Character ( Perl ) Class Operation Support

[Shiftcoder94]

The regex crate in Rust is highly performant and well-suited for a wide range of use cases. However, one limitation ( not a serious one, but can be useful in many cases )is the lack of support for character class set operations, such as union, intersection, and subtraction (e.g., [a-z&&[^aeiou]]). These features are supported in several other regex engines (e.g., Java, PCRE) and allow for more expressive and concise pattern definitions.

Their absence in Rust's regex engine can make certain patterns more cumbersome to express and may require additional logic or preprocessing to work around. Given the crate’s focus on performance and correctness, it would be valuable to understand whether support for character class operations is planned or has been intentionally omitted due to technical or design constraints.

Any insights or updates on this limitation would be greatly appreciated. Thank you for the excellent work on maintaining such a robust and efficient regex engine.

Reference - PCRE2 Project Syntax Page ( The operator section is a little down )

And to be honest, I've made a module which really does it, but the problem is, it does it in a different AST structure defined by me. Well I'm most willing to share it.

Yours Faithfully,
Shift

[BurntSushi]

This has been supported for years.

[BurntSushi]

Yes. That's correct. What do you expect it to be? The intersection of two disjoint classes is an empty class (which matches nothing).

[Shiftcoder94]

Yes. That's correct. What do you expect it to be? The intersection of two disjoint classes is an empty class (which matches nothing).

So! It does calculate the sets, in the SyntaxParser! That was what I didn't expect the regex_syntax crate to do. Thanks for clearing my foolish doubt.

[BurntSushi]

On the off chance you're not an AI, in the future, please just start with an MRE. I would strongly urge you to read this on how to create an MRE. For Rust, it should include a Cargo.toml, a complete source file, any relevant inputs, expected output, actual output and the commands you use to build and run the code.

[BurntSushi]

Correct output ( Not built on the same AST as of regex_syntax) -
Ok(Operation { data: Intersection(Chars { data: ['a', 'b', 'c'] }, Chars { data: ['d', 'e', 'f'] }) })

The top-level parser gives you an Hir, which is a level of abstraction above the AST. It does a ton of heavy lifting for you. If you really do want the AST in a way that matches the concrete syntax given, then you can drop down a layer of abstraction and use the AST parser directly. For example: https://play.rust-lang.org/?version=stable&mode=debug&edition=2024&gist=9c60d56180958485a2f7957de58584a2

[Shiftcoder94]

On the off chance you're not an AI, in the future, please just start with an MRE. I would strongly urge you to read this on how to create an MRE. For Rust, it should include a Cargo.toml, a complete source file, any relevant inputs, expected output, actual output and the commands you use to build and run the code.

Well! I'm not as much experienced as you are! Just a high-school student, but still very passionate and skillful. The suggestion will really help me in future. Thanks!