streamsx.topology.tester¶

Testing support for streaming applications.

Overview¶

Allows testing of a streaming application by creation conditions on streams that are expected to become valid during the processing. Tester is designed to be used with Python’s unittest module.

A complete application may be tested or fragments of it, for example a sub-graph can be tested in isolation that takes input data and scores it using a model.

Supports execution of the application on STREAMING_ANALYTICS_SERVICE, DISTRIBUTED or STANDALONE.

A Tester instance is created and associated with the Topology to be tested. Conditions are then created against streams, such as a stream must receive 10 tuples using tuple_count().

Here is a simple example that tests a filter correctly only passes tuples with values greater than 5:

import unittest
from streamsx.topology.topology import Topology
from streamsx.topology.tester import Tester

class TestSimpleFilter(unittest.TestCase):

    def setUp(self):
        # Sets self.test_ctxtype and self.test_config
        Tester.setup_streaming_analytics(self)

    def test_filter(self):
        # Declare the application to be tested
        topology = Topology()
        s = topology.source([5, 7, 2, 4, 9, 3, 8])
        s = s.filter(lambda x : x > 5)

        # Create tester and assign conditions
        tester = Tester(topology)
        tester.contents(s, [7, 9, 8])

        # Submit the application for test
        # If it fails an AssertionError will be raised.
       tester.test(self.test_ctxtype, self.test_config)

A stream may have any number of conditions and any number of streams may be tested.

A local_check() is supported where a method of the unittest class is executed once the job becomes healthy. This performs checks from the context of the Python unittest class, such as checking external effects of the application or using the REST api to monitor the application.

A test fails-fast if any of the following occur:

Any condition fails. E.g. a tuple failing a tuple_check().
The local_check() (if set) raises an error.
The job for the test:
- Fails to become healthy.
- Becomes unhealthy during the test run.
- Any processing element (PE) within the job restarts.

A test timeouts if it does not fail but its conditions do not become valid. The timeout is not fixed as an absolute test run time, but as a time since “progress” was made. This can allow tests to pass when healthy runs are run in a constrained environment that slows execution. For example with a tuple count condition of ten, progress is indicated by tuples arriving on a stream, so that as long as gaps between tuples are within the timeout period the test remains running until ten tuples appear.

Note

The test timeout value is not configurable.

Note

The submitted job (application under test) has additional elements (streams & operators) inserted to implement the conditions. These are visible through various APIs including the Streams console raw graph view. Such elements are put into the Tester category.

Note

The package streamsx.testing provides nose plugins to provide control over tests without having to modify their source code.

Changed in version 1.9: - Python 2.7 supported (except with Streaming Analytics service).

Module contents¶

Classes

Tester

Testing support for a Topology.

class streamsx.topology.tester.Tester(topology)¶

Bases: object

Testing support for a Topology.

Allows testing of a Topology by creating conditions against the contents of its streams.

Conditions may be added to a topology at any time before submission.

If a topology is submitted directly to a context then the graph is not modified. This allows testing code to be inserted while the topology is being built, but not acted upon unless the topology is submitted in test mode.

If a topology is submitted through the test method then the topology may be modified to include operations to ensure the conditions are met.

Warning

For future compatibility applications under test should not include intended failures that cause a processing element to stop or restart. Thus, currently testing is against expected application behavior.

Parameters: topology – Topology to be tested.

add_condition(stream, condition)¶

Add a condition to a stream.

Conditions are normally added through tuple_count(), contents() or tuple_check().

This allows an additional conditions that are implementations of Condition.

Parameters

stream (Stream) – Stream to be tested.
condition (Condition) – Arbitrary condition.

Returns

stream

Return type

Stream

contents(stream, expected, ordered=True)¶

Test that a stream contains the expected tuples.

Parameters

stream (Stream) – Stream to be tested.
expected (list) – Sequence of expected tuples.
ordered (bool) – True if the ordering of received tuples must match expected.

Returns

stream

Return type

Stream

eventual_result(stream, checker)¶

Test a stream reaches a known result or state.

Creates a test condition that the tuples on a stream eventually reach a known result or state. Each tuple on stream results in a call to checker(tuple_).

The return from checker is handled as:

None - The condition requires more tuples to become valid.
true value - The condition has become valid.
false value - The condition has failed. Once a condition has failed it can never become valid.

Thus checker is typically stateful and allows ensuring that condition becomes valid from a set of input tuples. For example in a financial application the application under test may need to achieve a final known balance, but due to timings of windows the number of tuples required to set the final balance may be variable.

Once the condition becomes valid any false value, except None, returned by processing of subsequent tuples will cause the condition to fail.

Returning None effectively never changes the state of the condition.

Parameters

stream (Stream) – Stream to be tested.
checker (callable) – Callable that returns evaluates the state of the stream with result to the result.

New in version 1.11.

static get_streams_version(test)¶

Returns IBM Streams product version string for a test.

Returns the product version corresponding to the test’s setup. For STANDALONE and DISTRIBUTED the product version corresponds to the version defined by the environment variable STREAMS_INSTALL.

Parameters: test (unittest.TestCase) – Test case setup to run IBM Streams tests.

local_check(callable)¶

Perform local check while the application is being tested.

A call to callable is made after the application under test is submitted and becomes healthy. The check is in the context of the Python runtime executing the unittest case, typically the callable is a method of the test case.

The application remains running until all the conditions are met and callable returns. If callable raises an error, typically through an assertion method from unittest then the test will fail.

Used for testing side effects of the application, typically with STREAMING_ANALYTICS_SERVICE or DISTRIBUTED. The callable may also use the REST api for context types that support it to dynamically monitor the running application.

The callable can use submission_result and streams_connection attributes from Tester instance to interact with the job or the running Streams instance. These REST binding classes can be obtained as follows:

Job - tester.submission_result.job

Instance - tester.submission_result.job.get_instance()

StreamsConnection - tester.streams_connection

Simple example of checking the job is healthy:

import unittest
from streamsx.topology.topology import Topology
from streamsx.topology.tester import Tester

class TestLocalCheckExample(unittest.TestCase):
    def setUp(self):
        Tester.setup_distributed(self)

    def test_job_is_healthy(self):
        topology = Topology()
        s = topology.source(['Hello', 'World'])

        self.tester = Tester(topology)
        self.tester.tuple_count(s, 2)

        # Add the local check
        self.tester.local_check = self.local_checks

        # Run the test
        self.tester.test(self.test_ctxtype, self.test_config)


    def local_checks(self):
        job = self.tester.submission_result.job
        self.assertEqual('healthy', job.health)

Warning

A local check must not cancel the job (application under test).

Warning

A local check is not supported in standalone mode.

Parameters: callable – Callable object.

static minimum_streams_version(test, required_version)¶

Checks test setup matches a minimum required IBM Streams version.

Parameters

test (unittest.TestCase) – Test case setup to run IBM Streams tests.
required_version (str) – VRMF of the minimum version the test requires. Examples are '4.3', 4.2.4.

Returns

True if the setup fulfills the minimum required version, false otherwise.

Return type

bool

punct_count(stream, count, exact=True)¶

Test that a stream contains a number of window punctuations.

If exact is True, then condition becomes valid when count punctuations are seen on stream during the test. Subsequently if additional punctuations are seen on stream then the condition fails and can never become valid.

If exact is False, then the condition becomes valid once count punctuations are seen on stream and remains valid regardless of any additional punctuations.

Note

Punctuation marks are in-band signals that are inserted between tuples in a stream. If sources or stream transforms insert window markers at all, and when they insert them depends on the source or the semantic of the stream transformation. One example is the aggregate(), which inserts a window marker into the output stream after each aggregation.

Parameters

stream (Stream) – Stream to be tested.
count (int) – Number of punctuations expected.
exact (bool) – True if the stream must contain exactly count punctuations, False if the stream must contain at least count punctuations.

Returns

stream

Return type

Stream

static require_streams_version(test, required_version)¶

Require a test has minimum IBM Streams version.

Skips the test if the test’s setup is not at the required minimum IBM Streams version.

Parameters

test (unittest.TestCase) – Test case setup to run IBM Streams tests.
required_version (str) – VRMF of the minimum version the test requires. Examples are '4.3', 4.2.4.

resets(minimum_resets=10)¶

Create a condition that randomly resets consistent regions. The condition becomes valid when each consistent region in the application under test has been reset minimum_resets times by the tester.

The resets are performed at arbitrary intervals scaled to the period of the region (if it is periodically triggered).

Note

A region is reset by initiating a request though the Job Control Plane. The reset is not driven by any injected failure, such as a PE restart.

Parameters: minimum_resets (int) – Minimum number of resets for each region.

New in version 1.11.

run_for(duration)¶

Run the test for a minimum number of seconds.

Creates a test wide condition that becomes valid when the application under test has been running for duration seconds. Maybe be called multiple times, the test will run as long as the maximum value provided.

Can be used to test applications without any externally visible streams, or streams that do not have testable conditions. For example a complete application may be tested by runnning it for for ten minutes and use local_check() to test any external impacts, such as messages published to a message queue system.

Parameters: duration (float) – Minimum number of seconds the test will run for.

static setup_distributed(test, verbose=None)¶

Set up a unittest.TestCase to run tests using IBM Streams distributed mode.

Two attributes are set in the test case:

test_ctxtype - Context type the test will be run in.

test_config - Test configuration.

Parameters

test (unittest.TestCase) – Test case to be set up to run tests using Tester
verbose (bool) – If true then the streamsx.topology.test logger is configured at DEBUG level with output sent to standard error.

Returns: None

Cloud Pak for Data integrated instance configuration

These environment variables define how the test is built and submitted.

CP4D_URL - Cloud Pak for Data deployment URL, e.g. https://cp4d_server:31843.

STREAMS_INSTANCE_ID - Streams service instance name.

STREAMS_USERNAME - (optional) User name to submit the test as, defaulting to the current operating system user name.

STREAMS_PASSWORD - Password for authentication.

Cloud Pak for Data standalone instance configuration

These environment variables define how the test is built and submitted.

STREAMS_BUILD_URL - Endpoint for the Streams build service.

STREAMS_REST_URL - Endpoint for the Streams SWS (REST) service.

STREAMS_USERNAME - (optional) User name to submit the test as, defaulting to the current operating system user name.

STREAMS_PASSWORD - Password for authentication.

Streams 4.2 & 4.3 instance configuration

Requires a local IBM Streams install define by the STREAMS_INSTALL environment variable. If STREAMS_INSTALL is not set then the test is skipped.

The Streams instance to use is defined by the environment variables:

STREAMS_ZKCONNECT - Zookeeper connection string (optional)

STREAMS_DOMAIN_ID - Domain identifier

STREAMS_INSTANCE_ID - Instance identifier

The user used to submit and monitor the job is set by the optional environment variables:

STREAMS_USERNAME - User name defaulting to streamsadmin.

STREAMS_PASSWORD - User password defaulting to passw0rd.

The defaults match the setup for testing on a IBM Streams Quick Start Edition (QSE) virtual machine.

Warning

streamtool is used to submit the job and requires that streamtool does not prompt for authentication. This is achieved by using streamtool genkey.

static setup_standalone(test, verbose=None)¶

Set up a unittest.TestCase to run tests using IBM Streams standalone mode.

Requires a local IBM Streams install define by the STREAMS_INSTALL environment variable. If STREAMS_INSTALL is not set, then the test is skipped.

A standalone application under test will run until a condition fails or all the streams are finalized or when the run_for() time (if set) elapses. Applications that include infinite streams must include set a run for time using run_for() to ensure the test completes

Two attributes are set in the test case:

test_ctxtype - Context type the test will be run in.

test_config- Test configuration.

Parameters

test (unittest.TestCase) – Test case to be set up to run tests using Tester
verbose (bool) – If true then the streamsx.topology.test logger is configured at DEBUG level with output sent to standard error.

Returns: None

static setup_streaming_analytics(test, service_name=None, force_remote_build=False, verbose=None)¶

Set up a unittest.TestCase to run tests using Streaming Analytics service on IBM Cloud.

The service to use is defined by:

VCAP_SERVICES environment variable containing streaming_analytics entries.

service_name which defaults to the value of STREAMING_ANALYTICS_SERVICE_NAME environment variable.

If VCAP_SERVICES is not set or a service name is not defined, then the test is skipped.

Two attributes are set in the test case:

test_ctxtype - Context type the test will be run in.

test_config - Test configuration.

Parameters

test (unittest.TestCase) – Test case to be set up to run tests using Tester
service_name (str) – Name of Streaming Analytics service to use. Must exist as an entry in the VCAP services. Defaults to value of STREAMING_ANALYTICS_SERVICE_NAME environment variable.
force_remote_build (bool) – Force use of the Streaming Analytics build service. If false and STREAMS_INSTALL is set then a local build will be used if the local environment is suitable for the service, otherwise the Streams application bundle is built using the build service.
verbose (bool) – If true then the streamsx.topology.test logger is configured at DEBUG level with output sent to standard error.

If run with Python 2 the test is skipped,.

Returns: None

test(ctxtype, config=None, assert_on_fail=True, username=None, password=None, always_collect_logs=False)¶

Test the topology.

Submits the topology for testing and verifies the test conditions are met and the job remained healthy through its execution.

The submitted application (job) is monitored for the test conditions and will be canceled when all the conditions are valid or at least one failed. In addition if a local check was specified using local_check() then that callable must complete before the job is cancelled.

The test passes if all conditions became valid and the local check callable (if present) completed without raising an error.

The test fails if the job is unhealthy, any condition fails or the local check callable (if present) raised an exception. In the event that the test fails when submitting to the STREAMING_ANALYTICS_SERVICE context, the application logs are retrieved as a tar file and are saved to the current working directory. The filesystem path to the application logs is saved in the tester’s result object under the application_logs key, i.e. tester.result[‘application_logs’]

Parameters

ctxtype (str) – Context type for submission.
config – Configuration for submission.
assert_on_fail (bool) – True to raise an assertion if the test fails, False to return the passed status.
username (str) – Deprecated
password (str) – Deprecated
always_collect_logs (bool) – True to always collect the console log and PE trace files of the test.

result¶: The result of the test. This can contain exit codes, application log paths, or other relevant test information.

submission_result¶: Result of the application submission from submit().

streams_connection¶

Connection object that can be used to interact with the REST API of the Streaming Analytics service or instance.

Type: StreamsConnection

Returns: True if test passed, False if test failed if assert_on_fail is False.
Return type: bool

Deprecated since version 1.8.3: username and password parameters. When required for a distributed test use the environment variables STREAMS_USERNAME and STREAMS_PASSWORD to define the Streams user.

tuple_check(stream, checker)¶

Check each tuple on a stream.

For each tuple t on stream checker(t) is called.

If the return evaluates to False then the condition fails. Once the condition fails it can never become valid. Otherwise the condition becomes or remains valid. The first tuple on the stream makes the condition valid if the checker callable evaluates to True.

The condition can be combined with tuple_count() with exact=False to test a stream map or filter with random input data.

An example of combining tuple_count and tuple_check to test a filter followed by a map is working correctly across a random set of values:

def rands():
    r = random.Random()
    while True:
        yield r.random()

class TestFilterMap(unittest.testCase):
# Set up omitted

    def test_filter(self):
        # Declare the application to be tested
        topology = Topology()
        r = topology.source(rands())
        r = r.filter(lambda x : x > 0.7)
        r = r.map(lambda x : x + 0.2)

        # Create tester and assign conditions
        tester = Tester(topology)
        # Ensure at least 1000 tuples pass through the filter.
        tester.tuple_count(r, 1000, exact=False)
        tester.tuple_check(r, lambda x : x > 0.9)


        # Submit the application for test
        # If it fails an AssertionError will be raised.
        tester.test(self.test_ctxtype, self.test_config)

Parameters

stream (Stream) – Stream to be tested.
checker (callable) – Callable that must evaluate to True for each tuple.

tuple_count(stream, count, exact=True)¶

Test that a stream contains a number of tuples.

If exact is True, then condition becomes valid when count tuples are seen on stream during the test. Subsequently if additional tuples are seen on stream then the condition fails and can never become valid.

If exact is False, then the condition becomes valid once count tuples are seen on stream and remains valid regardless of any additional tuples.

Parameters

stream (Stream) – Stream to be tested.
count (int) – Number of tuples expected.
exact (bool) – True if the stream must contain exactly count tuples, False if the stream must contain at least count tuples.

Returns

stream

Return type

Stream