dtrentin
|
53ce8ab4da
|
feat: per-machine data-item selection (v0.3)
Users pick which data items each machine monitors. Catalog comes from the
protocol probe; selection is opt-in and filters persistence (only selected
items are read/kept/saved). Whole solution green, 138 tests + 2 docker integration.
Domain:
- Machine.MonitoredItemIds (opt-in, empty = monitor nothing); backward-compatible
optional ctor param + With override.
- DataItemDescriptor (protocol-agnostic catalog entry).
- IProtocolDriver.ProbeAsync → full unfiltered item catalog.
MTConnect:
- MtconnectDriver.ProbeAsync (GET /probe → parser → descriptors).
- ReadCurrentAsync filters snapshot items to selected ids (ConnectionState
preserved); factory passes selection into driver.
Persistence:
- machines.MonitoredItemIdsJson column + idempotent ALTER-if-missing migration.
- Repository maps selection (System.Text.Json); round-tripped.
Core:
- IMachineMonitor.ProbeAsync(machine) exposes catalog to the UI via the plugin
factory.
App:
- Config: "Load items" probes the machine, shows a checklist (select all/none),
pre-selects existing choices in edit mode, offline fallback, saves selection.
- Detail: empty-state hint when a machine has no monitored items.
PAUL: v0.3 Phase 3 shipped; Phase 3.1 (theming/UX) next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
2026-07-22 00:28:47 +02:00 |
|
dtrentin
|
cf394bab31
|
refactor: v0.2.1 refinements — HttpClient lifetime, disconnection UX, DataGrid
Post-v0.2 polish (no behavior regressions; 127 tests + 2 docker integration green).
- MTConnect: shared static HttpClient (Timeout=Infinite) + per-request timeout
via linked CancellationTokenSource. Fixes socket-exhaustion risk from
new-HttpClient-per-driver on config edits / dynamic monitor reload.
- Disconnection UX: dashboard + detail show connection state with color
(green/red/grey dot + badge) via StatusKindToBrush converter; LastSeen
timestamp preserved across disconnects. VMs stay framework-agnostic.
- Delete confirm: VM-driven overlay naming the machine (replaces two-state
button) to prevent accidental deletion.
- Dashboard table: switched to DataGrid — aligned columns + Details gets its
own column (fixes row misalignment). Adds Avalonia.Controls.DataGrid 11.3.10.
PAUL: roadmap re-sequenced — v0.3 Data-Item Selection + UX, OPC UA → v0.4,
Fanuc → v0.5, History → v0.6.
Touches: Junction.Protocols.MTConnect, Junction.App, tests, .paul/.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
2026-07-22 00:09:48 +02:00 |
|
dtrentin
|
507753f82e
|
init: scaffold Junction multi-protocol machine monitor (M1 walking skeleton)
Greenfield C# app reading industrial machine data over pluggable protocols.
M1 walking skeleton: MTConnect plugin polls a machine, Avalonia dashboard
shows live last datum. Build 0 warn/0 err on net48 + net8.0 (Linux),
115 unit tests + 2 docker-gated integration tests.
Projects:
- Junction.Domain (netstandard2.0): Result<T>, models, IProtocolDriver,
IMachineRepository, plugin manifest. Zero package deps.
- Junction.Core (netstandard2.0): PollingEngine, PluginLoader
(Assembly.LoadFrom), MachineMonitor, DI extensions.
- Junction.Persistence (netstandard2.0): SqliteMachineRepository (Dapper),
schema, connection factory. Provider-swap seam to SQL Server.
- Junction.Protocols.MTConnect (netstandard2.0): HTTP driver + namespace-
version-agnostic parser (MTConnect 1.7 + 2.0). Runtime plugin.
- Junction.App (net48;net8.0): Avalonia MVVM, live dashboard, NLog, DI root.
- Junction.Tests (net8.0): xUnit + Moq, 117 tests.
- mock/: docker-compose MTConnect agent (ladder99/agent).
Target net48 for Windows 7/8 fleet compatibility. Avalonia pinned 11.3.x.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
2026-07-21 23:32:56 +02:00 |
|