jupyterlite / xeus-sqlite-kernel Goto Github PK

Problem

There's no place to try out this kernel yet.

Aside from off main, as new features are added, it would be nice to have something that reviewers can try out.

Suggested Improvement

• add a dev-focused binder with jupyter-server-proxy
• get set up for readthedocs (doesn't even require the mamb flag anymore to get a self-consistent environment)

Description

Opening a notebook with the jupyterlite-xeus-sqlite kernel gives the following error in the console:

RuntimeError: Aborted(Assertion failed: sz < N && "setting a small size", at: /install/include/xtl/xbasic_fixed_string.hpp,109,set_size). Build with -s ASSERTIONS=1 for more info.

Reproduce

Use jupyterlite-xeus-sqlite==0.2.0 in a JupyterLite deployment.

Expected behavior

The kernel should start without error.

Context

Noticed when updating to jupyterlite-sqlite for the Try Jupyter deployment: jupyter/try-jupyter#10

cc @SylvainCorlay @DerThorsten if you can have a look, thanks!

The jupyterlite/xeus-sqlite-kernel currently seems to require users to create a database as part of a script before they can query it, rather than connect to a sqlite file shipped as part of a JuptyerLite distribution or uploaded to it, or connect to a remote database file.

The phiresky/sql.js-httpvfs Javascript package wraps a SQLite emscripten distribution and also provides a virtual filesystem that allows the SQLite database to access very large remote SQLite databases (if configured appropriately) in an efficient way.

I wonder if this approach could be used to support remote connections from the xeus-sqlite-kernel? Users could then use a simple web server to publish both the JupyterLite environment, and even a large database, and then query the database efficiently from their browser?

The SQLime browser based SQlite playground (nalgeon/sqlime) also offers some interesting "web integrations", as described on its About SQLime page. For example, SQL files can be hosted Github gists, the gist id passed via a URL parameter to SQLime, then loaded and run by the playground when the page is loaded. (Databases files can also be uploaded from the desktop to the playground environment.)

The Basthon in-browser notebook [repo] & SQLite kernel loads a database from a file path specified via a URL parameter:

(I also note here examples of using custom javascript functions to create custom sqlite functions.)

vega is built in a static and non-static way. This doubles the compile times and only the static version is needed