, Intel MKL build) on win-amd64-py2.7 against a few other packages that were built against numpy-MKL-1.8.x.
While numpy and scipy pass all tests, some other packages (matplotlib, statsmodels, skimage, pandas, pytables, sklearn...) show a few new test failures (compared to testing with numpy-MKL-1.8.1). Many test errors are of kind:
ValueError: shape mismatch: value array of shape (24,) could not be broadcast to indexing result of shape (8,3)
I have attached a list of failing tests. The full test results are at < http://www.lfd.uci.edu/~gohlke/pythonlibs/tests/20140609-win-amd64-py2.7- numpy-1.9.0b1/> (compare to <http://www.lfd.uci.edu/~ gohlke/pythonlibs/tests/20140609-win-amd64-py2.7/>)
I have not investigated any further...
One of the matplotlib failures, and I suspect the others, comes from the assignment
found_index[refi_triangles[ancestor_mask, :] ] = np.repeat(ancestors[ancestor_mask], 3)
This fails with the error
ValueError: shape mismatch: value array of shape (13824,)
could not be broadcast to indexing result of shape (4608,3)
I confess I find the construction odd, but the error probably results from stricter indexing rules. Indeed, (13824,) does not broadcast to (4608,3). Apart from considerations of backward compatibility, should it?
Other errors are of the type: TypeError: NumPy boolean array indexing assignment requires a 0 or 1-dimensionalinput, input has 2 dimensions This one looks to arise is from stricter rules for boolean indexing. Chuck