plydata.helper_verbs.group_by_at

class plydata.helper_verbs.group_by_at(*args, **kwargs)[source]

Group by select columns

Parameters
datadataframe, optional

Useful when not using the >> operator.

namestuple or dict

Names of columns in dataframe. If a tuple, they should be names of columns. If a dict, they keys must be in.

  • startswithstr or tuple, optional

    All column names that start with this string will be included.

  • endswithstr or tuple, optional

    All column names that end with this string will be included.

  • containsstr or tuple, optional

    All column names that contain with this string will be included.

  • matchesstr or regex or tuple, optional

    All column names that match the string or a compiled regex pattern will be included. A tuple can be used to match multiple regexs.

  • dropbool, optional

    If True, the selection is inverted. The unspecified/unmatched columns are returned instead. Default is False.

functionscallable() or tuple or dict or str, optional

Functions to alter the columns:

  • function (any callable) - Function is applied to the column and the result columns replace the original columns.

  • tuple of functions - Each function is applied to all of the columns and the name (__name__) of the function is postfixed to resulting column names.

  • dict of the form {'name': function} - Allows you to apply one or more functions and also control the postfix to the name.

  • str - String can be used for more complex statements, but the resulting names will be terrible.

argstuple

Arguments to the functions. The arguments are pass to all functions.

kwargsdict

Keyword arguments to the functions. The keyword arguments are passed to all functions.

Examples

>>> import pandas as pd
>>> import numpy as np
>>> from plydata import *
>>> df = pd.DataFrame({
...     'alpha': list('aaabbb'),
...     'beta': list('babruq'),
...     'theta': list('cdecde'),
...     'x': [1, 2, 3, 4, 5, 6],
...     'y': [6, 5, 4, 3, 2, 1],
...     'z': [7, 9, 11, 8, 10, 12]
... })

In the simplest form it is not too different from group_by.

>>> df >> group_by_at(('x', 'y'))
groups: ['x', 'y']
  alpha beta theta  x  y   z
0     a    b     c  1  6   7
1     a    a     d  2  5   9
2     a    b     e  3  4  11
3     b    r     c  4  3   8
4     b    u     d  5  2  10
5     b    q     e  6  1  12

The power comes from the ability to do dynamic column selection. For example, regex match column names and apply function to get the group columns.

>>> def double(s): return s + s
>>> df >> group_by_at(dict(matches=r'\w+eta$'), double)
groups: ['beta', 'theta']
  alpha beta theta  x  y   z
0     a   bb    cc  1  6   7
1     a   aa    dd  2  5   9
2     a   bb    ee  3  4  11
3     b   rr    cc  4  3   8
4     b   uu    dd  5  2  10
5     b   qq    ee  6  1  12