Optimize the start position of bounded look-behinds

[shilangyu]

No description provided.

[Copilot]

Pull Request Overview

This PR optimizes the starting position calculation for bounded look-behinds in the regex automata engine. Key changes include:

• Replacing the flat list of look-behind start states with a tree structure to capture nesting.
• Constructing and sorting a vector of look-behind tuples in PikeVM to ensure correct evaluation order.
• Updating the compiler and builder components to handle the new look-behind tree structure.

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
regex-automata/src/nfa/thompson/pikevm.rs	Updated look-behind evaluation logic and added sorting of look-behind states.
regex-automata/src/nfa/thompson/nfa.rs	Refactored look-behind representation from a vector of state IDs to a tree structure.
regex-automata/src/nfa/thompson/compiler.rs	Modified look-around compilation to incorporate look-behind offset and nesting.
regex-automata/src/nfa/thompson/builder.rs	Updated builder API to construct the look-behind tree according to nesting paths.

Comments suppressed due to low confidence (3)

regex-automata/src/nfa/thompson/nfa.rs:1579

• [nitpick] The condition using preorder in the Debug implementation is not immediately clear. Adding an inline comment to clarify its intent would help future maintainers understand the logic.

                .iter().any(|i| !i.preorder(&|e| e.start_id() != sid))

regex-automata/src/nfa/thompson/compiler.rs:1060

• [nitpick] Clarify in a comment the assumption behind combining relative_start and maximum_len to calculate start_offset, ensuring that the resulting value correctly bounds the look-around start position.

let start_offset = match (relative_start, maximum_len) {

regex-automata/src/nfa/thompson/builder.rs:722

• [nitpick] The new start_lookbehind API relies on a nesting path. Adding a brief example or more detailed inline comment about the structure and expected format of 'path' would improve clarity.

pub fn start_lookbehind(&mut self, start_id: StateID, offset_from_start: Option<usize>, path: &[usize]) {

[shilangyu]

Can/should be moved to the Builder?