benchmark-action-playtest-target

Target repo for testing benchmark actions when deployed to another repo

Original repo

https://github.com/willGraham01/benchmark-action-playtest

Pages Deployment

https://willgraham01.github.io/benchmark-action-playtest-target/

What's going on?

The idea is that the original repo will push "profiling results" to the source branch, target, from which this repository will then build a website using the website_build/build_site.py script. This script can be run locally to test the website that is created - just be sure to have the target branch (or a test branch you want to draw the source from) up-to-date and checked-out in your local clone/fork. The src directory contains the website source that does not need to be updated - certain files have placeholder markers in them so that the script knows where to insert markdown. There are currently two (well, three if you count the index page) pages generated by the script, plus a bunch of extra files that are linked to.

• The index.md page, which is just rendered markdown that points to the other pages.
• The profiling results lookup table
• The summary statistics page

The profiling results lookup page

This lookup table on this page is auto-generated from the pyisession files that are pushed to the source branch. The contents of this page are some flavour text, surrounding a lookup table that will redirect viewers to the results (in rendered HTML) of the given profile run. The results of each run are themselves HTML files that are dumped into the build/pyis_html directory, and under the hood the script uses a pandas.DataFrame object to keep track of which pyis session corresponds to which HTML file. This association is then parsed into a markdown table, and rendered in the placeholder section of the src/profiling_index.md template page.

The run statistics page

The plots on this page are generated by reading the pyisession files and any associated meta-data files (currently these share the same name as their pyisession counterparts, but carry a .stats.json extension). The plots are placed into the build/plots directory as svgs and image includes are written into the placeholder section of the src/run_statistics.md page.

It is also possible for us to export a per-profiling session table of statistics we are interested in, implementing this in much the same way as the table in the profiling lookup page. This is currently not implemented.