Summary

This week, I focused on DMD’s diagnostic and error reporting infrastructure as part of enabling Language Server Protocol (LSP) integration. The main goal was to produce structured diagnostic data in JSON format that can be easily consumed by editors or tools such as serve-d and code-d.

1. Studied DMD’s Diagnostic System

• Understood how ErrorSink acts as an abstraction layer for all compiler messages (error, warning, deprecation, etc.).
• Traced how diagnostics propagate from parsing and semantic analysis to the sink layer.

2. Analysed Existing Tooling

• Examined D’s existing ecosystem components: serve-d and code-d for LSP support.
• Studied sarif.d and json.d in the DMD codebase.
• Found that sarif.d is designed for static analysis report generation, not for real-time LSP diagnostics.

3. Implemented errorsinkjson.d

• Created a new module errorsinkjson.d defining the class ErrorSinkJson.
• ErrorSinkJson inherits from ErrorSink and overrides all relevant virtual methods.
• Each diagnostic is serialized into JSON with fields like:

  {
      "type": "error",
      "file": "main.d",
      "line": 42,
      "column": 5,
  }
  

Work in : https://github.com/Iskaban10/dmd/tree/lsp

On Monday, 6 October 2025 at 14:10:48 UTC, Iskaban10 wrote:

Thanks for improving these things!

From my perspective this task is crucial for language adoption

Even big and successful languages are improving on this.
Maybe you will find some inspiration from new Python as well

https://realpython.com/python314-error-messages/