Implement high-performance sorted writing support in IcebergIO #38406
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,266 @@ | ||
| /* | ||
| * Licensed to the Apache Software Foundation (ASF) under one | ||
| * or more contributor license agreements. See the NOTICE file | ||
| * distributed with this work for additional information | ||
| * regarding copyright ownership. The ASF licenses this file | ||
| * to you under the Apache License, Version 2.0 (the | ||
| * "License"); you may not use this file except in compliance | ||
| * with the License. You may obtain a copy of the License at | ||
| * | ||
| * http://www.apache.org/licenses/LICENSE-2.0 | ||
| * | ||
| * Unless required by applicable law or agreed to in writing, software | ||
| * distributed under the License is distributed on an "AS IS" BASIS, | ||
| * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| * See the License for the specific language governing permissions and | ||
| * limitations under the License. | ||
| */ | ||
| package org.apache.beam.sdk.io.iceberg; | ||
|
|
||
| import java.io.ByteArrayInputStream; | ||
| import java.io.ByteArrayOutputStream; | ||
| import java.io.IOException; | ||
| import java.io.Serializable; | ||
| import java.nio.ByteBuffer; | ||
| import java.nio.charset.StandardCharsets; | ||
| import java.time.Instant; | ||
| import java.util.Date; | ||
| import java.util.Iterator; | ||
| import java.util.List; | ||
| import org.apache.beam.sdk.coders.RowCoder; | ||
| import org.apache.beam.sdk.extensions.sorter.BufferedExternalSorter; | ||
| import org.apache.beam.sdk.values.KV; | ||
| import org.apache.beam.sdk.values.Row; | ||
| import org.apache.iceberg.NullOrder; | ||
| import org.apache.iceberg.Schema; | ||
| import org.apache.iceberg.SortDirection; | ||
| import org.apache.iceberg.SortField; | ||
| import org.apache.iceberg.SortOrder; | ||
| import org.joda.time.ReadableInstant; | ||
|
|
||
| /** | ||
| * A utility class to sort Beam {@link Row}s based on an Iceberg {@link SortOrder}. Leverages {@link | ||
| * BufferedExternalSorter} to spill to local disk when elements exceed memory limit. | ||
| */ | ||
| class IcebergRowSorter implements Serializable { | ||
|
|
||
| public static Iterable<Row> sortRows( | ||
| Iterable<Row> rows, | ||
| SortOrder sortOrder, | ||
| Schema icebergSchema, | ||
| org.apache.beam.sdk.schemas.Schema beamSchema) { | ||
|
|
||
| if (sortOrder == null || !sortOrder.isSorted()) { | ||
| return rows; | ||
| } | ||
|
|
||
| BufferedExternalSorter.Options sorterOptions = BufferedExternalSorter.options(); | ||
| BufferedExternalSorter sorter = BufferedExternalSorter.create(sorterOptions); | ||
| RowCoder rowCoder = RowCoder.of(beamSchema); | ||
|
|
||
| List<SortField> fields = sortOrder.fields(); | ||
| String[] columnNames = new String[fields.size()]; | ||
| for (int i = 0; i < fields.size(); i++) { | ||
| columnNames[i] = icebergSchema.findColumnName(fields.get(i).sourceId()); | ||
| } | ||
|
|
||
| try { | ||
| for (Row row : rows) { | ||
| byte[] keyBytes = encodeSortKey(row, sortOrder, columnNames, icebergSchema, beamSchema); | ||
| ByteArrayOutputStream baos = new ByteArrayOutputStream(); | ||
| rowCoder.encode(row, baos); | ||
| byte[] valBytes = baos.toByteArray(); | ||
| sorter.add(KV.of(keyBytes, valBytes)); | ||
| } | ||
|
|
||
| Iterable<KV<byte[], byte[]>> sortedKVs = sorter.sort(); | ||
| return new Iterable<Row>() { | ||
| @Override | ||
| public Iterator<Row> iterator() { | ||
| final Iterator<KV<byte[], byte[]>> it = sortedKVs.iterator(); | ||
| return new Iterator<Row>() { | ||
| @Override | ||
| public boolean hasNext() { | ||
| return it.hasNext(); | ||
| } | ||
|
|
||
| @Override | ||
| public Row next() { | ||
| KV<byte[], byte[]> next = it.next(); | ||
| try { | ||
| ByteArrayInputStream bais = new ByteArrayInputStream(next.getValue()); | ||
| return rowCoder.decode(bais); | ||
| } catch (IOException e) { | ||
| throw new RuntimeException("Failed to decode Row during sorting", e); | ||
| } | ||
| } | ||
| }; | ||
| } | ||
| }; | ||
|
|
||
| } catch (IOException e) { | ||
| throw new RuntimeException("Failed to sort rows with external sorter", e); | ||
| } | ||
| } | ||
|
|
||
| @SuppressWarnings("nullness") | ||
| public static byte[] encodeSortKey( | ||
| Row row, | ||
| SortOrder sortOrder, | ||
| String[] columnNames, | ||
| Schema icebergSchema, | ||
| org.apache.beam.sdk.schemas.Schema beamSchema) | ||
| throws IOException { | ||
|
|
||
| ByteArrayOutputStream baos = new ByteArrayOutputStream(); | ||
|
Contributor
There was a problem hiding this comment. Same thing here, let's create |
||
| List<SortField> fields = sortOrder.fields(); | ||
| org.apache.iceberg.data.Record icebergRecord = null; | ||
|
|
||
| for (int i = 0; i < fields.size(); i++) { | ||
| SortField field = fields.get(i); | ||
| String colName = columnNames[i]; | ||
| Object val = row.getValue(colName); | ||
|
|
||
| if (!field.transform().isIdentity()) { | ||
| if (icebergRecord == null) { | ||
| icebergRecord = IcebergUtils.beamRowToIcebergRecord(icebergSchema, row); | ||
| } | ||
| Object icebergVal = icebergRecord.getField(colName); | ||
| if (icebergVal != null) { | ||
| val = field.transform().apply(icebergVal); | ||
|
Contributor
There was a problem hiding this comment. This is pretty heavy to extract the transformed value. Will make writes pretty slow Can we apply the transform to the Beam field object directly? If not, maybe we need a "beam object to iceberg object" conversion. IcebergUtils.copyFieldIntoRecord does something similar, but we may need to refactor it a little to fit this use case. |
||
| } else { | ||
| val = null; | ||
| } | ||
| } | ||
|
|
||
| boolean isNull = (val == null); | ||
| boolean isDesc = (field.direction() == SortDirection.DESC); | ||
| boolean nullsFirst = (field.nullOrder() == NullOrder.NULLS_FIRST); | ||
|
|
||
| // Determine correct header prefix to fulfill the NullOrder contracts | ||
| byte prefixByte; | ||
| if (isNull) { | ||
| prefixByte = nullsFirst ? (byte) 0x00 : (byte) 0xFF; | ||
| } else { | ||
| prefixByte = nullsFirst ? (byte) 0x01 : (byte) 0x00; | ||
| } | ||
|
|
||
| baos.write(prefixByte); | ||
|
|
||
| if (!isNull) { | ||
| writeValue(val, baos, isDesc); | ||
| } | ||
| } | ||
|
|
||
| return baos.toByteArray(); | ||
| } | ||
|
|
||
| private static void writeInt(int v, ByteArrayOutputStream baos, boolean invert) { | ||
| byte b3 = (byte) (v >>> 24); | ||
| byte b2 = (byte) (v >>> 16); | ||
| byte b1 = (byte) (v >>> 8); | ||
| byte b0 = (byte) v; | ||
| if (invert) { | ||
| baos.write(~b3); | ||
| baos.write(~b2); | ||
| baos.write(~b1); | ||
| baos.write(~b0); | ||
| } else { | ||
| baos.write(b3); | ||
| baos.write(b2); | ||
| baos.write(b1); | ||
| baos.write(b0); | ||
| } | ||
| } | ||
|
|
||
| private static void writeLong(long v, ByteArrayOutputStream baos, boolean invert) { | ||
| byte b7 = (byte) (v >>> 56); | ||
| byte b6 = (byte) (v >>> 48); | ||
| byte b5 = (byte) (v >>> 40); | ||
| byte b4 = (byte) (v >>> 32); | ||
| byte b3 = (byte) (v >>> 24); | ||
| byte b2 = (byte) (v >>> 16); | ||
| byte b1 = (byte) (v >>> 8); | ||
| byte b0 = (byte) v; | ||
| if (invert) { | ||
| baos.write(~b7); | ||
| baos.write(~b6); | ||
| baos.write(~b5); | ||
| baos.write(~b4); | ||
| baos.write(~b3); | ||
| baos.write(~b2); | ||
| baos.write(~b1); | ||
| baos.write(~b0); | ||
| } else { | ||
| baos.write(b7); | ||
| baos.write(b6); | ||
| baos.write(b5); | ||
| baos.write(b4); | ||
| baos.write(b3); | ||
| baos.write(b2); | ||
| baos.write(b1); | ||
| baos.write(b0); | ||
| } | ||
| } | ||
|
|
||
| @SuppressWarnings("JavaUtilDate") | ||
| private static void writeValue(Object val, ByteArrayOutputStream baos, boolean invert) | ||
| throws IOException { | ||
| if (val instanceof String) { | ||
| writeString((String) val, baos, invert); | ||
| } else if (val instanceof Integer) { | ||
| int v = (Integer) val; | ||
| writeInt(v ^ Integer.MIN_VALUE, baos, invert); | ||
| } else if (val instanceof Long) { | ||
| long v = (Long) val; | ||
| writeLong(v ^ Long.MIN_VALUE, baos, invert); | ||
| } else if (val instanceof Float) { | ||
| int bits = Float.floatToIntBits((Float) val); | ||
| bits = (bits >= 0) ? (bits ^ Integer.MIN_VALUE) : ~bits; | ||
| writeInt(bits, baos, invert); | ||
| } else if (val instanceof Double) { | ||
| long bits = Double.doubleToLongBits((Double) val); | ||
| bits = (bits >= 0) ? (bits ^ Long.MIN_VALUE) : ~bits; | ||
| writeLong(bits, baos, invert); | ||
| } else if (val instanceof Boolean) { | ||
| byte b = ((Boolean) val) ? (byte) 0x01 : (byte) 0x00; | ||
| baos.write(invert ? ~b : b); | ||
| } else if (val instanceof byte[]) { | ||
| writeByteArray((byte[]) val, baos, invert); | ||
| } else if (val instanceof ByteBuffer) { | ||
| writeByteArray(((ByteBuffer) val).array(), baos, invert); | ||
| } else if (val instanceof ReadableInstant) { | ||
| long enc = ((ReadableInstant) val).getMillis() ^ Long.MIN_VALUE; | ||
| writeLong(enc, baos, invert); | ||
| } else if (val instanceof Instant) { | ||
| long enc = ((Instant) val).toEpochMilli() ^ Long.MIN_VALUE; | ||
| writeLong(enc, baos, invert); | ||
| } else if (val instanceof Date) { | ||
| long enc = ((Date) val).getTime() ^ Long.MIN_VALUE; | ||
| writeLong(enc, baos, invert); | ||
| } else { | ||
| writeString(val.toString(), baos, invert); | ||
| } | ||
|
Comment on lines
+243
to
+246
Contributor
There was a problem hiding this comment. Does Iceberg support sorting on complex types (lists, maps, structs) ? We should either 1) add support for that or 2) throw an Unsupported Exception early on. I'd rather throw an error than fall back on String because it may lead to unexpected sorting behavior for some types.
Contributor
Author
There was a problem hiding this comment. It is not supported. It now raises an |
||
| } | ||
|
|
||
| private static void writeString(String s, ByteArrayOutputStream baos, boolean invert) | ||
| throws IOException { | ||
| byte[] bytes = s.getBytes(StandardCharsets.UTF_8); | ||
| writeByteArray(bytes, baos, invert); | ||
| } | ||
|
|
||
| private static void writeByteArray(byte[] bytes, ByteArrayOutputStream baos, boolean invert) { | ||
| for (byte b : bytes) { | ||
| if (b == 0x00) { | ||
| baos.write(invert ? ~(byte) 0x01 : (byte) 0x01); | ||
| baos.write(invert ? ~(byte) 0x01 : (byte) 0x01); | ||
| } else if (b == 0x01) { | ||
| baos.write(invert ? ~(byte) 0x01 : (byte) 0x01); | ||
| baos.write(invert ? ~(byte) 0x02 : (byte) 0x02); | ||
| } else { | ||
| baos.write(invert ? ~b : b); | ||
| } | ||
| } | ||
| baos.write(invert ? ~(byte) 0x00 : (byte) 0x00); | ||
| } | ||
| } | ||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Let's to clear and reuse the output stream object instead of creating a new one each iteration