corda/docs/source/api-testing.rst
Christian Sailer 3260d9f2c4
CORDA-1489 Exposure of node internals in mock network (#4130)
* Introduce public subset of config to tweak config via mock net work without exposing internal node config.

* Removal of functions exposing (internal) NodeConfiguration from the public test API

* Code review fixes

* Blank lines removed

* Documented mock network API change in upgrade notes.

* Updated documentation and API doc.

* More documentation/API doc
2018-11-05 09:29:05 +00:00

23 KiB

API: Testing

Flow testing

MockNetwork

Flow testing can be fully automated using a MockNetwork composed of StartedMockNode nodes. Each StartedMockNode behaves like a regular Corda node, but its services are either in-memory or mocked out.

A MockNetwork is created as follows:

class FlowTests {
    private lateinit var mockNet: MockNetwork

    @Before
    fun setup() {
        network = MockNetwork(listOf("my.cordapp.package", "my.other.cordapp.package"))
    }
}
public class IOUFlowTests {
    private MockNetwork network;

    @Before
    public void setup() {
        network = new MockNetwork(ImmutableList.of("my.cordapp.package", "my.other.cordapp.package"));
    }
}

The MockNetwork requires at a minimum a list of packages. Each package is packaged into a CorDapp JAR and installed as a CorDapp on each StartedMockNode.

Configuring the MockNetwork

The MockNetwork is configured automatically. You can tweak its configuration using a MockNetworkParameters object, or by using named parameters in Kotlin:

val network = MockNetwork(
        // A list of packages to scan. Any contracts, flows and Corda services within these
        // packages will be automatically available to any nodes within the mock network
        cordappPackages = listOf("my.cordapp.package", "my.other.cordapp.package"),
        // If true then each node will be run in its own thread. This can result in race conditions in your
        // code if not carefully written, but is more realistic and may help if you have flows in your app that
        // do long blocking operations.
        threadPerNode = false,
        // The notaries to use on the mock network. By default you get one mock notary and that is usually
        // sufficient.
        notarySpecs = listOf(MockNetworkNotarySpec(DUMMY_NOTARY_NAME)),
        // If true then messages will not be routed from sender to receiver until you use the
        // [MockNetwork.runNetwork] method. This is useful for writing single-threaded unit test code that can
        // examine the state of the mock network before and after a message is sent, without races and without
        // the receiving node immediately sending a response.
        networkSendManuallyPumped = false,
        // How traffic is allocated in the case where multiple nodes share a single identity, which happens for
        // notaries in a cluster. You don't normally ever need to change this: it is mostly useful for testing
        // notary implementations.
        servicePeerAllocationStrategy = InMemoryMessagingNetwork.ServicePeerAllocationStrategy.Random())

val network2 = MockNetwork(
        // A list of packages to scan. Any contracts, flows and Corda services within these
        // packages will be automatically available to any nodes within the mock network
        listOf("my.cordapp.package", "my.other.cordapp.package"), MockNetworkParameters(
        // If true then each node will be run in its own thread. This can result in race conditions in your
        // code if not carefully written, but is more realistic and may help if you have flows in your app that
        // do long blocking operations.
        threadPerNode = false,
        // The notaries to use on the mock network. By default you get one mock notary and that is usually
        // sufficient.
        notarySpecs = listOf(MockNetworkNotarySpec(DUMMY_NOTARY_NAME)),
        // If true then messages will not be routed from sender to receiver until you use the
        // [MockNetwork.runNetwork] method. This is useful for writing single-threaded unit test code that can
        // examine the state of the mock network before and after a message is sent, without races and without
        // the receiving node immediately sending a response.
        networkSendManuallyPumped = false,
        // How traffic is allocated in the case where multiple nodes share a single identity, which happens for
        // notaries in a cluster. You don't normally ever need to change this: it is mostly useful for testing
        // notary implementations.
        servicePeerAllocationStrategy = InMemoryMessagingNetwork.ServicePeerAllocationStrategy.Random())
)
MockNetwork network = MockNetwork(
        // A list of packages to scan. Any contracts, flows and Corda services within these
        // packages will be automatically available to any nodes within the mock network
        ImmutableList.of("my.cordapp.package", "my.other.cordapp.package"),
        new MockNetworkParameters()
                // If true then each node will be run in its own thread. This can result in race conditions in
                // your code if not carefully written, but is more realistic and may help if you have flows in
                // your app that do long blocking operations.
                .setThreadPerNode(false)
                // The notaries to use on the mock network. By default you get one mock notary and that is
                // usually sufficient.
                .setNotarySpecs(ImmutableList.of(new MockNetworkNotarySpec(DUMMY_NOTARY_NAME)))
                // If true then messages will not be routed from sender to receiver until you use the
                // [MockNetwork.runNetwork] method. This is useful for writing single-threaded unit test code
                // that can examine the state of the mock network before and after a message is sent, without
                // races and without the receiving node immediately sending a response.
                .setNetworkSendManuallyPumped(false)
                // How traffic is allocated in the case where multiple nodes share a single identity, which
                // happens for notaries in a cluster. You don't normally ever need to change this: it is mostly
                // useful for testing notary implementations.
                .setServicePeerAllocationStrategy(new InMemoryMessagingNetwork.ServicePeerAllocationStrategy.Random()));

Adding nodes to the network

Nodes are created on the MockNetwork using:

class FlowTests {
    private lateinit var mockNet: MockNetwork
    lateinit var nodeA: StartedMockNode
    lateinit var nodeB: StartedMockNode

    @Before
    fun setup() {
        network = MockNetwork(listOf("my.cordapp.package", "my.other.cordapp.package"))
        nodeA = network.createPartyNode()
        // We can optionally give the node a name.
        nodeB = network.createPartyNode(CordaX500Name("Bank B", "London", "GB"))
    }
}
public class IOUFlowTests {
    private MockNetwork network;
    private StartedMockNode a;
    private StartedMockNode b;

    @Before
    public void setup() {
        network = new MockNetwork(ImmutableList.of("my.cordapp.package", "my.other.cordapp.package"));
        nodeA = network.createPartyNode(null);
        // We can optionally give the node a name.
        nodeB = network.createPartyNode(new CordaX500Name("Bank B", "London", "GB"));
    }
}

Nodes added using createPartyNode are provided a default set of node parameters. However, it is also possible to provide different parameters to each node using the following methods on MockNetwork:

/**
 * Create a started node with the given parameters.
 *
 * @param legalName The node's legal name.
 * @param forcedID A unique identifier for the node.
 * @param entropyRoot The initial entropy value to use when generating keys. Defaults to an (insecure) random value,
 * but can be overridden to cause nodes to have stable or colliding identity/service keys.
 * @param configOverrides Add/override the default configuration/behaviour of the node
 * @param extraCordappPackages Extra CorDapp packages to add for this node.
 */
@JvmOverloads
fun createNode(legalName: CordaX500Name? = null,
               forcedID: Int? = null,
               entropyRoot: BigInteger = BigInteger.valueOf(random63BitValue()),
               configOverrides: MockNodeConfigOverrides? = null,
               extraCordappPackages: List<String> = emptyList()
): StartedMockNode

/** Create a started node with the given parameters. **/
fun createNode(parameters: MockNodeParameters = MockNodeParameters()): StartedMockNode

As you can see above, parameters can be added individually or encapsulated within a MockNodeParameters object. Of particular interest are configOverrides which allow you to override some of the default node configuration options. Please refer to the MockNodeConfigOverrides class for details what can currently be overridden. Also, the extraCordappPackages parameter allows you to add extra CorDapps to a specific node. This is useful when you wish for all nodes to load a common CorDapp but for a subset of nodes to load CorDapps specific to their role in the network.

Running the network

When using a MockNetwork, you must be careful to ensure that all the nodes have processed all the relevant messages before making assertions about the result of performing some action. For example, if you start a flow to update the ledger but don't wait until all the nodes involved have processed all the resulting messages, your nodes' vaults may not be in the state you expect.

When networkSendManuallyPumped is set to false, you must manually initiate the processing of received messages. You manually process received messages as follows:

  • StartedMockNode.pumpReceive() processes a single message from the node's queue
  • MockNetwork.runNetwork() processes all the messages in every node's queue until there are no further messages to process

When networkSendManuallyPumped is set to true, nodes will automatically process the messages they receive. You can block until all messages have been processed using MockNetwork.waitQuiescent().

Warning

If threadPerNode is set to true, networkSendManuallyPumped must also be set to true.

Running flows

A StartedMockNode starts a flow using the StartedNodeServices.startFlow method. This method returns a future representing the output of running the flow.

val signedTransactionFuture = nodeA.services.startFlow(IOUFlow(iouValue = 99, otherParty = nodeBParty))
CordaFuture<SignedTransaction> future = startFlow(a.getServices(), new ExampleFlow.Initiator(1, nodeBParty));

The network must then be manually run before retrieving the future's value:

val signedTransactionFuture = nodeA.services.startFlow(IOUFlow(iouValue = 99, otherParty = nodeBParty))
// Assuming network.networkSendManuallyPumped == false.
network.runNetwork()
val signedTransaction = future.get();
CordaFuture<SignedTransaction> future = startFlow(a.getServices(), new ExampleFlow.Initiator(1, nodeBParty));
// Assuming network.networkSendManuallyPumped == false.
network.runNetwork();
SignedTransaction signedTransaction = future.get();

Accessing StartedMockNode internals

Querying a node's vault

Recorded states can be retrieved from the vault of a StartedMockNode using:

val myStates = nodeA.services.vaultService.queryBy<MyStateType>().states
List<MyStateType> myStates = node.getServices().getVaultService().queryBy(MyStateType.class).getStates();

This allows you to check whether a given state has (or has not) been stored, and whether it has the correct attributes.

Examining a node's transaction storage

Recorded transactions can be retrieved from the transaction storage of a StartedMockNode using:

val transaction = nodeA.services.validatedTransactions.getTransaction(transaction.id)
SignedTransaction transaction = nodeA.getServices().getValidatedTransactions().getTransaction(transaction.getId())

This allows you to check whether a given transaction has (or has not) been stored, and whether it has the correct attributes.

This allows you to check whether a given state has (or has not) been stored, and whether it has the correct attributes.

Further examples

  • See the flow testing tutorial here <flow-testing>
  • See the oracle tutorial here <oracles> for information on testing @CordaService classes
  • Further examples are available in the Example CorDapp in Java and Kotlin

Contract testing

The Corda test framework includes the ability to create a test ledger by calling the ledger function on an implementation of the ServiceHub interface.

Test identities

You can create dummy identities to use in test transactions using the TestIdentity class:

../../docs/source/example-code/src/test/kotlin/net/corda/docs/kotlin/tutorial/testdsl/TutorialTestDSL.kt

../../docs/source/example-code/src/test/java/net/corda/docs/java/tutorial/testdsl/TutorialTestDSL.java

TestIdentity exposes the following fields and methods:

val identityParty: Party = bigCorp.party
val identityName: CordaX500Name = bigCorp.name
val identityPubKey: PublicKey = bigCorp.publicKey
val identityKeyPair: KeyPair = bigCorp.keyPair
val identityPartyAndCertificate: PartyAndCertificate = bigCorp.identity
Party identityParty = bigCorp.getParty();
CordaX500Name identityName = bigCorp.getName();
PublicKey identityPubKey = bigCorp.getPublicKey();
KeyPair identityKeyPair = bigCorp.getKeyPair();
PartyAndCertificate identityPartyAndCertificate = bigCorp.getIdentity();

You can also create a unique TestIdentity using the fresh method:

val uniqueTestIdentity: TestIdentity = TestIdentity.fresh("orgName")
TestIdentity uniqueTestIdentity = TestIdentity.Companion.fresh("orgName");

MockServices

A mock implementation of ServiceHub is provided in MockServices. This is a minimal ServiceHub that suffices to test contract logic. It has the ability to insert states into the vault, query the vault, and construct and check transactions.

../../docs/source/example-code/src/test/kotlin/net/corda/docs/kotlin/tutorial/testdsl/TutorialTestDSL.kt

../../docs/source/example-code/src/test/java/net/corda/docs/java/tutorial/testdsl/TutorialTestDSL.java

Alternatively, there is a helper constructor which just accepts a list of TestIdentity. The first identity provided is the identity of the node whose ServiceHub is being mocked, and any subsequent identities are identities that the node knows about. Only the calling package is scanned for cordapps and a test IdentityService is created for you, using all the given identities.

../../docs/source/example-code/src/test/kotlin/net/corda/docs/kotlin/tutorial/testdsl/TutorialTestDSL.kt

../../docs/source/example-code/src/test/java/net/corda/docs/java/tutorial/testdsl/TutorialTestDSL.java

Writing tests using a test ledger

The ServiceHub.ledger extension function allows you to create a test ledger. Within the ledger wrapper you can create transactions using the transaction function. Within a transaction you can define the input and output states for the transaction, alongside any commands that are being executed, the timeWindow in which the transaction has been executed, and any attachments, as shown in this example test:

../../docs/source/example-code/src/test/kotlin/net/corda/docs/kotlin/tutorial/testdsl/TutorialTestDSL.kt

../../docs/source/example-code/src/test/java/net/corda/docs/java/tutorial/testdsl/TutorialTestDSL.java

Once all the transaction components have been specified, you can run verifies() to check that the given transaction is valid.

Checking for failure states

In order to test for failures, you can use the failsWith method, or in Kotlin the fails with helper method, which assert that the transaction fails with a specific error. If you just want to assert that the transaction has failed without verifying the message, there is also a fails method.

../../docs/source/example-code/src/test/kotlin/net/corda/docs/kotlin/tutorial/testdsl/TutorialTestDSL.kt

../../docs/source/example-code/src/test/java/net/corda/docs/java/tutorial/testdsl/TutorialTestDSL.java

Note

The transaction DSL forces the last line of the test to be either a verifies or fails with statement.

Testing multiple scenarios at once

Within a single transaction block, you can assert several times that the transaction constructed so far either passes or fails verification. For example, you could test that a contract fails to verify because it has no output states, and then add the relevant output state and check that the contract verifies successfully, as in the following example:

../../docs/source/example-code/src/test/kotlin/net/corda/docs/kotlin/tutorial/testdsl/TutorialTestDSL.kt

../../docs/source/example-code/src/test/java/net/corda/docs/java/tutorial/testdsl/TutorialTestDSL.java

You can also use the tweak function to create a locally scoped transaction that you can make changes to and then return to the original, unmodified transaction. As in the following example:

../../docs/source/example-code/src/test/kotlin/net/corda/docs/kotlin/tutorial/testdsl/TutorialTestDSL.kt

../../docs/source/example-code/src/test/java/net/corda/docs/java/tutorial/testdsl/TutorialTestDSL.java

Chaining transactions

The following example shows that within a ledger, you can create more than one transaction in order to test chains of transactions. In addition to transaction, unverifiedTransaction can be used, as in the example below, to create transactions on the ledger without verifying them, for pre-populating the ledger with existing data. When chaining transactions, it is important to note that even though a transaction verifies successfully, the overall ledger may not be valid. This can be verified separately by placing a verifies or fails statement within the ledger block.

../../docs/source/example-code/src/test/kotlin/net/corda/docs/kotlin/tutorial/testdsl/TutorialTestDSL.kt

../../docs/source/example-code/src/test/java/net/corda/docs/java/tutorial/testdsl/TutorialTestDSL.java

Further examples

  • See the flow testing tutorial here <tutorial-test-dsl>
  • Further examples are available in the Example CorDapp in Java and Kotlin