我从 Kafka 读取 json,FieldExtractionBolt 读取 json 将数据提取到元组值并将它们传递给 CassandraWriterBolt,CassandraWriterBolt 又在 Cassandra 中写入一条记录,将所有这些元组值写入单独的列。
Kafka 上的 JSON 消息 -
{"pair":"GBPJPY","bid":134.4563,"ask":134.4354}
FieldExtractionBolt -
String message = tuple.getStringByField("message");
Map values = new Gson().fromJson(message, Map.class);
basicOutputCollector.emit(new Values(values.get("pair"), values.get("bid"), values.get("ask")));
CassandraWriterBolt -
return (CassandraWriterBolt) new CassandraWriterBolt(async(simpleQuery("INSERT INTO currency(pair, ask, bid) VALUES (?, ?, ?);").with(fields("pair", "ask", "bid")))
我尝试根据此处给出的答案编写测试 -如何通过以编程方式插入消息来 E2E 测试 Storm Topology 的功能
在我的项目中,我在 Spring 配置中定义了所有的螺栓、喷口和流。这使得编写/阅读我的拓扑非常容易。我通过从 ApplicationContext 获取 bolt、spouts 和 stream beans 来构建拓扑。在我的 Spring 配置中,KafkaSpout 和 CassandraWriterBolt 是在“prod”配置文件下定义的,因此它们只能在 prod 和“test”配置文件下使用,我为 KafkaSpout 和 CassandraWriterBolt 定义了存根。对于 KafkaSpout,我使用了 FixedToupleSpout,对于 CassandraWriterBolt,我使用了 TestWordCounter。
这是我的测试
@Test
public void testTopology(){
StormTopology topology = SpringBasedTopologyBuilder.getInstance().buildStormTopologyUsingApplicationContext(applicationContext);
TestJob COMPLETE_TOPOLOGY_TESTJOB = (cluster) -> {
MockedSources mocked = new MockedSources();
mocked.addMockData("kafkaSpout",
new Values("{\"pair\":\"GBPJPY\",\"bid\":134.4563,\"ask\":134.4354}"),
new Values("{\"pair\":\"GBPUSD\",\"bid\":1.4563,\"ask\":1.4354}"));
Config topoConf = new Config();
topoConf.setNumWorkers(2);
CompleteTopologyParam ctp = new CompleteTopologyParam();
ctp.setMockedSources(mocked);
ctp.setStormConf(topoConf);
Map<String, List<FixedTuple>> results = Testing.completeTopology(cluster, topology, ctp);
List<List<Object>> cassandraTuples = Testing.readTuples(results, "cassandraWriterBolt");
List<List<Object>> expectedCassandraTuples = Arrays.asList(Arrays.asList("GBPJPY", 1), Arrays.asList("GBPUSD", 1),
Arrays.asList("134.4563", 1), Arrays.asList("1.4563", 1), Arrays.asList("134.4354", 2));
assertTrue(expectedCassandraTuples + " expected, but found " + cassandraTuples,
Testing.multiseteq(expectedCassandraTuples, cassandraTuples));
MkClusterParam param = new MkClusterParam();
param.setSupervisors(4);
Testing.withSimulatedTimeLocalCluster(param, COMPLETE_TOPOLOGY_TESTJOB);
}
@Configuration
@Import(MainApplication.class)
public static class TestConfig
{
@Bean
public IRichSpout kafkaSpout(){
return new FixedTupleSpout(Arrays.asList(new FixedTuple(Arrays.asList("{\"pair\":\"GBPJPY\",\"bid\":134.4563,\"ask\":134.4354"))), new Fields(new String[]{"message"}));
}
@Bean
public IBasicBolt cassandraWriterBolt(){
return new TestWordCounter();
}
}
我得到的结果不是我所期望的。我收到以下错误 -
java.lang.AssertionError: [[GBPJPY, 1], [GBPUSD, 1], [134.4563, 1], [1.4563, 1], [134.4354, 2]] expected, but found [[GBPJPY, 1], [GBPUSD, 1]]
看起来,TestWordCounter 只是将第一个值作为元组读取(仅限货币对并跳过出价和要价)。似乎 TestWordCounter 在这里不是一个正确的选择。CassandraWriterBolt 的正确存根是什么,以便我可以断言它将收到 2 条记录,一条为 GBPJPY,另一条为 GBPUSD,以及他们的买入价和卖出价?