panproto-project

Assembles a directory of source files into a single unified schema graph.

What it does

When you have a project with dozens of files across multiple languages, each file is its own island of type information. panproto-project scans a directory tree, parses every file it recognizes (TypeScript, Python, Rust, and 245 other languages via tree-sitter), and joins the results into one schema that spans the whole project. Files that are not a recognized language are stored as raw text or binary nodes so nothing is dropped.

The joining works by prefixing each file's vertex names with its path: a function called add in src/utils.ts becomes src/utils.ts::add, so it never collides with an add in lib/math.py. After prefixing, the crate walks import statements and emits cross-file edges wherever one file references a name exported by another. The result is one schema where intra-file structure and inter-file dependencies are both visible.

An optional panproto.toml manifest lets you exclude directories (like node_modules or build), override which parser to use for a subtree, and configure per-package settings. An incremental parsing cache (keyed on file path, mtime, size, and content hash) skips re-parsing files that have not changed since the last build.

Quick example

use panproto_project::{ProjectBuilder, ProjectConfig};
use std::path::Path;

// Scan a directory and build a unified schema.
let mut builder = ProjectBuilder::new();
builder.add_directory(Path::new("my-project"))?;
let project = builder.build()?;

println!("{} files, {} vertices",
    project.file_map.len(),
    project.schema.vertices.len());

// Which protocol parsed each file?
for (path, proto) in &project.protocol_map {
    println!("{}: {}", path.display(), proto);
}

API overview

License

MIT