perf(types): идемпотентная регистрация member-source в MetadataCollectionSpecializer by nixel2007 · Pull Request #4189 · 1c-syntax/bsl-language-server

nixel2007 · 2026-06-22T22:11:09Z

Что и зачем

Обход дерева метаданных в MetadataCollectionSpecializer приходит к одному и тому же per-owner synthetic-типу многократно (общие имена табличных частей, общий element-type), а TypeRegistry.registerMemberSource добавляет источник без дедупликации (computeIfAbsent(...).add(...)). В итоге ~429K списков member-source держат ~6.7M лямбд (~14 дублей на тип), а getMembers материализует ~14-кратно избыточный граф членов, удерживаемый в membersCache.

Профилирование (JFR + heap dump + Eclipse MAT) показало, что TypeRegistry удерживает ~52% живого хипа, а весь 6-млн кластер MemberDescriptor/TypeSet/BilingualString/лямбд — это и есть эта дублирующая материализация.

Изменение

registerPerOwner помечает уже обработанные ref в рамках одного specialize() и не регистрирует источники повторно. Регистрация детерминирована по path-кодированному имени ref, поэтому повторная обработка была чисто избыточной — выход getMembers и так дедуплицируется putIfAbsent по имени.

Замер (analyze на `nixel2007/cpm`)

структура (было ~6M)	baseline	этот PR
MemberDescriptor	6.01M	442K
TypeSet	6.04M	475K
BilingualString	6.19M	617K
member-source лямбды	~6.7M	~450K

метрика	baseline	этот PR
live heap	5.04 ГБ	2.95 ГБ (−41.5%)
`LambdaForm.invokeExact` (CPU на дублях)	7.6%	исчезает из топа
`Object[]` allocation pressure	74%	49%

Соотношение 6.01M/442K ≈ 13.6× — ровно фактор дублирования. JSON-отчёт analyze побайтно идентичен baseline.

✅ Валидация

MetadataCollectionSpecializerTest (12 тестов, HBK-gated через BSL_LANGUAGE_SERVER_RUN_HBK_TESTS=true) прогнан локально на реальном HBK — 12/12 PASS на этой ветке и в комбинации A+B+C. Тесты покрывают per-MDO/per-collection специализацию, табличные части и рекурсию по под-владельцам (tabularSectionGetsRecursivePerSectionType) — ровно там, где работает гард дедупа. Плюс JSON-отчёт analyze побайтно идентичен baseline.

Крупнейший из трёх независимых перф-патчей; #A и #C складываются поверх.

🤖 Generated with Claude Code

…tionSpecializer Обход дерева метаданных приходит к одному и тому же per-owner synthetic-типу многократно (общие имена табличных частей, общий element-type), а TypeRegistry.registerMemberSource добавляет источник без дедупликации (computeIfAbsent(...).add(...)). В итоге ~429K списков member-source держали ~6.7M лямбд (~14 дублей на тип), а getMembers материализовал ~14-кратно избыточный граф членов, удерживаемый в membersCache. registerPerOwner теперь помечает уже обработанные ref в рамках одного specialize() и не регистрирует источники повторно. Регистрация детерминирована по path-кодированному имени ref, поэтому повторная обработка была чисто избыточной — выход getMembers и так дедуплицируется putIfAbsent по имени. Замер на nixel2007/cpm (analyze): MemberDescriptor 6.01M → 442K, TypeSet 6.04M → 475K, BilingualString 6.19M → 617K (ровно ~13.6×, совпало с оценкой дублей); live heap 5.04 ГБ → 2.95 ГБ (-41.5%); CPU на дублирующих member-source лямбдах (LambdaForm.invokeExact 7.6%) исчезает; Object[] allocation pressure 74% → 49%. JSON-отчёт побайтно идентичен baseline. Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]> Claude-Session: https://claude.ai/code/session_01EWA4hy5YvUBU1CXkdr4npc

coderabbitai · 2026-06-22T22:11:24Z

📝 Walkthrough

Walkthrough

MetadataCollectionSpecializer gains a Set<TypeRef> registeredOwners (backed by HashSet) that is cleared at the start of each specialize() call. In registerPerOwner, a set-membership check now causes an early return when the same per-owner type ref has already been registered during the current traversal, preventing duplicate TypeRegistry#registerMemberSource calls.

Changes

Per-owner registration deduplication

Layer / File(s)	Summary
`registeredOwners` field declaration and import `...types/registry/MetadataCollectionSpecializer.java`	Adds `java.util.HashSet` import and declares a private `Set<TypeRef> registeredOwners` field with a comment describing its deduplication purpose.
Clear on `specialize()` and guard in `registerPerOwner` `...types/registry/MetadataCollectionSpecializer.java`	Calls `registeredOwners.clear()` before the specialization traversal and adds an early-return guard in `registerPerOwner` using `registeredOwners.add(perOwnerRef)`, skipping re-registration of already-seen per-owner refs.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

Poem

🐇 A set to remember each owner I've seen,
No double-registering — keeps the code clean!
One clear() at the start, one add() as the gate,
Repeats hop right past, they arrive a bit late.
The registry rests, no duplicates remain! 🌿

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 66.67% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and specifically summarizes the main change: introducing idempotent registration of member sources in MetadataCollectionSpecializer to eliminate redundant processing.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ 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 perf/membersource-idempotent

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

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

coderabbitai

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In
`@src/main/java/com/github/_1c_syntax/bsl/languageserver/types/registry/MetadataCollectionSpecializer.java`:
- Around line 316-324: The `registeredOwners` field is mutable shared state in a
WorkspaceScope component that lacks thread-safety guarantees. Since
WorkspaceScope does not serialize method invocations, the `specialize()` method
can be called concurrently by multiple threads, causing race conditions when
`clear()` and `add()` operations execute on this HashSet. Either add
synchronization and document that `specialize()` must be invoked serially within
a workspace, or refactor `registeredOwners` from a class-level field to a local
variable declared within the `specialize()` method scope so each invocation has
its own dedup state.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

Push a commit to this branch (recommended)
Create a new PR with the fixes

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 5264b748-5668-4f78-8f56-7fa82eb72bf8

📥 Commits

Reviewing files that changed from the base of the PR and between 49181d7 and 4392509.

📒 Files selected for processing (1)

src/main/java/com/github/_1c_syntax/bsl/languageserver/types/registry/MetadataCollectionSpecializer.java

nixel2007 · 2026-06-22T22:30:12Z

Прогнал MetadataCollectionSpecializerTest локально на реальном HBK (BSL_LANGUAGE_SERVER_RUN_HBK_TESTS=true): 12/12 PASS на этой ветке. Включая tabularSectionGetsRecursivePerSectionType и per-MDO/per-collection кейсы — рекурсия по под-владельцам, где работает гард, отрабатывает корректно. Метаданная модель членов не сломана.

nixel2007 mentioned this pull request Jun 22, 2026

perf(types): интернирование TypeSet и BilingualString #4190

Closed

coderabbitai Bot reviewed Jun 22, 2026

View reviewed changes

Comment thread ...a/com/github/_1c_syntax/bsl/languageserver/types/registry/MetadataCollectionSpecializer.java

nixel2007 merged commit b2ae376 into develop Jun 22, 2026
37 of 38 checks passed

nixel2007 deleted the perf/membersource-idempotent branch June 22, 2026 22:35

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

perf(types): идемпотентная регистрация member-source в MetadataCollectionSpecializer#4189

perf(types): идемпотентная регистрация member-source в MetadataCollectionSpecializer#4189
nixel2007 merged 1 commit into
developfrom
perf/membersource-idempotent

nixel2007 commented Jun 22, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot commented Jun 22, 2026 •

edited

Loading

Walkthrough

Changes

Estimated code review effort

Poem

❌ Failed checks (1 warning)

Uh oh!

coderabbitai Bot left a comment

Uh oh!

Uh oh!

nixel2007 commented Jun 22, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

nixel2007 commented Jun 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Что и зачем

Изменение

Замер (analyze на nixel2007/cpm)

✅ Валидация

Uh oh!

coderabbitai Bot commented Jun 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Estimated code review effort

Poem

❌ Failed checks (1 warning)

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

nixel2007 commented Jun 22, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

nixel2007 commented Jun 22, 2026 •

edited

Loading

Замер (analyze на `nixel2007/cpm`)

coderabbitai Bot commented Jun 22, 2026 •

edited

Loading