refactor(opcuaclient): move NamespaceMap to a .Session leaf so Contracts is dependency-light

G-7 wired OpcUaClientTagDefinitionFactory.Inspect into the deploy-time TagConfig
gate, which gave ControlPlane a ProjectReference to
Driver.OpcUaClient.Contracts — and with it, transitively, the whole OPC UA client
stack (5 OPCFoundation packages) into every host that loads the control plane.

NamespaceMap was the only reason Contracts carried
OPCFoundation.NetStandard.Opc.Ua.Client: its entire public surface is
ISession/NamespaceTable/NodeId-typed. It moves to a new leaf project,
ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Session, referenced by the runtime .Driver
and by .Browser. Contracts is left with Core.Abstractions and NOTHING else —
the same shape as Calculation.Contracts.

The namespace stays ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient, so no consumer's
`using` changed. The .Session -> .Contracts direction is deliberate: the blob
read NamespaceMap needs is OpcUaClientTagDefinitionFactory.TryReadNodeId, the
one authority it shares with the deploy gate, and that must stay in the
dependency-light half.

Why a new project rather than folding NamespaceMap into the runtime .Driver and
having .Browser reference it (the Driver.Mqtt.Browser #512-style deliberate
exception): the AdminUI project-references .Browser and NOT .Driver, so that
shape would pull the OpcUaClient driver runtime — plus Core,
Opc.Ua.Configuration and Secrets.Abstractions — into the AdminUI's build/deploy
graph. That is exactly the violation #512 spent two extractions removing from
Mqtt.Browser; re-introducing it to save one csproj is a bad trade.

Guard: ContractsIsDependencyLightTests (Driver.OpcUaClient.Tests) asserts the
Contracts manifest names no Opc.Ua* assembly, plus a counterpart asserting
.Session does. An ABSENCE assertion is sound on the manifest where a presence
one is not — the compiler elides an unused reference, but it MUST emit one the
moment Opc.Ua-typed code is added. Falsified by temporarily re-adding the
package plus an Opc.Ua.NodeId-typed member: red on Opc.Ua.Types, green again on
removal.

Evidence — `dotnet list ControlPlane package --include-transitive | grep -i opc`:
  before: OPCFoundation.NetStandard.Opc.Ua.{Client,Configuration,Core,
          Security.Certificates,Types} 1.5.378.106
  after:  none (and 0 Opc.Ua*.dll in ControlPlane's output directory)
This commit is contained in:
Joseph Doherty
2026-08-09 21:56:29 -04:00
parent fefcd2c872
commit b4f8cb8de7
11 changed files with 161 additions and 14 deletions
+1
View File
@@ -48,6 +48,7 @@
<Project Path="src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.FOCAS.Contracts/ZB.MOM.WW.OtOpcUa.Driver.FOCAS.Contracts.csproj" />
<Project Path="src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.csproj" />
<Project Path="src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts.csproj" />
<Project Path="src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Session/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Session.csproj" />
<Project Path="src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Browser/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Browser.csproj" />
<Project Path="src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.Mqtt/ZB.MOM.WW.OtOpcUa.Driver.Mqtt.csproj" />
<Project Path="src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.Mqtt.Contracts/ZB.MOM.WW.OtOpcUa.Driver.Mqtt.Contracts.csproj" />
@@ -8,6 +8,10 @@
<ItemGroup>
<ProjectReference Include="..\..\Core\ZB.MOM.WW.OtOpcUa.Commons\ZB.MOM.WW.OtOpcUa.Commons.csproj" />
<ProjectReference Include="..\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts.csproj" />
<!-- NamespaceMap. Deliberately the .Session leaf and NOT the runtime .Driver project: the AdminUI
references THIS project, so a .Driver reference would pull the driver runtime into the
AdminUI's deploy graph (the Gitea #512 shape). See the .Session csproj block comment. -->
<ProjectReference Include="..\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Session\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Session.csproj" />
</ItemGroup>
<ItemGroup>
<PackageReference Include="OPCFoundation.NetStandard.Opc.Ua.Client" />
@@ -31,7 +31,8 @@ public sealed record OpcUaClientDriverOptions
/// with the cluster's raw OPC UA Client tags. Each <see cref="RawTagEntry"/> carries the tag's
/// <b>RawPath</b> identity plus a driver <c>TagConfig</c> blob whose <c>nodeId</c> key holds the
/// upstream OPC UA node id string the driver dials. The driver builds a <c>RawPath → nodeId</c>
/// resolution table (via <see cref="NamespaceMap"/>) from this list at connect time; a
/// resolution table (via <c>NamespaceMap</c>, in the sibling <c>.Session</c> project) from this
/// list at connect time; a
/// read/write/subscribe/history reference is always a RawPath resolved through that table.
/// </summary>
public IReadOnlyList<RawTagEntry> RawTags { get; init; } = [];
@@ -4,13 +4,19 @@ namespace ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient;
/// <summary>
/// The authored-raw-tag reader for the OPC UA Client (gateway) driver: one authority, consumed by both
/// the runtime (<see cref="NamespaceMap"/>'s <c>RawPath → upstream node id</c> table build) and the
/// the runtime (<c>NamespaceMap</c>'s <c>RawPath → upstream node id</c> table build) and the
/// deploy-time TagConfig gate (<see cref="Inspect"/>, deferment-register G-7).
/// <para>
/// <b>This read used to be private to <see cref="NamespaceMap"/>.</b> It moved here — unchanged —
/// <b>This read used to be private to <c>NamespaceMap</c>.</b> It moved here — unchanged —
/// rather than being mirrored: a mirror of an accept/reject boundary drifts the moment either side
/// is touched, and the drift is silent, with the gate describing a shape the driver no longer
/// reads. <see cref="NamespaceMap"/> now calls <see cref="TryReadNodeId"/>.
/// reads. <c>NamespaceMap</c> now calls <see cref="TryReadNodeId"/>.
/// </para>
/// <para>
/// <b><c>NamespaceMap</c> is named by text, not <c>cref</c>, on purpose.</b> It lives in the
/// sibling <c>Driver.OpcUaClient.Session</c> project, which references THIS one — the dependency
/// runs that way round so this assembly stays dependency-light for ControlPlane, which
/// project-references it for the G-7 gate. A <c>cref</c> here would need the reverse reference.
/// </para>
/// <para>
/// <b><see cref="Inspect"/> is never stricter than <see cref="TryReadNodeId"/></b> — it asks the
@@ -6,11 +6,12 @@
<TreatWarningsAsErrors>true</TreatWarningsAsErrors>
</PropertyGroup>
<ItemGroup>
<PackageReference Include="OPCFoundation.NetStandard.Opc.Ua.Client" />
</ItemGroup>
<ItemGroup>
<!-- v3: NamespaceMap resolves a RawPath (RawTagEntry identity) to the upstream node id.
RawTagEntry lives in the Core.Abstractions leaf beside EquipmentTagRefResolver. -->
<!-- DEPENDENCY-LIGHT ON PURPOSE: Core.Abstractions and nothing else, the same shape as
Calculation.Contracts. ControlPlane project-references this for the G-7 deploy-time
TagConfig gate (OpcUaClientTagDefinitionFactory.Inspect), so anything added here lands in
every host that loads the control plane. The OPC UA client SDK used to be here, for
NamespaceMap; NamespaceMap now lives in the sibling .Session leaf. Pinned by
ContractsIsDependencyLightTests (Driver.OpcUaClient.Tests). -->
<ProjectReference Include="..\..\Core\ZB.MOM.WW.OtOpcUa.Core.Abstractions\ZB.MOM.WW.OtOpcUa.Core.Abstractions.csproj"/>
</ItemGroup>
</Project>
@@ -52,6 +52,17 @@ public readonly record struct RawTagBinding(string NodeId, bool WriteIdempotent)
/// driver then re-binds against the live session via the static
/// <see cref="TryResolve(ISession?, string, out NodeId)"/>.
/// </para>
/// <para>
/// <b>Why this type sits in <c>.Session</c> and not <c>.Contracts</c>.</b> Its whole public
/// surface is OPC UA SDK-typed (<see cref="ISession"/>, <see cref="NamespaceTable"/>,
/// <see cref="NodeId"/>), so while it lived in <c>.Contracts</c> that project had to carry
/// <c>OPCFoundation.NetStandard.Opc.Ua.Client</c> — and once ControlPlane project-referenced
/// <c>.Contracts</c> for the G-7 deploy-time TagConfig gate, the whole OPC UA client stack
/// came with it into every host that loads the control plane. The blob read this class needs
/// stayed behind in <c>.Contracts</c> as
/// <c>OpcUaClientTagDefinitionFactory.TryReadNodeId</c> (the ONE authority the deploy gate and
/// this table share), which is why the reference runs <c>.Session → .Contracts</c>.
/// </para>
/// </remarks>
public sealed class NamespaceMap
{
@@ -0,0 +1,46 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<TargetFramework>net10.0</TargetFramework>
<Nullable>enable</Nullable>
<ImplicitUsings>enable</ImplicitUsings>
<TreatWarningsAsErrors>true</TreatWarningsAsErrors>
<!-- Namespace stays ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient (no .Session suffix) so moving
NamespaceMap here changed no consumer's `using`. -->
<RootNamespace>ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient</RootNamespace>
<AssemblyName>ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Session</AssemblyName>
</PropertyGroup>
<!--
The live-OPC-UA-session layer of the OpcUaClient driver: everything the driver AND the AdminUI's
browser both need that is typed in ISession / NodeId / NamespaceTable. Today that is exactly
NamespaceMap.
WHY IT IS NOT IN .Contracts. It used to be, and .Contracts therefore carried
OPCFoundation.NetStandard.Opc.Ua.Client. When G-7 wired OpcUaClientTagDefinitionFactory.Inspect into
the deploy-time TagConfig gate, ControlPlane gained a ProjectReference to .Contracts — and with it
the whole OPC UA client stack, transitively, into every host that loads the control plane. A
Contracts project is meant to be the dependency-light parse/shape layer (see Calculation.Contracts:
Core.Abstractions and nothing else), so the SDK-typed half moved out rather than the gate doing
without.
WHY NOT JUST PUT IT IN THE .Driver PROJECT (the Mqtt.Browser-style deliberate exception). The
AdminUI project-references .Browser and NOT the runtime .Driver. Having .Browser reference .Driver
to reach NamespaceMap would pull the whole driver runtime — plus Core, Opc.Ua.Configuration and
Secrets.Abstractions — into the AdminUI's build/deploy graph. That is precisely the violation Gitea
#512 spent two extractions removing from Driver.Mqtt.Browser; re-introducing the same shape here to
save one csproj would be trading a measured regression for a little less churn.
KEEP THIS LEAF THIN. Anything addable here that is NOT session-typed belongs in .Contracts, or the
dependency-light property this split exists to buy erodes one file at a time. The pin is
ContractsIsDependencyLightTests in Driver.OpcUaClient.Tests.
-->
<ItemGroup>
<PackageReference Include="OPCFoundation.NetStandard.Opc.Ua.Client" />
</ItemGroup>
<ItemGroup>
<!-- RawTagEntry (the v3 authored-tag identity NamespaceMap indexes) lives in the Core.Abstractions
leaf; OpcUaClientTagDefinitionFactory is the ONE authority for reading a TagConfig blob's
nodeId, shared with the deploy gate — NamespaceMap calls it rather than re-parsing. -->
<ProjectReference Include="..\..\Core\ZB.MOM.WW.OtOpcUa.Core.Abstractions\ZB.MOM.WW.OtOpcUa.Core.Abstractions.csproj"/>
<ProjectReference Include="..\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts.csproj"/>
</ItemGroup>
</Project>
@@ -14,6 +14,9 @@
<ItemGroup>
<ProjectReference Include="..\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts.csproj"/>
<!-- NamespaceMap (ISession-typed) — shared with .Browser, which is why it is its own leaf rather
than living here. See the block comment in the .Session csproj. -->
<ProjectReference Include="..\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Session\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Session.csproj"/>
<ProjectReference Include="..\..\Core\ZB.MOM.WW.OtOpcUa.Core.Abstractions\ZB.MOM.WW.OtOpcUa.Core.Abstractions.csproj"/>
<ProjectReference Include="..\..\Core\ZB.MOM.WW.OtOpcUa.Core\ZB.MOM.WW.OtOpcUa.Core.csproj"/>
</ItemGroup>
@@ -54,10 +54,8 @@
TagConfigValidator checked them. Adding the reference is what makes
EquipmentTagConfigInspectorCoverageTests demand a dispatch-map entry for them — the guard scans
the output directory, so an unreferenced Contracts assembly is invisible to it.
Note OpcUaClient.Contracts brings OPCFoundation.NetStandard.Opc.Ua.Client transitively: its
NamespaceMap re-binds a stable nsu= reference against a live session. Every host that loads
ControlPlane (Host, AdminUI) already carries the OPC UA stack, so this adds no new deployment
surface. -->
OpcUaClient.Contracts is dependency-light (Core.Abstractions only) — the OPC UA client SDK it
used to pull in transitively left with NamespaceMap, to Driver.OpcUaClient.Session. -->
<ProjectReference Include="..\..\Drivers\ZB.MOM.WW.OtOpcUa.Driver.Sql.Contracts\ZB.MOM.WW.OtOpcUa.Driver.Sql.Contracts.csproj"/>
<ProjectReference Include="..\..\Drivers\ZB.MOM.WW.OtOpcUa.Driver.MTConnect.Contracts\ZB.MOM.WW.OtOpcUa.Driver.MTConnect.Contracts.csproj"/>
<ProjectReference Include="..\..\Drivers\ZB.MOM.WW.OtOpcUa.Driver.Calculation.Contracts\ZB.MOM.WW.OtOpcUa.Driver.Calculation.Contracts.csproj"/>
@@ -22,7 +22,9 @@
<ItemGroup>
<ProjectReference Include="..\..\..\src\Drivers\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Browser\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Browser.csproj"/>
<ProjectReference Include="..\..\..\src\Drivers\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts.csproj"/>
<!-- NamespaceMap.FromTable — it moved from .Contracts to the .Session leaf when .Contracts was
made dependency-light (ControlPlane references .Contracts for the G-7 deploy gate). -->
<ProjectReference Include="..\..\..\src\Drivers\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Session\ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Session.csproj"/>
<ProjectReference Include="..\..\..\src\Core\ZB.MOM.WW.OtOpcUa.Commons\ZB.MOM.WW.OtOpcUa.Commons.csproj"/>
</ItemGroup>
@@ -0,0 +1,74 @@
using Shouldly;
using Xunit;
namespace ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Tests;
/// <summary>
/// Pins the dependency shape of <c>ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts</c>: it must not
/// reference the OPC UA client SDK.
/// <para>
/// <b>Why this matters beyond tidiness.</b> Deferment-register G-7 wired
/// <see cref="OpcUaClientTagDefinitionFactory.Inspect"/> into the deploy-time TagConfig gate, which
/// made <c>ControlPlane</c> project-reference this Contracts assembly. ControlPlane is loaded by
/// every admin- and driver-role host, so whatever Contracts drags in, they all carry. Contracts
/// used to hold <c>NamespaceMap</c> — <c>ISession</c>/<c>NodeId</c>-typed throughout — and with it a
/// <c>OPCFoundation.NetStandard.Opc.Ua.Client</c> PackageReference; NamespaceMap moved to the
/// sibling <c>.Session</c> leaf so the parse-only half could stay light (the same shape as
/// <c>Calculation.Contracts</c>: Core.Abstractions and nothing else).
/// </para>
/// <para>
/// <b>Direction matters: this is an ABSENCE assertion, and that is what makes it sound here.</b>
/// The assembly manifest is an unreliable witness for PRESENCE — the C# compiler elides a reference
/// nothing in the IL actually uses, so "assembly X is referenced" can be false while the csproj says
/// otherwise (the trap <c>EquipmentTagConfigInspectorCoverageTests</c> documents, which is why that
/// one scans the output directory instead). Absence has no such hole in the direction that matters:
/// the moment someone adds Opc.Ua-typed code to Contracts, the compiler MUST emit the reference and
/// this test goes red. A csproj PackageReference that nothing consumes is elided and slips past —
/// accepted deliberately, because an unused package reference is not the failure being guarded.
/// </para>
/// <para>
/// Falsified 2026-08-09 by temporarily adding an <c>Opc.Ua.NodeId</c>-typed member to Contracts:
/// the assertion failed naming <c>Opc.Ua.Core</c>, and passed again once removed.
/// </para>
/// </summary>
public sealed class ContractsIsDependencyLightTests
{
/// <summary>The Contracts assembly's manifest must name no OPC UA SDK assembly.</summary>
[Fact]
public void The_Contracts_assembly_does_not_reference_the_OPC_UA_SDK()
{
var contracts = typeof(OpcUaClientTagDefinitionFactory).Assembly;
// Positive control: name the assembly under test, so a rename/retarget cannot silently point this
// at something else that trivially satisfies the assertion below.
contracts.GetName().Name.ShouldBe("ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts");
var opcUa = contracts.GetReferencedAssemblies()
.Select(a => a.Name ?? string.Empty)
.Where(n => n.StartsWith("Opc.Ua", StringComparison.OrdinalIgnoreCase))
.OrderBy(n => n, StringComparer.Ordinal)
.ToArray();
opcUa.ShouldBeEmpty(
"OpcUaClient.Contracts must stay free of the OPC UA client SDK — ControlPlane project-references "
+ "it for the G-7 deploy-time TagConfig gate, so every host that loads the control plane would "
+ "carry the stack. Put SDK-typed code in Driver.OpcUaClient.Session instead. Found: "
+ string.Join(", ", opcUa));
}
/// <summary>
/// The counterpart the assertion above is meaningful against: the SDK-typed half really did land
/// somewhere, and that somewhere does reference the SDK. Without this, "no Opc.Ua reference
/// anywhere" would read as a pass.
/// </summary>
[Fact]
public void The_Session_assembly_is_where_the_OPC_UA_SDK_reference_lives()
{
var session = typeof(NamespaceMap).Assembly;
session.GetName().Name.ShouldBe("ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Session");
session.GetReferencedAssemblies()
.Select(a => a.Name ?? string.Empty)
.ShouldContain(n => n.StartsWith("Opc.Ua", StringComparison.OrdinalIgnoreCase));
}
}