SIEVE-website

mkdoc source code for SIEVE webpage

Local build

1. Set up a virtual environment

python3 -m venv .venv
source .venv/bin/activate

2. Open up a terminal and install Material for MkDocs with (starting from 9.2.0b0, Material supports blog feature):

pip3 install mkdocs-material=="9.2.0b0"

3. Install minify plugin

pip3 install mkdocs-minify-plugin

4. Previewing the website

mkdocs serve

How the landing page is built

• material/home.html is the landing page template. It extends main.html, which adds the fonts, the link-preview tags and analytics to every page.
• material/home.css styles the landing page. The colors are --sv-* tokens defined for light and dark in docs/stylesheets/extra.css, which also styles the header, the footer and the blog.
• material/partials/header.html is Material's header with two changes: it shows the short name (extra.short_name in mkdocs.yml), and on wide screens the Home and Blog tabs sit in the same row.
• material/sim.js runs the SIEVE simulation in "How it works", following the pseudocode in the blog post. Run node -e "require('./material/sim.js')" to load its cache model without a browser.
• material/charts.js draws the results charts with Chart.js and builds their data tables. The values were read from the vector figures in the paper and blog (docs/assets/images/illustrations).

Editing the adoption list

Adopters are data at the top of material/home.html, one row per entry:

("ClickHouse", "https://github.com/ClickHouse/ClickHouse/pull/123435", 54801242),

The fields are the name, the link, and the logo: a GitHub account id, or a path to an image under docs/assets. Section headings count the rows, so adding a row is all it takes.

• Link to the code, merged PR or docs that show SIEVE in use, not just the project homepage.
• Use the account id rather than the name for GitHub logos: the name-based avatar URL returns a blank default for many organizations. curl -sI https://github.com/<org>.png shows the id in the redirect (/u/<id>).
• Open pull requests go under in_progress until they merge.
• Keep the list in sync with the cacheMon homepage.