%%timeit could use local scope.

You can’t use timeit in a function (which is useful to test variable inputs of several function doing the same things, and plot graphs.

I believe a well placed @needs_local_scope and modification to timeit should allow that.

[Carreau]

I was likely refering to the line magic %timeit not the cell magic. I tried to make a %timeit, in a function in a for loop. For example:

def fib(n):
    return 1 if n < 2 else fib(n-1)+fib(n-2)

def timefib(P):
    t = []
    for i in range(P):
        res = %timeit -o fib(i)
        t.append(res)
    return t

And this timeit would need local scope.

I believe that what I was referring to and inadvertantly have put 2 % sign.

[takluyver]

Yes, you are right. The genuine global var j is getting overwritten.

That’s unfortunate. I would have been inclined to hold off from merging the PR in that case. Hopefully not too many people will hit it, though, because it only affects using %timeit inside a function, which I think is fairly unusual.

But, I have got another way of solving this, How about nested calls to exec(). Inner call will be made from inner() and we could pass local_ns to inner() first and then exec() will use them to execute stmt.

That’s a decent idea, but I think we probably shouldn’t pursue it. Timeit goes to some lengths to eliminate as much code as possible between the time measurements and the code to execute, so that the timings are as accurate as possible.