refex VS tree-sitter-swift

As someone who worked on a similar tool (https://github.com/ssbr/refex/tree/main/refex/fix/fixers, I did a bunch of the work to prep this for open-sourcing, though I think all my contributions are hidden behind the "Google-internal" anonymization), having auto-applied or auto-appliable fixers like this is super useful.

They can be auto-applied by post-commit (e.g. a generic `git fixcommit` style command that runs all the relevant lint tools and fixes them in the working copy, letting you review before push), or applied during code review (automatic comments with a "click here to apply fix" interface), both of which are nice.

Plus the same underlying tooling can be used to write more complex one-off fixes that may be used for migrations or cleanups.

You might like Refex, which automates AST transformations like this:

There's lots of confusion about what semgrep does here, which is kind of unfortunate. I haven't touched it much, but I have built a very similar tool (I'm one of the contributors to refex[1], which is a very similar project).

The starting point of semantic grep is very useful. When you have a big codebase, you often want to detect antipatterns, or not even antipatterns, but just uses of a thing, say you're renaming a method and want to track down the callers.

Being able to act on the AST, instead of hoping you searched up all of the variants of whitespace and line breaks and, depending on the specific example, different uses of argument passing, is really useful.

But often when you're semantically grepping, your goal is to replace something with something else (this is what refex was initially built for: to aide in large scale changes in python, as a sort of equivalent to the C++ tools that Google uses).

But then you want to shift left even further: once you have a pattern that you want to replace once, you can just enforce that a linter yell at you when anyone does it again. So it's very natural to develop a linter-style thing on top of one of these[2].

This is, as I understand it sort of the same thing that happens in C++: clang-tidy and clang-format are written on top of AST libraries that can be used for ad-hoc analysis and transformations, but you can also just plug them into a linter.

The thing is, for most organizations, enforcing code style and best practices is more valuable than apply a refactoring to 10M lines of code, because most organizations don't have 10M lines of code to refactor. That doesn't mean that these tools aren't also useful for ad-hoc transforms and exploratory analysis. They absolutely are!

[1]: https://github.com/ssbr/refex

[2]: https://github.com/ssbr/refex/tree/main/refex/fix

You could probably set up something to work on Swift packages, however the tree sitter grammar for Swift hasn’t been updated since 2019. sourcekit-lsp also isn’t likely to help you out either because it’s not made to understand Xcode projects. You’d need an LSP capable of parsing a .pbxproj (I highly doubt there is one, but I haven’t looked around), which has the terrible downside of not being something Apple considers a public API and can and will introduce breaking changes on any Xcode update. An Apple engineer confirmed this with me during a lab in 2020 when I was asking about how doing some hacky things with binary Swift packages.

https://github.com/returntocorp/ocaml-tree-sitter/blob/maste... appears to be the general answer to your question, but navigating to the tree-sitter docs shows that tree-sitter has one in progress: https://github.com/tree-sitter/tree-sitter-swift so hopefully the machinery to incorporate it into semgrep will not be horrific