ScadaBridge/tests/ZB.MOM.WW.ScadaBridge.Host.Tests/SiteLocalDbLegacyMigratorTests.cs

using Microsoft.Data.Sqlite;
using Microsoft.Extensions.Configuration;
using Microsoft.Extensions.DependencyInjection;
using ZB.MOM.WW.LocalDb;
using ZB.MOM.WW.ScadaBridge.Host;

namespace ZB.MOM.WW.ScadaBridge.Host.Tests;

/// <summary>
/// LocalDb Phase 1 (Task 6) — the one-time copy of the pre-Phase-1 site databases into
/// the consolidated one.
/// </summary>
/// <remarks>
/// This is a data migration that runs during host startup, so its failure modes are
/// expensive: a duplicate-on-rerun bug silently doubles a site's event history, and a
/// half-committed copy leaves an operator with no clean state to recover to. Every bullet
/// of the plan's semantics gets its own test.
/// </remarks>
public class SiteLocalDbLegacyMigratorTests : IDisposable
{
    private readonly string _root;
    private readonly List<ServiceProvider> _providers = [];

    public SiteLocalDbLegacyMigratorTests()
    {
        _root = Path.Combine(Path.GetTempPath(), $"localdb-migrator-{Guid.NewGuid():N}");
        Directory.CreateDirectory(_root);
    }

    public void Dispose()
    {
        foreach (var provider in _providers)
        {
            try { provider.Dispose(); } catch { /* best effort */ }
        }

        try { Directory.Delete(_root, recursive: true); } catch { /* best effort */ }
        GC.SuppressFinalize(this);
    }

    private string Path_(string name) => System.IO.Path.Combine(_root, name);

    /// <summary>A consolidated database with the tables created and registered, as the host has it.</summary>
    /// <remarks>
    /// <c>sf_messages</c> is registered here even though production <c>OnReady</c> does not
    /// register it until the Task 14 cutover. The oplog assertion below needs live capture
    /// triggers to mean anything, and the property under test — that the migrator runs after
    /// registration — is the same either way. The corollary is a real constraint on Task 14:
    /// <c>Migrate</c> must stay the LAST call in <c>OnReady</c>, after every registration.
    /// </remarks>
    private ILocalDb CreateConsolidated(IConfiguration config)
    {
        var provider = new ServiceCollection()
            .AddZbLocalDb(config, db =>
            {
                using var connection = db.CreateConnection();
                ZB.MOM.WW.ScadaBridge.SiteRuntime.Tracking.OperationTrackingSchema.Apply(connection);
                ZB.MOM.WW.ScadaBridge.SiteEventLogging.SiteEventLogSchema.Apply(connection);
                ZB.MOM.WW.ScadaBridge.StoreAndForward.StoreAndForwardSchema.Apply(connection);
                db.RegisterReplicated("OperationTracking");
                db.RegisterReplicated("site_events");
                db.RegisterReplicated("sf_messages");
            })
            .BuildServiceProvider();
        _providers.Add(provider);

        return provider.GetRequiredService<ILocalDb>();
    }

    private IConfiguration Config(
        string? trackingPath = null,
        string? eventsPath = null,
        string nodeName = "node-a",
        string? storeAndForwardPath = null)
    {
        var values = new Dictionary<string, string?>
        {
            ["LocalDb:Path"] = Path_("consolidated.db"),
            ["ScadaBridge:Node:NodeName"] = nodeName,

            // Always pinned inside the test's own directory, even when a test does not care
            // about it. Left unset, the resolver falls back to the CWD-relative code default
            // "./data/store-and-forward.db" — and the test run's CWD is the test binary's
            // output directory, so an unlucky run could migrate (and RENAME) a real file.
            ["ScadaBridge:StoreAndForward:SqliteDbPath"] =
                storeAndForwardPath ?? Path_("absent-store-and-forward.db"),
        };

        if (trackingPath is not null)
            values["ScadaBridge:OperationTracking:ConnectionString"] = $"Data Source={trackingPath}";
        if (eventsPath is not null)
            values["ScadaBridge:SiteEventLog:DatabasePath"] = eventsPath;

        return new ConfigurationBuilder().AddInMemoryCollection(values).Build();
    }

    /// <summary>Seeds a legacy store-and-forward file with the CURRENT column set.</summary>
    private static void SeedLegacyStoreAndForward(string path, params string[] ids)
    {
        using var connection = new SqliteConnection($"Data Source={path}");
        connection.Open();
        ZB.MOM.WW.ScadaBridge.StoreAndForward.StoreAndForwardSchema.Apply(connection);

        foreach (var id in ids)
        {
            using var cmd = connection.CreateCommand();
            cmd.CommandText = """
                INSERT INTO sf_messages (
                    id, category, target, payload_json, retry_count, max_retries,
                    retry_interval_ms, created_at, status, execution_id)
                VALUES ($id, 0, 'ERP', '{"order":1}', 0, 50, 30000, $now, 0, 'exec-1');
                """;
            cmd.Parameters.AddWithValue("$id", id);
            cmd.Parameters.AddWithValue("$now", DateTime.UtcNow.ToString("o"));
            cmd.ExecuteNonQuery();
        }
    }

    private static void SeedLegacyTracking(string path, params string[] ids)
    {
        using var connection = new SqliteConnection($"Data Source={path}");
        connection.Open();
        ZB.MOM.WW.ScadaBridge.SiteRuntime.Tracking.OperationTrackingSchema.Apply(connection);

        foreach (var id in ids)
        {
            using var cmd = connection.CreateCommand();
            cmd.CommandText = """
                INSERT INTO OperationTracking (
                    TrackedOperationId, Kind, TargetSummary, Status, RetryCount,
                    CreatedAtUtc, UpdatedAtUtc)
                VALUES ($id, 'ApiCallCached', 'ERP.GetOrder', 'Submitted', 0, $now, $now);
                """;
            cmd.Parameters.AddWithValue("$id", id);
            cmd.Parameters.AddWithValue("$now", DateTime.UtcNow.ToString("o"));
            cmd.ExecuteNonQuery();
        }
    }

    /// <summary>Seeds the LEGACY event schema: an autoincrement integer id, not a GUID.</summary>
    private static void SeedLegacyEvents(string path, int count)
    {
        using var connection = new SqliteConnection($"Data Source={path}");
        connection.Open();

        using (var ddl = connection.CreateCommand())
        {
            ddl.CommandText = """
                CREATE TABLE IF NOT EXISTS site_events (
                    id INTEGER PRIMARY KEY AUTOINCREMENT,
                    timestamp TEXT NOT NULL, event_type TEXT NOT NULL, severity TEXT NOT NULL,
                    instance_id TEXT, source TEXT NOT NULL, message TEXT NOT NULL, details TEXT
                );
                """;
            ddl.ExecuteNonQuery();
        }

        for (var i = 0; i < count; i++)
        {
            using var cmd = connection.CreateCommand();
            cmd.CommandText = """
                INSERT INTO site_events (timestamp, event_type, severity, source, message)
                VALUES ($ts, 'script', 'Info', 'src', $msg);
                """;
            cmd.Parameters.AddWithValue("$ts", DateTimeOffset.UtcNow.AddMinutes(-i).ToString("o"));
            cmd.Parameters.AddWithValue("$msg", $"legacy message {i}");
            cmd.ExecuteNonQuery();
        }
    }

    private static long CountRows(ILocalDb db, string table)
    {
        using var connection = db.CreateConnection();
        using var cmd = connection.CreateCommand();
        cmd.CommandText = $"SELECT COUNT(*) FROM {table}";
        return (long)cmd.ExecuteScalar()!;
    }

    private static List<string> SelectIds(ILocalDb db, string table, string idColumn)
    {
        using var connection = db.CreateConnection();
        using var cmd = connection.CreateCommand();
        cmd.CommandText = $"SELECT {idColumn} FROM {table} ORDER BY {idColumn}";
        using var reader = cmd.ExecuteReader();

        var ids = new List<string>();
        while (reader.Read()) ids.Add(reader.GetString(0));
        return ids;
    }

    [Fact]
    public void MissingLegacyFiles_AreANoOp()
    {
        // The expected case on the docker rig: neither legacy key is set anywhere, and
        // the CWD-relative defaults point outside the data volume, so there is usually
        // nothing there at all. That must be a clean no-op, not a startup failure.
        var config = Config(Path_("absent-tracking.db"), Path_("absent-events.db"));
        var db = CreateConsolidated(config);

        SiteLocalDbLegacyMigrator.Migrate(db, config);

        Assert.Equal(0, CountRows(db, "OperationTracking"));
        Assert.Equal(0, CountRows(db, "site_events"));
    }

    [Fact]
    public void LegacyTrackingRows_AreCopiedAndTheFileIsRenamed()
    {
        var trackingPath = Path_("site-tracking.db");
        SeedLegacyTracking(trackingPath, "op-1", "op-2", "op-3");
        var config = Config(trackingPath, Path_("absent-events.db"));
        var db = CreateConsolidated(config);

        SiteLocalDbLegacyMigrator.Migrate(db, config);

        Assert.Equal(3, CountRows(db, "OperationTracking"));
        Assert.Equal(["op-1", "op-2", "op-3"], SelectIds(db, "OperationTracking", "TrackedOperationId"));

        // Renamed, not deleted — the operator can still inspect the original.
        Assert.False(File.Exists(trackingPath));
        Assert.True(File.Exists(trackingPath + ".migrated"));
    }

    [Fact]
    public void LegacyEvents_GetDeterministicIds_NotFreshGuids()
    {
        // The whole reason a rerun cannot duplicate. Fresh GUIDs would make
        // INSERT OR IGNORE useless and double the history on every retry.
        var eventsPath = Path_("site_events.db");
        SeedLegacyEvents(eventsPath, count: 3);
        var config = Config(Path_("absent-tracking.db"), eventsPath, nodeName: "node-b");
        var db = CreateConsolidated(config);

        SiteLocalDbLegacyMigrator.Migrate(db, config);

        Assert.Equal(
            ["mig-node-b-1", "mig-node-b-2", "mig-node-b-3"],
            SelectIds(db, "site_events", "id"));
    }

    [Fact]
    public void SecondBoot_IsANoOp_BecauseTheFileIsAlreadyMarkedMigrated()
    {
        var trackingPath = Path_("site-tracking.db");
        var eventsPath = Path_("site_events.db");
        SeedLegacyTracking(trackingPath, "op-1", "op-2");
        SeedLegacyEvents(eventsPath, count: 2);
        var config = Config(trackingPath, eventsPath);
        var db = CreateConsolidated(config);

        SiteLocalDbLegacyMigrator.Migrate(db, config);
        SiteLocalDbLegacyMigrator.Migrate(db, config);

        Assert.Equal(2, CountRows(db, "OperationTracking"));
        Assert.Equal(2, CountRows(db, "site_events"));
    }

    [Fact]
    public void RerunAgainstAnUnrenamedFile_DoesNotDuplicate()
    {
        // Simulates the crash window: the copy committed but the rename did not. On the
        // next boot the migrator sees the legacy file again and re-copies it. Both tables
        // must absorb that with no duplicates — tracking via its natural TEXT key, events
        // via the deterministic mig- id.
        var trackingPath = Path_("site-tracking.db");
        var eventsPath = Path_("site_events.db");
        SeedLegacyTracking(trackingPath, "op-1", "op-2");
        SeedLegacyEvents(eventsPath, count: 2);
        var config = Config(trackingPath, eventsPath);
        var db = CreateConsolidated(config);

        SiteLocalDbLegacyMigrator.Migrate(db, config);

        // Undo the rename, exactly as a crash between commit and rename would leave it.
        File.Move(trackingPath + ".migrated", trackingPath);
        File.Move(eventsPath + ".migrated", eventsPath);

        SiteLocalDbLegacyMigrator.Migrate(db, config);

        Assert.Equal(2, CountRows(db, "OperationTracking"));
        Assert.Equal(2, CountRows(db, "site_events"));
    }

    [Fact]
    public void MigratedRows_EnterTheOplog_SoTheyActuallyReplicate()
    {
        // The reason the migrator runs AFTER RegisterReplicated. Capture is trigger-based:
        // migrate before registration and the rows are invisible to the peer forever, with
        // no error anywhere. Asserting on the oplog is the only way to catch that ordering
        // being reversed — every other test here would still pass.
        var trackingPath = Path_("site-tracking.db");
        SeedLegacyTracking(trackingPath, "op-1", "op-2");
        var config = Config(trackingPath, Path_("absent-events.db"));
        var db = CreateConsolidated(config);

        SiteLocalDbLegacyMigrator.Migrate(db, config);

        using var connection = db.CreateConnection();
        using var cmd = connection.CreateCommand();
        cmd.CommandText =
            "SELECT COUNT(*) FROM __localdb_oplog WHERE table_name = 'OperationTracking'";
        Assert.Equal(2L, (long)cmd.ExecuteScalar()!);
    }

    [Fact]
    public void SfMessages_AreCopiedFromTheLegacyFile()
    {
        // Phase 2 (Task 8). Unlike the two Phase 1 files, the store-and-forward default path
        // is inside the data volume, so on a real deployment this actually moves data:
        // undelivered messages that a lost migration would silently discard.
        var sfPath = Path_("store-and-forward.db");
        SeedLegacyStoreAndForward(sfPath, "msg-1", "msg-2", "msg-3");
        var config = Config(storeAndForwardPath: sfPath);
        var db = CreateConsolidated(config);

        SiteLocalDbLegacyMigrator.Migrate(db, config);

        Assert.Equal(3, CountRows(db, "sf_messages"));
        Assert.Equal(["msg-1", "msg-2", "msg-3"], SelectIds(db, "sf_messages", "id"));

        // Renamed, not deleted — same contract as the Phase 1 files.
        Assert.False(File.Exists(sfPath));
        Assert.True(File.Exists(sfPath + ".migrated"));
    }

    [Fact]
    public void SfMessages_Migration_IsIdempotent_WhenRerunAfterACrashBeforeRename()
    {
        // The crash window: copy committed, rename did not. sf_messages.id is a natural
        // TEXT key, so INSERT OR IGNORE absorbs the re-copy with no id synthesis needed.
        var sfPath = Path_("store-and-forward.db");
        SeedLegacyStoreAndForward(sfPath, "msg-1", "msg-2");
        var config = Config(storeAndForwardPath: sfPath);
        var db = CreateConsolidated(config);

        SiteLocalDbLegacyMigrator.Migrate(db, config);
        File.Move(sfPath + ".migrated", sfPath);
        SiteLocalDbLegacyMigrator.Migrate(db, config);

        Assert.Equal(2, CountRows(db, "sf_messages"));
    }

    [Fact]
    public void MigratedSfMessages_EnterTheOplog_SoTheyActuallyReplicate()
    {
        // Same ordering trap as the Phase 1 tables: migrate before RegisterReplicated and
        // the rows never enter the oplog, never reach the peer, and nothing errors. Only an
        // assertion on __localdb_oplog catches the ordering being reversed.
        var sfPath = Path_("store-and-forward.db");
        SeedLegacyStoreAndForward(sfPath, "msg-1", "msg-2");
        var config = Config(storeAndForwardPath: sfPath);
        var db = CreateConsolidated(config);

        SiteLocalDbLegacyMigrator.Migrate(db, config);

        using var connection = db.CreateConnection();
        using var cmd = connection.CreateCommand();
        cmd.CommandText = "SELECT COUNT(*) FROM __localdb_oplog WHERE table_name = 'sf_messages'";
        Assert.Equal(2L, (long)cmd.ExecuteScalar()!);
    }

    [Fact]
    public void SfMessages_FromAnOlderBuild_MigrateDespiteMissingColumns()
    {
        // A legacy file written before execution_id / parent_execution_id / last_attempt_at_ms
        // existed. Naming a missing column in the SELECT throws "no such column", and the
        // reader treats that as an unrecognised shape — which would silently discard every
        // buffered message. The copy intersects the column sets instead.
        var sfPath = Path_("store-and-forward.db");
        using (var legacy = new SqliteConnection($"Data Source={sfPath}"))
        {
            legacy.Open();
            using var ddl = legacy.CreateCommand();
            ddl.CommandText = """
                CREATE TABLE sf_messages (
                    id TEXT PRIMARY KEY, category INTEGER NOT NULL, target TEXT NOT NULL,
                    payload_json TEXT NOT NULL, retry_count INTEGER NOT NULL DEFAULT 0,
                    max_retries INTEGER NOT NULL DEFAULT 50,
                    retry_interval_ms INTEGER NOT NULL DEFAULT 30000,
                    created_at TEXT NOT NULL, last_attempt_at TEXT,
                    status INTEGER NOT NULL DEFAULT 0, last_error TEXT, origin_instance TEXT
                );
                INSERT INTO sf_messages (id, category, target, payload_json, created_at)
                VALUES ('old-1', 0, 'ERP', '{}', '2026-01-01T00:00:00Z');
                """;
            ddl.ExecuteNonQuery();
        }

        var config = Config(storeAndForwardPath: sfPath);
        var db = CreateConsolidated(config);

        SiteLocalDbLegacyMigrator.Migrate(db, config);

        Assert.Equal(1, CountRows(db, "sf_messages"));
        Assert.Equal(["old-1"], SelectIds(db, "sf_messages", "id"));
    }

    [Fact]
    public void LegacyFileWithoutTheExpectedTable_IsTreatedAsEmpty_NotAFailure()
    {
        // A file predating the table (or an unrecognised shape) must not fail host boot.
        var trackingPath = Path_("site-tracking.db");
        using (var connection = new SqliteConnection($"Data Source={trackingPath}"))
        {
            connection.Open();
            using var ddl = connection.CreateCommand();
            ddl.CommandText = "CREATE TABLE something_else (x INTEGER);";
            ddl.ExecuteNonQuery();
        }

        var config = Config(trackingPath, Path_("absent-events.db"));
        var db = CreateConsolidated(config);

        SiteLocalDbLegacyMigrator.Migrate(db, config);

        Assert.Equal(0, CountRows(db, "OperationTracking"));
        Assert.True(File.Exists(trackingPath + ".migrated"));
    }

    [Fact]
    public void InMemoryLegacyConnectionStrings_AreSkipped()
    {
        // Test/dev configurations point the legacy stores at in-memory databases. There is
        // nothing durable to migrate, and Path.GetFullPath on ":memory:" would be nonsense.
        var config = new ConfigurationBuilder().AddInMemoryCollection(new Dictionary<string, string?>
        {
            ["LocalDb:Path"] = Path_("consolidated.db"),
            ["ScadaBridge:Node:NodeName"] = "node-a",
            ["ScadaBridge:OperationTracking:ConnectionString"] =
                "Data Source=file:legacy?mode=memory&cache=shared",
            ["ScadaBridge:SiteEventLog:DatabasePath"] = ":memory:",
        }).Build();

        Assert.Equal(string.Empty, SiteLocalDbLegacyMigrator.ResolveTrackingPath(config));
        Assert.Equal(string.Empty, SiteLocalDbLegacyMigrator.ResolveEventLogPath(config));

        var db = CreateConsolidated(config);
        SiteLocalDbLegacyMigrator.Migrate(db, config);

        Assert.Equal(0, CountRows(db, "site_events"));
    }

    [Fact]
    public void UnsetLegacyKeys_FallBackToTheCwdRelativeCodeDefaults()
    {
        // Neither key is set in any docker appsettings, so the OLD code used these
        // CWD-relative defaults. Resolving them anywhere else (e.g. against the data
        // volume) would look for the legacy data in a directory it was never written to.
        var config = new ConfigurationBuilder().AddInMemoryCollection(new Dictionary<string, string?>
        {
            ["LocalDb:Path"] = Path_("consolidated.db"),
        }).Build();

        Assert.Equal(
            System.IO.Path.GetFullPath("site-tracking.db"),
            SiteLocalDbLegacyMigrator.ResolveTrackingPath(config));
        Assert.Equal(
            System.IO.Path.GetFullPath("site_events.db"),
            SiteLocalDbLegacyMigrator.ResolveEventLogPath(config));
    }
}