re module implementation

[Embers-of-the-Fire]

This PR introduces a subset of Python’s re module. It might close #132 as a partial implementation.

Full API list

• re.Match
  • bool, type, __eq__, __repr__
  • .group(), .groups()
  • .span(), .end()
  • .string
• re.Pattern
  • .pattern
  • .flags
  • Shares logic with module-level functions
• re.IGNORECASE, re.MULTILINE, re.DOTALL
• re.compile(), re.match(), re.fullmatch(), re.findall(), re.sub()
• re.PatternError (notes)

Unresolved questions

• re.PatternError behavior
• Regex behavior

re.PatternError behavior

The monty implementation introduces re.PatternError as a plain error with only a test.
The monty-python binding falls back to RuntimeError because retrieving re.PatternError is awkward unless we do something like py.import("re")?.get_attr("PatternError")?.call1((msg,)).

Regex behavior

By default, this PR uses the regex crate, which does not support backreferences, lookahead, etc.
Switching to fancy_regex could more closely match CPython’s default behavior, but might introduce ReDoS risk.

[codspeed-hq]

Merging this PR will not alter performance

✅ 13 untouched benchmarks

---

_{Comparing Embers-of-the-Fire:re (9b7b706) with main (a07e336)}

[codecov]

Codecov Report

❌ Patch coverage is 69.35229% with 194 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
crates/monty/src/types/re_pattern.rs	60.46%	69 Missing and 33 partials ⚠️
crates/monty/src/types/re_match.rs	77.65%	29 Missing and 13 partials ⚠️
crates/monty/src/modules/re.rs	75.69%	13 Missing and 22 partials ⚠️
crates/monty/src/heap.rs	65.00%	7 Missing ⚠️
crates/monty/src/exception_private.rs	57.14%	3 Missing ⚠️
crates/monty/src/types/type.rs	0.00%	2 Missing ⚠️
crates/monty/src/bytecode/vm/call.rs	0.00%	1 Missing ⚠️
crates/monty/src/modules/mod.rs	90.00%	1 Missing ⚠️
crates/monty/src/object.rs	0.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[samuelcolvin]

The other option is regex-pcre2 which should be most compatible with python.

No idea about it's dos risks or whether the c code is harder to compile for all platforms than pure rust?

[Embers-of-the-Fire]

@samuelcolvin ReDoS stems from backtracking features, so just one more implementation won’t work. IIRC google's re2 avoids those constructs entirely like regex. So it’s a trade‑off between a fast, stable engine and full compatibility.

[samuelcolvin]

Partial review since I have to go, but this looks great overall.

I would suggest we use fancy-regex with backtrack_limit() set to something reasonable, with that the dos risk seems acceptable.

[Embers-of-the-Fire]

So iirc re.PatternError was added and replaced re.error in some version, maybe that's why CI/CD complain about it :(. Could anyone show me how to create a proper alias?

[samuelcolvin]

So iirc re.PatternError was added and replaced re.error in some version, maybe that's why CI/CD complain about it :(. Could anyone show me how to create a proper alias?

A few options:

• use pyo3 macros (like this to have different logic for older python
• try to import re.PatternError and fallback to importing re.error
• try to import re.PatternError and fallback to RuntimeError

[Embers-of-the-Fire]

@samuelcolvin apologies for disturbing, but I wish to confirm the codecov. Certain sections of code prove difficult to construct during testing, so I am uncertain whether they require coverage.