Dolphin Explorer GitHub

Architecture

DE / PLATFORM SYSTEM

Built to scale from one line to the whole survey.

Dolphin Explorer is organized into five strict layers. The result is a platform where readers, geodesy, rendering, and processing can evolve independently without turning the application into a fragile monolith.

01Strict boundaries

Dependencies point down. Product policy stays out of the rendering and data layers.

02Durable artifacts

Indexed data and processing outputs are preserved with the project across sessions.

03Visible-first loading

The application becomes useful before every byte of a source file is decoded.

System map

Five layers. Clear responsibilities.

Read from the interface down to the shared data model. Higher layers use lower layers; lower layers stay independent of the application above them.

APPLICATION SURFACE↓ DEPENDENCY DIRECTION
  1. 01

    ui

    src/ui

    Self-contained Qt features and the application composition root.

  2. 02

    app

    src/app

    Project state, sources, layers, contacts, import classification, processing services, and workers.

  3. 03

    pipeline

    src/pipeline

    Node registry, graph structure, serialization, dirty tracking, and the topological runner.

  4. 04

    util · io · geo · render

    src/io, src/geo, src/render

    Format probing, readers, indexing, decoding, parsed artifacts, GDAL raster I/O, CRS handling, and display logic.

  5. 05

    core

    src/core

    Header-only survey and spatial data model: pings, traces, navigation, contacts, artifact indexes, raster grids.

SHARED FOUNDATION Business policy belongs in services and coordinators.

User interface

Self-contained features.

Application-wide state and events are coordinated through AppState, ProjectEventBus, and WindowRegistry. src/ui/mainwindow/ is the composition root.

map
2D/3D spatial display, survey geometry, raster surfaces, and overlays
waterfall
Sidescan waterfall rendering, tools, and contact interaction
subbottom
Sub-bottom profile viewing and interpretation
nodegraph
Visual processing-graph editing
processing
Processing setup, progress, and execution
import
Project entry points and survey import setup
metadata
Source and layer metadata presentation
geodesy
Coordinate-system and geospatial workflows
contacts
Contact browsing, reporting, snapshots, and measurements
datalibrary
Project data and layer organization
export
Image, raster, and related export workflows

Repository

Repository layout.

CMake subdirectories are ordered util → core → io → geo → pipeline → render → app → ui.

Tree
Dolphin-Explorer/
|-- cmake/            CMake support and safety checks
|-- resources/        Icons and Qt resources
|-- src/
|   |-- core/         Shared data model
|   |-- util/         General utilities
|   |-- io/           Survey readers, raster I/O, parsed artifacts
|   |-- geo/          CRS and sonar geospatial algorithms
|   |-- render/       Rendering support
|   |-- pipeline/     Processing engine and nodes
|   |-- app/          Projects, layers, services, and workers
|   `-- ui/           Qt UI features and composition
|-- tests/            Unit, integration, smoke, perf tests
|-- CMakeLists.txt    Root CMake configuration
|-- build_mingw.bat   Primary configure/build script
|-- build_quick.bat   Incremental build helper
|-- build.bat         Alternative MSBuild workflow
`-- launch.bat        Runtime deployment and launcher

Conventions

Development conventions.

Dataset identity is based on the normalized absolute path plus a file size and modification-time fingerprint, not a content hash.

  • Use C++20 and the existing Qt and CMake patterns.
  • Preserve the strict dependency direction: io and pipeline never depend on app, and app never depends on ui.
  • Keep business policy out of MainWindow; prefer coordinators or application services.
  • Preserve index-first, visible-first loading. Never require full-file decoding just to activate a layer.
  • Treat .dlpd files as durable artifacts, not temporary caches.
  • Keep unfinished UI actions hidden or disabled instead of shipping clickable placeholders.
  • Keep the default concurrency cap of two for heavy import and decode jobs.
  • Add or update focused tests when changing readers, persistence, processing nodes, geospatial behavior, or UI policy.

From structure to source

Build it. Explore it.

Set up the toolchain and see how the layers work together.

Open setup guide