Wrong doc for RuleRuntime.getQueryResults - throwing RuntimeException instead of empty results ---------------------------------------------------------------------------------------------- Key: DROOLS-579 URL: https://issues.jboss.org/browse/DROOLS-579 Project: Drools Issue Type: Bug Security Level: Public(Everyone can see) Affects Versions: 6.1.0.Final Reporter: Matteo Mortari Assignee: Mark Proctor Attachments: 20140826.drools6test-missingquery.zip The [javadoc|http://docs.jboss.org/drools/release/6.1.0.Final/kie-api-javadoc/..., java.lang.Object...)] for public api {{RuleRuntime.getQueryResults}} mentions: _If the query does not exist or no results match the query it is empty_ ; however if query does not exist, actually RuntimeException is thrown instead. Details follows. I'm reporting this as I assume this will become even more relevant for the execution server, ref 2nd test below. h5. Proposals My proposals are that either: * javadoc shall be at least corrected - I know there is no obligation to document unchecked exceptions, but [here the doc is explicitly saying something different|http://docs.jboss.org/drools/release/6.1.0.Final/kie-api-javado..., java.lang.Object...)] . I will attach pull request accordingly. * the public api method shall reflect the doc and {{QueryResult}} returned should resemble Java 8's Optional then - if I understood some good api design styles to mitigate this scenario from Mario, instead of throwing the RuntimeException, the {{QueryResult}} returned could stand in place of Java 8's Optional. Without dependency on Java 8 API, the {{QueryResult}} in case 'query does not exist' could be empty, to guarantee retro-compatibility, and have specific methods to distinguish root cause of being empty between query-not-found or no-results-match-criteria. ... or possibly ack the former option for the time being, and welcome the latter option as a suggestion for enhancement on the new releases, please? Thanks h5. Details If you consider the following code {code} @Test public void missingQueryViaApi() { String drl = "query \"my query A\"\n" + " $r : String()\n" + "end"; KieHelper helper = new KieHelper(); helper.addContent( drl, ResourceType.DRL ); KieSession ksession = helper.build().newKieSession(); // If ... no results match the query it is empty QueryResults validQueryNameResults = ksession.getQueryResults("my query A"); assertNotNull("The query exists so result should NOT be null", validQueryNameResults); assertEquals("The query exists and result should be empty", 0, validQueryNameResults.size()); // If the query does not exist ... it is empty boolean hasThrownRTE = false; try { ksession.getQueryResults("an non-existent query"); } catch (RuntimeException e) { hasThrownRTE = true; } assertFalse("Should not throw RTE, should be sort-of 'empty' by doc", hasThrownRTE); } {code} The test fails on the last assert, as accordingly to [javadoc|http://docs.jboss.org/drools/release/6.1.0.Final/kie-api-javadoc/..., java.lang.Object...)], but instead it actually throws the RuntimeException. This becomes even more of an issue if you consider invoking via Commands: {code} @Test public void missingQueryViaCommands() { String drl = "query \"my query A\"\n" + " $r : String()\n" + "end"; String someText = "This is some text"; KieHelper helper = new KieHelper(); helper.addContent( drl, ResourceType.DRL ); KieSession ksession = helper.build().newKieSession(); List<Command<?>> cmds = new ArrayList<Command<?>>(); cmds.add( CommandFactory.newInsert( someText )); cmds.add( CommandFactory.newQuery( "results A", "my query A" )); cmds.add( CommandFactory.newQuery( "results NE", "an non-existent query" )); ExecutionResults results = ksession.execute( CommandFactory.newBatchExecution( cmds ) ); QueryResults queryResults = ( QueryResults ) results.getValue( "results A" ); assertEquals(someText, queryResults.iterator().next().get("$r") ); } {code} The test fails on the {{ksession.execute()}} and that's why I wanted to report this I spotted, because I'm afraid might have more deep implications for the drools execution server.