Issue with likelihood function being a method

[johannesbuchner]

Issue with likelihood function being a method

[JohannesBuchner]

Thank you for the report.
Another solution would be to pass a lambda function that calls the method.

[jtlz2]

@timothydmorton Would you mind posting your fix?

[timothydmorton]

line 191 of the current run.py I changed to:

if nargs == 4 and not inspect.ismethod(LogLikelihood):

This is crude because it ignores the current purpose of this test, which is presumably to test whether lnew is passed, but that's the idea. A more complete fix would first test whether LogLikelihood is a method and then would test nargs (3 vs. 4 or 4 vs. 5 depending on whether it's a method).

[JohannesBuchner]

You pass myobject.LogLikelihood not MyClass.LogLikelihood to run(), right?

It is strange that python treats methods differently. I expected if you pass myobject.LogLikelihood, that it acts just like a function which calls MyClass.LogLikelihood with additionally myobject as the first parameter.

[JohannesBuchner]

Would

 nargs = len(inspect.getargspec(LogLikelihood).args) - inspect.ismethod(LogLikelihood)

also work? (line 187)

[timothydmorton]

right, I'm passing myobject.LogLikelihood. Also, I just made that pull request for a quick fix, but didn't test it. I don't actually know whether it's a generally acceptable solution or not.

[timothydmorton]

yes, that sounds like a much cleaner fix, and should work the same.

[JohannesBuchner]

Hey,
I just released an update to PyMultiNest, which includes a fix to this problem. Could you upgrade and test it?
Thank you, Johannes

[JohannesBuchner]

Closing, I think this is fixed.