refactor(runtime): split field_set_by_name.rs (#7402)

[proggeramlug]

Closes #7402. The file sat at exactly 2000 of 2000 allowed lines, so the next line anyone added turned lint red — and lint is required, so every merge after that bypasses a required gate. That already happened once today.

file	lines
field_set_by_name.rs	567
field_set_by_name/tail.rs	963
field_set_by_name/fast_paths.rs	284
field_set_by_name/write_helpers.rs	170
field_set_by_name/attr_variants.rs	87

Pure move, proven. Every relocated region diffs IDENTICAL against origin/main — tail.rs is 937 lines verbatim. Only 5 lines differ across the whole 2000: four file-private unsafe fns promoted to pub(super), and one now-dead #[allow(unused_assignments)] whose assignments left with the macro.

The split point is the RuntimeHandleScope creation, chosen deliberately so the refresh_roots_after_alloc!() macro (#7341), all 16 of its call sites, and both GC_STORE_AUDIT(INIT) markers stay in one file with their order untouched. No rooted step crosses a module boundary — that ordering was the subject of two PRs today and must not move.

Verification

• Lib tests: 1697 passed / 0 failed / 3 ignored — byte-identical to an origin/main baseline run in the same session
• Gap set (93 tests): 88 pass / 3 node_fail / 2 run_fail on both arms, with per-test verdicts diffing identical, not merely matching counts
• check_file_size.sh exit 0, proven able to fail via PERRY_FILE_SIZE_THRESHOLD=900
• addr_class_inventory.py passes; sums preserved across the split (handle-floor 11 → 1+3+6+1, lone-valid-obj-ptr 3 → 2+1, 544 ratcheted sites before and after)
• Warnings 4 → 4, identical set; cargo check proven able to fail via a deliberate error

Latent hazard this exposed

Both addr-class gate inputs are path-keyed, so any file split trips them even with zero new sites. addr_class_allowlist.txt matches via rel_path.startswith(prefix) with an entry ending in .rs, and addr_class_ratchet_baseline.txt is a per-(rule, file) count. Re-keyed here; it will bite the next person told to split a file.

Note for #7435

Its hunk targets the URL-setter arm, now at field_set_by_name/tail.rs:317 — the patch is written against field_set_by_name.rs:1240 and needs retargeting. tail.rs has 1032 lines of headroom for its 5.

Summary by CodeRabbit

• Refactor

  • Reorganized named-property assignment internals into focused components without changing behavior.
  • Preserved support for object, array, closure, class, namespace, descriptor, and prototype-related property writes.
  • Maintained garbage-collection safety and property publication ordering during assignments.

• Documentation

  • Added release documentation describing the internal restructuring.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 2575a279-1403-48e1-92a5-9d2243c236fb

📥 Commits

Reviewing files that changed from the base of the PR and between 73c7118 and a33a895.

📒 Files selected for processing (8)

• changelog.d/7402-split-field-set-by-name.md
• crates/perry-runtime/src/object/field_set_by_name.rs
• crates/perry-runtime/src/object/field_set_by_name/attr_variants.rs
• crates/perry-runtime/src/object/field_set_by_name/fast_paths.rs
• crates/perry-runtime/src/object/field_set_by_name/tail.rs
• crates/perry-runtime/src/object/field_set_by_name/write_helpers.rs
• scripts/addr_class_allowlist.txt
• scripts/addr_class_ratchet_baseline.txt

---

📝 Walkthrough

Walkthrough

The field-setting implementation is split into topical Rust modules. The entry point delegates rooted assignment to the tail module, while fast paths, attribute setters, write helpers, and path-based metadata are updated.

Changes

Field setter split

Layer / File(s)	Summary
Module wiring and metadata crates/perry-runtime/src/object/field_set_by_name.rs, changelog.d/..., scripts/*	The entry point declares and re-exports the new modules, delegates the assignment tail, and updates allowlist and ratchet paths.
Attribute and write helper contracts crates/perry-runtime/src/object/field_set_by_name/attr_variants.rs, .../write_helpers.rs	The split adds attribute-specific setters, key utilities, closure handling, class-static mirroring, and native namespace overrides.
Validated fast paths crates/perry-runtime/src/object/field_set_by_name/fast_paths.rs	The split adds overwrite and transition-cache paths with receiver, key, shape, storage, and GC validation.
Rooted assignment tail crates/perry-runtime/src/object/field_set_by_name/tail.rs	The relocated path handles special objects, descriptors, prototype setters, shape transitions, storage, root refreshes, publication ordering, and mirroring.

Estimated code review effort: 4 (Complex) | ~45 minutes

Possibly related PRs

• PerryTS/perry#6811: Optimizes the fast paths relocated by this split.
• PerryTS/perry#6928: Modifies the native-module namespace override moved into write_helpers.rs.
• PerryTS/perry#7401: Updates the root-refresh logic retained in the relocated assignment paths.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch refactor/7402-split-field-set-by-name

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}