Rewrite schema merge for efficiency and referential equality

[kmcginnes]

Description

Rewrites updateSchemaFromEntities and its supporting functions to be more efficient and preserve referential equality at every level of the schema when nothing changes. This reduces unnecessary React re-renders when entities are merged into an already-up-to-date schema.

Changes

• Single-pass merge instead of pre-check + merge — The old code ran shouldUpdateSchemaFromEntities + hasNewPrefixNamespaces to decide if an update was needed, then updateSchemaFromEntities repeated the same work to do the merge. The new code does one merge pass that preserves referential equality when nothing changes, making the pre-check redundant and removing it entirely.

• Merge directly from entities — mergeVertices and mergeEdges work directly from Vertex[]/Edge[] instead of first converting to intermediate VertexTypeConfig[]/EdgeTypeConfig[] via mapVertexToTypeConfigs/mapEdgeToTypeConfig, avoiding throwaway allocations.

• Merge functions return discovered IRIs — mergeVertices and mergeEdges return { configs, newIris } where newIris contains only the type/attribute IRIs that were actually new. This keeps prefix scanning focused on what changed.

• Optimized prefix scanning — mergePrefixes only scans newly-discovered IRIs and entity IDs, rather than re-scanning all existing schema types on every call.

• Fixed activeSchemaSelector setter churn — The setter allocated a new Map unconditionally, even when the update was a no-op (e.g., returning the same value, or deleting a key that was not present). Now it checks newValue === prev before cloning and guards the RESET/undefined path against no-op deletes.

• Added createPrefixTypeConfig factory — Accepts plain strings { prefix, uri, inferred? } and produces the branded RdfPrefix/IriNamespace types, reducing manual as casts.

• Added debug logging — Logs when the schema is updated, when it is already up to date, and when new vertex types, edge types, attributes, or prefixes are discovered.

Benchmarks (70k vertex types, 3 attributes each)

The old code called getSchemaUris on every merge, which iterated all 70k types and 210k attributes to build a Set<string> — even when nothing changed.

Scenario	Old	New	Speedup
No-op (50 entities already in schema)	935ms	6.5ms	144x
10 new types added	1,306ms	6.4ms	203x
Mixed (40 existing + 10 new)	1,293ms	6.9ms	188x
Empty schema, 100 new entities	0.99ms	1.18ms	~1x

The empty schema case is roughly equivalent since there is no existing schema to skip.

New tests

• Multi-label vertex adds all types to schema
• Prefix deduplication stability across incremental updates (country, country2 → country3)
• New attribute namespace on existing type generates a prefix
• New types get prefixes but existing types do not
• schemaAtom not churned when active config has no schema entry
• schemaAtom not churned when entities already match schema
• Vertices/edges/prefixes array references preserved when only the other changes

Validation

• pnpm checks passes (lint, format, types)
• pnpm test passes (1646 tests across 152 files)
• Manually tested by emptying the schema via debug action and verifying incremental rediscovery through search

Check List

• I confirm that my contribution is made under the terms of the Apache 2.0 license.
• I have verified pnpm checks passes with no errors.
• I have verified pnpm test passes with no failures.
• I have covered new added functionality with unit tests if necessary.
• I have updated documentation if necessary.