data-science-guide

Cheat sheets across data science, published as a static site with GitHub Pages.

https://jshiriyev.github.io/data-science-guide/

One new sheet goes up each week.

Subjects

#	Subject	Folder
1	Data Science Overview	data-science-overview/
2	Probability and Statistics	probability-and-statistics/
3	Python Programming	python-programming/
4	Data Analytics Tools	data-analytics-tools/
5	Machine Learning Algorithms	machine-learning-algorithms/
6	Big Data Principles	big-data-principles/
7	Deep Learning Methods	deep-learning-methods/

Each folder holds one .html file per topic — data-analytics-tools/sql-cheat-sheet.html, data-analytics-tools/matplotlib.html, and so on. File names are free-form; the landing page reads each sheet's <title>, not its filename.

Adding a cheat sheet

Drop the .html file into the right subject folder, commit and push:

cp ~/Downloads/Pandas_Cheat_Sheet.html python-programming/
git add -A && git commit -m "Add pandas cheat sheet" && git push

That is enough. The deploy normalises the file, regenerates the index and publishes it. A sheet keeps its own markup, styling and title — nothing has to match a house style.

To preview before pushing, run the same steps locally:

python tools/adopt_sheet.py      # add doctype/charset/viewport if missing
python tools/build_catalog.py    # regenerate the landing page index
python tools/check_site.py       # links and structure
python -m http.server 8000       # preview at http://localhost:8000

Sheets written elsewhere often arrive as a fragment — no <!DOCTYPE>, no <head>, no charset — which means quirks-mode rendering and mojibake wherever the file uses an em dash or an arrow. adopt_sheet.py adds that scaffolding without touching the page's own markup, and is safe to re-run.

The landing page card takes its name from the sheet's <title> and its summary from <meta name="description">. A sheet with no description still publishes; its card just has no summary line. To add one:

python tools/adopt_sheet.py python-programming/Pandas_Cheat_Sheet.html \
  --description "Selection, joins, groupby and reshaping."

There are no dependencies to install — plain Python 3, and plain HTML/CSS/JS.

Repository layout

index.html              landing page, rendered from data/catalog.js
assets/css, assets/js   shared styles and behaviour for every sheet
data/catalog.js         generated index of the sheets on disk
templates/              starting point for a new sheet
tools/                  sheet adopter, catalog generator, site checker
.github/workflows/      GitHub Pages deployment

Writing a sheet against the shared styles

Optional — a sheet is free to be entirely self-contained. But if you start from templates/cheatsheet.html, then writing <section class="section" id="..."> with an <h2>, and .card blocks inside it, is enough to get a table of contents, scroll tracking, a search filter (/ to focus), copy buttons on code blocks, a light/dark toggle and a print stylesheet, with no per-sheet wiring.