Tree-sitter: a new parsing system for programming tools - GitHub Universe 2017

They use tree-sitter link It's ok with errors in AST and just marks such nodes as invalid. There is a long (42 minutes) but worth to watch talk about tree-sitter ▶ this video Or the shorter version (20 minutes) link

`tree-sitter` has been post to the website a bunch of times so I remembered from those submissions: link Also these resources: link ▶ this video

100% agree. If starting now I think you should consider targeting mlir[1]. It fixes many problems you will have with llvm, and allows you to create language specific dialects and transformations. Impressive new work by a top-notch team led by LLVM creator Chris Lattner. It has LLVM as an optional target, so you get that for free. For a parser, you should consider using tree-sitter[2]. Tree-sitter gives you live editor support for free. Impressive work by Max Brunsfeld. [1] link [2] ▶ this video

Tree-sitter is not just for code folding, it's a full-on incremental parser that's fast enough to run on every keystroke. See their expand selection example for another really compelling use-case. If you're interested in more details, there's also this talk from the developer behind tree-sitter ( ▶ this video ), where he showcases how they could use tree-sitter to quickly parse entire multi-megabyte minified JS files to provide complete syntax highlighting for them that updates as you edit, and how the accuracy and usefulness of the actual syntax highlighting can improve over regex based implementations due to having access to the fully specified grammar of the language. I think tree-sitter is also a game changer for extension development once it stabilizes and the APIs are opened up. Extensions like swackets will no longer have arbitrary performance constraints that break user experience ( link ), and it opens the door to a whole new set of intelligent extensions that understand the languages you write at a much deeper level that were previously unfeasible due to parsing performance, or required some external heavy-weight language server.