This is an automated email from the ASF dual-hosted git repository. xxyu pushed a commit to branch kylin5 in repository https://gitbox.apache.org/repos/asf/kylin.git
commit 141eb4a7c306712ee5525789fdcb8cd8dbc68cc1 Author: Ruixuan Zhang <ruixuan.zh...@kyligence.io> AuthorDate: Wed Feb 1 15:25:05 2023 +0800 KYLIN-5454 add UT for Async Query Result --- .../kylin/rest/service/AysncQueryServiceTest.java | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) diff --git a/src/query-service/src/test/java/org/apache/kylin/rest/service/AysncQueryServiceTest.java b/src/query-service/src/test/java/org/apache/kylin/rest/service/AysncQueryServiceTest.java index 241ffd3d10..d5303e8345 100644 --- a/src/query-service/src/test/java/org/apache/kylin/rest/service/AysncQueryServiceTest.java +++ b/src/query-service/src/test/java/org/apache/kylin/rest/service/AysncQueryServiceTest.java @@ -59,6 +59,7 @@ import org.apache.hadoop.fs.Path; import org.apache.kylin.common.KapConfig; import org.apache.kylin.common.QueryContext; import org.apache.kylin.common.exception.KylinException; +import org.apache.kylin.common.util.HadoopUtil; import org.apache.kylin.common.util.RandomUtil; import org.apache.kylin.metadata.query.QueryMetricsContext; import org.apache.kylin.metadata.querymeta.SelectedColumnMeta; @@ -220,6 +221,12 @@ public class AysncQueryServiceTest extends ServiceTestBase { new QueryExec(PROJECT, getTestConfig()).executeQuery(sql); assertSame(AsyncQueryService.QueryStatus.SUCCESS, asyncQueryService.queryStatus(PROJECT, queryId)); + String workingDir = getTestConfig().getHdfsWorkingDirectory(PROJECT); + FileSystem fs = HadoopUtil.getWorkingFileSystem(); + Path path = new Path(workingDir + "/async_query_result" + "/" + queryId); + Assert.assertTrue(fs.exists(path)); + Path file = new Path(path + "/" + queryId + "." + "csv"); + Assert.assertTrue(fs.exists(file)); List<org.apache.spark.sql.Row> rowList = ss.read() .csv(asyncQueryService.getAsyncQueryResultDir(PROJECT, queryId).toString()).collectAsList(); @@ -466,6 +473,12 @@ public class AysncQueryServiceTest extends ServiceTestBase { queryContext.setProject(PROJECT); SparkSqlClient.executeSql(ss, sql, UUID.fromString(queryId), PROJECT); assertSame(AsyncQueryService.QueryStatus.SUCCESS, asyncQueryService.queryStatus(PROJECT, queryId)); + String workingDir = getTestConfig().getHdfsWorkingDirectory(PROJECT); + FileSystem fs = HadoopUtil.getWorkingFileSystem(); + Path path = new Path(workingDir + "/async_query_result" + "/" + queryId); + Assert.assertTrue(fs.exists(path)); + Path resultFile = new Path(path + "/" + queryId + "." + "xlsx"); + Assert.assertTrue(fs.exists(resultFile)); HttpServletResponse response = mock(HttpServletResponse.class); ByteArrayOutputStream outputStream = mockOutputStream(response); asyncQueryService.retrieveSavedQueryResult(PROJECT, queryId, response, "xlsx", encodeDefault); @@ -938,6 +951,12 @@ public class AysncQueryServiceTest extends ServiceTestBase { new QueryExec(PROJECT, getTestConfig()).executeQuery(sql); assertSame(AsyncQueryService.QueryStatus.SUCCESS, asyncQueryService.queryStatus(PROJECT, queryId)); + String workingDir = getTestConfig().getHdfsWorkingDirectory(PROJECT); + FileSystem fs = HadoopUtil.getWorkingFileSystem(); + Path path = new Path(workingDir + "/async_query_result" + "/" + queryId); + Assert.assertTrue(fs.exists(path)); + Path file = new Path(path + "/" + queryId + "." + "csv"); + Assert.assertTrue(fs.exists(file)); QueryMetricsContext.start(queryId, ""); Assert.assertTrue(QueryMetricsContext.isStarted());