building a class from an existing one

Question

I am trying to build a class from pandas DataFrame. I just want to add an attribute 'name' to DataFrame class. But the codes below yield errors in recursion maximum depth reached. Which way to make it work? Thanks

import pandas as pd
class DatFrame(pd.DataFrame):
    def __init__(self, name, data=None, index=None, columns=None,
                 dtype=None, copy=False):
        self.name = name
        pd.DataFrame.__init__(self, data=None, index=None, 
                              columns=None, dtype=None, copy=False)


x = array([[9, 7, 5],
          [7, 3, 1],
          [8, 8, 3],
          [7, 4, 3]])
cols = ['a', 'b', 'c']
index = ['D', 'E', 'F', 'G']

s = DatFrame('huy', x, index, cols)

Error: RecursionError: maximum recursion depth exceeded while calling a Python object

Just so you know, you can dynamically add attributes to Python objects. So, let's say you have a pd.DataFrame referenced by variable df, then you can just do df.name = 'huy' — juanpa.arrivillaga
Also, there are special considerations when subclassing pandas data structures. See the docs — juanpa.arrivillaga

Jean-François Fabre Jean-François Fabre · Accepted Answer · 2017-07-28T21:26:02

First, you're passing None to all parameters of your dataframe

Second, the fix it is to call the __init__ method first, before setting name.

It's probably because the pd.DataFrame object expects an empty variable dictionary when initializing (recursion occurs in __getattr__), and your name member confuses it:

class DatFrame(pd.DataFrame):
    def __init__(self, name, data=None, index=None, columns=None,
                 dtype=None, copy=False):
        pd.DataFrame.__init__(self, data=data, index=index,
                              columns=columns, dtype=dtype, copy=copy)
        self.name = name

As a general rule, it's always better to call the parent constructor first thing, then set your specifics.

As comments said, if it's only to add name (without adding methods) you could create a factory method that adds dynamically the attribute, no need to inherit:

def create_datframe(name,*args,**kwargs):
    r = pd.DataFrame(*args,**kwargs)
    r.name = name
    return r

building a class from an existing one

1 Answers