Map showing the cpython repositiory, highlighting the files that Guido van Rossum changed the most
Map showing the cpython repositiory, highlighting the files that Guido van Rossum changed the most

Basic use guide

  • Generate database with ./ {path_to_repo_dir}
  • Run web server with python (flask must be installed, can be install from pip)
  • Connect on
  • Available repos will be displayed, select the one you want to view
  • Select email to search for in the form at the bottom and click submit

This project consists of two parts:

  1. Git log -> database
  2. Database -> treemap

Git log -> database

Scans through an entire git history using git log, and creates a database using three tables:

  • Files, which just keeps track of filenames
  • Commits, which stores commit hash, author, committer
  • CommitFile, which stores an instance of a certain file being changed by a certain commit, and tracks how many lines were added/removed by that commit
  • Author, which stores an author name and email
  • CommitAuthor, which links commits and Author in order to support coauthors on commits

Using these we can keep track of which files/commits changed the repository the most, which in itself can provide useful insight

Database -> treemap

Taking the database above, uses an SQL query to generate a JSON object with the following structure:

“children”: [, …]

“val”: “>

  "children": [, ...]


then uses this to generate an inline svg image representing a treemap of the file system, with the size of each rectangle being the val described above.

Then generates a second JSON object in a similar manner to above, but filtering for the things we want (only certain emails, date ranges, etc), then uses this to highlight the rectangles in varying intensity based on the vals returned eg highlighting the files changed most by a certain author.


These speeds were attained on my personal computer.

Database generation

Repo Number of commits Git log time Git log size Database time Database size Total time
linux 1,154,884 60 minutes 444MB 462.618 seconds 733MB 68 minutes
cpython 115,874 4.6 minutes 44.6MB 36.607 seconds 74.3MB 5.2 minutes

Time taken seems to scale linearly, going through approximately 300 commits/second, or requiring 0.0033 seconds/commit.
Database size also scales linearly, with approximately 2600 commits/MB, or requiring 384 B/commit.

Querying database and displaying treemap

For this test I filtered each repo by its most prominent authors:

Currently treemap.js uses a global variable MIN_AREA to not render smallest files for better performance.

While these performances are not as fast as desired, a more typically sized repo should perform fine.

Wanted features

Submodule tracking

Currently the only submodule changes that can be seen are the top level commit pointer changes. In the future would like to recursively explore submodules and add their files to the database.

Faster database generation

Currently done using git log which can take a very long time for large repos. Will look into any other ways of getting needed information on files.

Asynchronous javascript

Currently no async functions are used. I believe the performance of the webpage could be imporved if things such as file loading and svg drawing was done asynchronously.

Remembering filters

Filters must be re-entered every time the page is loaded. Ideally filters would be remembered either through cookies or by storing the filters as a url query, which would allow users to bookmark queries.

Filter builder sidebar

Current interface is very barebones. Want to have a sidebar that will allow users to select authors, date ranges, etc, to control the highlighting.

Selectable colours per author

Currently red is hardcoded for all results. In order to show multiple authors we want to highlight in different colours, will need to decide how to colour files edited by both authors.

Read More