Repository navigation
Profiling inner functions #300
Description
In my code, I have numerous inner functions which, for various reasons, I would like to leave as inner functions if at all possible. However, line_profiler does not appear to profile these functions by default. For example, the following code
from line_profiler import LineProfiler
import numpy as np
m = 2
def main():
n = 3
def called_func(x):
y = x**m
z = x + y + n
return z
for i in range(10):
x = np.random.rand()
z = called_func(x)
print(z)
lp = LineProfiler()
lp_wrapper = lp(main)
lp_wrapper()
lp.print_stats()
returns
4.335277136950235
Timer unit: 1e-07 s
Total time: 0.0007159 s
File: c:\Users\...\AppData\Local\Programs\Python\Python312\...\line_profiler_test.py
Function: main at line 5
Line # Hits Time Per Hit % Time Line Contents
==============================================================
5 def main():
6 1 54.0 54.0 0.8 n = 3
7 1 10.0 10.0 0.1 def called_func(x):
8 y = x**m
9 z = x + y + n
10 return z
11 11 83.0 7.5 1.2 for i in range(10):
12 10 218.0 21.8 3.0 x = np.random.rand()
13 10 263.0 26.3 3.7 z = called_func(x)
14 1 6531.0 6531.0 91.2 print(z)
Is it possible to modify the code above so that called_func is also profiled in a second block of output (while leaving it as an inner function)? A command-line approach would be fine if that would be better in some respect than the above in-code approach.
Several notes. I am working on Windows 10. So far, I have made several attempts to include a line like lp.add_function(main.called_func) and have tried to incorporate ideas from this Medium article and this Stack Overflow post without success. I also saw the comment below on the main GitHub page, but could not figure out how to implement it.
Nesting is common when using LineProfiler as a decorator. In order to support nesting, use
enable_by_count()anddisable_by_count(). These functions will increment and decrement a counter and only actually enable or disable the profiler when the count transitions from or to 0.
Any help would be appreciated.
Activity
ta946 commentedon Oct 26, 2024
you have 2 options that require you to modify the code, and 1 that doesnt:
you can do what you were doing before, just create new wrapped function
called_func_wrapped = lp(called_func)and use that inseadz = called_func_wrapped (x)you can use decorators instead of returning a wrapped function
https://kernprof.readthedocs.io/en/latest/line_profiler.autoprofile.html#module-line_profiler.autoprofile