Source code extraction directive (tree-sitter)

[yumike]

Problem

Documentation that lists values from code (enum members, constants, configuration keys) goes stale when the code changes. Keeping markdown tables in sync with source files is manual and error-prone.

Proposed solution

A directive that extracts structured data from source files using tree-sitter and renders it as a markdown table (or list) at serve time.

Example

Given a Rust enum:

/// HTTP status codes used by the API.
pub enum StatusCode {
    /// Request succeeded
    Ok = 200,
    /// Resource not found
    NotFound = 404,
    /// Internal server error
    InternalError = 500,
}

A directive in markdown:

```sourcetable
file: ../src/status.rs
query: enum_variant
columns:
  - name: Code
    field: value
  - name: Name
    field: name
  - name: Description
    field: comment
```

Would render:

Code	Name	Description
200	Ok	Request succeeded
404	NotFound	Resource not found
500	InternalError	Internal server error

Why tree-sitter

• Language-agnostic — works with Rust, Python, Go, C#, TypeScript, etc.
• Parses real syntax, not fragile regex
• Mature grammars for most languages
• Fits rw's "no build step" philosophy — parse on demand at serve time

Use cases

• Enum reference tables (status codes, error codes, feature flags)
• Configuration key inventories from constants
• CLI flag/subcommand lists extracted from struct definitions