Thank you for your comment

Beau­tiful Racket / tuto­rials

Extend a data format: jsonic

jsonic proves that a DSL can be a simple & prac­tical tool for solving a program­ming problem. We needed fewer than 50 lines of code for jsonic—making it even shorter than bf. Undoubt­edly, writing a whole JSON-gener­a­tion library in our favorite program­ming language would’ve been more work.

Of course, one reason jsonic can be small is that it relies heavily on Racket: as an imple­men­ta­tion language, as the language for the embedded expres­sions, and for the json library, which handles the heavy lifting.

But that’s a good thing. All we need to do is design and imple­ment the inter­face for our little language.

The other reason jsonic is small is that we’ve treated it as a tool for personal use. If we were sharing this DSL with others, we’d want to include tests, docu­men­ta­tion, better error messages, and inte­gra­tion with DrRacket (about which, more shortly).

Mostly, this jsonic tuto­rial was meant to nail down the basic two-part frame­work of a Racket-imple­mented language that we saw in stacker and bf:

  1. A reader that converts the source code into paren­the­sized S-expres­sions. We made our reader by chaining a tokenizer to a grammar-based parser.

  2. An expander that adds bind­ings to the code produced by the reader, so the code can be eval­u­ated as a Racket program. We made our expander by relying on macros and helper func­tions.

The next tuto­rials are going to assume that we’re completely cool with these ingre­di­ents. If any of these concepts are unclear, this is a good moment to review.

Choose your own adven­ture

As a junior JSON engi­neer at Mega­Corp, your inven­tion of jsonic provided a huge boost to your produc­tivity. Sure, it’s a toy language. But a very useful toy. In fact, by creating a JSON preprocessor, you discov­ered that you were able to complete your weekly JSON assign­ments in 1/3 the usual time.

What’s your next thought?

← prev next →