[PHPCraftdream]

Summary

• Add native/pure mutation benchmarks with exact source and independent semantic/order checks, including nested verbatim insertion.
• Pin the corrected Rust core 1f7c717597cceecb8da84aaf8f47e62945b1db98, which fixes lossless verbatim reindent. The successful Native benchmark retains the original positive oracle; no case is skipped.
• Preallocate the Pure remove expected-object list to the known entry count. Preserve output parsing, semantic verification, source/error behavior, and order. Compare base and candidate on one runner in six alternating pairs without a timing gate.

Validation and measurement

Exact head: f6f69f67f551f7945b6dd5c45d4250714058d229. CI 40535955772 passed 13/13 jobs, including all native/pure regressions and the paired benchmark.

On this runner, Pure wide-remove median increased from 131,346 to 154,101 ns/op (17.3% slower), while unchanged wide-rename increased from 64,391 to 73,943 ns/op (14.8% slower). Run-to-run variation is substantial; the candidate has not demonstrated an improvement. The benchmark does not isolate list allocation cost; further profiling is needed before claiming a benefit.

Dependency and scope

This draft depends on the unmerged Rust PR #24, whose exact core commit must remain reachable. No local product builds, tests, or benchmarks were run by the parent. No dependency version, package version, ABI, or release change. The pre-existing local spec checkout change is excluded. No merge is requested.

🤖 Generated with Claude Code