Redeclaration within different scopes

[AbdelrahmanKhaledd]

$ cat a.py
from math import log2, floor
def ff():
    i: i32
    for i in range(1, 1000):
        x: i32 
        x = i
        y: f32
        y = float(x)
        assert x.bit_length() == floor(log2(y)) + 1

    for i in range((1 << 30), (1 << 30) + 100 , 2):
        x: i32
        x = i
        y: f32
        y = float(x)
        assert x.bit_length() == floor(log2(y)) + 1
ff()

result:

$ ./src/bin/lpython -o out a.py
semantic error: Symbol is already declared in the same scope
 --> a.py:5:9
  |
5 |         x: i32 
  |         ~~~~~~ original declaration
  |
12 |         x: i32
  |         ^^^^^^ redeclaration


Note: if any of the above error or warning messages are not clear or are lacking
context please report it to us (we consider that a bug that needs to be fixed).

[certik]

This is an interesting one. In ASR we currently do not create a new scope in a for loop. There are two ways out:

• Either always create a new scope in a for loop
• Or the LPython frontend has to insert a "Block" ASR node (which has a scope) if there are any local variables in the loop, and declare them there

Probably the second option seems the way to start.

[Smit-create]

CPython does something like the following, it allows to declare variables inside for loop but they are also accessible outside the loop:

def f():
    i: int
    for i in range(10):
        x: int
        x = i
    print(x)
    for i in range(15):
        x: int
        x = i
    print(x)
f()

The output is:

9
14

[czgdp1807]

To enable the above, one might need to move the variables declared inside the loop at a line just before the starting of the loop. Seems like CPython has a runtime scope and loops don't have any scope of their own. So, CPython adds x into the current runtime scope and when the loop ends x is still present inside the scope and it is accessible.