,
13 tweets,
3 min read
Read on Twitter
What I wanted Babel to be: A thread
At the beginning of 2015 when Babel was still called 6to5, I created the issue "Project Scope and Future" to solicit ideas and feedback from the community about my desire to generalize 6to5. github.com/babel/babel/is…
In it, I say this:
> A JS transpiler based on future standards and specifications, whether ECMAScript or not, I think is an extremely noble goal. [...] Having an interopable platform to base code transformations on I think is a very powerful idea.
> A JS transpiler based on future standards and specifications, whether ECMAScript or not, I think is an extremely noble goal. [...] Having an interopable platform to base code transformations on I think is a very powerful idea.
After this, we changed the name of 6to5 to Babel, to remove the ES6 to 5 connotation, and made ourselves future proof. The general platform concept manifested in the release of Babel 6, which moved out all of the core transforms and made it a general-purpose JS compiler.
This was only supposed to be the beginning of my vision though. I released Babel 6 live on stage at "EmberCamp London 2015", where in the closing slides, I spoke of my excitement for new tools to be built on top of Babel, including a linter, autoformatter, minifier and more.
After Babel 6, I eventually left the project due to mental health reasons and never really got involved again. Since then I've been experimenting with a lot of these ideas and have arrived at some conclusions.
The idea of these tools being built on top of Babel isn't very sound. An ad-hoc collection of tools is a fragile ecosystem. Atomic releases of all the tools to support new syntax, behaviour, configuration, and more is extremely difficult.
Today's JavaScript programs uses several different parsers spread across the engines the code is executing in, all the way to the linters, compilers, bundlers and more that are a part of the development process.
Consolidation of parsers isn't very interesting though. Sure we can all share the same AST, and syntax support, but that doesn't give us any new capabilities. What I'm excited about are things like:
- a testing framework that uses the same bundler for your test files as the one in production
- a linter built on top of compiler so you can do complex AST autofixes
- a single config file
- a bundler that understands how dependencies are used, and optimizes your bundle
- a linter built on top of compiler so you can do complex AST autofixes
- a single config file
- a bundler that understands how dependencies are used, and optimizes your bundle
- a minifier that does constant folding and inlining
- a compiler transforms that don't rely on order
- a compiler that uses an immutable AST to perform transforms, with accurate scope tracking
- a linter that gives me suggestions on how to fix my code
- ...
- a compiler transforms that don't rely on order
- a compiler that uses an immutable AST to perform transforms, with accurate scope tracking
- a linter that gives me suggestions on how to fix my code
- ...
What possibilities are you excited about?
This is a more technical take on from last year
