[FEA] Support logical operators such as & and | for cudf.dataframe.series.Series

[paulhendricks]

Is your feature request related to a problem? Please describe.

The ability to use logical operators & and | with cudf.dataframe.series.Series data structures.

For example:

mask_and = mask_1 & mask_2
mask_or = mask_1 | mask_2

The whole reproducible code is below:

import cudf

# Create a test csv file
filename = 'foo.csv'
lines = [
  "num1,datetime,text",
  "123,2018-11-13T12:00:00,abc",
  "456,2018-11-14T12:35:01,def",
  "789,2018-11-15T18:02:59,ghi"
]
with open(filename, 'w') as fp:
    fp.write('\n'.join(lines)+'\n')

# Read the file with cudf using only certain columns (omitting strings)
names = ['num1', 'datetime', 'text']
dtypes = ['int', 'date', 'str']
gdf = cudf.io.csv.read_csv(filename, delimiter=',',
                        names=names, dtype=dtypes,
                        skiprows=1, usecols=['num1', 'datetime'])
print(gdf.head())

mask_1 = gdf['num1'] == 456
print(mask_1)
print(type(mask_1))

specific_date = gdf['datetime'][2]
mask_2 = gdf['datetime'] == specific_date
print(mask_2)
print(type(mask_2))

mask_and = mask_1 & mask_2  # TypeError: unsupported operand type(s) for &: 'Series' and 'Series'
mask_or = mask_1 | mask_2  # TypeError: unsupported operand type(s) for |: 'Series' and 'Series'

import pandas as pd

df = pd.read_csv(filename)

mask_1 = df['num1'] == 456
print(mask_1)
print(type(mask_1))

specific_date = df['datetime'][2]
mask_2 = df['datetime'] == specific_date
print(mask_2)
print(type(mask_2))

mask_and = mask_1 & mask_2  # Works!
mask_or = mask_1 | mask_2  # Works!

Describe the solution you'd like

mask_and = mask_1 & mask_2  # Works!
mask_or = mask_1 | mask_2  # Works!

Describe alternatives you've considered

The following uses NumPy and ought to work:

import numpy as np

mask_and = np.logical_and(mask_1, mask_2)
mask_or = np.logical_or(mask_1, mask_2)

print(gdf[mask_and])
print(gdf[mask_or])

Additional context

Pandas documentation on boolean indexing here: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#boolean-indexing

[mrocklin]

Looks like a copy of #49 . Closing this in order to consolidate things there.