Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Variable precision timestamp support for Hive write operations #5283

Merged
merged 6 commits into from
Nov 3, 2020
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
Expand Up @@ -390,7 +390,7 @@ public HiveTableHandle getTableHandle(ConnectorSession session, SchemaTableName
tableName.getTableName(),
table.get().getParameters(),
getPartitionKeyColumnHandles(table.get(), typeManager),
getRegularColumnHandles(table.get(), typeManager, getTimestampPrecision(session).getPrecision()),
getRegularColumnHandles(table.get(), typeManager, getTimestampPrecision(session)),
getHiveBucketHandle(session, table.get(), typeManager));
}

Expand Down Expand Up @@ -566,7 +566,7 @@ private ConnectorTableMetadata doGetTableMetadata(ConnectorSession session, Sche

Function<HiveColumnHandle, ColumnMetadata> metadataGetter = columnMetadataGetter(table);
ImmutableList.Builder<ColumnMetadata> columns = ImmutableList.builder();
for (HiveColumnHandle columnHandle : hiveColumnHandles(table, typeManager, getTimestampPrecision(session).getPrecision())) {
for (HiveColumnHandle columnHandle : hiveColumnHandles(table, typeManager, getTimestampPrecision(session))) {
columns.add(metadataGetter.apply(columnHandle));
}

Expand Down Expand Up @@ -710,7 +710,7 @@ public Map<String, ColumnHandle> getColumnHandles(ConnectorSession session, Conn
SchemaTableName tableName = ((HiveTableHandle) tableHandle).getSchemaTableName();
Table table = metastore.getTable(new HiveIdentity(session), tableName.getSchemaName(), tableName.getTableName())
.orElseThrow(() -> new TableNotFoundException(tableName));
return hiveColumnHandles(table, typeManager, getTimestampPrecision(session).getPrecision()).stream()
return hiveColumnHandles(table, typeManager, getTimestampPrecision(session)).stream()
.collect(toImmutableMap(HiveColumnHandle::getName, identity()));
}

Expand Down Expand Up @@ -849,7 +849,7 @@ public void createTable(ConnectorSession session, ConnectorTableMetadata tableMe
throw new PrestoException(NOT_SUPPORTED, "Bucketing/Partitioning columns not supported when Avro schema url is set");
}

validateTimestampColumns(tableMetadata.getColumns());
validateTimestampColumns(tableMetadata.getColumns(), getTimestampPrecision(session));
List<HiveColumnHandle> columnHandles = getColumnHandles(tableMetadata, ImmutableSet.copyOf(partitionedBy));
HiveStorageFormat hiveStorageFormat = getHiveStorageFormat(tableMetadata.getProperties());
Map<String, String> tableProperties = getEmptyTableProperties(tableMetadata, bucketProperty, new HdfsContext(session, schemaName, tableName));
Expand Down Expand Up @@ -1236,11 +1236,11 @@ public void finishStatisticsCollection(ConnectorSession session, ConnectorTableH
List<String> partitionColumnNames = partitionColumns.stream()
.map(Column::getName)
.collect(toImmutableList());
// TODO: revisit when handling write path
List<HiveColumnHandle> hiveColumnHandles = hiveColumnHandles(table, typeManager, TimestampType.DEFAULT_PRECISION);
HiveTimestampPrecision timestampPrecision = getTimestampPrecision(session);
List<HiveColumnHandle> hiveColumnHandles = hiveColumnHandles(table, typeManager, timestampPrecision);
Map<String, Type> columnTypes = hiveColumnHandles.stream()
.filter(columnHandle -> !columnHandle.isHidden())
.collect(toImmutableMap(HiveColumnHandle::getName, column -> column.getHiveType().getType(typeManager)));
.collect(toImmutableMap(HiveColumnHandle::getName, column -> column.getHiveType().getType(typeManager, timestampPrecision)));

Map<List<String>, ComputedStatistics> computedStatisticsMap = createComputedStatisticsToPartitionMap(computedStatistics, partitionColumnNames, columnTypes);

Expand Down Expand Up @@ -1568,7 +1568,7 @@ public HiveInsertTableHandle beginInsert(ConnectorSession session, ConnectorTabl
}
}

List<HiveColumnHandle> handles = hiveColumnHandles(table, typeManager, getTimestampPrecision(session).getPrecision()).stream()
List<HiveColumnHandle> handles = hiveColumnHandles(table, typeManager, getTimestampPrecision(session)).stream()
.filter(columnHandle -> !columnHandle.isHidden())
.collect(toList());

Expand Down Expand Up @@ -2447,7 +2447,7 @@ else if (hiveTableHandle.isInAcidTransaction()) {
@Override
public Optional<ConnectorNewTableLayout> getNewTableLayout(ConnectorSession session, ConnectorTableMetadata tableMetadata)
{
validateTimestampColumns(tableMetadata.getColumns());
validateTimestampColumns(tableMetadata.getColumns(), getTimestampPrecision(session));
validatePartitionColumns(tableMetadata);
validateBucketColumns(tableMetadata);
validateColumns(tableMetadata);
Expand Down Expand Up @@ -2502,11 +2502,12 @@ public TableStatisticsMetadata getStatisticsCollectionMetadata(ConnectorSession

private TableStatisticsMetadata getStatisticsCollectionMetadata(List<ColumnMetadata> columns, List<String> partitionedBy, Optional<Set<String>> analyzeColumns, boolean includeRowCount)
{
validateTimestampColumns(columns);
Set<ColumnStatisticMetadata> columnStatistics = columns.stream()
.filter(column -> !partitionedBy.contains(column.getName()))
.filter(column -> !column.isHidden())
.filter(column -> analyzeColumns.isEmpty() || analyzeColumns.get().contains(column.getName()))
// TODO: we only support stats collection at millis precision for now (https://github.com/prestosql/presto/issues/5170)
.filter(column -> !(column.getType() instanceof TimestampType) || column.getType() == TIMESTAMP_MILLIS)
.map(this::getColumnStatisticMetadata)
.flatMap(List::stream)
.collect(toImmutableSet());
Expand Down Expand Up @@ -2720,14 +2721,14 @@ private static void validateColumns(ConnectorTableMetadata tableMetadata)
}
}

// temporary, until variable precision timestamps are supported on write
private static void validateTimestampColumns(List<ColumnMetadata> columns)
// TODO validate timestamps in structural types (https://github.com/prestosql/presto/issues/5195)
private static void validateTimestampColumns(List<ColumnMetadata> columns, HiveTimestampPrecision timestampPrecision)
aalbu marked this conversation as resolved.
Show resolved Hide resolved
{
for (ColumnMetadata column : columns) {
Type type = column.getType();
if (type instanceof TimestampType) {
if (type != TIMESTAMP_MILLIS) {
throw new PrestoException(NOT_SUPPORTED, "CREATE TABLE, INSERT and ANALYZE are not supported with requested timestamp precision: " + type);
if (((TimestampType) type).getPrecision() != timestampPrecision.getPrecision()) {
throw new PrestoException(NOT_SUPPORTED, format("Incorrect timestamp precision for %s; the configured precision is %s", type, timestampPrecision));
}
}
}
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -127,12 +127,12 @@ public Type getType(TypeManager typeManager)
return typeManager.getType(getTypeSignature());
}

public Type getType(TypeManager typeManager, int timestampPrecision)
public Type getType(TypeManager typeManager, HiveTimestampPrecision timestampPrecision)
{
Type tentativeType = typeManager.getType(getTypeSignature());
// TODO: handle timestamps in structural types (https://github.com/prestosql/presto/issues/5195)
if (tentativeType instanceof TimestampType) {
return TimestampType.createTimestampType(timestampPrecision);
return TimestampType.createTimestampType(timestampPrecision.getPrecision());
}
return tentativeType;
}
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -84,6 +84,7 @@
import static io.prestosql.plugin.hive.HiveSessionProperties.getCompressionCodec;
import static io.prestosql.plugin.hive.HiveSessionProperties.getInsertExistingPartitionsBehavior;
import static io.prestosql.plugin.hive.HiveSessionProperties.getTemporaryStagingDirectoryPath;
import static io.prestosql.plugin.hive.HiveSessionProperties.getTimestampPrecision;
import static io.prestosql.plugin.hive.HiveSessionProperties.isTemporaryStagingDirectoryEnabled;
import static io.prestosql.plugin.hive.LocationHandle.WriteMode.DIRECT_TO_TARGET_EXISTING_DIRECTORY;
import static io.prestosql.plugin.hive.metastore.MetastoreUtil.getHiveSchema;
Expand Down Expand Up @@ -544,7 +545,7 @@ else if (insertExistingPartitionsBehavior == InsertExistingPartitionsBehavior.ER
}

List<Type> types = dataColumns.stream()
.map(column -> column.getHiveType().getType(typeManager))
.map(column -> column.getHiveType().getType(typeManager, getTimestampPrecision(session)))
.collect(toImmutableList());

Map<String, Integer> columnIndexes = new HashMap<>();
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,7 @@
import static io.prestosql.plugin.hive.HiveErrorCode.HIVE_WRITE_VALIDATION_FAILED;
import static io.prestosql.plugin.hive.HiveMetadata.PRESTO_QUERY_ID_NAME;
import static io.prestosql.plugin.hive.HiveMetadata.PRESTO_VERSION_NAME;
import static io.prestosql.plugin.hive.HiveSessionProperties.getTimestampPrecision;
import static io.prestosql.plugin.hive.HiveSessionProperties.isRcfileOptimizedWriterValidate;
import static io.prestosql.plugin.hive.rcfile.RcFilePageSourceFactory.createTextVectorEncoding;
import static io.prestosql.plugin.hive.util.HiveUtil.getColumnNames;
Expand Down Expand Up @@ -122,7 +123,7 @@ else if (ColumnarSerDe.class.getName().equals(storageFormat.getSerDe())) {
// an index to rearrange columns in the proper order
List<String> fileColumnNames = getColumnNames(schema);
List<Type> fileColumnTypes = getColumnTypes(schema).stream()
.map(hiveType -> hiveType.getType(typeManager))
.map(hiveType -> hiveType.getType(typeManager, getTimestampPrecision(session)))
.collect(toList());

int[] fileInputColumnIndexes = fileColumnNames.stream()
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -45,6 +45,7 @@
import static com.google.common.collect.ImmutableList.toImmutableList;
import static io.prestosql.plugin.hive.HiveErrorCode.HIVE_WRITER_CLOSE_ERROR;
import static io.prestosql.plugin.hive.HiveErrorCode.HIVE_WRITER_DATA_ERROR;
import static io.prestosql.plugin.hive.HiveSessionProperties.getTimestampPrecision;
import static io.prestosql.plugin.hive.util.HiveUtil.getColumnNames;
import static io.prestosql.plugin.hive.util.HiveUtil.getColumnTypes;
import static io.prestosql.plugin.hive.util.HiveWriteUtils.createRecordWriter;
Expand Down Expand Up @@ -90,7 +91,7 @@ public RecordFileWriter(
// existing tables may have columns in a different order
List<String> fileColumnNames = getColumnNames(schema);
List<Type> fileColumnTypes = getColumnTypes(schema).stream()
.map(hiveType -> hiveType.getType(typeManager))
.map(hiveType -> hiveType.getType(typeManager, getTimestampPrecision(session)))
.collect(toList());

fieldCount = fileColumnNames.size();
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -64,6 +64,7 @@
import static io.prestosql.plugin.hive.HiveSessionProperties.getOrcOptimizedWriterMinStripeSize;
import static io.prestosql.plugin.hive.HiveSessionProperties.getOrcOptimizedWriterValidateMode;
import static io.prestosql.plugin.hive.HiveSessionProperties.getOrcStringStatisticsLimit;
import static io.prestosql.plugin.hive.HiveSessionProperties.getTimestampPrecision;
import static io.prestosql.plugin.hive.HiveSessionProperties.isOrcOptimizedWriterValidate;
import static io.prestosql.plugin.hive.acid.AcidSchema.ACID_COLUMN_NAMES;
import static io.prestosql.plugin.hive.acid.AcidSchema.createAcidColumnPrestoTypes;
Expand Down Expand Up @@ -148,7 +149,7 @@ public Optional<FileWriter> createFileWriter(
// an index to rearrange columns in the proper order
List<String> fileColumnNames = getColumnNames(schema);
List<Type> fileColumnTypes = getColumnTypes(schema).stream()
.map(hiveType -> hiveType.getType(typeManager))
.map(hiveType -> hiveType.getType(typeManager, getTimestampPrecision(session)))
.collect(toList());

int[] fileInputColumnIndexes = fileColumnNames.stream()
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -42,6 +42,7 @@
import java.util.concurrent.Callable;

import static io.prestosql.plugin.hive.HiveErrorCode.HIVE_WRITER_OPEN_ERROR;
import static io.prestosql.plugin.hive.HiveSessionProperties.getTimestampPrecision;
import static io.prestosql.plugin.hive.util.HiveUtil.getColumnNames;
import static io.prestosql.plugin.hive.util.HiveUtil.getColumnTypes;
import static java.util.Objects.requireNonNull;
Expand Down Expand Up @@ -91,7 +92,7 @@ public Optional<FileWriter> createFileWriter(

List<String> fileColumnNames = getColumnNames(schema);
List<Type> fileColumnTypes = getColumnTypes(schema).stream()
.map(hiveType -> hiveType.getType(typeManager))
.map(hiveType -> hiveType.getType(typeManager, getTimestampPrecision(session)))
.collect(toList());

int[] fileInputColumnIndexes = fileColumnNames.stream()
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,7 @@
import io.prestosql.plugin.hive.HiveColumnHandle;
import io.prestosql.plugin.hive.HiveConfig;
import io.prestosql.plugin.hive.HivePageSourceFactory;
import io.prestosql.plugin.hive.HiveTimestampPrecision;
import io.prestosql.plugin.hive.ReaderProjections;
import io.prestosql.plugin.hive.acid.AcidTransaction;
import io.prestosql.plugin.hive.util.FSDataInputStreamTail;
Expand Down Expand Up @@ -184,7 +185,7 @@ else if (deserializerClassName.equals(ColumnarSerDe.class.getName())) {

try {
ImmutableMap.Builder<Integer, Type> readColumns = ImmutableMap.builder();
int timestampPrecision = getTimestampPrecision(session).getPrecision();
HiveTimestampPrecision timestampPrecision = getTimestampPrecision(session);
for (HiveColumnHandle column : projectedReaderColumns) {
readColumns.put(column.getBaseHiveColumnIndex(), column.getHiveType().getType(typeManager, timestampPrecision));
}
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -24,8 +24,10 @@
import io.prestosql.spi.type.DecimalType;
import io.prestosql.spi.type.DoubleType;
import io.prestosql.spi.type.IntegerType;
import io.prestosql.spi.type.LongTimestamp;
import io.prestosql.spi.type.RealType;
import io.prestosql.spi.type.SmallintType;
import io.prestosql.spi.type.TimestampType;
import io.prestosql.spi.type.TinyintType;
import io.prestosql.spi.type.Type;
import io.prestosql.spi.type.VarbinaryType;
Expand Down Expand Up @@ -56,10 +58,13 @@
import static io.prestosql.plugin.hive.util.HiveUtil.isMapType;
import static io.prestosql.plugin.hive.util.HiveUtil.isRowType;
import static io.prestosql.plugin.hive.util.HiveWriteUtils.getHiveDecimal;
import static io.prestosql.spi.type.TimestampType.TIMESTAMP_MILLIS;
import static io.prestosql.spi.type.Timestamps.MICROSECONDS_PER_MILLISECOND;
import static io.prestosql.spi.type.Timestamps.MICROSECONDS_PER_SECOND;
import static io.prestosql.spi.type.Timestamps.MILLISECONDS_PER_SECOND;
import static io.prestosql.spi.type.Timestamps.PICOSECONDS_PER_MICROSECOND;
import static io.prestosql.spi.type.Timestamps.PICOSECONDS_PER_NANOSECOND;
import static java.lang.Float.intBitsToFloat;
import static java.lang.Math.floorDiv;
import static java.lang.Math.floorMod;
import static java.lang.Math.toIntExact;
import static java.util.Objects.requireNonNull;

Expand Down Expand Up @@ -118,8 +123,8 @@ public FieldSetter create(SettableStructObjectInspector rowInspector, Object row
return new DateFieldSetter(rowInspector, row, field);
}

if (type.equals(TIMESTAMP_MILLIS)) {
return new TimestampFieldSetter(rowInspector, row, field, timeZone);
if (type instanceof TimestampType) {
return new TimestampFieldSetter(rowInspector, row, field, (TimestampType) type, timeZone);
}

if (type instanceof DecimalType) {
Expand Down Expand Up @@ -365,22 +370,49 @@ private static class TimestampFieldSetter
extends FieldSetter
{
private final DateTimeZone timeZone;
private final TimestampType type;
private final TimestampWritableV2 value = new TimestampWritableV2();

public TimestampFieldSetter(SettableStructObjectInspector rowInspector, Object row, StructField field, DateTimeZone timeZone)
public TimestampFieldSetter(SettableStructObjectInspector rowInspector, Object row, StructField field, TimestampType type, DateTimeZone timeZone)
{
super(rowInspector, row, field);
this.type = requireNonNull(type, "type is null");
this.timeZone = requireNonNull(timeZone, "timeZone is null");
}

@Override
public void setField(Block block, int position)
{
long epochMilli = floorDiv(TIMESTAMP_MILLIS.getLong(block, position), MICROSECONDS_PER_MILLISECOND);
epochMilli = timeZone.convertLocalToUTC(epochMilli, false);
value.set(Timestamp.ofEpochMilli(epochMilli));
long epochMicros;
aalbu marked this conversation as resolved.
Show resolved Hide resolved
int picosOfMicro;
if (type.isShort()) {
epochMicros = type.getLong(block, position);
picosOfMicro = 0;
}
else {
LongTimestamp longTimestamp = (LongTimestamp) type.getObject(block, position);
epochMicros = longTimestamp.getEpochMicros();
picosOfMicro = longTimestamp.getPicosOfMicro();
}

long epochSeconds = floorDiv(epochMicros, MICROSECONDS_PER_SECOND);
long picosOfSecond = (long) floorMod(epochMicros, MICROSECONDS_PER_SECOND) * PICOSECONDS_PER_MICROSECOND + picosOfMicro;

epochSeconds = convertLocalEpochSecondsToUtc(epochSeconds);
// no rounding since the the data has nanosecond precision, at most
int nanosOfSecond = toIntExact(picosOfSecond / PICOSECONDS_PER_NANOSECOND);

Timestamp timestamp = Timestamp.ofEpochSecond(epochSeconds, nanosOfSecond);
value.set(timestamp);
rowInspector.setStructFieldData(row, field, value);
}

private long convertLocalEpochSecondsToUtc(long epochSeconds)
{
long epochMillis = epochSeconds * MILLISECONDS_PER_SECOND;
epochMillis = timeZone.convertLocalToUTC(epochMillis, false);
return epochMillis / MILLISECONDS_PER_SECOND;
}
}

private static class DecimalFieldSetter
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,7 @@
import io.prestosql.plugin.hive.HiveBucketProperty;
import io.prestosql.plugin.hive.HiveColumnHandle;
import io.prestosql.plugin.hive.HiveTableHandle;
import io.prestosql.plugin.hive.HiveTimestampPrecision;
import io.prestosql.plugin.hive.HiveType;
import io.prestosql.plugin.hive.metastore.Column;
import io.prestosql.plugin.hive.metastore.Table;
Expand Down Expand Up @@ -184,7 +185,7 @@ public static Optional<HiveBucketHandle> getHiveBucketHandle(ConnectorSession se
return Optional.empty();
}

int timestampPrecision = getTimestampPrecision(session).getPrecision();
HiveTimestampPrecision timestampPrecision = getTimestampPrecision(session);
Map<String, HiveColumnHandle> map = getRegularColumnHandles(table, typeManager, timestampPrecision).stream()
.collect(Collectors.toMap(HiveColumnHandle::getName, identity()));

Expand Down
Loading