fix(historian): clustered existence reads + tag-predicate tests (T4 review)

Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
This commit is contained in:
Joseph Doherty
2026-10-11 02:16:14 -04:00
parent 4c053cc4d0
commit 9d90dc9457
2 changed files with 122 additions and 9 deletions
@@ -98,19 +98,57 @@ internal sealed class QuestDbHistorianValueWriter : IHistorianValueWriter
}
}
// T0 Q11 range form: tombstones are NOT filtered out (a deleted row must not be resurrected).
/// <summary>A gap between consecutive batch times larger than this starts a new existence-query range.</summary>
internal static readonly TimeSpan ExistenceClusterGap = TimeSpan.FromMinutes(1);
/// <summary>
/// Splits ascending <paramref name="sortedTimes"/> into inclusive <c>[Lo, Hi]</c> ranges: a new range starts where the gap to the
/// previous time exceeds <paramref name="gap"/> or the range already holds <paramref name="maxPerCluster"/> times. One outlier
/// time therefore costs its own tiny query instead of widening one range over every stored row in between.
/// </summary>
/// <param name="sortedTimes">Times in ascending order.</param>
/// <param name="gap">Largest gap kept inside one range.</param>
/// <param name="maxPerCluster">Most times in one range (&gt;= 1).</param>
/// <returns>The ranges, ascending.</returns>
internal static IReadOnlyList<(DateTime Lo, DateTime Hi)> Cluster(IReadOnlyList<DateTime> sortedTimes, TimeSpan gap, int maxPerCluster)
{
ArgumentOutOfRangeException.ThrowIfLessThan(maxPerCluster, 1);
var result = new List<(DateTime, DateTime)>();
if (sortedTimes.Count == 0) return result;
var lo = sortedTimes[0];
var prev = lo;
var n = 1;
for (var i = 1; i < sortedTimes.Count; i++)
{
var t = sortedTimes[i];
if (t - prev > gap || n >= maxPerCluster)
{
result.Add((lo, prev));
lo = t;
n = 0;
}
prev = t;
n++;
}
result.Add((lo, prev));
return result;
}
// T0 Q11 range form, one query per cluster: tombstones are NOT filtered out (a deleted row must not be resurrected).
private async Task<HashSet<long>> ExistingTicksAsync(string tag, List<HistorianSampleRow> rows, CancellationToken ct)
{
var min = rows.Min(r => QuestDbConnection.Round(r.TimeUtc));
var max = rows.Max(r => QuestDbConnection.Round(r.TimeUtc));
var times = rows.Select(r => QuestDbConnection.Round(r.TimeUtc)).OrderBy(t => t).ToList();
var set = new HashSet<long>();
await using var conn = await _conn.OpenAsync(ct).ConfigureAwait(false);
await using var cmd = _conn.CreateCommand(conn, $"SELECT ts FROM {_conn.SampleTable} WHERE tag = $1 AND ts >= $2 AND ts <= $3");
QuestDbConnection.BindText(cmd, tag);
QuestDbConnection.BindTimestamp(cmd, min);
QuestDbConnection.BindTimestamp(cmd, max);
await using var r = await cmd.ExecuteReaderAsync(ct).ConfigureAwait(false);
while (await r.ReadAsync(ct).ConfigureAwait(false)) set.Add(QuestDbConnection.ReadUtc(r, 0).Ticks);
foreach (var (lo, hi) in Cluster(times, ExistenceClusterGap, BatchRows))
{
await using var cmd = _conn.CreateCommand(conn, $"SELECT ts FROM {_conn.SampleTable} WHERE tag = $1 AND ts >= $2 AND ts <= $3");
QuestDbConnection.BindText(cmd, tag);
QuestDbConnection.BindTimestamp(cmd, lo);
QuestDbConnection.BindTimestamp(cmd, hi);
await using var r = await cmd.ExecuteReaderAsync(ct).ConfigureAwait(false);
while (await r.ReadAsync(ct).ConfigureAwait(false)) set.Add(QuestDbConnection.ReadUtc(r, 0).Ticks);
}
return set;
}
@@ -218,4 +218,79 @@ public sealed class WriteTests(QuestDbFixture fx)
cts.Cancel();
(await w.WriteLiveValuesAsync("x", [new(T0, HistorianValue.FromDouble(1), 0)], cts.Token)).ShouldBeFalse();
}
private static async Task SeedRowAsync(QuestDbConnection c, QuestDbWalVisibility wal, string tag, DateTime t, double v)
{
await using (var conn = await c.OpenAsync(default))
{
await using var cmd = c.CreateCommand(conn,
$"INSERT INTO {c.SampleTable} (ts, tag, kind, num_value, int_value, str_value, status, deleted, server_ts) VALUES ($1,$2,1,$3,NULL,NULL,0,false,$4)");
QuestDbConnection.BindTimestamp(cmd, t);
QuestDbConnection.BindText(cmd, tag);
cmd.Parameters.Add(new Npgsql.NpgsqlParameter { NpgsqlDbType = NpgsqlTypes.NpgsqlDbType.Double, Value = v });
QuestDbConnection.BindTimestamp(cmd, t);
await cmd.ExecuteNonQueryAsync();
}
await wal.WaitAsync(c.SampleTable, default);
}
[Fact]
public async Task A_row_on_another_tag_at_the_same_timestamp_does_not_suppress_the_write()
{
fx.SkipIfUnavailable();
var (w, c) = Make();
var wal = new QuestDbWalVisibility(c, TimeSpan.FromSeconds(5));
(await w.WriteLiveValuesAsync("tag.A", [new(T0, HistorianValue.FromDouble(1), 0)], CancellationToken.None)).ShouldBeTrue();
(await w.WriteLiveValuesAsync("tag.B", [new(T0, HistorianValue.FromDouble(2), 0)], CancellationToken.None)).ShouldBeTrue();
(await RowsAsync(c, "tag.A")).Single().N.ShouldBe(1d);
(await RowsAsync(c, "tag.B")).Single().N.ShouldBe(2d);
}
[Fact]
public async Task An_absent_interior_timestamp_is_written_while_both_present_ends_are_left_alone()
{
fx.SkipIfUnavailable();
var (w, c) = Make();
var wal = new QuestDbWalVisibility(c, TimeSpan.FromSeconds(5));
await w.WriteLiveValuesAsync("tag.ends", [new(T0.AddSeconds(100), HistorianValue.FromDouble(0), 0)], CancellationToken.None); // creates the tag/tables
await SeedRowAsync(c, wal, "tag.ends", T0, 10);
await SeedRowAsync(c, wal, "tag.ends", T0.AddSeconds(20), 30);
(await w.WriteLiveValuesAsync("tag.ends",
[new(T0, HistorianValue.FromDouble(-1), 0), new(T0.AddSeconds(10), HistorianValue.FromDouble(20), 0), new(T0.AddSeconds(20), HistorianValue.FromDouble(-3), 0)],
CancellationToken.None)).ShouldBeTrue();
var rows = (await RowsAsync(c, "tag.ends")).Where(r => r.T <= T0.AddSeconds(20)).ToList();
rows.Select(r => r.N).ShouldBe([10d, 20d, 30d]);
}
[Fact]
public async Task An_outlier_timestamp_does_not_widen_the_existence_read_over_the_stored_rows_between()
{
fx.SkipIfUnavailable();
var (w, c) = Make();
// Behavioural check of the whole path with a far-apart batch: both samples land, a redelivery adds nothing.
var far = T0.AddYears(1);
HistorizationValue[] vals = [new(T0, HistorianValue.FromDouble(1), 0), new(far, HistorianValue.FromDouble(2), 0)];
(await w.WriteLiveValuesAsync("tag.far", vals, CancellationToken.None)).ShouldBeTrue();
(await w.WriteLiveValuesAsync("tag.far", vals, CancellationToken.None)).ShouldBeTrue();
(await RowsAsync(c, "tag.far")).Count.ShouldBe(2);
}
[Fact]
public void Cluster_splits_on_a_gap_over_the_limit_and_at_the_per_cluster_cap()
{
var g = TimeSpan.FromMinutes(1);
QuestDbHistorianValueWriter.Cluster([], g, 10).ShouldBeEmpty();
QuestDbHistorianValueWriter.Cluster([T0], g, 10).ShouldBe([(T0, T0)]);
// exactly the gap stays together; one tick over splits
QuestDbHistorianValueWriter.Cluster([T0, T0 + g], g, 10).ShouldBe([(T0, T0 + g)]);
QuestDbHistorianValueWriter.Cluster([T0, T0 + g + TimeSpan.FromTicks(10)], g, 10)
.ShouldBe([(T0, T0), (T0 + g + TimeSpan.FromTicks(10), T0 + g + TimeSpan.FromTicks(10))]);
// outlier gets its own range
var far = T0.AddYears(1);
QuestDbHistorianValueWriter.Cluster([T0, T0.AddSeconds(1), far], g, 10).ShouldBe([(T0, T0.AddSeconds(1)), (far, far)]);
// cap: 5 close times, max 2 per cluster
var five = Enumerable.Range(0, 5).Select(i => T0.AddSeconds(i)).ToList();
QuestDbHistorianValueWriter.Cluster(five, g, 2)
.ShouldBe([(five[0], five[1]), (five[2], five[3]), (five[4], five[4])]);
}
}