Give semantic, vector and colour members their default in generated C# [patch]

[matt-edmondson]

Fixes #218

What was wrong

CSharpCodeGenerator.DefaultInitialiserFor only had arms for Int, Long, Float, Double, Bool, String and Enum. Validation also accepts a NumberDefault on a Semantic member (reading through to its representation) and on Vector2/3/4 and the colours. For those members the SchemaDefault attribute was written but no initialiser was, so new Body().Mass came out as 0 in C# while the C++ output started it at 2.5. The round-trip tests only read the attribute back, so they didn't catch it.

Change

DefaultInitialiserFor now delegates to DefaultLiteralFor(value, type), which builds the default as an expression of the member's type:

• Semantic: builds the representation's literal with the existing rules, then wraps it in the semantic type's explicit conversion, e.g. (Kilograms)(2.5f). The inner value is parenthesised so a negative default doesn't parse as subtraction. A refinement chain such as Heavy : Kilograms : Float works because Representation() already walks it. A chain that never reaches a real type falls through, as validation already reports that case.
• Vector2/3/4 and ColorRGB/RGBA: the number goes in every component, which matches the C++ side's reading. Examples: new System.Numerics.Vector3(1f, 1f, 1f), new ktsu.Schema.Runtime.Vector2<int>(2, 2), new ktsu.Schema.Runtime.ColorRgba(0.5f, …).
• Existing arms: unchanged, apart from now returning a bare expression that DefaultInitialiserFor wraps in = …;.
• TravelsAsBytes structs: they still get their parameterless constructor, because that check keys off InitialiserFor(member) being non-empty, and the new arms are included in that.

Out of scope: a NumberDefault on a Quantity member still gets no C# initialiser. Constructing a ktsu.Semantics.Quantities value is a separate question from this issue.

Tests

CodeGenerationRoundTripTests.TestSemanticVectorAndColourMembersStartAtTheirDefaults runs once for a class and once for a TravelsAsBytes struct. It compiles a class with these defaulted members:

• a semantic over Float (2.5 and −1.5)
• a semantic over Int
• a refining semantic (Heavy : Kilograms)
• a float Vector3
• an int Vector2
• a ColorRGBA

It then constructs an instance and checks each member's runtime value, not just the attribute.

With the CSharpCodeGenerator.cs change reverted, both rows fail with Expected values to be equal. With the change, the full Schema.Test suite passes on net10.0 (504/504). I couldn't run the net8.0/net9.0 legs because those runtimes aren't installed in this environment. The only dotnet format findings in the changed range are IDE0001 on the two pre-existing Models.Types.String/Enum arms I moved. That qualification style is used throughout the file.

This PR and #226 (for #217) both touch CSharpCodeGenerator.cs, in different methods. The String arm here still calls Quote, so whichever merges second should rebase cleanly.

🤖 Generated with Claude Code

https://claude.ai/code/session_011PMSp3sjbza8Td6fxS2Kjr

---

Generated by Claude Code

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
88.2% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud