try-else clauses measured at 1 more than correct complexity about mccabe HOT 4 OPEN

Also see discussion on #26

from mccabe.

Sorry for the confusion on #26. I agree that this is the right graph (with the exception of where the edges from the try body start for the except clauses.

from mccabe.

Should we get more buy-in before I submit a patch?
The quick hack I put in to make that diagram is pretty ugly.

from mccabe.

Note that this problem is not at all restricted to try-except constructs with an else clause. Where there is no 'else' the graph simply contains an edge going directly from the 'try' statement to the end of the block without passing through either the body of the try or any of the exception handlers.

This is a regression in 0.3 (and all later versions) - 0.2.1 generates the correct graph. I suspect but have not bothered to verify that this is due to 0a42f06

i.e. this function:

def test_try(s):
    try:
        return int(s)
    except TypeError:
        return -1

returns a cyclomatic complexity of 3 instead of 2 (consistent with 0.2.1, but only because of the off-by-one error #5 that was fixed); and this function:

def test_try2(s1, s2):
    try:
        i1 = int(s1)
    except TypeError:
        i1 = -1
    try:
        i2 = int(s2)
    except TypeError:
        i2 = -1
    return i1, i2

returns a cyclomatic complexity of 5 instead of 3 (vs. 4 in 0.2.1 due to the aforementioned off-by-one error fix.)

Ergo functions whose complexity is dominated by try/except have values that are off by up to a factor of 2.

from mccabe.