Blockly integration

[dkuku]

I was thinking how to change the chart smart cell to make a grid. The code is very similar but it would need probably a separate cell to create grids, or something more flexible than the current smart cell.
This reminded me of scratch where you can build code using puzzle tiles that fit into each other. Scratch uses blockly under the hood and you can use it to code python in jupyter:
https://blog.jupyter.org/visual-programming-in-jupyterlab-with-blockly-7731ec3e113c
I made a poc to use it in livebook:

This may be useful for some other tasks like building neural network layers by drag and drop and lower the bar to start playing with livebook for non tech folks.

[josevalim]

Oh, this is indeed very neat and it could be very useful for smart cells (since the smart cells just generate code). We don't have bandwidth to tackle this ourselves right now so we encourage you and others to continue exploring it. Btw, can you model recursion with blockly?

/cc @michalslaski

[dkuku]

Looking at the scratch docs you can:

In Scratch 2.0 there is the ability to create custom stack blocks. These can be recursive since they can have inputs. This allows more advanced recursion than Scratch 1.4, allowing things like creating fractals and finding the factorial of a number more easily.

in scratch 1.4 you could only do tail recursion

Scratch 1.4 provides a limited type of recursive call, namely tail recursive calls, in which the recursive call is the last thing done in the script.

https://en.scratch-wiki.info/wiki/Recursion

[michalslaski]

Hi, very interesting idea. Thanks for CCing me.

There is also Snap! visual language, similar to Scratch. According to the authors, the Snap! project focuses more on advanced programming language features such as first class functions. For example it has blocks like call, pipe and stream, which allow constructs familiar to functional programmers. Some examples can be found in the Snap Reference Manual.

I'm not sure if the Snap JavaScript library is as easily embeddable as the blockly one, but the chapter about running Snap offline may give an overview of how it works.