Feature: AI-powered implementation stub generation from MAP/CLASS declarations

[msarson]

Overview

When a procedure is declared in a MAP or a CLASS interface, the implementation body doesn't exist yet. The extension already knows the full signature, parameter types, return type, CLASS hierarchy, and surrounding scope. This is ideal context for AI-powered stub generation — producing a skeleton implementation that the developer can fill in.

Example

Given a MAP declaration:

KanbanProcess_Kanban  PROCEDURE(KanbanWrapperClass pKanban)

Generate:

ThisWindow.KanbanProcess_Kanban  PROCEDURE(KanbanWrapperClass pKanban)
  CODE
    ! TODO: Implement KanbanProcess_Kanban
    RETURN

Or for a CLASS with an interface, generate all required method stubs at once.

Why the Clarion Extension Is Well-Positioned

The LSP already has everything needed as AI context:

• Full procedure signature (parameters, return type, attributes like PROC, VIRTUAL, DERIVED)
• CLASS hierarchy — knows which interface methods are required
• Surrounding scope — knows what variables/types are in play
• DocumentStructure scope tree — knows where the implementation section should go
• SolutionManager — knows the full project structure

This makes AI generation significantly more accurate than generic Copilot completions, which have no Clarion-specific knowledge.

Use Cases

1. MAP procedure stub — declare in MAP, generate implementation body in the CODE section
2. CLASS interface stub — declare a CLASS that implements an interface, generate all required method stubs at once
3. Signature-driven generation — infer likely body from procedure name + parameter types + return type

Implementation Approaches

These are not mutually exclusive:

• VS Code Chat Participant (@clarion) — user asks @clarion generate stubs for ThisWindow and the participant uses LSP knowledge to produce them
• VS Code Language Model Tool — extension registers a clarion_generateStub tool that Copilot can call during inline completions
• Code Action — a lightbulb/quick-fix on a MAP declaration that offers "Generate implementation stub"

The Code Action approach is the most discoverable and requires no separate AI integration — it could use the VS Code LM API (vscode.lm.selectChatModels) to call the model with the signature as context.

Relationship to Other Issues

• Builds on the existing GoTo Implementation infrastructure (MemberLocatorService, SymbolFinderService)
• Would benefit from Feature: Include dependency graph for accurate cross-file symbol resolution #52 (include dependency graph) for accurate CLASS hierarchy resolution