ac3 in csp.py

[onursurme]

at line 171 in csp.py, in the function AC3, it is written as :
if Xk != Xi:

there is no need to exclude Xi because Xi already can't be a neighbour to itself.
but I guess it should be
if Xk != Xj:
because Xi is already revised with Xj, so there is no need to add Xi, Xj to the queue again.

we can also see this in the book at Figure6.3 as :
for each Xk in Xi.NEIGHBORS - {Xj} do
here again, Xj is excluded.

[darius]

You appear to be right -- thank you!

We're also missing a test that would provoke this bug. Would you like to make a pull request with a new test case, or should I?

[onursurme]

I'll try to create a test. I'll inform you about the result.

Sent from Yahoo Mail on Android

On Fri, 2 Sep, 2016 at 6:14, Darius Baconnotifications@github.com wrote:
You appear to be right -- thank you!

We're also missing a test that would provoke this bug. Would you like to make a pull request with a new test case, or should I?

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or mute the thread.

[onursurme]

Here is a test case which I think reveals the bug :

from csp import *

def t1f(A,a,B,b):
    return int(a)>int(b)

test1=CSP(["a","b"],{"a":["2","4"],"b":["3","5"]},{"a":"b","b":"a"},t1f)
AC3(test1)
print(test1.infer_assignment())

a must be bigger than b, so b can't be 5, and a must be 4.
when I run this test with the wrong version of AC3 (if Xk != Xi: version), the test says b=5, which is wrong.
but when I run the test with the right version of AC3 (if Xk != Xj: version), the test says a=4, which is true.