Intelliblocks for the JupyterLab/Blockly Extension

20 Jan 2020 . Category . Comments #research #education-research #datawhys #fsharp #data-science #machine-learning #programming #statistics #ldi

My Fable Jupyterlab Blockly Extension now has intelliblocks - blocks that use intellisense to call properties and functions on any Python library.

The intelliblocks address one of the key challenges with Blockly blocks - each block must be created somehow. Since it takes me 10-30 minutes to create a block using Blockly’s developer tools, it could take weeks if not months to create blocks for a single library with hundreds of classes and thousands of members. Since there are thousands of libraries out there, and they are constantly changing, creating blocks manually is effectively impossible.

I created the intelliblocks for my Applied Computational Linguistics class, so current usage focuses on calling high-level NLP libraries with these blocks, but the mechanism is sufficiently general that they should be useful for any application.

Below is a demonstration of the intelliblocks in action. It is important to note that since intellisense comes from the kernel, the blocks cannot configure themselves until the code has been executed on the kernel. This is necessarily true for dynamically typed languages like Python, though not for statically typed languages.