Drill includes several data sets for testing, and also provides some tools for generating test data sets.
There are three primary data sets bundled with drill for testing:
- Sample Data: These are parquet files in the sample-data folder.
- Resource Data: These are data files in the exec/java-exec/src/test/resources folder.
- TPCH Data: These are trimmed down versions of the tpch data sets. They are retrieved and bundled
in the contrib/data maven submodule. They are also accessible on [Apache Drill's S3 bucket](http://apache-drill.s3.amazonaws.com/files/sf-0
.01_tpc-h_parquet.tgz).
When unit tests are running all of the files in these data set are available from the classpath storage plugin. The tpch
files include:
- customer.parquet
- lineitem.parquet
- nation.parquet
- orders.parquet
- part.parquet
- partsup.parquet
- region.parquet
- supplier.parquet
See seventhTest() in ExampleTest for an example of how to do this.
BaseTestQuery (Deprecated Use ClusterFixture or ClusterTest Instead)
When using the BaseDirTestWatcher you
can make sample-data accessible from the dfs
storage plugin by doing the following:
public class TestMyClass {
@ClassRule
public static final BaseDirTestWatcher dirTestWatcher = new BaseDirTestWatcher();
@BeforeClass
public static void setupFiles() {
dirTestWatcher.copyFileToRoot(Paths.get("sample-data", "region.parquet"));
}
@Test
public void simpleTest() {
// dfs.root.`sample-data/region.parquet` will be accessible from my test
}
}
Or if you are extending BaseTestQuery
public class TestMyClass extends BaseTestQuery {
@BeforeClass
public static void setupFiles() {
dirTestWatcher.copyFileToRoot(Paths.get("sample-data", "region.parquet"));
}
@Test
public void simpleTest() {
// dfs.root.`sample-data/region.parquet` will be accessible from my test
}
}
See sixthTest() in ExampleTest for an example of how to do this.
BaseTestQuery (Deprecated Use ClusterFixture or ClusterTest Instead)
When using the BaseDirTestWatcher you
can make data from exec/java-exec/src/test/resources accessible from the dfs
storage plugin by doing the following:
public class TestMyClass {
@ClassRule
public static final BaseDirTestWatcher dirTestWatcher = new BaseDirTestWatcher();
@BeforeClass
public static void setupFiles() {
dirTestWatcher.copyResourceToRoot(Paths.get("join", "empty_part"));
}
@Test
public void simpleTest() {
// src/test/resources/join/empty_part is acessible at dfs.root.`join/empty_part` from my test
}
}
Or if you are extending BaseTestQuery
public class TestMyClass extends BaseTestQuery {
@BeforeClass
public static void setupFiles() {
dirTestWatcher.copyResourceToRoot(Paths.get("join", "empty_part"));
}
@Test
public void simpleTest() {
// src/test/resources/join/empty_part is acessible at dfs.root.`join/empty_part` from my test
}
}
TPCH data is accessible via the classpath storage plugin
cp.`tpch/customer.parquet`
There are a few ways to generate data for testing:
- JsonFileBuilder
- Inline Mock Scanner
- MockRecordReader
The JsonFileBuilder can be used to create json data files. It's useful for creating data files for both integration and unit tests. An example of using the JsonFileBuilder can be found in secondTest() in ExampleTest.
The MockScanner is a special scanner that generates data for a query. It can only be used for integration testing.
The MockScanner can be used directly from sql queries. An example can be found in thirdTest() in ExampleTest.
The example uses the following query
SELECT id_i, name_s10 FROM `mock`.`employees_5`
The select columns encode the name and type of the columns to generate. The table name also encodes the number of records to generate.
- Columns: Columns in the sql query have the form (name)_(type). (name) is the name that is assigned to a column. (type) is the data type
of the column. Valid types are:
- Use i for specifying an integer column.
- Use s(n) for specifying a string column. Replace (n) with an integer which specifies the number of character in each varchar record.
- Table: Tables in the sql query have the form (name)_(n). Where (name) is the name of the source table. (n) specifies the number of records to generate for the source table.