In python 3 I found that class attribute can be used as a argument in __init__()
function, like below:
file test.py:
class Foo:
var1 = 23333
def __init__(self, var=var1):
self.var = var
run in cmd:
C:\Users\rikka\Desktop>py -3 -i test.py
>>> f1=Foo()
>>> f1.var
23333
but by using a dot.expression, when init this class, interpreter will report an error:
file test2.py:
class Foo:
var1 = 23333
def __init__(self, var=Foo.var1):
self.var = var
run in cmd:
C:\Users\rikka\Desktop>py -3 -i test2.py
Traceback (most recent call last):
File "test2.py", line 1, in <module>
class Foo:
File "test2.py", line 3, in Foo
def __init__(self, var=Foo.var1):
NameError: name 'Foo' is not defined
I just don't know why interpreter cannot find name 'Foo' since Foo is a name in the global frame in the environment. is there something scope related concept about python class that I don't fully understand?
Function defaults are set at function definition time, not when being called. As such, it is not the expression var1
that is stored but the value that variable represents, 23333
. var1
happens to be a local variable when the function is defined, because all names in a class body are treated as locals in a function when the class is built, but the name Foo
does not yet exist because the class hasn't finished building yet.
Use a sentinel instead, and in the body of the function then determine the current value of Foo.var1
:
def __init__(self, var=None):
if var is None:
var = Foo.var1
self.var = var
I used None
as the sentinel here because it is readily available and not often needed as an actual value. If you do need to be able to set var
as a distinct (i.e. non-default) value, use a different singleton sentinel:
_sentinel = object()
class Foo:
var = 23333
def __init__(self, var=_sentinel):
if var is _sentinel:
var = Foo.var1
self.var = var