Structures and extension objects

The suite exposes structured data in two distinct styles. Both end up describing similar shapes (3D points, ranges, records) but exercise different parts of an OPC UA client.

Structures (Object + child variables)

Path: TestServer / Structures

Plain OPC UA pattern — an Object node with named children. Each field is a separate Variable node that you read independently.

TestPoint

Structures/TestPoint/
├── X  (Double, RW, 1.0)
├── Y  (Double, RW, 2.0)
└── Z  (Double, RW, 3.0)

TestRange

Structures/TestRange/
├── Min    (Double, RW, 0.0)
├── Max    (Double, RW, 100.0)
└── Value  (Double, RW, 50.0)

TestPerson

Structures/TestPerson/
├── Name    (String, RW, "John Doe")
├── Age     (UInt32, RW, 30)
└── Active  (Boolean, RW, true)

TestNested

A structure containing another structure (1-level nesting):

Structures/TestNested/
├── Label       (String, RW, "origin")
├── Timestamp   (DateTime, RW, startup time)
└── Point       (Object)
    ├── X       (Double, RW, 10.0)
    ├── Y       (Double, RW, 20.0)
    └── Z       (Double, RW, 30.0)

Point is reached via the HasComponent reference. The browse needs two hops from TestNested to read X/Y/Z.

PointCollection

Five identically-shaped Point_N objects. The initial coordinate values are seeded with new Random(42) and assigned by repeated rng.NextDouble() * 100 calls in StructuresBuilder.cs. Because the seed is fixed, the values are deterministic across server restarts, but they are not the round multiples sometimes documented in test specs — they fall somewhere in [0, 100) for each coordinate. Read them to discover the actual values for your environment, or treat them as opaque test data and only assert structural shape:

Structures/PointCollection/
├── Point_0 → (X, Y, Z each a Double in [0, 100))
├── Point_1 → (X, Y, Z each a Double in [0, 100))
├── Point_2 → (X, Y, Z each a Double in [0, 100))
├── Point_3 → (X, Y, Z each a Double in [0, 100))
└── Point_4 → (X, Y, Z each a Double in [0, 100))

Useful for testing "browse all children, then read all descendants" patterns.

DeepNesting

A chain of 10 nested objects:

Structures/DeepNesting/
└── Level_1 (Depth=1, Name="Level 1")
    └── Level_2 (Depth=2, Name="Level 2")
        └── Level_3 (Depth=3, Name="Level 3")
            └── …
                └── Level_10 (Depth=10, Name="Level 10")

Each level has a Depth (UInt32) and Name (String).

Extension objects (binary-encoded)

Path: TestServer / ExtensionObjects

Two variables whose Value attribute is a binary ExtensionObject — a single typed blob with all fields encoded together. The types are defined in the third namespace (urn:opcua:test-server:custom-types); resolve its index from Server.NamespaceArray at runtime — it has historically been ns=3 but is not pinned by the SDK.

In the table below <ct> is that resolved index. The DataType node uses one numeric id and the Default Binary encoding node uses a different one — ExtensionObject values carry the encoding NodeId in their typeId field.

Binary body

Both types are 3 × 8 bytes (3 IEEE-754 doubles, little-endian):

PointValue body (24 bytes):
  X (8 bytes) | Y (8 bytes) | Z (8 bytes)
  TypeId (encoding):  ns=<ct>;i=3010

RangeValue body (24 bytes):
  Min (8 bytes) | Max (8 bytes) | Value (8 bytes)
  TypeId (encoding):  ns=<ct>;i=3011

Reading

ev = read(node)
assert ev.typeId == NodeId(custom_types_ns, 3010)
fields = decode(ev.body, TestPointXYZ)
assert fields.X == 1.5

Writing

body = encode({X: 10.0, Y: 20.0, Z: 30.0}, TestPointXYZ)
write(node, ExtensionObject(typeId=NodeId(custom_types_ns, 3010), body=body))

Type discovery

The DataType node is reachable via the custom-types namespace:

Browse the custom-types namespace → find TestPointXYZ DataType node (i=3000)
                                 → child encoding `Default Binary` (i=3010)

The server registers DataTypeState nodes with SuperTypeId = DataTypeIds.Structure. There is no DataTypeDefinition attribute populated, so type-dictionary-only clients will need to know the layout (3 × Double, little-endian) out of band.

Read-only RangeValue

RangeValue is R only. Writes return Bad_NotWritable.

Difference recap

Test patterns

Structures

• Browse TestNested → expect 2 vars + 1 object.
• Browse PointCollection → expect 5 objects with same shape.
• Recursive browse from DeepNesting → reach Level_10.

Extension objects

• Read PointValue → verify the typeId matches the resolved encoding NodeId (i=3010 in the custom-types namespace) and decode the body.
• Write a new value → read back → verify round-trip.
• Read RangeValue → assert Bad_NotWritable on write.
• Browse the custom-types namespace → discover TestPointXYZ (i=3000) and TestRangeStruct (i=3001) as DataType nodes.

Where to read next

• Access control — the 50 access-mode test variables.
• Methods — methods that act on structures and arrays.