🐛 Bug

Compiling against the C++ API on macOS using GCC-9.3, and cmake seems to use a bad flag:
... -fopenmp -D_GLIBCXX_USE_CXX11_ABI= -std=c++14 ... -- note how it "blanks out" the _GLIBCXX_USE_CXX11_ABI variable. This causes the compiler to fail in the stdlib:

/usr/local/Cellar/gcc@9/9.3.0/include/c++/9.3.0/x86_64-apple-darwin18/bits/c++config.h:273:27: error: #if with no expr

The main __init__.pyi file currently contains a module-level __getattr__() function and
,while quite convenient due to the incomplete nature of the stubs, its use is rather unsafe:
any getattr(numpy, str(...)) operation is currently considered valid by static type checkers.

For this reason a set of placeholder stubs was recently added to generic/ndarray (https://github.com/numpy/

When merging a dask dataframe, the resulting index is duplicated - seems to be because of the number of partitions. See example below:

import pandas as pd
import dask.dataframe as dd

a = dd.from_pandas(pd.DataFrame({'a': [1,2,3,4]}), npartitions=2)
b = pd.DataFrame({'a': [1,2,3,4], 'b': [2,3,4,5]})

a.merge(b, on='a').compute()

Returns

|

• I have tried using the latest released version of Numba
• I have included below a minimal working reproducer

In trying to replace numpy's setxor1d (which is unfortunately not supported by Numba), I came across an error that does occur for (integer) ndarrays but not for lists.
In my case the setxor1d can be replaced by [i for i in a if i not in b], but for the sake of testing, here i

Well, Gumbel Distribution is magical. Basically, given a sequence of K logits, i.e., "\log a_1, \log a_2, ..., \log a_K" and K independent gumbel random variables, i.e., "g_1, g_2, ..., g_K". We have

\argmax_i \log a_i + g_i ~ Categorical({a_i / sum(a)})

This gives you a very simple way to sampl

Support Series.median()

What happened:
xr.DataArray([1], coords=[('onecoord', [2])]).sel(onecoord=2).to_dataframe(name='name') raise an exception ValueError: no valid index for a 0-dimensional object

What you expected to happen:

the same behavior as: xr.DataArray([1], coords=[('onecoord', [2])]).to_dataframe(name='name')

Anything else we need to know?:

I see that the array after the select