[
https://issues.apache.org/jira/browse/BEAM-7013?focusedWorklogId=293413&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-293413
]
ASF GitHub Bot logged work on BEAM-7013:
----------------------------------------
Author: ASF GitHub Bot
Created on: 12/Aug/19 21:39
Start Date: 12/Aug/19 21:39
Worklog Time Spent: 10m
Work Description: robinyqiu commented on pull request #9144: [BEAM-7013]
Integrating ZetaSketch's HLL++ algorithm with Beam
URL: https://github.com/apache/beam/pull/9144#discussion_r313140450
##########
File path:
sdks/java/extensions/zetasketch/src/main/java/org/apache/beam/sdk/extensions/zetasketch/HllCountMergePartialFn.java
##########
@@ -0,0 +1,151 @@
+/*
+ * 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.extensions.zetasketch;
+
+import com.google.zetasketch.HyperLogLogPlusPlus;
+import java.io.IOException;
+import java.io.InputStream;
+import java.io.OutputStream;
+import javax.annotation.Nullable;
+import org.apache.beam.sdk.coders.AtomicCoder;
+import org.apache.beam.sdk.coders.Coder;
+import org.apache.beam.sdk.coders.CoderRegistry;
+import org.apache.beam.sdk.coders.NullableCoder;
+import
org.apache.beam.sdk.extensions.zetasketch.HllCountMergePartialFn.HyperLogLogPlusPlusWrapper;
+import org.apache.beam.sdk.transforms.Combine;
+
+/**
+ * {@link Combine.CombineFn} for the {@link HllCount.MergePartial} combiner.
+ *
+ * @param <T> type of the HLL++ sketch to be merged
+ */
+class HllCountMergePartialFn<T>
+ extends Combine.CombineFn<byte[], HyperLogLogPlusPlusWrapper<T>, byte[]> {
+
+ // Call HllCountMergePartialFn.create() to instantiate
+ private HllCountMergePartialFn() {}
+
+ static HllCountMergePartialFn<?> create() {
+ return new HllCountMergePartialFn();
+ }
+
+ /**
+ * Accumulator for the {@link HllCount.MergePartial} combiner. Cannot use
{@link
+ * HyperLogLogPlusPlus} directly because we need an "identity element" for
the {@code merge}
+ * operation, and we are not able to create such "identity element" of type
{@link
+ * HyperLogLogPlusPlus} without knowing the sketch precision and value type.
+ */
+ static final class HyperLogLogPlusPlusWrapper<T> {
+ @Nullable private HyperLogLogPlusPlus<T> hll;
+
+ private HyperLogLogPlusPlusWrapper(@Nullable HyperLogLogPlusPlus<T> hll) {
+ this.hll = hll;
+ }
+ }
+
+ /**
+ * Coder for the {@link HyperLogLogPlusPlusWrapper} class with generic type
parameter {@code T}.
+ */
+ private static final class HyperLogLogPlusPlusWrapperCoder<T>
+ extends AtomicCoder<HyperLogLogPlusPlusWrapper<T>> {
+
+ private static final HyperLogLogPlusPlusWrapperCoder<?> INSTANCE =
+ new HyperLogLogPlusPlusWrapperCoder<>();
+
+ private final Coder<HyperLogLogPlusPlus<T>> nullableHllCoder =
+ NullableCoder.of(HyperLogLogPlusPlusCoder.of());
+
+ // Generic singleton factory pattern; the coder works for all
HyperLogLogPlusPlusWrapper objects
+ // at runtime regardless of type T
+ @SuppressWarnings("unchecked")
+ static <T> HyperLogLogPlusPlusWrapperCoder<T> of() {
+ return (HyperLogLogPlusPlusWrapperCoder<T>) INSTANCE;
+ }
+
+ @Override
+ public void encode(HyperLogLogPlusPlusWrapper<T> value, OutputStream
outStream)
+ throws IOException {
+ nullableHllCoder.encode(value.hll, outStream);
+ }
+
+ @Override
+ public HyperLogLogPlusPlusWrapper<T> decode(InputStream inStream) throws
IOException {
+ return new
HyperLogLogPlusPlusWrapper<T>(nullableHllCoder.decode(inStream));
+ }
+
+ @Override
+ public void verifyDeterministic() throws NonDeterministicException {
+ nullableHllCoder.verifyDeterministic();
+ }
+ }
+
+ @Override
+ public Coder<HyperLogLogPlusPlusWrapper<T>> getAccumulatorCoder(
+ CoderRegistry registry, Coder<byte[]> inputCoder) {
+ return HyperLogLogPlusPlusWrapperCoder.of();
+ }
+
+ @Override
+ public HyperLogLogPlusPlusWrapper<T> createAccumulator() {
+ return new HyperLogLogPlusPlusWrapper<>(null);
+ }
+
+ @Override
+ public HyperLogLogPlusPlusWrapper<T> addInput(
+ HyperLogLogPlusPlusWrapper<T> accumulator, byte[] input) {
+ if (accumulator.hll == null) {
+ @SuppressWarnings("unchecked")
+ HyperLogLogPlusPlus<T> hll = (HyperLogLogPlusPlus<T>)
HyperLogLogPlusPlus.forProto(input);
+ accumulator.hll = hll;
+ } else {
+ accumulator.hll.merge(input);
+ }
+ return accumulator;
+ }
+
+ @Override
+ public HyperLogLogPlusPlusWrapper<T> mergeAccumulators(
+ Iterable<HyperLogLogPlusPlusWrapper<T>> accumulators) {
+ HyperLogLogPlusPlusWrapper<T> merged = createAccumulator();
+ for (HyperLogLogPlusPlusWrapper<T> accumulator : accumulators) {
+ if (accumulator.hll == null) {
+ continue;
+ }
+ if (merged.hll == null) {
+ // Cannot set merged.hll to accumulator.hll directly because we
shouldn't mutate accumulator
Review comment:
> CombineFn#mergeAccumulators javadoc says:
"May modify any of the argument accumulators. May return a fresh
accumulator, or may return one of the (modified) argument accumulators."
That seems to contradict the "we shouldn't mutate accumulator"?
Yes they are contradictory, and the doc you read is outdated. I did not find
the javadoc you mentioned in `Combine.java`. Could you please share a link to
it? I think we should correct it to avoid future confusion. Thank you!
> If we indeed cannot mutate them, I would move this comment to right on top
of line 143
Done! (I think you mean line 134)
> P.S. the serialize-then-deserialize step is fairly expensive, if we ever
find a workaround for it in the future.
Yes I agree. But it is good that this serialize-then-deserialize step will
not be called very often (number of calls = number of "identity elements").
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 293413)
Time Spent: 13h 10m (was: 13h)
> A new count distinct transform based on BigQuery compatible HyperLogLog++
> implementation
> ----------------------------------------------------------------------------------------
>
> Key: BEAM-7013
> URL: https://issues.apache.org/jira/browse/BEAM-7013
> Project: Beam
> Issue Type: New Feature
> Components: extensions-java-sketching, sdk-java-core
> Reporter: Yueyang Qiu
> Assignee: Yueyang Qiu
> Priority: Major
> Fix For: 2.16.0
>
> Time Spent: 13h 10m
> Remaining Estimate: 0h
>
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
