init method gets a value while none is passed

Question

I have a class TextBuffer. It has an __init__ function which can be used to pass an initial value for the content of TextBuffer. If no value is passed, then it should use an empty list as default value.

The class has a method add, in order to add text to the TextBuffer:

a = TextBuffer()
a += "TEXT"

For some reason, calling the __init__ method without parameters does not use the default value of [] for lst, but a previously added value to another instance of TextBuffer, which is not in the __init__ call

a = TextBuffer()
a += "BUFFER A"
b = TextBuffer()  ### lst in the __init__ call has the value of ["BUFFER A"]
b += "BUFFER B"

I have no idea why this happens. Maybe I'm doing something wrong in the __add__ method?

Please consider the complete example code:

import pprint

### Two functions for byte/string conversion in Python3
def b(x):
    if ( type(x) == type(b'')):
        return x
    else:
        return x.encode(encoding='UTF-8')
def s(x):
    if ( type(x) == type(b'') ):
        return x.decode(encoding='UTF-8')
    else:
        return x


class TextBuffer():
    def __init__( self, lst = [], maxlines = None ):
        self.content = []
        if ( type(lst) == type([])):
            if (len(lst) > 0):
                print("INIT got a nonempty lst value:")
                pprint.pprint(lst)
                print("------------------------------")
            self.content = lst


    def __getitem__( self, i ):
        try:
            return self.content[i]
        except IndexError:
            return None

    def __iter__( self ):
        for line in self.content:
            yield line

    def __contains__( self, item ):
        return item in self.content

    def __len__( self ):
        return len(self.content)

    def __add__( self, item ):
        self.content.append(item)
        return self

    def __radd__( self, item ):
        return self.__add__(item)

    def __repr__( self ):
        result = ""
        for line in self.content:
            if ( type(line) == type(b"") ):
                result+=s(line)
            else:
                result+=line
        return result

    def __str__( self ):
        return repr(self)

    def __unicode__( self ):
        return repr(self.content)

### TextBuffer INIT with empty list creates an empty textbuffer
a = TextBuffer(lst=[])
print("a = TextBuffer(lst=[])")

a += "BUFFER A"


### TextBuffer INIT with empty list creates an empty textbuffer
b = TextBuffer(lst=[])
print("b = TextBuffer(lst=[])")

b += "BUFFER B"

print("Content of TextBuffer a:")
print(a)
print("Content of TextBuffer b:")
print(b)

print("-------------------------")

### TextBuffer INIT without any parameters should use default value for lst of []
### So an empty list. Should be the same as TextBuffer(lst=[])
a = TextBuffer()
print("a = TextBuffer()")

a += "BUFFER A"

### TextBuffer INIT without any parameters should use default value for lst of []
### So an empty list. Should be the same as TextBuffer(lst=[])
### But now, the value of lst is not [], it is the string added to TextBuffer 'a': ['BUFFER A']
b = TextBuffer()
print("b = TextBuffer()")

b += "BUFFER B"

print("Content of TextBuffer a:")
print(a)
print("Content of TextBuffer b:")
print(b)

Because you have a mutable thing as the default argument (`[]`). All instances share _that_ list, so if one of them adds an item to it, they all have that added. This is a duplicate of many questions. — RemcoGerlich, Jan 17 '17 at 08:29
http://python.net/~goodger/projects/pycon/2007/idiomatic/handout.html#default-parameter-values — RemcoGerlich, Jan 17 '17 at 08:30
There _are_ situations where you can safely use a default mutable argument, this is not one of them. — PM 2Ring, Jan 17 '17 at 08:33

Willem Van Onsem · Accepted Answer · 2017-01-17T08:45:43.773

3

You cannot use lst = [] as default argument in this case, since Python will makes one list in memory and passes each time a reference to the same object (list) to the constructor. This means that your two objects share the same lst and thus modifications made by one object are reflected in the other and vice versa.

You can use the following trick:

def __init__( self, lst = None, maxlines = None ):
    if lst is None:
        lst = []
    #... (remainder of the constructor)

since here you force Python to construct a new list. Evidently you could use another value than None; but I think for your case this will be sufficient.

In general it is a good idea to not feed mutable objects as default values, always use immutable objects. There are however, as @hiroprotagonist describes cases where this behavior is wanted (like memoization), but I would advice to be very careful with these.

edited Jan 17 '17 at 08:45

answered Jan 17 '17 at 08:29

Willem Van Onsem

443,496
30
428
555

4

you *can* use a mutable type as default argument. if you know what you are doing it may be a good idea (e.g. meomize). – hiro protagonist Jan 17 '17 at 08:32
when is the list passed to the instace? while loading the module or when calling the `__init__` mehtod? – Netwave Jan 17 '17 at 08:32
@hiroprotagonist yes, but generally that's not what you want to do. – Jean-François Fabre Jan 17 '17 at 08:33
@hiroprotagonist: yeah you are right, my point was merely that for the scenario the OP describes, this is not a good idea. I softened my claim. – Willem Van Onsem Jan 17 '17 at 08:33
1

@DanielSanchez: well each time you call the constructor with default value, the reference is passed (if you would feed `lst` another list, evidently the other list will be used). – Willem Van Onsem Jan 17 '17 at 08:35
1

@Jean-FrançoisFabre: i agree. hence the 'if you know what you are doing'-part. – hiro protagonist Jan 17 '17 at 08:37

init method gets a value while none is passed

1 Answers1

Linked

__init__ method gets a value while none is passed

1 Answers1

Linked

init method gets a value while none is passed