Flame analyzer

This package is an often used part of our debug environment at Publons. It helps benchmark and explain inefficiencies in pieces of code as well as our dependencies on different service response times.

There are four kinds of Context managers we support with this package

• FileFlame
• InlineFlame
• DjangoFileFlame
• DjangoInlineFlame

They all serve the same use case outputting a Flame graph to your machine for you to dive into and debug your code. FileFlame/DjangoFileFlame save the graph to an SVG for you to share, while InlineFlame/DjangoInlineFlame will render it in your IPython browser.

Examples

Saving a Flame graph to an SVG can be done with the following benchmarking

from flame_analyzer import FileFlame


with FileFlame('./file_flame_test.svg'):
    # Some expensive piece of code.
    [len(u.email) for u in  User.objects.all()]

Or directly to the IPython notebook

from flame_analyzer import InlineFlame


with InlineFlame():
    # Some expensive piece of code.
    [len(u.email) for u in  User.objects.all()]

You can also optionally configure the width by adding the width kwarg

with FileFlame(
    './file_flame_test.svg', flame_width=1200,
    options={'title': 'This is my test title'}
):
    # some expensive piece of code
    [len(u.email) for u in  User.objects.all()]

Extensions

By default both IPython and Django are optional imports meaning you can install this library and use it in the terminal to debug your app code without them installed. Support can be added for other Database frameworks or if your wanting to hook into the context enter/exit methods by creating your own hooks and adding to the output flame type your wanting for example

from flame_analyzer import InlineFlame

class CustomHook:
    """
    Append the time taken to execute to the flame graphs title.
    """
    def before(self):
        self.called_before = '< Called before code execution >'

    def after(self):
        self.called_after = '< Called after code execution >'

    def modify_flame_options(self, flame_options):
        title = flame_options.get('title', '')
        flame_options['title'] = self.called_before + ' --- ' + self.called_after
        return flame_options


class CustomInlineFlame(InlineFlame):
    hook_classes = (CustomHook,)

with CustomInlineFlame(flame_width=500):
    total_email_length = 0
    for u in User.objects.all():
        total_email_length += len(u.email)
    print(total_email_length)

Outputs the IPython viewed Graph

Credits to the following projects:

• https://github.com/brendangregg/FlameGraph
• https://github.com/23andMe/djdt-flamegraph