I try to understand memorization, particularly regarding nonlocal mem with the code below:
def mem_last(f):
mem = 0
def helper(x):
print("mem = ", mem)
mem = f(x + mem)
return mem
return helper
def mem_last1(f):
mem = {"0": 0}
def helper(x):
print("mem = ", mem)
mem["0"] = f(x + mem["0"])
return mem["0"]
return helper
def f(x):
print("x = ", x)
return x
f = mem_last(f)
#f = mem_last1(f)
print(f(0))
When I use f = mem_last(f), where mem is a scalar, there is an error:
py mem.py
Traceback (most recent call last):
File "mem.py", line 15, in <module>
print(f(0))
File "mem.py", line 4, in helper
print("mem = ", mem)
UnboundLocalError: local variable 'mem' referenced before assignment
If I use f = mem_last1(f), where mem is a dict, there is no error and it works well.
Please explain why there is a difference between a dict and a scalar var. Thank you very much.
PS: I am using 3.7.7 64 bit on Windows 10. I have read the thread you provide and I fix it by add a line "nonlocal mem" before 'print("mem = ", mem)'.
However, my question is why I do not need to add line "nonlocal mem" when mem is a dict or a list, and no error occur. So why python treats scalar different from dict or list?
