Design outline

[agoose77]

My current approach is to use a two-pronged approach:

1. A "passive" markdown renderer that expands {{ expr }} calls into <input type="hidden" value="expr"> nodes. This is done using a jupyterlab-markup renderer. These nodes are then replaced by the Markdown cell with the result of rendering mime-bundles stored in the cell attachments. There are issues here with formatting etc, but we can work out the details later!
2. An executor that stores the result of these expressions into the cell attachments.

There are going to be some tricky issues w.r.t to UX to sort out here, but for now this is quick to get up and running.

[agoose77]

This design has been further delineated into separate components:

graph LR;

markup[Markdown Source]-->markup_expr[jupyterlab-markup-expr*]
markup_expr-->html[HTML Output]
html-->imarkdown[Interactive Markdown]

Loading

There is now an optional jupyterlab-markup-expr extension that parses the Markdown for {{ expr }} syntax, and outputs special HTML input nodes of the form

<input type=hidden class=eval-expr value=" expr ">

jupyterlab-imarkdown only concerns itself with rendering these HTML elements, meaning that you can bring your own parser for different syntax variants without needing to disable the built-in extension.