shoyer

from numba import guvectorize, jit
import numpy as np
import pandas as pd

@guvectorize(['void(float64[:], int64[:], float64[:], float64[:])'],
             '(x),(x),(y)->()')
def _grouped_sum_guvec_simple(values, labels, target, out):
    for i in range(len(values)):
        idx = labels[i]
        target[idx] += values[i]

shoyer

import inspect
import types

    
def injected(df, thunk):
    """Evaluate a thunk in the context of DataFrame
    
    >>> df = pd.DataFrame({'x': [0, 1, 2]}, index=['a', 'b', 'c'])
    >>> injected(df, lambda: x ** 2)
    a    0

shoyer

import functools

def nd_groupby_apply(ds, dims, func):
    if isinstance(dims, str):
        dims = [dims]
    if len(dims) > 1:
        func = functools.partial(nd_groupby_apply, dims=dims[1:], func=func) 
    return ds.groupby(dims[0]).apply(func)

def nd_group_over_apply(ds, dims, func):

shoyer

shoyer

         4926489 function calls (4835695 primitive calls) in 11.314 seconds

   Ordered by: internal time

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
    21928    1.839    0.000    4.441    0.000 slicing.py:202(slice_slices_and_integers)
        4    1.521    0.380    1.610    0.402 {sum}
   109869    1.494    0.000    1.494    0.000 {method 'update' of 'dict' objects}
   131496    1.173    0.000    1.333    0.000 slicing.py:241(_slice_1d)
    65748    0.624    0.000    1.551    0.000 slicing.py:544(new_blockdim)

shoyer

shoyer

@jit(nopython=True)
def g0(x, y):
    s = 0
    for i in range(len(x)):
        s += x[i] * y[i]
    return s

@jit(nopython=True)
def g1(x, y):

shoyer

shoyer

shoyer