LeanCorpus-only capabilities

These capabilities have no direct equivalent in Lucene.NET 4.8.0-beta00018 or Java Lucene 10.5.1. Lucene may expose lower-level building blocks from which an application could construct comparable behaviour.

This page was audited against the checked-out LeanCorpus source on 25 August 2026. Return to the feature comparison overview for the complete comparison.

.NET application model

Capability	LeanCorpus API	Why it differs
Native AOT support	AOT-safe core and smoke application	Trim-safe paths avoid runtime code generation and reflection-dependent serialisation.
Asynchronous document writes	`AddDocumentAsync` and `AddDocumentsAsync`	Native `ValueTask` APIs rather than wrappers around synchronous indexing.
Streaming bulk ingestion	`AddDocumentsAsync(IAsyncEnumerable<LeanDocument>)`	Bounded batches consume asynchronous producers directly.
Asynchronous result streaming	`SearchAsync`	Results are exposed as `IAsyncEnumerable<ScoreDoc>`.
Source-generated document mapping	`LeanDocumentMap<T>` and `[LeanDocument]`	Compile-time field descriptors and mapping code.
Typed LINQ queries	`LeanQueryable<T>`	Expressions are translated through the generated document map.
JSON document mapping	`JsonDocumentMapper`	Maps `JsonElement` trees into indexed field paths.
Source-generated JSON metadata	`LeanCorpusJsonContext`	Reflection-free metadata for LeanCorpus serialisation contracts.

Resource controls and indexing safety

Capability	LeanCorpus API	Why it differs
Cancellation-token search	`SearchOptions.CancellationToken`	Uses standard .NET cooperative cancellation throughout segment search.
Per-query memory budget	`SearchOptions.MaxResultBytes`	Caps intermediate result storage.
Explicit partial-result state	`TopDocs.IsPartial`	Records when timeout or budget controls truncate a result.
Indexing queue backpressure	`IndexWriterConfig.MaxQueuedDocs`	Blocks producers when the pending document queue is full.
Segment-count backpressure	`IndexWriterConfig.MergeThrottleSegments`	Blocks writes until background merging reduces the segment count.
First-class schema validation	`IndexSchema`	Enforces field types and required fields at write time.
Per-field stored compression	`FieldCompressionPolicy`	Chooses stored-field compression for each field rather than for the whole codec or segment.

Operations and diagnostics

Capability	LeanCorpus API	Why it differs
Manifest-backed backup and restore	`IndexBackup`	CRC manifests, incremental parent chains, and validated restore are one public workflow.
Pre-open compatibility verdict	`IndexCompatibility`	Reports read and write compatibility before opening an index.
Structured format inspection	`IndexFormatInspector`	Inventories codec roles, versions, checksums, sidecars, and orphaned files.
Staged codec migration	`IndexCodecMigrator`	Plans, rewrites, validates, and publishes format migrations without a full reindex.
OpenTelemetry-oriented tracing	`ActivitySource` name `Rowles.LeanCorpus`	Index, search, backup, and migration operations emit .NET activities.
.NET metrics integration	`MeterMetricsCollector`	Exposes counters and histograms through `System.Diagnostics.Metrics`.
Slow-query and recent-search records	`SlowQueryLog` and `SearchAnalytics`	Bounded in-process diagnostic buffers are provided by the core library.
Structured CLI JSON output	`leancorpus-cli --json`	Check, inspect, compatibility, backup, restore, and migration workflows share machine-readable output.

Search and ranking extensions

Capability	LeanCorpus API	Why it differs
Built-in BM25L and BM25+	`Bm25LSimilarity` and `Bm25PlusSimilarity`	Additional BM25-family models are included rather than requiring a custom similarity.
Additional TF-IDF variants	`TfIdfAugmentedSimilarity`, `TfIdfDoubleNormSimilarity`, and `TfIdfPivotedSimilarity`	Alternative term-frequency and length-normalisation models are built in.
Fixed-bucket numeric histograms	`AggregationType.Histogram`	A direct aggregation request is available in the search API.
Named ranking profiles and query rules	`RankingProfile` and `QueryRuleSet`	Immutable identities bind similarities, bounded ranking stages, filters, boosts, and pins to a result.
Snapshot-bound cursor sessions	`SearchSessionManager`	Versioned cursors remain bound to a retained commit, query, sort, and optional ranking identity.