perf(ui): shared KPI cache, live-cache-backed alarm summary, coalesced debug renders

This commit is contained in:
Joseph Doherty
2026-08-14 20:53:38 -04:00
parent ee193cd2bb
commit 8c0b36b2fa
19 changed files with 1554 additions and 64 deletions
@@ -0,0 +1,169 @@
using Microsoft.Extensions.DependencyInjection;
using NSubstitute;
using ZB.MOM.WW.ScadaBridge.CentralUI.Services;
using ZB.MOM.WW.ScadaBridge.Commons.Entities.Instances;
using ZB.MOM.WW.ScadaBridge.Commons.Entities.Sites;
using ZB.MOM.WW.ScadaBridge.Commons.Interfaces.Repositories;
using ZB.MOM.WW.ScadaBridge.Commons.Messages.DebugView;
using ZB.MOM.WW.ScadaBridge.Commons.Messages.Streaming;
using ZB.MOM.WW.ScadaBridge.Commons.Types.Enums;
using ZB.MOM.WW.ScadaBridge.Communication;
namespace ZB.MOM.WW.ScadaBridge.CentralUI.Tests.Services;
/// <summary>
/// Unit tests for <see cref="SharedAlarmSummaryService"/> (arch-review WP2.4). The Alarm
/// Summary page polls per circuit; the shared façade must collapse those into ONE per-site
/// fan-out per freshness window, and must widen that window while the live alarm cache is
/// serving the site (where the poll only still supplies the not-reporting list).
/// </summary>
public class SharedAlarmSummaryServiceTests : IDisposable
{
private const int SiteId = 7;
private const string SiteIdentifier = "plant-a";
private static readonly DateTimeOffset T0 = new(2026, 8, 14, 9, 0, 0, TimeSpan.Zero);
private readonly ITemplateEngineRepository _instanceRepo = Substitute.For<ITemplateEngineRepository>();
private readonly ISiteRepository _siteRepo = Substitute.For<ISiteRepository>();
private readonly IInstanceSnapshotClient _snapshotClient = Substitute.For<IInstanceSnapshotClient>();
private readonly FakeLiveCache _liveCache = new();
private readonly ServiceProvider _provider;
private DateTimeOffset _now = T0;
public SharedAlarmSummaryServiceTests()
{
_siteRepo.GetSiteByIdAsync(SiteId, Arg.Any<CancellationToken>())
.Returns(new Site("Plant A", SiteIdentifier) { Id = SiteId });
_instanceRepo.GetInstancesBySiteIdAsync(SiteId, Arg.Any<CancellationToken>())
.Returns(new List<Instance>
{
new("inst-a") { Id = 1, SiteId = SiteId, State = InstanceState.Enabled },
});
_snapshotClient.GetSnapshotAsync(SiteIdentifier, "inst-a", Arg.Any<CancellationToken>())
.Returns(new DebugViewSnapshot(
"inst-a",
Array.Empty<AttributeValueChanged>(),
new[] { new AlarmStateChanged("inst-a", "A-alarm", AlarmState.Active, 500, T0) },
T0));
var services = new ServiceCollection();
services.AddSingleton(_instanceRepo);
services.AddSingleton(_siteRepo);
services.AddSingleton(_snapshotClient);
services.AddScoped<AlarmSummaryService>();
_provider = services.BuildServiceProvider();
}
private SharedAlarmSummaryService CreateSut(TimeSpan liveCacheTtl) =>
new(_provider.GetRequiredService<IServiceScopeFactory>(), _liveCache, liveCacheTtl, () => _now);
[Fact]
public async Task ConcurrentCircuits_ShareOneFanOut()
{
var sut = CreateSut(TimeSpan.FromSeconds(60));
var results = await Task.WhenAll(Enumerable.Range(0, 8).Select(_ => sut.GetSiteAlarmsAsync(SiteId)));
await _instanceRepo.Received(1).GetInstancesBySiteIdAsync(SiteId, Arg.Any<CancellationToken>());
Assert.All(results, r => Assert.Single(r.Alarms));
}
[Fact]
public async Task ColdLiveCache_RefreshesWithinThePageTick()
{
var sut = CreateSut(TimeSpan.FromSeconds(60));
_liveCache.Live = false;
await sut.GetSiteAlarmsAsync(SiteId);
// The page polls every 15s and, while the cache is cold, the poll is its
// authoritative rebuild — so the memo must have expired by then.
_now = T0.AddSeconds(15);
await sut.GetSiteAlarmsAsync(SiteId);
await _instanceRepo.Received(2).GetInstancesBySiteIdAsync(SiteId, Arg.Any<CancellationToken>());
}
[Fact]
public async Task LiveCacheServing_WidensTheWindowToTheReconcileInterval()
{
var sut = CreateSut(TimeSpan.FromSeconds(60));
_liveCache.Live = true;
await sut.GetSiteAlarmsAsync(SiteId);
_now = T0.AddSeconds(30);
await sut.GetSiteAlarmsAsync(SiteId);
// Live deltas own the rows; only the not-reporting list still comes from the
// fan-out, so a 30s-old answer is fine and costs no second fan-out.
await _instanceRepo.Received(1).GetInstancesBySiteIdAsync(SiteId, Arg.Any<CancellationToken>());
_now = T0.AddSeconds(61);
await sut.GetSiteAlarmsAsync(SiteId);
await _instanceRepo.Received(2).GetInstancesBySiteIdAsync(SiteId, Arg.Any<CancellationToken>());
}
[Fact]
public async Task DifferentSites_DoNotShareAMemoSlot()
{
const int otherSite = 8;
_siteRepo.GetSiteByIdAsync(otherSite, Arg.Any<CancellationToken>())
.Returns(new Site("Plant B", "plant-b") { Id = otherSite });
_instanceRepo.GetInstancesBySiteIdAsync(otherSite, Arg.Any<CancellationToken>())
.Returns(new List<Instance>());
var sut = CreateSut(TimeSpan.FromSeconds(60));
await sut.GetSiteAlarmsAsync(SiteId);
await sut.GetSiteAlarmsAsync(otherSite);
await _instanceRepo.Received(1).GetInstancesBySiteIdAsync(SiteId, Arg.Any<CancellationToken>());
await _instanceRepo.Received(1).GetInstancesBySiteIdAsync(otherSite, Arg.Any<CancellationToken>());
}
[Fact]
public void PureMethods_MatchTheDirectImplementation()
{
var sut = CreateSut(TimeSpan.FromSeconds(60));
var direct = new AlarmSummaryService(_instanceRepo, _siteRepo, _snapshotClient);
var alarms = new List<AlarmStateChanged>
{
new("inst-b", "B-alarm", AlarmState.Active, 900, T0),
new("inst-a", "A-alarm", AlarmState.Normal, 100, T0),
};
var shared = sut.BuildFromLiveAlarms(alarms);
var expected = direct.BuildFromLiveAlarms(alarms);
Assert.Equal(
expected.Alarms.Select(r => r.Alarm.AlarmName),
shared.Alarms.Select(r => r.Alarm.AlarmName));
var expectedRollup = direct.ComputeRollup(expected.Alarms);
var sharedRollup = sut.ComputeRollup(shared.Alarms);
Assert.Equal(expectedRollup.TotalActive, sharedRollup.TotalActive);
Assert.Equal(expectedRollup.WorstSeverity, sharedRollup.WorstSeverity);
Assert.Equal(expectedRollup.UnackedCount, sharedRollup.UnackedCount);
Assert.Equal(expectedRollup.CountsByKind, sharedRollup.CountsByKind);
}
public void Dispose() => _provider.Dispose();
/// <summary>Liveness-only stub — the façade consults nothing else on the live cache.</summary>
private sealed class FakeLiveCache : ISiteAlarmLiveCache
{
public bool Live { get; set; }
public IDisposable Subscribe(int siteId, Action onChanged) => new NoOp();
public IReadOnlyList<AlarmStateChanged> GetCurrentAlarms(int siteId) =>
Array.Empty<AlarmStateChanged>();
public bool IsLive(int siteId) => Live;
private sealed class NoOp : IDisposable
{
public void Dispose() { }
}
}
}
@@ -0,0 +1,166 @@
using ZB.MOM.WW.ScadaBridge.CentralUI.Services;
namespace ZB.MOM.WW.ScadaBridge.CentralUI.Tests.Services;
/// <summary>
/// Unit tests for <see cref="SingleFlightMemo{T}"/> — the primitive behind the process-level
/// KPI snapshot cache and the shared Alarm Summary fan-out (arch-review WP2.4). The two
/// properties the whole optimisation rests on are tested here: <b>single-flight</b> (N
/// concurrent circuits asking the same question produce ONE query) and <b>TTL</b> (a query
/// answered inside the freshness window is served from memory, and one outside it is not).
/// </summary>
public class SingleFlightMemoTests
{
private static readonly DateTimeOffset T0 = new(2026, 8, 14, 9, 0, 0, TimeSpan.Zero);
[Fact]
public async Task ConcurrentCallers_ShareOneFlight()
{
var released = new TaskCompletionSource<int>(TaskCreationOptions.RunContinuationsAsynchronously);
var invocations = 0;
var memo = new SingleFlightMemo<int>(TimeSpan.FromSeconds(10), () => T0);
// Ten circuits ask while the first flight is still in progress.
var waiters = Enumerable.Range(0, 10)
.Select(_ => memo.GetAsync(() =>
{
Interlocked.Increment(ref invocations);
return released.Task;
}))
.ToArray();
released.SetResult(42);
var results = await Task.WhenAll(waiters);
Assert.Equal(1, invocations);
Assert.Equal(1, memo.FlightCount);
Assert.All(results, r => Assert.Equal(42, r));
}
[Fact]
public async Task WithinTtl_ServesTheMemoizedValue_WithoutRequerying()
{
var now = T0;
var invocations = 0;
var memo = new SingleFlightMemo<int>(TimeSpan.FromSeconds(8), () => now);
var first = await memo.GetAsync(() => Task.FromResult(++invocations));
now = T0.AddSeconds(7);
var second = await memo.GetAsync(() => Task.FromResult(++invocations));
Assert.Equal(1, first);
Assert.Equal(1, second);
Assert.Equal(1, memo.FlightCount);
}
[Fact]
public async Task AfterTtl_RequeriesOnce()
{
var now = T0;
var invocations = 0;
var memo = new SingleFlightMemo<int>(TimeSpan.FromSeconds(8), () => now);
await memo.GetAsync(() => Task.FromResult(++invocations));
now = T0.AddSeconds(9);
var refreshed = await memo.GetAsync(() => Task.FromResult(++invocations));
Assert.Equal(2, refreshed);
Assert.Equal(2, memo.FlightCount);
}
[Fact]
public async Task ForceRefresh_BypassesAFreshValue()
{
var now = T0;
var invocations = 0;
var memo = new SingleFlightMemo<int>(TimeSpan.FromSeconds(60), () => now);
await memo.GetAsync(() => Task.FromResult(++invocations));
var forced = await memo.GetAsync(() => Task.FromResult(++invocations), forceRefresh: true);
Assert.Equal(2, forced);
Assert.Equal(2, memo.FlightCount);
}
[Fact]
public async Task ForceRefresh_StillJoinsAnInFlightRound()
{
var released = new TaskCompletionSource<int>(TaskCreationOptions.RunContinuationsAsynchronously);
var invocations = 0;
var memo = new SingleFlightMemo<int>(TimeSpan.FromSeconds(60), () => T0);
var first = memo.GetAsync(() =>
{
Interlocked.Increment(ref invocations);
return released.Task;
});
// A burst of operator "Refresh" clicks must not multiply the query.
var forcedA = memo.GetAsync(() => Task.FromResult(-1), forceRefresh: true);
var forcedB = memo.GetAsync(() => Task.FromResult(-1), forceRefresh: true);
released.SetResult(7);
Assert.Equal(7, await first);
Assert.Equal(7, await forcedA);
Assert.Equal(7, await forcedB);
Assert.Equal(1, invocations);
}
[Fact]
public async Task Failure_IsNotMemoized_AndTheNextCallerRetries()
{
var now = T0;
var invocations = 0;
var memo = new SingleFlightMemo<int>(TimeSpan.FromSeconds(60), () => now);
await Assert.ThrowsAsync<InvalidOperationException>(() =>
memo.GetAsync(() =>
{
invocations++;
return Task.FromException<int>(new InvalidOperationException("KPI query failed"));
}));
// Same instant, well inside the TTL — a cached fault would starve the tiles for a
// whole window, so the next caller must produce a new flight.
var recovered = await memo.GetAsync(() => Task.FromResult(++invocations));
Assert.Equal(2, recovered);
Assert.Equal(2, memo.FlightCount);
}
[Fact]
public async Task TtlOverride_AppliesToThatFlight()
{
var now = T0;
var invocations = 0;
var memo = new SingleFlightMemo<int>(TimeSpan.FromSeconds(5), () => now);
// Produced under a 60s override — still fresh 30s later despite the 5s default.
await memo.GetAsync(() => Task.FromResult(++invocations), ttlOverride: TimeSpan.FromSeconds(60));
now = T0.AddSeconds(30);
var second = await memo.GetAsync(() => Task.FromResult(++invocations));
Assert.Equal(1, second);
Assert.Equal(1, memo.FlightCount);
}
[Fact]
public async Task CallerCancellation_DoesNotCancelTheSharedFlight()
{
var released = new TaskCompletionSource<int>(TaskCreationOptions.RunContinuationsAsynchronously);
var memo = new SingleFlightMemo<int>(TimeSpan.FromSeconds(60), () => T0);
using var giveUp = new CancellationTokenSource();
var impatient = memo.GetAsync(() => released.Task, cancellationToken: giveUp.Token);
var patient = memo.GetAsync(() => Task.FromResult(-1));
giveUp.Cancel();
await Assert.ThrowsAnyAsync<OperationCanceledException>(() => impatient);
// The abandoned wait must not have taken the round down with it.
released.SetResult(99);
Assert.Equal(99, await patient);
}
}