Let a construction hold a designated initialiser in the graph

[matt-edmondson]

Closes #52.

ConstructionExpression.Arguments was declared AstSlotKind.Expression, and a MemberInitialiser is not an expression — it names the member a value is for rather than evaluating to one. So Accepts refused it, which made the handling two cases further down unreachable. That case is written with no when child is Expression guard, unlike all five of its neighbours, and Arguments is typed Collection<AstNode> rather than Collection<Expression>: both say the intent was to accept it.

The AST, the serializer and all seven generators handle one in exactly this position — a designated initialiser in C++, an object initialiser in C#, a keyword argument in Python, an object literal in JavaScript. The graph is meant to be the uniform view of the AST, and here it could not express a shape every other projection could.

A kind, not a widening

The obvious fix is to let MemberInitialiser through the Expression arm. That would be wrong, and checking why was the useful part of this:

The slot is shared with CallExpression.Arguments. Every generator reads construction.Arguments for a member initialiser — CSharpGenerator:356, CppGenerator:502, PythonGenerator:228, JavaScriptGenerator:188, RustGenerator:1222, GoGenerator:1603, CGenerator:736 — and none reads a call's. One attached to a call would be written as whatever fell out of the generic dispatch rather than as a named argument.

So AstSlotKind.Element is its own kind and a construction gets its own slot: same name, because the editor labels the pin the same way and the attach cases switch on the name, different rule for what may stand in it. The kinds part exactly where the generators do, and the tests pin both halves — a construction takes one, a call refuses it.

Coverage

• DeclarationSlotsTests — attach, read back, detach; plus the call refusing it.
• AstGraphTests — the whole chain rather than the schema's answer alone, because AstGraph.ValidateConnection is where the editor felt this: drag a link, validate against the slot, and the AST ends up holding the node.

Both were checked against the bug rather than written after it was gone. With the Element arm reverted to IsExpression(candidate), the schema test fails on Accepts and the graph test on result.Success.

850/850 tests pass.

🤖 Generated with Claude Code

https://claude.ai/code/session_01QGCMUrT3jBgmANHNHcPmBf

---

Generated by Claude Code

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud