https://github.com/dnephin/pstatsinteractive

A web service to interactively inspect large python profiles.
https://github.com/dnephin/pstatsinteractive

Last synced: about 2 months ago
JSON representation

A web service to interactively inspect large python profiles.

• Host: GitHub
• URL: https://github.com/dnephin/pstatsinteractive
• Owner: dnephin
• Created: 2012-01-31T09:22:16.000Z (over 13 years ago)
• Default Branch: master
• Last Pushed: 2015-08-19T22:24:58.000Z (over 9 years ago)
• Last Synced: 2025-01-26T00:13:00.520Z (3 months ago)
• Language: Python
• Homepage:
• Size: 219 KB
• Stars: 1
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
PstatsInteractive

=================

PstatsInteractive is a mini python web service for viewing python profiles.

The goal of this project is to create an interactive profile inspector that works

with very large profiles.

There are already a few ways to view python profiles, but each has it's own

limitations or shortcomings.

* Interactively with `ipython` using the `pstats` module

  This is standard method for viewing profiles. It works with profiles of any size

  but it can take a while to find exactly what you're looking for. It's missing

  nice visuals, and a way to tag and share interesting views.

* [RunSnakeRun](http://www.vrplumber.com/programming/runsnakerun/)

  Includes a nice visual view of calls, but it can be hard to fit all the data in

  a single screen.  Large profiles end up looking like many thin layers of an

  onion. Also lacks a way to tag and share interesting views.

* [gprofdof](http://code.google.com/p/jrfonseca/wiki/Gprof2Dot) and [Graphviz](http://www.graphviz.org/) to create an image

  Static image works well for smaller profiles, but can be difficult to navigate

  large profiles. Easy to share the entire image, but not a sub-section. Not

  interactive.

* [snakeviz](http://jiffyclub.github.io/snakeviz/) which is a web app and d3 graph

  Does not work with large profiles.

* [pstats_viewer.py](http://chadaustin.me/2008/05/open-sourced-our-pstats-viewer/) a single python file that starts a basic web server

  This is a step in right direction. Lacks styling and a visual representation.

  The work is done on the server side, so not as interactive as it could be.

Plan

----

* reuse elements of `pstats_viewer.py` and `snakeviz` to create a client-side web based viewer

* use [dagre](https://github.com/cpettitt/dagre) to create graphviz like diagrams of sub-sections of a profile

* use [d3js](http://d3js.org/) to create pie charts of callee time (including gc time)

* web server should serve json blobs of profiles

* include a page to allow selection of available profiles

Notes

-----

Pstats are stored using this structure. Primitive call count is the number of

non-recursive calls.

    { 

        (

            file_name,

            line_number,

            function_name,

        ): (

            primitive call count,

            total call count,

            total time,

            cumulative time,

            callers

        ),

        ...

    }

callers is a dict with this structure:

    {

        (

            file_name, 

            line_number, 

            function_name

        ): (

            total call count,

            primitive call count,

            total time attributed to this caller,

            cumulative time attributed to this caller

        ),

        ...

    }