Profiling & Visualization Tools in Python - Part 1

Original post: on medium Also posted: on

What in my python code is taking time ?

Sounds like a familiar question asked!

Well, I have that question pop up quite often, especially when dealing with legacy code bases and unfamiliar code. This is my effort in helping shed light on exploring an answer to the question


Consider the following simple piece of code in a file called

def print_method():
    print("hello world")

def test_print_method():
    for i in xrange(2):

for i in range(3):

When you execute the code using python executable, following output of hello world is printed 6 times

$ python
hello world
hello world
hello world
hello world
hello world
hello world

Probing Parameters

Let’s look in to different tools which we could use to determine

  • Code path
  • Number of calls and
  • Which method took the most time !


This blog has been created after executing the commands on a Linux distro. If on a different platform, I assume one knows how to get to the required (or similar) software packages

Basis for profiling

We basically rely on profile module, more specifically cProfile module to generate the data needed for different visualization tools. ( Easiest way to run cProfile on a python code is to run it as a module with python executable by passing the actual script as an argument to cProfile


python -m cProfile

Along with the expected output of hello world lines ,we see additional information about the time it took to execute each method.

$ python -m cProfile
hello world
hello world
hello world
hello world
hello world
hello world
         12 function calls in 0.000 seconds
Ordered by: standard name
ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        1    0.000    0.000    0.000    0.000<module>)
        6    0.000    0.000    0.000    0.000
        3    0.000    0.000    0.000    0.000
        1    0.000    0.000    0.000    0.000 {method 'disable' of '_lsprof.Profiler' objects}
        1    0.000    0.000    0.000    0.000 {range}

Output might not be super digestible instantly at the first glance. Lets dive into different visualization tools available which can make the timing information more perceivable.

A picture is worth a thousand words


My favorite one among all the available tools is gprof2dot

Install gprof2dot

pip install gprof2dot --user

Generate .pstats file

Execute the script this time with the timing information being directed to an external output file, rather than standard console output. Observe the -o flag with the output filename being test.pstats

python -m cProfile -o test.pstats

Visualize profiling information

Assuming you have dot and eog installed, run the following command in the terminal where the profiling output file test.pstats is located

gprof2dot -f pstats test.pstats | dot -Tpng -o output.png && eog output.png

Bingo, you get a window which shows something like the following

gprof2dot output gprof2dot output


Above graph shows you that 91.67% was spent in test_print_method and the same method was called 3 times (3x) , which in turn calls print_method (overall 6 times).

Other method calls like range is a tiny amount of total execution time, but it is also visible in the graph

This was a simple python code. This method works equally well with complex code. Pasting a sample image from gprof2dot’s github repo

gprof2dot sample output gprof2dot sample output


SnakeViz is a browser based visualization tool. It needs the output in a .profile format, rather than .pstats when the profiling output is generated using cProfile module

Install snakeviz

pip install snakeviz --user

Generate .profile file

Execute the script this time with the timing information being redirected using -o flag to output file named test.profile

python -m cProfile -o test.profile

Visualize profiling information

Run the following command in the terminal where the profiling output file test.profile is located

snakeviz test.profile


There will be some information printed in the console when the command is run and a new window which pops in a browser session

$ snakeviz test.profile
snakeviz web server started on; enter Ctrl-C to exit
START /usr/bin/google-chrome-stable ""
Opening in existing browser session.
WARNING:tornado.access:404 GET /images/sort_both.png ( 1.04ms
WARNING:tornado.access:404 GET /images/sort_desc.png ( 0.49ms

Snakeviz icicle visulaization Snakeviz icicle visulaization

Snakeviz suburst visulaization Snakeviz suburst visulaization


This seems to be an abandoned project, but hey ! still works.

Install pycallgraph

pip install pycallgraph --user

Generate png visualization file

Execute the script using pycallgraph executable, rather than going via python executable.


The following examples specify graphviz as the outputter, so it’s required to be installed

pycallgraph graphviz -- ./

Visualize profiling information

Above command generates a pycallgraph.png image. Open the image using any image viewer. Using eog as an example on my host

$ eog pycallgraph.png

pycallgraph png pycallgraph png

Shows the timing and number of calls information.


I still believe gprof2dot does a better job at giving % and eye tracks naturally the flow in the graph much easier than other visualization methods.

Latest Blogposts

Secureboot + Ubuntu + VirtualBox Signing kernel modules

Set of steps required for dealing with secureboot on Ubuntu where VirutalBox service has issues

9 May 2021

Fixing the postfix error dpkg

Steps to fix the postfix error happening during apt upgrade ubuntu.

21 April 2021

Running a react app on Local Kubernetes cluster on Windows 10

Process and steps for running react app on local k8s cluster using minikube on windows 10

16 March 2021

Gatsby site hosted on AWS Amplify redirecting to homepage always

Using the rewrites and redirects on AWS Amplify for the depolyed personal website

2 November 2020

Connecting AWS Amplify for deployment of website

Explains how to connect the gatsby website hosted on github to AWS Amplify for deployment

1 November 2020

Evolution of this website

will be capturing the evolution of my website chronologically (latest first)

31 October 2020