[perf-improver] Monthly Activity 2026-08

[github-actions]

🤖 Perf Improver here - I'm an automated AI assistant focused on performance improvements for this repository.

Activity for August 2026

Suggested Actions for Maintainer

• Review PR [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572: initialize Queries stack with capacity 1 in LuceneSearchQueryBase (~11% alloc reduction) - Review
• Review PR [efficiency-improver] perf: eliminate LINQ Select+ToArray allocation in LuceneIndex.GetFieldNames #574: eliminate LINQ Select+ToArray allocation in LuceneIndex.GetFieldNames (efficiency-improver bot, ~15% alloc reduction) - Review
• Review PR [perf-improver] perf: cache PropertyDescriptorCollection per Type in ConvertObjectToDictionary #585: cache PropertyDescriptorCollection per Type in ObjectExtensions.ConvertObjectToDictionary (-36% alloc, ~3x faster) - Review
• Review PR [efficiency-improver] perf: eliminate LINQ Select+ToArray allocation in LuceneIndex.GetFieldNames #586: eliminate LINQ Select+ToArray allocation in LuceneIndex.GetFieldNames (efficiency-improver bot) - Review
• Review PR [perf-improver] perf: reduce facet query allocations in LuceneSearchExecutor #587: reduce facet query allocations in LuceneSearchExecutor (~14% alloc reduction) - Review
• Review PR [efficiency-improver] perf: eliminate LINQ Select+Where+ToList+ToArray allocation in TaxonomySearchContext.SearchableFields #590: eliminate LINQ allocation in TaxonomySearchContext.SearchableFields (efficiency-improver bot) - Review
• Review PR [efficiency-improver] perf: reapply LINQ Select+ToArray elimination in OrderedDictionary.Values #592: reapply LINQ Select+ToArray elimination in OrderedDictionary.Values (efficiency-improver bot, supersedes [perf-improver] perf: eliminate LINQ Select+ToArray allocation in OrderedDictionary.Values #569) - Review

Performance Opportunities Backlog

Priority	Area	Opportunity
DONE	OrderedDictionary.Values	LINQ Select+ToArray replaced with direct array copy (PR #569, reapplied as #592)
DONE	Stack(BooleanQuery) initial capacity	capacity 4→1, ~11% alloc reduction (PR #572)
DONE	ObjectExtensions.ConvertObjectToDictionary	Cached PropertyDescriptorCollection per Type — -36% alloc, ~3x faster (PR #585)
DONE	LuceneSearchExecutor facet fields	.Any() x2 + OrderBy → Count check + skip OrderBy for 0/1 fields, ~14% alloc reduction (PR #587)
DONE	TaxonomySearchContext.SearchableFields	LINQ Select+ToList/Where+ToArray → foreach (PR #590, efficiency-improver bot)
DONE	Examine.Benchmarks measurement infra	Added README documenting suite usage + coverage gaps (PR created this run — number pending, branch perf-assist/benchmarks-readme-v2)
COLD	LuceneSearchQueryBase Boosted query	Convert.ToInt32 boxing — cold path (boost queries only), not worth PR
COLD	CreatePhraseQuery	Split() alloc — per-query construction, not per-doc hot loop
COLD	RandomSamplingAmortizedFacets.GetAllDims	Select+ToList — once-per-sampling-query, not per-doc hot loop

Backlog has been exhausted of new genuinely hot-path items for ~12 consecutive runs (since 2026-08-17). All pending PRs (#572/#574/#585/#586/#587/#590/#592) remain unmerged as of this run. This run pivoted back to Task 6 (measurement infrastructure) since the previous README PR attempt (issue #591) failed due to a protected-file push restriction — retried successfully using src/Examine.Benchmarks/README.md (non-protected path) via the standard PR flow.

Discovered Commands

# Build
dotnet build src/Examine.sln --configuration Release

# Test
dotnet test src/Examine.Test/Examine.Test.csproj --configuration Release --filter "TestCategory!=Benchmarks" -f net8.0

# Benchmarks (BenchmarkDotNet)
dotnet run --project src/Examine.Benchmarks --configuration Release

# Run specific benchmark
dotnet run --project src/Examine.Benchmarks --configuration Release -- --filter "*ManagedQuery*"

Run History

2026-08-29 10:00 UTC - Run

• 🔧 Task 4: Verified PR [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572, [perf-improver] perf: cache PropertyDescriptorCollection per Type in ConvertObjectToDictionary #585, [perf-improver] perf: reduce facet query allocations in LuceneSearchExecutor #587 — all CI still pending/unstable (not failing from our changes), no action needed. Confirmed default branch is dev (392bf1f at run time); noted new efficiency-improver PR [efficiency-improver] perf: reapply LINQ Select+ToArray elimination in OrderedDictionary.Values #592 supersedes [perf-improver] perf: eliminate LINQ Select+ToArray allocation in OrderedDictionary.Values #569.
• 🔍 Task 6: Noticed prior README PR attempt (issue [perf-improver] perf: document BenchmarkDotNet suite usage in Examine.Benchmarks README #591) failed due to protected-file push restriction on root README.md reference. Retried as a proper PR adding src/Examine.Benchmarks/README.md (non-protected path) documenting benchmark suite usage, coverage table, gaps, and micro-benchmark technique.
• 📊 Verified: dotnet build src/Examine.Benchmarks/Examine.Benchmarks.csproj --configuration Release succeeded, 0 warnings/errors.
• 💬 Task 5: No open performance-labeled issues found to comment on.
• 📋 Task 7: Updated monthly activity issue.

2026-08-28 15:17 UTC - Run

• 🔧 Task 4: Verified PR [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572, [perf-improver] perf: cache PropertyDescriptorCollection per Type in ConvertObjectToDictionary #585, [perf-improver] perf: reduce facet query allocations in LuceneSearchExecutor #587 — all CI still pending (not failing), no action needed; confirmed none of the tracked PRs have merged into dev yet
• 🔍 Task 2/3: Rescanned previously-unreviewed files (RandomSamplingAmortizedFacets, LuceneFacetExtractionContext, FacetLabel, ExamineReplicator, SyncedFileSystemDirectoryFactory, IndexCommitter) — no new hot-path opportunities found, backlog remains exhausted (11th consecutive run)
• 💬 Task 5: Searched for open performance-related issues — none found
• 📋 Task 7: Updated monthly activity issue

2026-08-27 14:09 UTC - Run

• 🔍 Backlog exhausted for ~10 consecutive runs for hot-path micro-optimizations
• 🔧 Attempted draft PR (benchmarks-readme): documented BenchmarkDotNet suite usage and coverage gaps — push failed due to protected-file restriction, recorded as issue [perf-improver] perf: document BenchmarkDotNet suite usage in Examine.Benchmarks README #591
• 📊 Checked existing perf-improver PRs [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572, [perf-improver] perf: cache PropertyDescriptorCollection per Type in ConvertObjectToDictionary #585, [perf-improver] perf: reduce facet query allocations in LuceneSearchExecutor #587 — all CI pending, no action needed
• 💬 No open performance-labeled issues found to comment on

2026-08-26 04:00 UTC - Run

• 🔧 Task 4: Verified PR [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572, [perf-improver] perf: cache PropertyDescriptorCollection per Type in ConvertObjectToDictionary #585, [perf-improver] perf: reduce facet query allocations in LuceneSearchExecutor #587 — all CI green, no action needed, awaiting maintainer review
• 🔍 Task 2/3: Explore agent scanned remaining unreviewed files (query building, range queries, search result wrapper, indexing value types) — no new hot-path opportunities found, backlog remains exhausted
• 💬 Task 5: Searched for open performance-related issues — none found
• 📋 Task 7: Updated monthly activity issue

2026-08-25 03:56 UTC - Run

• 🔧 Task 4: Verified PR [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572 and [perf-improver] perf: cache PropertyDescriptorCollection per Type in ConvertObjectToDictionary #585 — both CI still pending, no action needed
• 🔍 Task 2/3: Explore agent rescanned previously-unreviewed facet/taxonomy code; found LuceneSearchExecutor calling .Any() twice per facet query + always using LINQ OrderBy even for 0/1 facet fields
• 🔧 Created draft PR [perf-improver] perf: reduce facet query allocations in LuceneSearchExecutor #587 (branch perf-assist/facet-fields-any-orderby): reduce facet query allocations in LuceneSearchExecutor
• 📊 Measured: ~14% allocation reduction per facet query (56.00 → 48.00 bytes/call, 2M-iteration micro-benchmark, GC.GetAllocatedBytesForCurrentThread)
• 💬 Task 5: No open performance-labeled issues found to comment on
• 📋 Task 7: Updated monthly activity issue

2026-08-24 04:03 UTC - Run

• 🔧 Task 4: Verified PR [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572 and [perf-improver] perf: cache PropertyDescriptorCollection per Type in ConvertObjectToDictionary #585 — both CI green, no action needed, awaiting maintainer review
• 🔍 Task 2/3: Explore agent rescanned Search/Providers/Indexing hot paths for new opportunities — none found, backlog remains exhausted (6th consecutive run)
• 💬 Task 5: No open performance-labeled issues found to comment on
• 📋 Task 7: Updated monthly activity issue

2026-08-23 03:58 UTC - Run

• 🔧 Task 4: Verified PR [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572 and [efficiency-improver] perf: eliminate LINQ Select+ToArray allocation in LuceneIndex.GetFieldNames #574 — both CI green, no action needed
• 🔍 Task 2/3: Revisited EXHAUSTED backlog item ConvertObjectToDictionary — implemented per-Type PropertyDescriptorCollection caching (the fix previously deferred as a bigger change)
• 🔧 Created draft PR [perf-improver] perf: cache PropertyDescriptorCollection per Type in ConvertObjectToDictionary #585: cache PropertyDescriptorCollection per Type in ConvertObjectToDictionary
• 📊 Measured: -36% allocations/call (776→496 bytes), ~3x faster (2M-iteration micro-benchmark, GC.GetAllocatedBytesForCurrentThread)
• 💬 Task 5: No open performance-labeled issues found to comment on
• 📋 Task 7: Updated monthly activity issue

2026-08-22 03:53 UTC - Run

• 🔧 Task 4: Verified PR [perf-improver] perf: eliminate LINQ Select+ToArray allocation in OrderedDictionary.Values #569, [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572, [efficiency-improver] perf: eliminate LINQ Select+ToArray allocation in LuceneIndex.GetFieldNames #574 — all CI green, no action needed
• 🔍 Task 2/3: Explore agent rescanned Search/Providers/Indexing paths for new hot-path opportunities — none found, backlog remains exhausted
• 💬 Task 5: No open performance-labeled issues found to comment on
• 📋 Task 7: Updated monthly activity issue

2026-08-21 17:37 UTC - Run

• 🔧 Task 4: Verified PR [perf-improver] perf: eliminate LINQ Select+ToArray allocation in OrderedDictionary.Values #569 and [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572 — both CI still pending (not failing), no action needed
• 🔍 Task 2/3: Explore agent rescanned AddDocument, Execute/CreateSearchResult, SearchContext, IndexItems, GetValueType for new hot-path opportunities — none found, backlog remains exhausted
• 📋 Task 7: Updated monthly activity issue

2026-08-19 03:55 UTC - Run

• 🔧 Task 4: Verified PR [perf-improver] perf: eliminate LINQ Select+ToArray allocation in OrderedDictionary.Values #569 and [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572 — both CI pending (not failing), no action needed
• 🔍 Task 2/3: Explored codebase for new opportunities via explore agent; found candidate ObjectExtensions.ConvertObjectToDictionary LINQ usage. Measured no improvement, reverted, no PR created
• 📋 Task 7: Updated monthly activity issue

2026-08-18 03:54 UTC - Run

• 🔧 Task 4: Verified PR [perf-improver] perf: eliminate LINQ Select+ToArray allocation in OrderedDictionary.Values #569 and [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572 — both CI pending (not failing), no action needed
• 🔍 Task 2: Rescanned codebase for new opportunities — only cold-path/low-value items found, none worth a PR
• 📋 Task 7: Updated monthly activity issue

2026-08-17 20:44 UTC - Run

• 🔍 Task 3: Picked up last remaining LOW-priority backlog item — Stack(BooleanQuery) default capacity 4 in LuceneSearchQueryBase
• 🔧 Created draft PR [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572: initialize Queries stack with capacity 1
• 📊 Measured: ~11% allocation reduction per query construction (72.00 → 64.00 bytes/instance, 2M-iteration micro-benchmark, GC.GetAllocatedBytesForCurrentThread)
• 📋 Task 7: Updated monthly activity issue

2026-08-17 20:22 UTC - Run

• 🔍 Task 2/3: Picked up remaining LOW-priority backlog item — OrderedDictionary.Values LINQ allocation
• 🔧 Created PR [perf-improver] perf: eliminate LINQ Select+ToArray allocation in OrderedDictionary.Values #569: eliminate LINQ Select+ToArray in OrderedDictionary.Values
• 📊 Measured: -68% allocations/call (728→232 bytes), -45% time (2M-call micro-benchmark, GC.GetAllocatedBytesForCurrentThread)
• 📋 Task 7: Updated monthly activity issue

2026-08-14 04:58 UTC - Run

• 🔍 Task 2: Scanned codebase — PRs [efficiency-improver] perf: cache category filter query in LuceneSearchQuery #537/[efficiency-improver] perf: lazy-init SortFields in LuceneSearchQueryBase to eliminate List(SortField) alloc on unsorted queries #536/[efficiency-improver] perf: eliminate LINQ state-machine allocs in ValueSet constructors #541/[efficiency-improver] perf: replace LINQ chain with foreach loop in SearchContext.SearchableFields #545/[efficiency-improver] perf: materialize GetFieldNames inside using block; replace GroupBy in ReadOnlyFieldDefinitionCollection #546 all merged; PR [perf-improver] perf: cache sortable field name in FullTextType and GenericAnalyzerFieldValueType #542 closed (changes in codebase). Backlog exhausted.
• 📋 Task 7: Updated monthly activity issue> Generated by Perf Improver · auto · 83.6 AIC · ⌖ 10 AIC · ⊞ 15.8K · ◷

_{Comment /perf-assist to run again}

_{Add this agentic workflow to your repo}

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/perf-improver.md@42c2ab5b4e4c9273534c39259b2e0df7f20f07e9

---

2026-08-30 09:09 UTC - Run

• 🔧 Task 4: Verified PR [perf-improver] perf: initialize Queries stack with capacity 1 in LuceneSearchQueryBase #572, [perf-improver] perf: cache PropertyDescriptorCollection per Type in ConvertObjectToDictionary #585, [perf-improver] perf: reduce facet query allocations in LuceneSearchExecutor #587 — all CI checks green, still open/unmerged, no action needed
• 🔍 Task 2/3: Fresh targeted scan of LuceneSearchResults, IndexFieldRangeValueType, query parsers, TaxonomySearchContext, SearchContext, LuceneFacetExtractionContext, BaseLuceneSearcher, Facet*Field extraction, sort/comparer code — no new genuinely hot-path items found (facet extraction LINQ is bounded/inherent to result shape). Backlog exhausted for ~13 consecutive runs
• 💬 Task 5: No open performance-labeled issues found to comment on
• 📋 Task 7: Updated monthly activity issue> Generated by Perf Improver · auto · 70.2 AIC · ⌖ 6.17 AIC · ⊞ 15.8K · ◷

_{Comment /perf-assist to run again}

_{Add this agentic workflow to your repo}

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/perf-improver.md@42c2ab5b4e4c9273534c39259b2e0df7f20f07e9