use of org.apache.drill.exec.physical.rowSet.RowSet in project drill by apache.
the class TestResultSetLoaderUnions method testListofListofScalar.
/**
* The semantics of the ListVector are such that it allows
* multi-dimensional lists. In this way, it is like a (slightly
* more normalized) version of the repeated list vector. This form
* allows arrays to be null.
* <p>
* This test verifies that the (non-repeated) list vector can
* be used to create multi-dimensional arrays in the result set
* loader layer. However, the rest of Drill does not support this
* functionality at present, so this test is more of a proof-of-
* concept than a necessity.
*/
@Test
public void testListofListofScalar() {
// JSON equivalent: {a: [[1, 2], [3, 4]]}
final ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator());
final RowSetLoader writer = rsLoader.writer();
// Can write a batch as if this was a repeated Varchar, except
// that any value can also be null.
rsLoader.startBatch();
writer.addColumn(MaterializedField.create("a", Types.optional(MinorType.LIST)));
final ArrayWriter outerArray = writer.array("a");
final VariantWriter outerVariant = outerArray.variant();
outerVariant.addMember(MinorType.LIST);
final ArrayWriter innerArray = outerVariant.array();
final VariantWriter innerVariant = innerArray.variant();
innerVariant.addMember(MinorType.INT);
writer.addSingleCol(listValue(listValue(1, 2), listValue(3, 4)));
final RowSet results = fixture.wrap(rsLoader.harvest());
// Verify metadata
final ListVector outer = (ListVector) results.container().getValueVector(0).getValueVector();
final MajorType outerType = outer.getField().getType();
assertEquals(1, outerType.getSubTypeCount());
assertEquals(MinorType.LIST, outerType.getSubType(0));
assertEquals(1, outer.getField().getChildren().size());
final ListVector inner = (ListVector) outer.getDataVector();
assertSame(inner.getField(), outer.getField().getChildren().iterator().next());
final MajorType innerType = inner.getField().getType();
assertEquals(1, innerType.getSubTypeCount());
assertEquals(MinorType.INT, innerType.getSubType(0));
assertEquals(1, inner.getField().getChildren().size());
final ValueVector data = inner.getDataVector();
assertSame(data.getField(), inner.getField().getChildren().iterator().next());
assertEquals(MinorType.INT, data.getField().getType().getMinorType());
assertEquals(DataMode.OPTIONAL, data.getField().getType().getMode());
assertTrue(data instanceof NullableIntVector);
// Note use of TupleMetadata: BatchSchema can't hold the
// structure of a list.
final TupleMetadata expectedSchema = new SchemaBuilder().addList("a").addList().addType(MinorType.INT).resumeUnion().resumeSchema().buildSchema();
final RowSet expected = new RowSetBuilder(fixture.allocator(), expectedSchema).addSingleCol(listValue(listValue(1, 2), listValue(3, 4))).build();
RowSetUtilities.verify(expected, results);
}
use of org.apache.drill.exec.physical.rowSet.RowSet in project drill by apache.
the class TestResultSetLoaderDictArray method testDictValue.
@Test
public void testDictValue() {
TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).addDictArray("d", MinorType.VARCHAR).dictValue().key(MinorType.VARCHAR).value(MinorType.VARCHAR).resumeDict().resumeSchema().buildSchema();
ResultSetLoaderImpl.ResultSetOptions options = new ResultSetOptionBuilder().readerSchema(schema).build();
ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
RowSetLoader rootWriter = rsLoader.writer();
// Write a couple of rows
rsLoader.startBatch();
rootWriter.addRow(10, objArray(map("a", map("a", "a1", "b", "a2", "c", "a3"), "b", map("d", "a4"), "c", map()), map("b", map("b", "a2")))).addRow(20, objArray()).addRow(30, objArray(map("a", map("a", "b1", "b", "b1")), map("b", map("e", "b2"), "a", map("h", "b1", "g", "b3"), "c", map("a", "b4")), map("b", map("a", "b3", "c", "c3"), "a", map())));
// Verify the batch
RowSet actual = fixture.wrap(rsLoader.harvest());
SingleRowSet expected = fixture.rowSetBuilder(schema).addRow(10, objArray(map("a", map("a", "a1", "b", "a2", "c", "a3"), "b", map("d", "a4"), "c", map()), map("b", map("b", "a2")))).addRow(20, objArray()).addRow(30, objArray(map("a", map("a", "b1", "b", "b1")), map("b", map("e", "b2"), "a", map("h", "b1", "g", "b3"), "c", map("a", "b4")), map("b", map("a", "b3", "c", "c3"), "a", map()))).build();
RowSetUtilities.verify(expected, actual);
rsLoader.close();
}
use of org.apache.drill.exec.physical.rowSet.RowSet in project drill by apache.
the class TestResultSetLoaderMapArray method testBasics.
@Test
public void testBasics() {
TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).addMapArray("m").add("c", MinorType.INT).add("d", MinorType.VARCHAR).resumeSchema().buildSchema();
ResultSetLoaderImpl.ResultSetOptions options = new ResultSetOptionBuilder().readerSchema(schema).build();
ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
RowSetLoader rootWriter = rsLoader.writer();
// Verify structure and schema
TupleMetadata actualSchema = rootWriter.tupleSchema();
assertEquals(2, actualSchema.size());
assertTrue(actualSchema.metadata(1).isArray());
assertTrue(actualSchema.metadata(1).isMap());
assertEquals(2, actualSchema.metadata("m").tupleSchema().size());
assertEquals(2, actualSchema.column("m").getChildren().size());
TupleWriter mapWriter = rootWriter.array("m").tuple();
assertSame(actualSchema.metadata("m").tupleSchema(), mapWriter.schema().tupleSchema());
assertSame(mapWriter.tupleSchema(), mapWriter.schema().tupleSchema());
assertSame(mapWriter.tupleSchema().metadata(0), mapWriter.scalar(0).schema());
assertSame(mapWriter.tupleSchema().metadata(1), mapWriter.scalar(1).schema());
// Write a couple of rows with arrays.
rsLoader.startBatch();
rootWriter.addRow(10, mapArray(mapValue(110, "d1.1"), mapValue(120, "d2.2"))).addRow(20, mapArray()).addRow(30, mapArray(mapValue(310, "d3.1"), mapValue(320, "d3.2"), mapValue(330, "d3.3")));
// Verify the first batch
RowSet actual = fixture.wrap(rsLoader.harvest());
RepeatedMapVector mapVector = (RepeatedMapVector) actual.container().getValueVector(1).getValueVector();
MaterializedField mapField = mapVector.getField();
assertEquals(2, mapField.getChildren().size());
Iterator<MaterializedField> iter = mapField.getChildren().iterator();
assertTrue(mapWriter.scalar(0).schema().schema().isEquivalent(iter.next()));
assertTrue(mapWriter.scalar(1).schema().schema().isEquivalent(iter.next()));
SingleRowSet expected = fixture.rowSetBuilder(schema).addRow(10, mapArray(mapValue(110, "d1.1"), mapValue(120, "d2.2"))).addRow(20, mapArray()).addRow(30, mapArray(mapValue(310, "d3.1"), mapValue(320, "d3.2"), mapValue(330, "d3.3"))).build();
RowSetUtilities.verify(expected, actual);
// In the second, create a row, then add a map member.
// Should be back-filled to empty for the first row.
rsLoader.startBatch();
rootWriter.addRow(40, mapArray(mapValue(410, "d4.1"), mapValue(420, "d4.2")));
mapWriter.addColumn(SchemaBuilder.columnSchema("e", MinorType.VARCHAR, DataMode.OPTIONAL));
rootWriter.addRow(50, mapArray(mapValue(510, "d5.1", "e5.1"), mapValue(520, "d5.2", null))).addRow(60, mapArray(mapValue(610, "d6.1", "e6.1"), mapValue(620, "d6.2", null), mapValue(630, "d6.3", "e6.3")));
// Verify the second batch
actual = fixture.wrap(rsLoader.harvest());
mapVector = (RepeatedMapVector) actual.container().getValueVector(1).getValueVector();
mapField = mapVector.getField();
assertEquals(3, mapField.getChildren().size());
TupleMetadata expectedSchema = new SchemaBuilder().add("a", MinorType.INT).addMapArray("m").add("c", MinorType.INT).add("d", MinorType.VARCHAR).addNullable("e", MinorType.VARCHAR).resumeSchema().buildSchema();
expected = fixture.rowSetBuilder(expectedSchema).addRow(40, mapArray(mapValue(410, "d4.1", null), mapValue(420, "d4.2", null))).addRow(50, mapArray(mapValue(510, "d5.1", "e5.1"), mapValue(520, "d5.2", null))).addRow(60, mapArray(mapValue(610, "d6.1", "e6.1"), mapValue(620, "d6.2", null), mapValue(630, "d6.3", "e6.3"))).build();
RowSetUtilities.verify(expected, actual);
rsLoader.close();
}
use of org.apache.drill.exec.physical.rowSet.RowSet in project drill by apache.
the class TestResultSetLoaderMapArray method testOmittedValues.
/**
* Check that the "fill-empties" logic descends down into
* a repeated map.
*/
@Test
public void testOmittedValues() {
TupleMetadata schema = new SchemaBuilder().add("id", MinorType.INT).addMapArray("m").addNullable("a", MinorType.INT).addNullable("b", MinorType.VARCHAR).resumeSchema().buildSchema();
ResultSetLoaderImpl.ResultSetOptions options = new ResultSetOptionBuilder().readerSchema(schema).rowCountLimit(ValueVector.MAX_ROW_COUNT).build();
ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
RowSetLoader rootWriter = rsLoader.writer();
int mapSkip = 5;
int entrySkip = 3;
int rowCount = 1000;
int entryCount = 10;
rsLoader.startBatch();
ArrayWriter maWriter = rootWriter.array("m");
TupleWriter mWriter = maWriter.tuple();
for (int i = 0; i < rowCount; i++) {
rootWriter.start();
rootWriter.scalar(0).setInt(i);
if (i % mapSkip != 0) {
for (int j = 0; j < entryCount; j++) {
if (j % entrySkip != 0) {
mWriter.scalar(0).setInt(i * entryCount + j);
mWriter.scalar(1).setString("b-" + i + "." + j);
}
maWriter.save();
}
}
rootWriter.save();
}
RowSet result = fixture.wrap(rsLoader.harvest());
assertEquals(rowCount, result.rowCount());
RowSetReader reader = result.reader();
ArrayReader maReader = reader.array("m");
TupleReader mReader = maReader.tuple();
for (int i = 0; i < rowCount; i++) {
assertTrue(reader.next());
assertEquals(i, reader.scalar(0).getInt());
if (i % mapSkip == 0) {
assertEquals(0, maReader.size());
continue;
}
assertEquals(entryCount, maReader.size());
for (int j = 0; j < entryCount; j++) {
assertTrue(maReader.next());
if (j % entrySkip == 0) {
assertTrue(mReader.scalar(0).isNull());
assertTrue(mReader.scalar(1).isNull());
} else {
assertFalse(mReader.scalar(0).isNull());
assertFalse(mReader.scalar(1).isNull());
assertEquals(i * entryCount + j, mReader.scalar(0).getInt());
assertEquals("b-" + i + "." + j, mReader.scalar(1).getString());
}
}
}
result.clear();
rsLoader.close();
}
use of org.apache.drill.exec.physical.rowSet.RowSet in project drill by apache.
the class TestResultSetLoaderMapArray method testNestedArray.
@Test
public void testNestedArray() {
TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).addMapArray("m").add("c", MinorType.INT).addArray("d", MinorType.VARCHAR).resumeSchema().buildSchema();
ResultSetLoaderImpl.ResultSetOptions options = new ResultSetOptionBuilder().readerSchema(schema).build();
ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
RowSetLoader rootWriter = rsLoader.writer();
// Write a couple of rows with arrays within arrays.
// (And, of course, the Varchar is actually an array of
// bytes, so that's three array levels.)
rsLoader.startBatch();
rootWriter.addRow(10, mapArray(mapValue(110, strArray("d1.1.1", "d1.1.2")), mapValue(120, strArray("d1.2.1", "d1.2.2")))).addRow(20, mapArray()).addRow(30, mapArray(mapValue(310, strArray("d3.1.1", "d3.2.2")), mapValue(320, strArray()), mapValue(330, strArray("d3.3.1", "d1.2.2"))));
// Verify the batch
RowSet actual = fixture.wrap(rsLoader.harvest());
SingleRowSet expected = fixture.rowSetBuilder(schema).addRow(10, mapArray(mapValue(110, strArray("d1.1.1", "d1.1.2")), mapValue(120, strArray("d1.2.1", "d1.2.2")))).addRow(20, mapArray()).addRow(30, mapArray(mapValue(310, strArray("d3.1.1", "d3.2.2")), mapValue(320, strArray()), mapValue(330, strArray("d3.3.1", "d1.2.2")))).build();
RowSetUtilities.verify(expected, actual);
rsLoader.close();
}
Aggregations