Profiling function that call another function

[tmwstw7]

Hi everyone, I've stuck with such kind of issue: I'm trying to profile function that calls couple of other functions, e.g.:

def foo(a, b):
    return a+b

@profile
def bar(d):
   f = foo(4,5)
   return d*f

and as was mentioned in other issues with similar topics, Memory metric is the same during all time. I got that this caused by calling one func from another, but still didn't understand how to fix it. Could someone explain how to make things work properly?

[viniciusd]

I will follow this issue because I have a similar situation.

I need to figure what which function is causing my memory peak, analyzing the memory differences line by line at the root level function doesn't do the job, as the peak happens at some deeper function call

Looking forward to a way of performing this kind of analysis.

I assume it should be possible, given it is possible to generate the plot of memory usage over time. However, if the profiling data generated over time is acquired via sampling, it would explain why there isn't an option to generate a whole code-base report of memory usage

edit

Just checked the code, it spawns the process and samples its memory usage periodically, indeed: https://github.com/pythonprofilers/memory_profiler/blob/master/memory_profiler.py#L362

[viniciusd]

As far as I read the code, it is completely possible to add the possibility of tracing consumption in the whole codebase using the same approach (sys.settrace), but it isn't possible to accomplish it with the current state of memory_profiler

Given all I wanted was finding which part of my code was generating a peak of ram usage, I combined the plotting feature with a binary search: deleted half of the code and ran the plot to check whether the peak was still there, then continued with the binary search.