use of org.apache.drill.exec.vector.accessor.ScalarWriter in project drill by apache.
the class TestResultSetLoaderRepeatedList method do2DTest.
private void do2DTest(TupleMetadata schema, ResultSetLoader rsLoader) {
final RowSetLoader writer = rsLoader.writer();
// Sanity check of writer structure
assertEquals(2, writer.size());
final ObjectWriter listObj = writer.column("list2");
assertEquals(ObjectType.ARRAY, listObj.type());
final ArrayWriter listWriter = listObj.array();
assertEquals(ObjectType.ARRAY, listWriter.entryType());
final ArrayWriter innerWriter = listWriter.array();
assertEquals(ObjectType.SCALAR, innerWriter.entryType());
final ScalarWriter strWriter = innerWriter.scalar();
assertEquals(ValueType.STRING, strWriter.valueType());
// Sanity test of schema
final TupleMetadata rowSchema = writer.tupleSchema();
assertEquals(2, rowSchema.size());
final ColumnMetadata listSchema = rowSchema.metadata(1);
assertEquals(MinorType.LIST, listSchema.type());
assertEquals(DataMode.REPEATED, listSchema.mode());
assertTrue(listSchema instanceof RepeatedListColumnMetadata);
assertEquals(StructureType.MULTI_ARRAY, listSchema.structureType());
assertNotNull(listSchema.childSchema());
final ColumnMetadata elementSchema = listSchema.childSchema();
assertEquals(listSchema.name(), elementSchema.name());
assertEquals(MinorType.VARCHAR, elementSchema.type());
assertEquals(DataMode.REPEATED, elementSchema.mode());
// Write values
rsLoader.startBatch();
writer.addRow(1, objArray(strArray("a", "b"), strArray("c", "d"))).addRow(2, objArray(strArray("e"), strArray(), strArray("f", "g", "h"))).addRow(3, objArray()).addRow(4, objArray(strArray(), strArray("i"), strArray()));
// Verify the values.
// (Relies on the row set level repeated list tests having passed.)
final RowSet expected = fixture.rowSetBuilder(schema).addRow(1, objArray(strArray("a", "b"), strArray("c", "d"))).addRow(2, objArray(strArray("e"), strArray(), strArray("f", "g", "h"))).addRow(3, objArray()).addRow(4, objArray(strArray(), strArray("i"), strArray())).build();
RowSetUtilities.verify(expected, fixture.wrap(rsLoader.harvest()));
}
use of org.apache.drill.exec.vector.accessor.ScalarWriter in project drill by apache.
the class TestResultSetLoaderUnions method testSimpleList.
/**
* Test for the case of a list defined to contain exactly one type.
* Relies on the row set tests to verify that the single type model
* works for lists. Here we test that the ResultSetLoader put the
* pieces together correctly.
*/
@Test
public void testSimpleList() {
// Schema with a list declared with one type, not expandable
final TupleMetadata schema = new SchemaBuilder().add("id", MinorType.INT).addList("list").addType(MinorType.VARCHAR).resumeSchema().buildSchema();
schema.metadata("list").variantSchema().becomeSimple();
final ResultSetLoaderImpl.ResultSetOptions options = new ResultSetOptionBuilder().readerSchema(schema).build();
final ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
final RowSetLoader writer = rsLoader.writer();
// Sanity check: should be an array of Varchar because we said the
// types within the list is not expandable.
final ArrayWriter arrWriter = writer.array("list");
assertEquals(ObjectType.SCALAR, arrWriter.entryType());
final ScalarWriter strWriter = arrWriter.scalar();
assertEquals(ValueType.STRING, strWriter.valueType());
// Can write a batch as if this was a repeated Varchar, except
// that any value can also be null.
rsLoader.startBatch();
writer.addRow(1, strArray("fred", "barney")).addRow(2, null).addRow(3, strArray("wilma", "betty", "pebbles"));
// Verify
final SingleRowSet expected = fixture.rowSetBuilder(schema).addRow(1, strArray("fred", "barney")).addRow(2, null).addRow(3, strArray("wilma", "betty", "pebbles")).build();
RowSetUtilities.verify(expected, fixture.wrap(rsLoader.harvest()));
}
use of org.apache.drill.exec.vector.accessor.ScalarWriter in project drill by apache.
the class TestResultSetLoaderUnions method testSimpleListDynamic.
/**
* Test a simple list created dynamically at load time.
* The list must include a single type member.
*/
@Test
public void testSimpleListDynamic() {
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("id", Types.required(MinorType.INT)));
final ColumnMetadata colSchema = MetadataUtils.newVariant("list", DataMode.REPEATED);
colSchema.variantSchema().addType(MinorType.VARCHAR);
colSchema.variantSchema().becomeSimple();
writer.addColumn(colSchema);
// Sanity check: should be an array of Varchar because we said the
// types within the list is not expandable.
final ArrayWriter arrWriter = writer.array("list");
assertEquals(ObjectType.SCALAR, arrWriter.entryType());
final ScalarWriter strWriter = arrWriter.scalar();
assertEquals(ValueType.STRING, strWriter.valueType());
writer.addRow(1, strArray("fred", "barney")).addRow(2, null).addRow(3, strArray("wilma", "betty", "pebbles"));
// Verify
final TupleMetadata schema = new SchemaBuilder().add("id", MinorType.INT).addList("list").addType(MinorType.VARCHAR).resumeSchema().buildSchema();
final SingleRowSet expected = fixture.rowSetBuilder(schema).addRow(1, strArray("fred", "barney")).addRow(2, null).addRow(3, strArray("wilma", "betty", "pebbles")).build();
RowSetUtilities.verify(expected, fixture.wrap(rsLoader.harvest()));
}
use of org.apache.drill.exec.vector.accessor.ScalarWriter in project drill by apache.
the class TestResultSetSchemaChange method testSchemaChangeFirstBatch.
/**
* Test the case where the schema changes in the first batch.
* Schema changes before the first record are trivial and tested
* elsewhere. Here we write some records, then add new columns, as a
* JSON reader might do.
*/
@Test
public void testSchemaChangeFirstBatch() {
ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator());
RowSetLoader rootWriter = rsLoader.writer();
rootWriter.addColumn(SchemaBuilder.columnSchema("a", MinorType.VARCHAR, DataMode.REQUIRED));
// Create initial rows
rsLoader.startBatch();
int rowCount = 0;
for (int i = 0; i < 2; i++) {
rootWriter.start();
rowCount++;
rootWriter.scalar(0).setString("a_" + rowCount);
rootWriter.save();
}
// Add a second column: nullable.
rootWriter.addColumn(SchemaBuilder.columnSchema("b", MinorType.INT, DataMode.OPTIONAL));
for (int i = 0; i < 2; i++) {
rootWriter.start();
rowCount++;
rootWriter.scalar(0).setString("a_" + rowCount);
rootWriter.scalar(1).setInt(rowCount);
rootWriter.save();
}
// Add a third column. Use variable-width so that offset
// vectors must be back-filled.
rootWriter.addColumn(SchemaBuilder.columnSchema("c", MinorType.VARCHAR, DataMode.OPTIONAL));
for (int i = 0; i < 2; i++) {
rootWriter.start();
rowCount++;
rootWriter.scalar(0).setString("a_" + rowCount);
rootWriter.scalar(1).setInt(rowCount);
rootWriter.scalar(2).setString("c_" + rowCount);
rootWriter.save();
}
// Fourth: Required Varchar. Previous rows are back-filled with empty strings.
// And a required int. Back-filled with zeros.
// May occasionally be useful. But, does have to work to prevent
// vector corruption if some reader decides to go this route.
rootWriter.addColumn(SchemaBuilder.columnSchema("d", MinorType.VARCHAR, DataMode.REQUIRED));
rootWriter.addColumn(SchemaBuilder.columnSchema("e", MinorType.INT, DataMode.REQUIRED));
for (int i = 0; i < 2; i++) {
rootWriter.start();
rowCount++;
rootWriter.scalar(0).setString("a_" + rowCount);
rootWriter.scalar(1).setInt(rowCount);
rootWriter.scalar(2).setString("c_" + rowCount);
rootWriter.scalar(3).setString("d_" + rowCount);
rootWriter.scalar(4).setInt(rowCount * 10);
rootWriter.save();
}
// Add an array. Now two offset vectors must be back-filled.
rootWriter.addColumn(SchemaBuilder.columnSchema("f", MinorType.VARCHAR, DataMode.REPEATED));
for (int i = 0; i < 2; i++) {
rootWriter.start();
rowCount++;
rootWriter.scalar(0).setString("a_" + rowCount);
rootWriter.scalar(1).setInt(rowCount);
rootWriter.scalar(2).setString("c_" + rowCount);
rootWriter.scalar(3).setString("d_" + rowCount);
rootWriter.scalar(4).setInt(rowCount * 10);
ScalarWriter arrayWriter = rootWriter.column(5).array().scalar();
arrayWriter.setString("f_" + rowCount + "-1");
arrayWriter.setString("f_" + rowCount + "-2");
rootWriter.save();
}
// Harvest the batch and verify.
RowSet actual = fixture.wrap(rsLoader.harvest());
TupleMetadata expectedSchema = new SchemaBuilder().add("a", MinorType.VARCHAR).addNullable("b", MinorType.INT).addNullable("c", MinorType.VARCHAR).add("d", MinorType.VARCHAR).add("e", MinorType.INT).addArray("f", MinorType.VARCHAR).buildSchema();
SingleRowSet expected = fixture.rowSetBuilder(expectedSchema).addRow("a_1", null, null, "", 0, strArray()).addRow("a_2", null, null, "", 0, strArray()).addRow("a_3", 3, null, "", 0, strArray()).addRow("a_4", 4, null, "", 0, strArray()).addRow("a_5", 5, "c_5", "", 0, strArray()).addRow("a_6", 6, "c_6", "", 0, strArray()).addRow("a_7", 7, "c_7", "d_7", 70, strArray()).addRow("a_8", 8, "c_8", "d_8", 80, strArray()).addRow("a_9", 9, "c_9", "d_9", 90, strArray("f_9-1", "f_9-2")).addRow("a_10", 10, "c_10", "d_10", 100, strArray("f_10-1", "f_10-2")).build();
RowSetUtilities.verify(expected, actual);
rsLoader.close();
}
use of org.apache.drill.exec.vector.accessor.ScalarWriter in project drill by apache.
the class TestMapAccessors method testDoubleNestedArray.
/**
* Test a doubly-nested array of maps.
*/
@Test
public void testDoubleNestedArray() {
TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).addMapArray("m1").add("b", MinorType.INT).addMapArray("m2").add("c", MinorType.INT).addArray("d", MinorType.VARCHAR).resumeMap().resumeSchema().buildSchema();
RowSetBuilder builder = fixture.rowSetBuilder(schema);
RowSetWriter rootWriter = builder.writer();
ScalarWriter aWriter = rootWriter.scalar("a");
ArrayWriter a1Writer = rootWriter.array("m1");
TupleWriter m1Writer = a1Writer.tuple();
ScalarWriter bWriter = m1Writer.scalar("b");
ArrayWriter a2Writer = m1Writer.array("m2");
TupleWriter m2Writer = a2Writer.tuple();
ScalarWriter cWriter = m2Writer.scalar("c");
ScalarWriter dWriter = m2Writer.array("d").scalar();
for (int i = 0; i < 5; i++) {
aWriter.setInt(i);
for (int j = 0; j < 4; j++) {
int a1Key = i + 10 + j;
bWriter.setInt(a1Key);
for (int k = 0; k < 3; k++) {
int a2Key = a1Key * 10 + k;
cWriter.setInt(a2Key);
for (int l = 0; l < 2; l++) {
dWriter.setString("d-" + (a2Key * 10 + l));
}
a2Writer.save();
}
a1Writer.save();
}
rootWriter.save();
}
RowSet results = builder.build();
RowSetReader reader = results.reader();
ScalarReader aReader = reader.scalar("a");
ArrayReader a1Reader = reader.array("m1");
TupleReader m1Reader = a1Reader.tuple();
ScalarReader bReader = m1Reader.scalar("b");
ArrayReader a2Reader = m1Reader.array("m2");
TupleReader m2Reader = a2Reader.tuple();
ScalarReader cReader = m2Reader.scalar("c");
ArrayReader dArray = m2Reader.array("d");
ScalarReader dReader = dArray.scalar();
for (int i = 0; i < 5; i++) {
assertTrue(reader.next());
assertEquals(i, aReader.getInt());
for (int j = 0; j < 4; j++) {
assertTrue(a1Reader.next());
int a1Key = i + 10 + j;
assertEquals(a1Key, bReader.getInt());
for (int k = 0; k < 3; k++) {
assertTrue(a2Reader.next());
int a2Key = a1Key * 10 + k;
assertEquals(a2Key, cReader.getInt());
for (int l = 0; l < 2; l++) {
assertTrue(dArray.next());
assertEquals("d-" + (a2Key * 10 + l), dReader.getString());
}
}
}
}
results.clear();
}
Aggregations