Skip to main content

Architecture

This section explains how DependencyAtlas is built internally.

Read This Section If​

  • you maintain the backend or frontend
  • you want to understand the JuliaSyntax-first pipeline
  • you need to know where identity, JET views, and query semantics are computed

Recommended Path

Overview

Start with the high-level backend/frontend map and the current design seam.

Pipeline

See the end-to-end path from Julia source to snapshots and query-time projection.

DepAtlasSource Internals

See the current phase boundaries, lowering files, apply layers, and maintenance rules inside the source-first engine.

Lightweight Static Dispatch

Understand the deliberately small static dispatch layer used by da_source for local narrowing.

Engine Integrations

Understand the default DepAtlasSource path and the on-demand JET layer.

Frontend and Store

Understand runtime storage, query caching, and the two frontend surfaces.

Current Backend Flow

01

Syntax Facts

JuliaSyntax scans files and emits raw syntax-level facts.

02

Lowering

Ownership, body ranges, and local bindings are lowered into typed structure and call facts.

03

Base Graph

Typed structure facts build canonical file/module/method identity.

04

Enrichment

JET is available on demand as an optional compile-time evidence layer.

05

Query

Projection, pagination, reachability, and node appearance are computed at query time.

What You Will Learn​

  • how syntax facts become typed structure and call facts
  • how DepAtlasSource is currently split into scan, lowering, orchestration, and apply phases
  • where canonical graph identity is defined
  • how the default engine and the optional JET layer fit together without competing for graph identity
  • how snapshots, projection, and the UI runtime fit together