Github user henryr commented on a diff in the pull request:
https://github.com/apache/spark/pull/21070#discussion_r181902045
--- Diff:
sql/core/src/main/java/org/apache/spark/sql/execution/datasources/parquet/VectorizedPlainValuesReader.java
---
@@ -63,115 +59,159 @@ public final void readBooleans(int total,
WritableColumnVector c, int rowId) {
}
}
+ private ByteBuffer getBuffer(int length) {
+ try {
+ return in.slice(length).order(ByteOrder.LITTLE_ENDIAN);
+ } catch (IOException e) {
+ throw new ParquetDecodingException("Failed to read " + length + "
bytes", e);
+ }
+ }
+
@Override
public final void readIntegers(int total, WritableColumnVector c, int
rowId) {
- c.putIntsLittleEndian(rowId, total, buffer, offset -
Platform.BYTE_ARRAY_OFFSET);
- offset += 4 * total;
+ int requiredBytes = total * 4;
+ ByteBuffer buffer = getBuffer(requiredBytes);
+
+ if (buffer.hasArray()) {
+ int offset = buffer.arrayOffset() + buffer.position();
+ c.putIntsLittleEndian(rowId, total, buffer.array(), offset -
Platform.BYTE_ARRAY_OFFSET);
+ } else {
+ for (int i = 0; i < total; i += 1) {
+ c.putInt(rowId + i, buffer.getInt());
+ }
+ }
}
@Override
public final void readLongs(int total, WritableColumnVector c, int
rowId) {
- c.putLongsLittleEndian(rowId, total, buffer, offset -
Platform.BYTE_ARRAY_OFFSET);
- offset += 8 * total;
+ int requiredBytes = total * 8;
+ ByteBuffer buffer = getBuffer(requiredBytes);
+
+ if (buffer.hasArray()) {
+ int offset = buffer.arrayOffset() + buffer.position();
+ c.putLongsLittleEndian(rowId, total, buffer.array(), offset -
Platform.BYTE_ARRAY_OFFSET);
+ } else {
+ for (int i = 0; i < total; i += 1) {
+ c.putLong(rowId + i, buffer.getLong());
+ }
+ }
}
@Override
public final void readFloats(int total, WritableColumnVector c, int
rowId) {
- c.putFloats(rowId, total, buffer, offset - Platform.BYTE_ARRAY_OFFSET);
- offset += 4 * total;
+ int requiredBytes = total * 4;
+ ByteBuffer buffer = getBuffer(requiredBytes);
+
+ if (buffer.hasArray()) {
+ int offset = buffer.arrayOffset() + buffer.position();
+ c.putFloats(rowId, total, buffer.array(), offset -
Platform.BYTE_ARRAY_OFFSET);
+ } else {
+ for (int i = 0; i < total; i += 1) {
+ c.putFloat(rowId + i, buffer.getFloat());
+ }
+ }
}
@Override
public final void readDoubles(int total, WritableColumnVector c, int
rowId) {
- c.putDoubles(rowId, total, buffer, offset -
Platform.BYTE_ARRAY_OFFSET);
- offset += 8 * total;
+ int requiredBytes = total * 8;
+ ByteBuffer buffer = getBuffer(requiredBytes);
+
+ if (buffer.hasArray()) {
+ int offset = buffer.arrayOffset() + buffer.position();
+ c.putDoubles(rowId, total, buffer.array(), offset -
Platform.BYTE_ARRAY_OFFSET);
+ } else {
+ for (int i = 0; i < total; i += 1) {
+ c.putDouble(rowId + i, buffer.getDouble());
+ }
+ }
+ }
+
+ private byte getByte() {
+ try {
+ return (byte) in.read();
+ } catch (IOException e) {
+ throw new ParquetDecodingException("Failed to read a byte", e);
+ }
}
@Override
public final void readBytes(int total, WritableColumnVector c, int
rowId) {
- for (int i = 0; i < total; i++) {
- // Bytes are stored as a 4-byte little endian int. Just read the
first byte.
- // TODO: consider pushing this in ColumnVector by adding a readBytes
with a stride.
- c.putByte(rowId + i, Platform.getByte(buffer, offset));
- offset += 4;
+ int requiredBytes = total * 4;
+ ByteBuffer buffer = getBuffer(requiredBytes);
+
+ for (int i = 0; i < total; i += 1) {
+ c.putByte(rowId + i, buffer.get());
--- End diff --
could you preserve the comment about "Bytes are stored as 4-byte little
endian int. Just read the first byte."?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
