An SQL notebook that runs entirely in the browser.
Sqlit is an SQL and Python notebook designed for database educators and students. The tool is completely web based and can be run on modern web browsers with no installation required. While similar alternatives exist, sqlit aims to be particularly convenient with a Markdown-based file format and browser-native execution.
| Type | Description |
|---|---|
sql |
Execute SQL queries, results shown as a table |
python |
Run Python; database accessible via sqlite3 |
markdown |
Documentation or instructions |
load |
Load an external .sqlite file from a URL |
In order to run SQL queries in the browser, sqlit uses SQLite, which has been compiled to WebAssembly (abbreviated Wasm), the low-level language that web browsers understand. Databases are loaded into SQLite memory, which users can query using SQL directly, or via the sqlite3 Python module.
Python is supported through Pyodide, which is a project that compiles the standard interpreter to Wasm so that it can run natively in browsers.
There are many benefits to keeping all execution in the browser. The main one is that it requires no installation or setup for users. They can just go to the website, load a sheet, then get to work. Another advantage is that the app is simple to maintain and host, as it only requires a web server to host a few files.
A feature which has been prioritized is keeping the file format of sheets human-readable and easy to manage externally. Users often want to version-track sheets, like lectures or lab instructions, and reviewing diffs is significantly easier in a Markdown-based format than in a JSON-based one (like Jupyter's .ipynb format).
The main drawback with a less complex file format is that it puts some limitations on what kind of data can be stored. For the purpose of having students go through SQL and Python assignments, or a lecturer demonstrating database concepts, this is okay.
As for editing a .sqlit.md sheet externally, it's 99% conventional Markdown. The only exception is for creating runnable cells, which is done by wrapping code blocks in four backticks instead of three. As an example, this is how you would create an executable SQL cell in your sqlit sheet:
````sql
SELECT sqlite_version() AS version
````
For Python or Load cells, just replace the sql flag with python or load respectively. This allows external editors like VS Code to still give you features like syntax highlighting for the given language when editing. It also allows for regular non-interactive code blocks to be added to your sheet by writing regular Markdown code blocks (three backticks).
Developing locally is straightforward. The project uses Vite for bundling and pnpm as its Node.js package manager (which can be installed from the default one, npm). Once these are installed on your computer, simply run the commands below.
pnpm installpnpm devpnpm dev # Fetch Pyodide + start dev server
pnpm build # Fetch Pyodide + type-check + production build
pnpm preview # Preview production build
pnpm test # Run tests
pnpm lint # Biome linter
pnpm format # Biome formatterAt the moment, Python's input() calls don't work in Safari. The reason is that the web server has the header Cross-Origin-Embedder-Policy set to credentialless, which allows for the use of SharedArrayBuffer (used for Python input handling) while also letting the app fetch resources (like .sqlite files or images) from external sources. Unlike Chromium-based browsers (Chrome, Edge, Brave, ...) and Firefox, Safari does not yet implement this header value.