org.apache.drill.exec.physical.impl.scan.project

Class TestNullColumnLoader

java.lang.Object

org.apache.drill.test.BaseTest

org.apache.drill.test.DrillTest

org.apache.drill.test.SubOperatorTest

org.apache.drill.exec.physical.impl.scan.project.TestNullColumnLoader

public class TestNullColumnLoader
extends SubOperatorTest

Test the mechanism that handles all-null columns during projection. An all-null column is one projected in the query, but which does not actually exist in the underlying data source (or input operator.)

In anticipation of having type information, this mechanism can create the classic nullable Int null column, or one of any other type and mode.

Nested Class Summary

Nested classes/interfaces inherited from class org.apache.drill.test.DrillTest

DrillTest.MemWatcher

Field Summary

Fields inherited from class org.apache.drill.test.SubOperatorTest

dirTestWatcher, fixture

Fields inherited from class org.apache.drill.test.DrillTest

logOutcome, objectMapper, REPEAT_RULE, thrownException, TIMEOUT

Constructor Summary

Constructors

Constructor and Description
TestNullColumnLoader()

Method Summary

Modifier and Type	Method and Description
void	testBasics() Test the simplest case: default null type, nothing in the vector cache.
void	testCachedTypesAllowRequired() Suppose, in the previous test, that one of the columns that goes missing is a required column.
void	testCachedTypesMapToNullable() Drill requires "schema persistence": if a scan operator reads two files, F1 and F2, then the scan operator must provide the same vectors from both readers.
void	testCustomNullType() Test the ability to use a type other than nullable INT for null columns.
void	testDefaultValue() Test the ability to provide a default value for a "null" column.
void	testNullColumnBuilder() Test the shim class that adapts between the null column loader and the projection mechanism.
void	testNullColumnBuilderWithSchema() Test using an output schema, along with a default value property, to define a default value for missing columns.
void	testSchemaWithConflicts() Test the various conflicts that can occur: Schema is required, but no default value for null column. Query wants a different type than that in the schema. Query wants a different mode than that in the schema. The type and mode provided to the builder is that which would result from schema smoothing.

Methods inherited from class org.apache.drill.test.SubOperatorTest

classSetup, classTeardown

Methods inherited from class org.apache.drill.test.DrillTest

escapeJsonString, finishDrillTest, initDrillTest

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

TestNullColumnLoader

public TestNullColumnLoader()

Method Detail

testBasics

public void testBasics()

Test the simplest case: default null type, nothing in the vector cache. Specify no column type, the special NULL type, or a predefined type. Output types should be set accordingly.

testCustomNullType

public void testCustomNullType()

Test the ability to use a type other than nullable INT for null columns. This occurs, for example, in the CSV reader where no column is ever INT (nullable or otherwise) and we want our null columns to be (non-nullable) VARCHAR.

testDefaultValue

public void testDefaultValue()

Test the ability to provide a default value for a "null" column. Default values are only allowed for required "null" columns. For nullable columns, NULL is already the default.

testCachedTypesMapToNullable

public void testCachedTypesMapToNullable()

Drill requires "schema persistence": if a scan operator reads two files, F1 and F2, then the scan operator must provide the same vectors from both readers. Not just the same types, the same value vector instances (but, of course, populated with different data.)

Test the case in which the reader for F1 found columns (a, b, c) but, F2 found only (a, b), requiring that we fill in column c, filled with nulls, but of the same type that it was in file F1. We use a vector cache to pull off this trick. This test ensures that the null column mechanism looks in that vector cache when asked to create a nullable column.

testCachedTypesAllowRequired

public void testCachedTypesAllowRequired()

Suppose, in the previous test, that one of the columns that goes missing is a required column. The null-column mechanism can create the "null" column as a required column, then fill it with empty values (zero or "") -- if the scan operator feels doing so would be helpful.

testNullColumnBuilder

public void testNullColumnBuilder()

Test the shim class that adapts between the null column loader and the projection mechanism. The projection mechanism uses this to pull in the null columns which the null column loader has created.

testNullColumnBuilderWithSchema

public void testNullColumnBuilderWithSchema()

Test using an output schema, along with a default value property, to define a default value for missing columns.

testSchemaWithConflicts

public void testSchemaWithConflicts()

Test the various conflicts that can occur:

• Schema is required, but no default value for null column.
• Query wants a different type than that in the schema.
• Query wants a different mode than that in the schema.
The type and mode provided to the builder is that which would result from schema smoothing. The types and modes should usually match, but verify the rules when they don't.

Defaults for null columns are ignored: null columns use NULL as the null value.