feat(mqtt): write-through (IWritable) — Plain publish + Sparkplug NCMD/DCMD (#508)

MqttDriver implemented no IWritable, so every MQTT node materialized read-only
regardless of what the operator authored. It now publishes.

  Plain      -> the tag's writeTopic; UTF-8 text (Raw/Scalar) or a bare JSON
                literal (Json)
  Sparkplug  -> DCMD for a device metric, NCMD for an edge-node metric, carrying
                the metric BY NAME with an explicit Datatype and no seq

Both at QoS 0, retain:false. Retain is deliberately not configurable — a retained
command is replayed by the broker on every future subscribe, including the
device's own reconnect, so one stale write would be re-applied forever.

writeTopic is a new, separate TagConfig key, and its ABSENCE is meaningful: the
write is refused BadNotWritable and the node is discovered ViewOnly. The issue's
"do not skip" warning drove the design — the one outcome worse than read-only
nodes is a node that looks writable, returns Good and publishes nothing. Reusing
the read topic would be worse still: it would echo back and read as a confirmed
device value. Browse-time writability is computed by the SAME encoder the write
path uses, so a node cannot advertise writability it would then refuse.

Good means published, not applied — MQTT has no application-level write ack, so
this is Galaxy's optimistic-Good posture and inherits its property: a failed
DEVICE write can never be surfaced. A failed PUBLISH is reported Bad; the
optimism covers what the device did with the command, never whether it was sent.
WriteIdempotent stays default-off.

The encoder is pure and lives in .Contracts, so both wire shapes are testable
without a broker. Unrepresentable values and types with no Sparkplug encoding are
refused rather than coerced — a Reference must not silently become a String on
the wire, and "not-a-number" must not become 0 on a live plant.

LIVE GATE PASSED against the real Mosquitto fixture on 10.100.0.35: 21/21
integration tests, including 6 new ones that observe the published message with
an INDEPENDENT broker subscriber — a fake asserting "publish was called" could
not tell the feature working from the feature publishing to the wrong topic.

The gate earned its keep: the NCMD test initially failed because NCMD is a SHARED
channel — the driver publishes its own rebirth request (Node Control/Rebirth) on
the very topic an edge-node metric write uses, and it arrived first. Correct
driver behaviour, wrong test; the watcher now matches on metric name. Documented,
because anything consuming that topic needs to know. DCMD has no contention.

Offline: 24 encoder + 9 driver tests. The four new status constants are NAMED so
StatusCodeParityTests can see them (#497/#519) — it went 258 -> 262 and validated
each against the pinned SDK.

Also corrects the Mqtt.md "Known gaps" table: prunes the six rows closed in this
round, and fixes the #507 row, which still described rediscovery as inert after
that issue was closed as superseded and its injection path deleted.

Claude-Session: https://claude.ai/code/session_015p7wGqy3YpZNCpDzTpGMKo
This commit is contained in:
Joseph Doherty
2026-07-31 06:33:23 -04:00
parent 7f53bd96ec
commit 72be19225d
11 changed files with 1441 additions and 24 deletions
@@ -0,0 +1,304 @@
using System.Globalization;
using System.Text;
using System.Text.Json;
using Google.Protobuf;
using Org.Eclipse.Tahu.Protobuf;
using ZB.MOM.WW.OtOpcUa.Core.Abstractions;
using ZB.MOM.WW.OtOpcUa.Driver.Mqtt.Sparkplug;
using TahuDataType = Org.Eclipse.Tahu.Protobuf.DataType;
namespace ZB.MOM.WW.OtOpcUa.Driver.Mqtt;
/// <summary>
/// Why a write could not be encoded. Maps 1:1 onto the status the driver returns, so the reason a
/// write was refused is decided here — once, purely — rather than at the transport.
/// </summary>
public enum MqttWriteEncodeError
{
/// <summary>Encoding succeeded.</summary>
None = 0,
/// <summary>
/// The tag declares no write destination: a Plain tag with no <c>writeTopic</c>. Deliberately a
/// hard refusal — see <see cref="MqttWriteEncoder"/>'s remarks on why a silent no-op would be
/// worse than the honest read-only behaviour this replaced.
/// </summary>
NoWriteDestination,
/// <summary>The value is null, or cannot be represented in the tag's declared data type.</summary>
ValueNotRepresentable,
/// <summary>The tag's declared data type has no Sparkplug command encoding.</summary>
UnsupportedDataType,
}
/// <summary>
/// Encodes an outbound MQTT write — the payload bytes and the topic to publish them on — for both
/// ingest shapes (Gitea #508). Pure: no I/O, no transport, no clock beyond the Sparkplug command
/// timestamp, so the whole encode contract is unit-testable without a broker.
/// </summary>
/// <remarks>
/// <para>
/// <b>A missing write destination is a refusal, never a no-op.</b> Before #508 the driver
/// implemented no <c>IWritable</c> at all, so every MQTT node materialised read-only — honest, if
/// limited. The one outcome that would be strictly worse is a node that <em>looks</em> writable,
/// accepts a write, returns Good and publishes nothing. So a Plain tag with no <c>writeTopic</c>
/// yields <see cref="MqttWriteEncodeError.NoWriteDestination"/> and the driver surfaces
/// <c>BadNotWritable</c>.
/// </para>
/// <para>
/// <b>The write topic is deliberately separate from the read topic.</b> A tag legitimately reads
/// <c>.../state</c> and writes <c>.../set</c>, so the destination is its own authored field rather
/// than being inferred from <see cref="MqttTagDefinition.Topic"/>. Publishing a command back onto
/// the tag's own read topic would, on a broker with no device behind it, echo straight back and
/// read as a confirmed device value.
/// </para>
/// <para>
/// <b>Sparkplug writes are commands, not data.</b> A metric under a device encodes as a DCMD, one
/// owned by the edge node itself as an NCMD, addressed by the tag's stable
/// <c>(group, edgeNode, device?, metric)</c> tuple — never by the per-birth alias, which may be
/// reused for a different metric across a rebirth. No <c>seq</c> is stamped, for the reason
/// <see cref="RebirthRequester"/> documents at length: the sequence stream belongs to the messages
/// an edge node itself publishes, and a strict node validating a host-supplied one against its own
/// counter would have grounds to reject the command.
/// </para>
/// <para>
/// <b>jsonPath is not applied on write.</b> It is a read-side extraction from the read topic's
/// document, and the write topic is a different topic whose payload shape nothing in the tag config
/// describes. A <see cref="MqttPayloadFormat.Json"/> write therefore emits the value as a bare JSON
/// literal. Reconstructing a wrapped document from an extraction path would be guesswork, and
/// guessing a payload shape is how a write lands somewhere nobody authored.
/// </para>
/// </remarks>
public static class MqttWriteEncoder
{
/// <summary>QoS for an outbound write. See the remarks on <see cref="TryEncode"/>.</summary>
public const int WriteQos = 0;
/// <summary>
/// Encodes one write for the tag <paramref name="def"/> under the driver's <paramref name="mode"/>.
/// </summary>
/// <param name="mode">The driver's ingest shape — chosen by driver config, never sniffed from the tag.</param>
/// <param name="def">The resolved tag definition.</param>
/// <param name="value">The value to write.</param>
/// <param name="encoded">The topic + payload to publish, when this returns <see cref="MqttWriteEncodeError.None"/>.</param>
/// <returns>The encode outcome.</returns>
/// <remarks>
/// <b>QoS 0, retain <see langword="false"/>.</b> Retain is the one that matters: a retained
/// command would be replayed by the broker to the device on every future subscribe — including
/// the device's own reconnect — so a single stale write would be re-applied forever. This is the
/// same hazard <see cref="RebirthRequester"/> avoids for NCMD, and the reason neither is
/// configurable.
/// </remarks>
public static MqttWriteEncodeError TryEncode(
MqttMode mode,
MqttTagDefinition def,
object? value,
out MqttEncodedWrite encoded)
{
ArgumentNullException.ThrowIfNull(def);
encoded = default;
return mode == MqttMode.SparkplugB
? TryEncodeSparkplug(def, value, out encoded)
: TryEncodePlain(def, value, out encoded);
}
/// <summary>Encodes a Plain-mode write: the value's text/JSON rendering onto the tag's write topic.</summary>
/// <param name="def">The resolved tag definition.</param>
/// <param name="value">The value to write.</param>
/// <param name="encoded">The topic + payload on success.</param>
/// <returns>The encode outcome.</returns>
private static MqttWriteEncodeError TryEncodePlain(
MqttTagDefinition def,
object? value,
out MqttEncodedWrite encoded)
{
encoded = default;
if (string.IsNullOrWhiteSpace(def.WriteTopic))
{
return MqttWriteEncodeError.NoWriteDestination;
}
if (value is null)
{
// No MQTT payload means "the value is null" — a zero-length payload is a legitimate
// clear-the-retained-message instruction, which is a different operation entirely.
return MqttWriteEncodeError.ValueNotRepresentable;
}
string text;
if (def.PayloadFormat == MqttPayloadFormat.Json)
{
try
{
text = JsonSerializer.Serialize(value);
}
catch (NotSupportedException)
{
return MqttWriteEncodeError.ValueNotRepresentable;
}
}
else
{
// Raw and Scalar are both "the payload's UTF-8 text" on the read side; the write side
// mirrors that. Invariant culture, or a comma decimal separator on a European host would
// publish a number the device cannot parse.
text = value is IFormattable f
? f.ToString(null, CultureInfo.InvariantCulture)
: value.ToString() ?? "";
}
encoded = new MqttEncodedWrite(def.WriteTopic!.Trim(), Encoding.UTF8.GetBytes(text));
return MqttWriteEncodeError.None;
}
/// <summary>Encodes a Sparkplug-mode write as an NCMD (node metric) or DCMD (device metric).</summary>
/// <param name="def">The resolved tag definition.</param>
/// <param name="value">The value to write.</param>
/// <param name="encoded">The topic + payload on success.</param>
/// <returns>The encode outcome.</returns>
private static MqttWriteEncodeError TryEncodeSparkplug(
MqttTagDefinition def,
object? value,
out MqttEncodedWrite encoded)
{
encoded = default;
if (def.GroupId is not { Length: > 0 } groupId
|| def.EdgeNodeId is not { Length: > 0 } edgeNodeId
|| def.MetricName is not { Length: > 0 } metricName)
{
// A Sparkplug tag's command address IS its binding tuple, so an incomplete tuple is the
// Sparkplug equivalent of a missing write topic rather than a different kind of failure.
return MqttWriteEncodeError.NoWriteDestination;
}
if (value is null)
{
return MqttWriteEncodeError.ValueNotRepresentable;
}
var metric = new Payload.Types.Metric { Name = metricName };
var error = ApplyValue(metric, def.DataType, value);
if (error != MqttWriteEncodeError.None)
{
return error;
}
var timestampMs = (ulong)DateTimeOffset.UtcNow.ToUnixTimeMilliseconds();
metric.Timestamp = timestampMs;
var payload = new Payload { Timestamp = timestampMs };
payload.Metrics.Add(metric);
// A device metric is commanded with DCMD addressed to that device; an edge-node metric with
// NCMD. Sending a device metric as NCMD would address the node, which does not own it.
var isDevice = def.DeviceId is { Length: > 0 };
var topic = SparkplugTopic.Format(
groupId,
isDevice ? SparkplugMessageType.DCMD : SparkplugMessageType.NCMD,
edgeNodeId,
isDevice ? def.DeviceId : null);
encoded = new MqttEncodedWrite(topic, payload.ToByteArray());
return MqttWriteEncodeError.None;
}
/// <summary>
/// Sets the metric's typed value field and its <c>Datatype</c> from the tag's declared type.
/// </summary>
/// <param name="metric">The metric being built.</param>
/// <param name="dataType">The tag's declared driver data type.</param>
/// <param name="value">The value to convert.</param>
/// <returns>The encode outcome.</returns>
/// <remarks>
/// <c>Datatype</c> is set explicitly rather than left for the node to infer, for the same reason
/// <see cref="RebirthRequester"/> sets it: a node that treats a missing or ambiguous field as
/// malformed drops the command silently, and a silently-dropped write is the failure mode this
/// whole feature must not have.
/// </remarks>
private static MqttWriteEncodeError ApplyValue(
Payload.Types.Metric metric,
DriverDataType dataType,
object value)
{
try
{
switch (dataType)
{
case DriverDataType.Boolean:
metric.Datatype = (uint)TahuDataType.Boolean;
metric.BooleanValue = Convert.ToBoolean(value, CultureInfo.InvariantCulture);
return MqttWriteEncodeError.None;
case DriverDataType.Int16:
metric.Datatype = (uint)TahuDataType.Int16;
metric.IntValue = unchecked((uint)Convert.ToInt16(value, CultureInfo.InvariantCulture));
return MqttWriteEncodeError.None;
case DriverDataType.Int32:
metric.Datatype = (uint)TahuDataType.Int32;
metric.IntValue = unchecked((uint)Convert.ToInt32(value, CultureInfo.InvariantCulture));
return MqttWriteEncodeError.None;
case DriverDataType.Int64:
metric.Datatype = (uint)TahuDataType.Int64;
metric.LongValue = unchecked((ulong)Convert.ToInt64(value, CultureInfo.InvariantCulture));
return MqttWriteEncodeError.None;
case DriverDataType.UInt16:
metric.Datatype = (uint)TahuDataType.Uint16;
metric.IntValue = Convert.ToUInt16(value, CultureInfo.InvariantCulture);
return MqttWriteEncodeError.None;
case DriverDataType.UInt32:
metric.Datatype = (uint)TahuDataType.Uint32;
metric.IntValue = Convert.ToUInt32(value, CultureInfo.InvariantCulture);
return MqttWriteEncodeError.None;
case DriverDataType.UInt64:
metric.Datatype = (uint)TahuDataType.Uint64;
metric.LongValue = Convert.ToUInt64(value, CultureInfo.InvariantCulture);
return MqttWriteEncodeError.None;
case DriverDataType.Float32:
metric.Datatype = (uint)TahuDataType.Float;
metric.FloatValue = Convert.ToSingle(value, CultureInfo.InvariantCulture);
return MqttWriteEncodeError.None;
case DriverDataType.Float64:
metric.Datatype = (uint)TahuDataType.Double;
metric.DoubleValue = Convert.ToDouble(value, CultureInfo.InvariantCulture);
return MqttWriteEncodeError.None;
case DriverDataType.String:
metric.Datatype = (uint)TahuDataType.String;
metric.StringValue = value as string ?? Convert.ToString(value, CultureInfo.InvariantCulture) ?? "";
return MqttWriteEncodeError.None;
case DriverDataType.DateTime:
metric.Datatype = (uint)TahuDataType.DateTime;
metric.LongValue = (ulong)new DateTimeOffset(
Convert.ToDateTime(value, CultureInfo.InvariantCulture).ToUniversalTime())
.ToUnixTimeMilliseconds();
return MqttWriteEncodeError.None;
default:
// DriverDataType.Reference and anything added later: refuse rather than guess a
// wire representation. An unsupported type must not become a String on the wire.
return MqttWriteEncodeError.UnsupportedDataType;
}
}
catch (Exception ex) when (ex is InvalidCastException or FormatException or OverflowException)
{
return MqttWriteEncodeError.ValueNotRepresentable;
}
}
}
/// <summary>One encoded outbound write: where it goes and what it carries.</summary>
/// <param name="Topic">The concrete topic to publish on.</param>
/// <param name="Payload">The encoded message body.</param>
public readonly record struct MqttEncodedWrite(string Topic, byte[] Payload);