ScadaBridge/tests/ZB.MOM.WW.ScadaBridge.SiteRuntime.Tests/Scripts/NodeLogCaptureTests.cs

using Akka.Actor;
using Akka.TestKit.Xunit2;
using Microsoft.CodeAnalysis.CSharp.Scripting;
using Microsoft.CodeAnalysis.Scripting;
using Microsoft.Extensions.Logging;
using Microsoft.Extensions.Logging.Abstractions;
using ZB.MOM.WW.Audit;
using ZB.MOM.WW.ScadaBridge.Commons.Messages.Instance;
using ZB.MOM.WW.ScadaBridge.Commons.Messages.ScriptExecution;
using ZB.MOM.WW.ScadaBridge.Commons.Observability;
using ZB.MOM.WW.ScadaBridge.Commons.Types.Audit;
using ZB.MOM.WW.ScadaBridge.Commons.Types.Debugging;
using ZB.MOM.WW.ScadaBridge.Commons.Types.Enums;
using ZB.MOM.WW.ScadaBridge.Commons.Types.Scripts;
using ZB.MOM.WW.ScadaBridge.SiteRuntime.Scripts;

namespace ZB.MOM.WW.ScadaBridge.SiteRuntime.Tests.Scripts;

/// <summary>
/// ScadaBridge#56: while a run is traced, node-log lines that OTHER actors write for it (tagged with
/// its <c>ExecutionId</c>) are copied into its <c>ScriptLog</c> rows as source <c>node</c> — under the
/// run's line cap, through the same collector (and so the same redaction) as every other row — and
/// an untraced run copies nothing.
/// </summary>
public class NodeLogCaptureTests : TestKit, IDisposable
{
    private readonly ScriptExecutionScheduler _scheduler = new(4);
    private readonly SharedScriptLibrary _sharedLibrary = new(
        new ScriptCompilationService(NullLogger<ScriptCompilationService>.Instance),
        NullLogger<SharedScriptLibrary>.Instance);

    /// <summary>The node's Serilog pipeline as the host builds it: LogContext enrichment plus the capture sink.</summary>
    private readonly Serilog.ILogger _node = new Serilog.LoggerConfiguration()
        .MinimumLevel.Information()
        .Enrich.FromLogContext()
        .WriteTo.Sink(new ExecutionLogCaptureSink())
        .CreateLogger();

    /// <inheritdoc />
    void IDisposable.Dispose()
    {
        Shutdown();
        _scheduler.Dispose();
    }

    /// <summary>A Microsoft logger over a Serilog logger (the host's <c>UseSerilog</c> shape), with a category.</summary>
    private sealed class SerilogMel(Serilog.ILogger inner, string category) : ILogger
    {
        private readonly Serilog.ILogger _inner = inner.ForContext("SourceContext", category);

        public IDisposable? BeginScope<TState>(TState state) where TState : notnull => null;

        public bool IsEnabled(LogLevel logLevel) => logLevel >= LogLevel.Information;

        public void Log<TState>(LogLevel logLevel, EventId eventId, TState state, Exception? exception,
            Func<TState, Exception?, string> formatter)
        {
            if (!IsEnabled(logLevel)) return;
            var level = logLevel switch
            {
                LogLevel.Information => Serilog.Events.LogEventLevel.Information,
                LogLevel.Warning => Serilog.Events.LogEventLevel.Warning,
                _ => Serilog.Events.LogEventLevel.Error,
            };
            _inner.Write(level, exception, "{Message:l}", formatter(state, exception));
        }
    }

    /// <summary>
    /// Stands in for the Instance Actor: re-pushes the message's execution id (as the real one does in
    /// AroundReceive), logs a line on ITS thread, and acknowledges the write.
    /// </summary>
    private sealed class LoggingInstanceActor : ReceiveActor
    {
        public LoggingInstanceActor(Serilog.ILogger node)
        {
            var log = node.ForContext("SourceContext", "ZB.MOM.WW.ScadaBridge.SiteRuntime.Actors.InstanceActor");
            Receive<SetStaticAttributeCommand>(cmd =>
            {
                using (ExecutionLogContext.PushFor(cmd))
                {
                    log.Warning("instance actor applied {Attribute} token=hunter2", cmd.AttributeName);
                }

                Sender.Tell(new SetStaticAttributeResponse(
                    cmd.CorrelationId, cmd.InstanceUniqueName, cmd.AttributeName, true, null, DateTimeOffset.UtcNow));
            });
        }
    }

    private static Script<object?> Compile(string code)
    {
        var options = ScriptOptions.Default
            .WithReferences(typeof(object).Assembly, typeof(Enumerable).Assembly, typeof(ScriptGlobals).Assembly)
            .WithImports("System", "System.Collections.Generic", "System.Linq", "System.Threading.Tasks");
        var script = CSharpScript.Create<object?>(code, options, typeof(ScriptGlobals));
        script.Compile();
        return script;
    }

    private ScriptCallResult Run(string body, bool trace, int maxLogLines = 200)
    {
        var caller = CreateTestProbe();
        var instance = Sys.ActorOf(Props.Create(() => new LoggingInstanceActor(_node)));
        ScriptRunLauncher.LaunchScript(
            scriptName: "Runner", instanceName: "Inst1", compiledScript: Compile(body),
            parameters: null, callDepth: 0, instanceActor: instance,
            sharedScriptLibrary: _sharedLibrary,
            options: new SiteRuntimeOptions
            {
                MaxScriptCallDepth = 10, ScriptExecutionTimeoutSeconds = 30, StuckScriptGraceMs = 30000,
                MaxLogLinesPerRun = maxLogLines,
            },
            replyTo: caller.Ref, correlationId: "corr-1", completionTarget: CreateTestProbe().Ref,
            siteCommunicationActor: null,
            logger: new SerilogMel(_node, "ZB.MOM.WW.ScadaBridge.SiteRuntime.Actors.ScriptActor"),
            scope: ScriptScope.Root, runId: 1, healthCollector: null, serviceProvider: null,
            parentExecutionId: null, executionTimeoutSeconds: null, scheduler: _scheduler,
            trace: trace, triggerExtra: null);
        return caller.ExpectMsg<ScriptCallResult>(TimeSpan.FromSeconds(20));
    }

    private static List<(AuditEvent Event, ScriptTraceExtra Extra)> LogRows(ScriptCallResult result)
        => result.TraceEvents!
            .Where(e => e.AsRow().Kind == AuditKind.ScriptLog)
            .Select(e => (e, ScriptTraceExtra.TryParse(e.AsRow().Extra)!))
            .ToList();

    /// <summary>
    /// Cross-actor: the Instance Actor's line, written on its own thread while serving the traced
    /// run's write, comes back inline as a <c>node</c> row of that run, with its category and level.
    /// </summary>
    [Fact]
    public void TracedRun_OtherActorsLine_IsCapturedAsANodeRow()
    {
        var result = Run("await Instance.SetAttribute(\"Setpoint\", \"42\"); return null;", trace: true);

        Assert.True(result.Success, result.ErrorMessage);
        var node = Assert.Single(LogRows(result), r => r.Extra.LogSource == "node");
        Assert.Equal("warn", node.Extra.LogLevel);
        Assert.Equal("ZB.MOM.WW.ScadaBridge.SiteRuntime.Actors.InstanceActor", node.Extra.LogCategory);
        Assert.Contains("instance actor applied \"Setpoint\"", node.Event.AsRow().RequestSummary);
        Assert.Equal(result.ExecutionId, node.Event.AsRow().ExecutionId);
        // The registration ends with the run: a later line for the same id is not captured.
        Assert.False(ExecutionLogCapture.TryGetTarget(result.ExecutionId!.Value, out _));
    }

    /// <summary>
    /// An inline shared script's write is stamped with the SHARED script's own execution id; its
    /// node line still lands in the caller's rows (where the shared script's Log.* lines go).
    /// </summary>
    [Fact]
    public void SharedScriptsWrite_NodeLine_LandsInTheCallersRows()
    {
        Assert.True(_sharedLibrary.CompileAndRegister(
            "WriteHelper", "await Instance.SetAttribute(\"FromShared\", \"1\"); return null;"));

        var result = Run("await Instance.Scripts.CallShared(\"WriteHelper\"); return null;", trace: true);

        Assert.True(result.Success, result.ErrorMessage);
        var node = Assert.Single(LogRows(result), r => r.Extra.LogSource == "node");
        Assert.Contains("FromShared", node.Event.AsRow().RequestSummary);
        Assert.Equal(result.ExecutionId, node.Event.AsRow().ExecutionId);
    }

    /// <summary>An untraced run never registers, so the same line is not captured (and the reply has no rows).</summary>
    [Fact]
    public void UntracedRun_CapturesNothing()
    {
        var result = Run("await Instance.SetAttribute(\"Setpoint\", \"42\"); return null;", trace: false);

        Assert.True(result.Success, result.ErrorMessage);
        Assert.True(result.TraceEvents is null || result.TraceEvents.Count == 0);
    }

    /// <summary>
    /// No double capture: the script's own <c>Log.Info</c> goes to the node logger (tagged with the run's
    /// id) and is a <c>script</c> row exactly once — never a second <c>node</c> row. The launcher's
    /// own failure line is the <c>ScriptRun</c> row, so it is not copied either.
    /// </summary>
    [Fact]
    public void ScriptLogLine_AndLauncherFailureLine_AreNotCopiedAsNodeRows()
    {
        var result = Run("Log.Info(\"hello\"); throw new InvalidOperationException(\"boom\");", trace: true);

        Assert.False(result.Success);
        var rows = LogRows(result);
        var script = Assert.Single(rows);
        Assert.Equal("script", script.Extra.LogSource);
        Assert.Equal("hello", script.Event.AsRow().RequestSummary);
    }

    /// <summary>
    /// Node rows share the run's line cap with the script's own lines: past it they are dropped and
    /// the closing runtime row names the count.
    /// </summary>
    [Fact]
    public void NodeRows_ShareTheRunsLineCap()
    {
        var result = Run(
            "Log.Info(\"one\"); for (var i = 0; i < 3; i++) await Instance.SetAttribute(\"A\" + i, \"1\"); return null;",
            trace: true, maxLogLines: 2);

        Assert.True(result.Success, result.ErrorMessage);
        var rows = LogRows(result);
        Assert.Equal(3, rows.Count); // "one" + 1 node row + the closing row
        Assert.Single(rows, r => r.Extra.LogSource == "node");
        var closing = Assert.Single(rows, r => r.Extra.LogSource == "runtime");
        Assert.Contains("2 further log line(s)", closing.Event.AsRow().RequestSummary);
    }

    /// <summary>
    /// Node rows ride the same inline collector as every other row, so its redactor applies: what Live
    /// shows is what Replay stores.
    /// </summary>
    /// <returns>A task that represents the asynchronous operation.</returns>
    [Fact]
    public async Task NodeRows_PassTheCollectorsRedactor()
    {
        var collector = new FlowTraceCollector(redactor: new TokenRedactor());
        var exec = Guid.NewGuid();
        var ctx = new ScriptRuntimeContext(
            Sys.ActorOf(Props.Create(() => new LoggingInstanceActor(_node))), _sharedLibrary,
            currentCallDepth: 0, maxCallDepth: 10, askTimeout: TimeSpan.FromSeconds(5),
            instanceName: "Inst1", logger: NullLogger.Instance, siteId: "site-a",
            sourceScript: "ScriptActor:Runner", executionId: exec, isTraced: true, traceCollector: collector);

        using (ctx.Log.CaptureNodeLines(exec))
        {
            await ctx.SetAttribute("Setpoint", "42");
        }

        var node = Assert.Single(collector.Complete(), e => e.AsRow().Kind == AuditKind.ScriptLog);
        Assert.DoesNotContain("hunter2", node.DetailsJson);
        Assert.Contains("token=***", node.AsRow().RequestSummary);
    }

    /// <summary>Scrubs <c>hunter2</c> from the details, standing in for the host's body-regex redaction.</summary>
    private sealed class TokenRedactor : IAuditRedactor
    {
        public AuditEvent Apply(AuditEvent rawEvent)
            => rawEvent with { DetailsJson = rawEvent.DetailsJson?.Replace("hunter2", "***", StringComparison.Ordinal) };
    }

    private (ScriptLogger Logger, List<(string Level, string Source, string Message, string? Category)> Rows) Traced(Guid id)
    {
        var rows = new List<(string, string, string, string?)>();
        var logger = new ScriptLogger(
            new SerilogMel(_node, "ScriptActor"),
            (level, source, message, _, _, category, _) => { lock (rows) rows.Add((level, source, message, category)); },
            maxLines: 200, "Inst1", "Runner", id);
        return (logger, rows);
    }

    /// <summary>An untraced logger never touches the registry.</summary>
    [Fact]
    public void UntracedLogger_CaptureNodeLines_DoesNotRegister()
    {
        var id = Guid.NewGuid();
        var untraced = new ScriptLogger(NullLogger.Instance, emit: null, maxLines: 200, "Inst1", "Runner", id);

        using (untraced.CaptureNodeLines(id))
        {
            Assert.False(ExecutionLogCapture.TryGetTarget(id, out _));
        }
    }

    /// <summary>FlushDropped (every run-end path) closes the copy, and a capture opened after it is refused.</summary>
    [Fact]
    public void FlushDropped_ClosesTheCopy()
    {
        var id = Guid.NewGuid();
        var (logger, rows) = Traced(id);
        var capture = logger.CaptureNodeLines(id);
        Assert.True(ExecutionLogCapture.TryGetTarget(id, out _));

        logger.FlushDropped();

        Assert.False(ExecutionLogCapture.TryGetTarget(id, out _));
        using (logger.CaptureNodeLines(id))
        using (ExecutionLogContext.Push(id))
        {
            _node.Warning("after the run");
        }

        Assert.Empty(rows);
        capture.Dispose();
    }

    /// <summary>
    /// The runtime decorator's forward is suppressed: a runtime warning logged while traced is one
    /// <c>runtime</c> row, never a second <c>node</c> row — while a genuine other-actor line still is one.
    /// </summary>
    [Fact]
    public void RuntimeLine_IsNotCopiedAgain_ButAnotherLoggersLineIs()
    {
        var id = Guid.NewGuid();
        var (logger, rows) = Traced(id);
        var runtime = new TraceCapturingLogger(new SerilogMel(_node, "ScriptActor"), logger);

        using (logger.CaptureNodeLines(id))
        using (ExecutionLogContext.Push(id))
        {
            runtime.LogWarning("runtime warning");
            _node.ForContext("SourceContext", "Acme.Other").Warning("other actor");
        }

        Assert.Equal(
            [("warn", "runtime", "runtime warning", (string?)null), ("warn", "node", "other actor", "Acme.Other")],
            rows);
    }
}