Increase code coverage in unit test pipeline

[aaronburtle]

Why make this change?

Related to: #3648
Closes #3717

What is this change?

Improves branch coverage by adding fast, database-free unit tests for pure-logic classes that were previously at 0% or very low coverage. No product/runtime behavior is changed.

Adds ~300 unit test methods (500+ cases including [DataRow] variations) across 35 files, all running in the existing unit-test job

• Config JSON converters & model: EmbeddingsOptions, DmlToolsConfig, AKVRetryPolicy, AzureKeyVault, Compression, Bool, RuntimeCache/Level2 converters; EmbeddingsEndpointOptions role resolution; DeserializationVariableReplacementSettings (env + local @akv file resolution).
• MCP tooling: non-DB validation paths for the built-in create/read/update/delete_record tools (tool-disabled, invalid args, metadata failures); McpArgumentParser, McpJsonHelper, McpResponseBuilder, McpErrorHelpers, McpAuthorizationHelper, McpMetadataHelper, and McpTelemetryHelper.
• Core helpers/parsers/extensions: TypeHelper, EnumExtensions, Utf8JsonReaderExtensions, HttpStatusCodeExtensions, ProductInfo, and MergeJsonProvider.
• GraphQL builder: GraphQLNaming, GraphQLUtils, GraphQLStoredProcedureBuilderHelpers, SchemaConverter type mapping, and EdmModelBuilder.
• REST / OpenAPI & request validation: OpenApiDocumentor, and RequestValidator (static + instance) paths.
• Runtime config validation: RuntimeConfigValidator and RuntimeConfigProvider logic.
• SQL pure-logic helpers: SqlPaginationUtil and SqlResponseHelpers (mutation) paths.

How was this tested?

• Integration Tests
• Unit Tests

Sample Request(s)

N/A

[Copilot]

Pull request overview

This PR increases unit-test/branch coverage by adding database-free tests for pure-logic helpers, config JSON converters, and MCP tooling validation paths, without changing product/runtime behavior.

Changes:

• Adds unit tests for multiple config JSON converters (embeddings, cache, compression, Key Vault, DML tools, bool parsing, etc.).
• Adds unit tests for core helper utilities/parsers (TypeHelper, enum parsing, JSON reader extensions, entity source parsing, merge-json).
• Adds non-DB unit tests for MCP utilities and early-return validation/error paths in built-in DML tools.

Reviewed changes

Copilot reviewed 23 out of 23 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
src/Service.Tests/UnitTests/Utf8JsonReaderExtensionsTests.cs	Covers string/null/non-string token handling for Utf8JsonReaderExtensions.DeserializeString.
src/Service.Tests/UnitTests/TypeHelperTests.cs	Exercises TypeHelper type mappings and GraphQL value-node conversions.
src/Service.Tests/UnitTests/RuntimeCacheOptionsConverterTests.cs	Verifies runtime cache (L1/L2) JSON read/write behavior and validation.
src/Service.Tests/UnitTests/ProductInfoTests.cs	Tests ProductInfo version formatting and user-agent env-var behavior.
src/Service.Tests/UnitTests/MergeJsonProviderTests.cs	Tests JSON merge semantics for objects/arrays/type mismatches/null overrides.
src/Service.Tests/UnitTests/HttpStatusCodeExtensionsTests.cs	Validates HttpStatusCodeExtensions.IsClientError classification.
src/Service.Tests/UnitTests/EnumExtensionsTests.cs	Tests enum parsing incl. EnumMember values and error messaging.
src/Service.Tests/UnitTests/EntitySourceNamesParserTests.cs	Covers parsing/validation for schema.table inputs with bracket rules.
src/Service.Tests/UnitTests/EmbeddingsOptionsConverterTests.cs	Tests embeddings options converter read/write, defaults, and invalid inputs.
src/Service.Tests/UnitTests/EmbeddingsEndpointOptionsTests.cs	Verifies embeddings endpoint role-resolution and role-check logic.
src/Service.Tests/UnitTests/DmlToolsConfigConverterTests.cs	Tests boolean/object forms for MCP DML tool configuration conversion.
src/Service.Tests/UnitTests/DeserializationVariableReplacementSettingsTests.cs	Covers env-var replacement and local @akv file resolution logic.
src/Service.Tests/UnitTests/CompressionOptionsConverterTests.cs	Tests compression options JSON read/write, defaults, and invalid values.
src/Service.Tests/UnitTests/BoolJsonConverterTests.cs	Covers bool converter parsing of string/number representations and serialization.
src/Service.Tests/UnitTests/AzureKeyVaultOptionsConverterTests.cs	Tests AKV options converter behavior incl. retry policy and validation.
src/Service.Tests/UnitTests/AKVRetryPolicyOptionsConverterTests.cs	Tests AKV retry policy converter defaults, validation, and serialization.
src/Service.Tests/Mcp/McpTelemetryHelperTests.cs	Tests MCP telemetry helper operation inference and exception→code mapping/wrapper behavior.
src/Service.Tests/Mcp/McpResponseBuilderTests.cs	Tests MCP response builder success/error payloads and IActionResult JSON extraction.
src/Service.Tests/Mcp/McpJsonHelperTests.cs	Tests JSON value conversion and engine-result extraction helpers.
src/Service.Tests/Mcp/McpErrorHelpersTests.cs	Verifies standardized MCP error response shapes/messages.
src/Service.Tests/Mcp/McpAuthorizationHelperTests.cs	Tests authorization helper behavior with mocked authorization resolver/HTTP context.
src/Service.Tests/Mcp/McpArgumentParserTests.cs	Tests MCP argument parsing/validation for built-in tools and execute parameters.
src/Service.Tests/Mcp/BuiltInDmlToolValidationTests.cs	Adds non-DB tests for built-in DML tool early validation/tool-disabled/metadata-failure branches.

[aaronburtle]

@copilot can you review this again please, new tests have been added for additional coverage in the unit test pipeline.

[Copilot]

@copilot can you review this again please, new tests have been added for additional coverage in the unit test pipeline.

Reviewed the new tests. Build succeeds with 0 errors/warnings and all 151 new tests pass locally. A few observations on the additions:

OpenApiDocumentorUnitTests – The new tests thoroughly cover REST-disabled, read-only permissions, stored procedure, view, multi-entity, custom path, non-strict request body, and base-route scenarios. All look correct.

RuntimeConfigValidatorUnitTests – Good coverage of ValidateRestURI, ValidateGraphQLURI, ValidateMcpUri, ValidateAppInsightsTelemetryConnectionString, ValidateAzureLogAnalyticsAuth, ValidateFileSinkPath, ValidateEmbeddingsOptions, and ValidateGlobalEndpointRouteConfig. Assertions align with the validator's documented behavior.

RequestValidatorInstanceUnitTests – Covers CheckFirstValidity, ValidateRequestContext, ValidateInsertRequestContext (valid, missing-required, null-non-nullable, strict/non-strict, read-only), ValidateStoredProcedureRequestContext, and ValidateEntity. All look correct.

SqlResponseHelpersMutationUnitTests – Tests for OkResponse, OkMutationResponse, ConstructPrimaryKeyRoute, ConstructOkMutationResponse, and ConstructCreatedResultResponse look sound.

EdmModelBuilderTests – Correctly verifies that stored procedures and linking entities are skipped, and that a table entity creates an EDM entity type with the right key and property count.

McpResponseBuilderTests (from first commit) – The GetJsonValue_Number_PreservesNumericValue test now uses Convert.ToDecimal to compare, which correctly handles both long (from TryGetInt64) and double return paths without being type-specific.

No issues found — the tests are well-structured and appropriate for a coverage-improvement PR.

[aaronburtle]

@copilot So now that we are considering looking at the union of all of the pipelines in order to calculate the real code coverage, from that position, did our changes to the unit test add any redundant tests or are the increased branching exclusive to the unit test pipeline?

[Copilot]

@copilot So now that we are considering looking at the union of all of the pipelines in order to calculate the real code coverage, from that position, did our changes to the unit test add any redundan...

Good question. Looking at the union of all pipeline coverage, the new tests are almost entirely exclusive — here's the breakdown:

How the pipelines partition tests:
The unit test pipeline runs everything with TestCategory!=CosmosDb_NoSql&TestCategory!=MsSql&TestCategory!=PostgreSql&TestCategory!=MySql&TestCategory!=DwSql (see build-pipelines.yml:Run Unit Tests). None of the new test classes carry a database [TestCategory], so they only execute in the unit test pipeline and are never redundantly run in the MsSql/PostgreSql/MySQL/CosmosDB pipelines.

Code-path coverage — likely exclusive to the unit test pipeline:
Most new branches target paths integration tests structurally cannot reach:

• Config converter error branches — integration tests always load valid configs; the converters' malformed-JSON, wrong-type, and null-handling branches are never hit.
• MCP tool early-return paths — disabled-tool, invalid-argument, and metadata-failure short-circuits are not exercised by any happy-path integration scenario.
• RuntimeConfigValidator / RequestValidator error branches — invalid URIs, missing telemetry auth, missing required body fields, writes to read-only columns, etc. Integration tests exercise only successful validation paths.
• OpenApiDocumentor conditional branches — REST-disabled entity skipping, read-only-permission operation exclusion, non-strict request body variations; integration tests generate docs for normal, fully-configured entities.
• JSON converter edge cases — things like AKVRetryPolicy, Compression, Bool, and RuntimeCacheOptions converters have branches for unrecognized values or missing keys that valid integration-test configs never trigger.

Code-path coverage — some overlap with integration tests possible:
A small set of utility happy-path branches in TypeHelper (for common CLR types present in integration-test schemas), standard SqlPaginationUtil cursor logic, and the success paths in SqlResponseHelpers may already be touched by MsSql/PostgreSql integration runs. These account for a minority of the new methods and, even if already covered by integration tests, the unit tests still serve as fast isolated regression protection.

Bottom line: When taking the union, the bulk of the new branch coverage is net-new (error paths, edge cases, and configurations that integration tests are not designed to exercise). The small amount of potential overlap is in utility function happy paths, not the targeted negative/edge-case branches the PR was designed to improve.