Skip to content

[SPARK-16334][BACKPORT] Reusing same dictionary column for decoding consecutive row groups shouldn't throw an error #14944

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.

Sign up for GitHub

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

Closed
wants to merge 1 commit into from
Closed

[SPARK-16334][BACKPORT] Reusing same dictionary column for decoding consecutive row groups shouldn't throw an error #14944

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
54 changes: 38 additions & 16 deletions .../main/java/org/apache/spark/sql/execution/datasources/parquet/VectorizedColumnReader.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -221,15 +221,21 @@ private void decodeDictionaryIds(int rowId, int num, ColumnVector column,
if (column.dataType() == DataTypes.IntegerType ||
DecimalType.is32BitDecimalType(column.dataType())) {
for (int i = rowId; i < rowId + num; ++i) {
column.putInt(i, dictionary.decodeToInt(dictionaryIds.getInt(i)));
if (!column.isNullAt(i)) {
column.putInt(i, dictionary.decodeToInt(dictionaryIds.getInt(i)));
}
}
} else if (column.dataType() == DataTypes.ByteType) {
for (int i = rowId; i < rowId + num; ++i) {
column.putByte(i, (byte) dictionary.decodeToInt(dictionaryIds.getInt(i)));
if (!column.isNullAt(i)) {
column.putByte(i, (byte) dictionary.decodeToInt(dictionaryIds.getInt(i)));
}
}
} else if (column.dataType() == DataTypes.ShortType) {
for (int i = rowId; i < rowId + num; ++i) {
column.putShort(i, (short) dictionary.decodeToInt(dictionaryIds.getInt(i)));
if (!column.isNullAt(i)) {
column.putShort(i, (short) dictionary.decodeToInt(dictionaryIds.getInt(i)));
}
}
} else {
throw new UnsupportedOperationException("Unimplemented type: " + column.dataType());
Expand All @@ -240,7 +246,9 @@ private void decodeDictionaryIds(int rowId, int num, ColumnVector column,
if (column.dataType() == DataTypes.LongType ||
DecimalType.is64BitDecimalType(column.dataType())) {
for (int i = rowId; i < rowId + num; ++i) {
column.putLong(i, dictionary.decodeToLong(dictionaryIds.getInt(i)));
if (!column.isNullAt(i)) {
column.putLong(i, dictionary.decodeToLong(dictionaryIds.getInt(i)));
}
}
} else {
throw new UnsupportedOperationException("Unimplemented type: " + column.dataType());
Expand All @@ -249,21 +257,27 @@ private void decodeDictionaryIds(int rowId, int num, ColumnVector column,

case FLOAT:
for (int i = rowId; i < rowId + num; ++i) {
column.putFloat(i, dictionary.decodeToFloat(dictionaryIds.getInt(i)));
if (!column.isNullAt(i)) {
column.putFloat(i, dictionary.decodeToFloat(dictionaryIds.getInt(i)));
}
}
break;

case DOUBLE:
for (int i = rowId; i < rowId + num; ++i) {
column.putDouble(i, dictionary.decodeToDouble(dictionaryIds.getInt(i)));
if (!column.isNullAt(i)) {
column.putDouble(i, dictionary.decodeToDouble(dictionaryIds.getInt(i)));
}
}
break;
case INT96:
if (column.dataType() == DataTypes.TimestampType) {
for (int i = rowId; i < rowId + num; ++i) {
// TODO: Convert dictionary of Binaries to dictionary of Longs
Binary v = dictionary.decodeToBinary(dictionaryIds.getInt(i));
column.putLong(i, ParquetRowConverter.binaryToSQLTimestamp(v));
if (!column.isNullAt(i)) {
Binary v = dictionary.decodeToBinary(dictionaryIds.getInt(i));
column.putLong(i, ParquetRowConverter.binaryToSQLTimestamp(v));
}
}
} else {
throw new UnsupportedOperationException();
Expand All @@ -275,26 +289,34 @@ private void decodeDictionaryIds(int rowId, int num, ColumnVector column,
// and reuse it across batches. This should mean adding a ByteArray would just update
// the length and offset.
for (int i = rowId; i < rowId + num; ++i) {
Binary v = dictionary.decodeToBinary(dictionaryIds.getInt(i));
column.putByteArray(i, v.getBytes());
if (!column.isNullAt(i)) {
Binary v = dictionary.decodeToBinary(dictionaryIds.getInt(i));
column.putByteArray(i, v.getBytes());
}
}
break;
case FIXED_LEN_BYTE_ARRAY:
// DecimalType written in the legacy mode
if (DecimalType.is32BitDecimalType(column.dataType())) {
for (int i = rowId; i < rowId + num; ++i) {
Binary v = dictionary.decodeToBinary(dictionaryIds.getInt(i));
column.putInt(i, (int) ParquetRowConverter.binaryToUnscaledLong(v));
if (!column.isNullAt(i)) {
Binary v = dictionary.decodeToBinary(dictionaryIds.getInt(i));
column.putInt(i, (int) ParquetRowConverter.binaryToUnscaledLong(v));
}
}
} else if (DecimalType.is64BitDecimalType(column.dataType())) {
for (int i = rowId; i < rowId + num; ++i) {
Binary v = dictionary.decodeToBinary(dictionaryIds.getInt(i));
column.putLong(i, ParquetRowConverter.binaryToUnscaledLong(v));
if (!column.isNullAt(i)) {
Binary v = dictionary.decodeToBinary(dictionaryIds.getInt(i));
column.putLong(i, ParquetRowConverter.binaryToUnscaledLong(v));
}
}
} else if (DecimalType.isByteArrayDecimalType(column.dataType())) {
for (int i = rowId; i < rowId + num; ++i) {
Binary v = dictionary.decodeToBinary(dictionaryIds.getInt(i));
column.putByteArray(i, v.getBytes());
if (!column.isNullAt(i)) {
Binary v = dictionary.decodeToBinary(dictionaryIds.getInt(i));
column.putByteArray(i, v.getBytes());
}
}
} else {
throw new UnsupportedOperationException();
Expand Down