Skip to content

[ML-Dataframe] Feature/fib multi aggs and sources #34525

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

Merged
merged 5 commits into from
Oct 19, 2018
Merged

[ML-Dataframe] Feature/fib multi aggs and sources #34525

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
4f61ebb
support multiple sources and aggregations
Oct 16, 2018
63c9247
extract result parsing into a separate class
Oct 16, 2018
d0284f3
simplify test by removing unnecessary aggregations
Oct 17, 2018
50669ca
address review comments
Oct 17, 2018
f331658
add some documentation/notes on inner design
Oct 17, 2018
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
62 changes: 62 additions & 0 deletions .../main/java/org/elasticsearch/xpack/ml/featureindexbuilder/job/AggregationResultUtils.java
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,62 @@
/*
* Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one
* or more contributor license agreements. Licensed under the Elastic License;
* you may not use this file except in compliance with the Elastic License.
*/

package org.elasticsearch.xpack.ml.featureindexbuilder.job;

import org.apache.log4j.Logger;
import org.elasticsearch.search.aggregations.Aggregation;
import org.elasticsearch.search.aggregations.AggregationBuilder;
import org.elasticsearch.search.aggregations.bucket.composite.CompositeAggregation;
import org.elasticsearch.search.aggregations.bucket.composite.CompositeValuesSourceBuilder;
import org.elasticsearch.search.aggregations.metrics.NumericMetricsAggregation;
import org.elasticsearch.search.aggregations.metrics.NumericMetricsAggregation.SingleValue;

import java.util.Collection;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.stream.Stream;

final class AggregationResultUtils {
private static final Logger logger = Logger.getLogger(AggregationResultUtils.class.getName());

/**
* Extracts aggregation results from a composite aggregation and puts it into a map.
*
* @param agg The aggregation result
* @param sources The original sources used for querying
* @param aggregationBuilders the aggregation used for querying
* @return a map containing the results of the aggregation in a consumable way
*/
public static Stream<Map<String, Object>> extractCompositeAggregationResults(CompositeAggregation agg,
List<CompositeValuesSourceBuilder<?>> sources, Collection<AggregationBuilder> aggregationBuilders) {
return agg.getBuckets().stream().map(bucket -> {
Map<String, Object> document = new HashMap<>();
for (CompositeValuesSourceBuilder<?> source : sources) {
String destinationFieldName = source.name();
document.put(destinationFieldName, bucket.getKey().get(destinationFieldName));
}
for (AggregationBuilder aggregationBuilder : aggregationBuilders) {
String aggName = aggregationBuilder.getName();

// TODO: support other aggregation types
Aggregation aggResult = bucket.getAggregations().get(aggName);

if (aggResult instanceof NumericMetricsAggregation.SingleValue) {
NumericMetricsAggregation.SingleValue aggResultSingleValue = (SingleValue) aggResult;
document.put(aggName, aggResultSingleValue.value());
} else {
// Execution should never reach this point!
// Creating jobs with unsupported aggregations shall not be possible
logger.error("Dataframe Internal Error: unsupported aggregation ["+ aggResult.getName() +"], ignoring");
assert false;
}
}
return document;
});
}
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Seems like we are needlessly creating a leaky/incomplete abstraction. We should return a stream of objects that knows how to transform themselves into an XContentObject instead of creating a HashMap and then requiring the caller to know how to do the transformation (even thought the caller is us). This will afford us a nice separation of duties and an OK abstraction, though definitely not perfect as aggregations are so flexible.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you elaborate on what is leaky/incomplete?

Using a map is intentional, will add some reasoning as documentation.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It seems to me that all of this building could be self contained. As it stands, we have multiple functions passing around not well defined objects and just happen to know the inner workings of it. Maybe that is just how aggregations are, but seems to me a very easy way to introduce latent bugs.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

To answer the questions:

We have no schema for the output, it could be anything, having that said, this is an internal helper class, not exposed to the outside world. Before it goes out XContent::map ensures it is proper.

Regarding using maps as intermediate format: I checked for a more high-level construct and it turns out, there isn't one. The concept of a document in ingest is Map<String, Object>, so I choose the same here. Eventually we want to support running post-pivot ingest transformations. So this allows us to easily hook in ingest. (LBNL the map makes testing easier, but that isn't the primary reason.)


}
32 changes: 17 additions & 15 deletions ...n/java/org/elasticsearch/xpack/ml/featureindexbuilder/job/FeatureIndexBuilderIndexer.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,22 +17,23 @@
import org.elasticsearch.search.aggregations.bucket.composite.CompositeAggregation;
import org.elasticsearch.search.aggregations.bucket.composite.CompositeAggregationBuilder;
import org.elasticsearch.search.aggregations.bucket.composite.CompositeValuesSourceBuilder;
import org.elasticsearch.search.aggregations.metrics.NumericMetricsAggregation;
import org.elasticsearch.search.builder.SearchSourceBuilder;
import org.elasticsearch.xpack.core.indexing.AsyncTwoPhaseIndexer;
import org.elasticsearch.xpack.core.indexing.IndexerState;
import org.elasticsearch.xpack.core.indexing.IterationResult;

import java.io.IOException;
import java.io.UncheckedIOException;
import java.util.Collection;
import java.util.List;
import java.util.Map;
import java.util.concurrent.Executor;
import java.util.concurrent.atomic.AtomicReference;
import java.util.stream.Collectors;
import java.util.stream.Stream;

import static org.elasticsearch.xpack.core.ml.job.persistence.ElasticsearchMappings.DOC_TYPE;
import static org.elasticsearch.common.xcontent.XContentFactory.jsonBuilder;
import static org.elasticsearch.xpack.core.ml.job.persistence.ElasticsearchMappings.DOC_TYPE;

public abstract class FeatureIndexBuilderIndexer extends AsyncTwoPhaseIndexer<Map<String, Object>, FeatureIndexBuilderJobStats> {

Expand All @@ -58,36 +59,37 @@ protected void onStartJob(long now) {
@Override
protected IterationResult<Map<String, Object>> doProcess(SearchResponse searchResponse) {
final CompositeAggregation agg = searchResponse.getAggregations().get("feature");
return new IterationResult<>(processBuckets(agg), agg.afterKey(), agg.getBuckets().isEmpty());
return new IterationResult<>(processBucketsToIndexRequests(agg).collect(Collectors.toList()), agg.afterKey(),
agg.getBuckets().isEmpty());
}

/*
* Mocked demo case
* Parses the result and creates a stream of indexable documents
*
* TODO: replace with proper implementation
* Implementation decisions:
*
* Extraction uses generic maps as intermediate exchange format in order to hook in ingest pipelines/processors
* in later versions, see {@link IngestDocument).
*/
private List<IndexRequest> processBuckets(CompositeAggregation agg) {
// for now only 1 source supported
String destinationFieldName = job.getConfig().getSourceConfig().getSources().get(0).name();
String aggName = job.getConfig().getAggregationConfig().getAggregatorFactories().iterator().next().getName();
private Stream<IndexRequest> processBucketsToIndexRequests(CompositeAggregation agg) {
String indexName = job.getConfig().getDestinationIndex();
List<CompositeValuesSourceBuilder<?>> sources = job.getConfig().getSourceConfig().getSources();
Collection<AggregationBuilder> aggregationBuilders = job.getConfig().getAggregationConfig().getAggregatorFactories();

return agg.getBuckets().stream().map(b -> {
NumericMetricsAggregation.SingleValue aggResult = b.getAggregations().get(aggName);
return AggregationResultUtils.extractCompositeAggregationResults(agg, sources, aggregationBuilders).map(document -> {
XContentBuilder builder;
try {
builder = jsonBuilder();
builder.startObject();
builder.field(destinationFieldName, b.getKey().get(destinationFieldName));
builder.field(aggName, aggResult.value());
builder.map(document);
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

As mentioned above, I think whatever object we return should know how to take a builder and add itself appropriately.

builder.endObject();
} catch (IOException e) {
throw new UncheckedIOException(e);
}

String indexName = job.getConfig().getDestinationIndex();
IndexRequest request = new IndexRequest(indexName, DOC_TYPE).source(builder);
return request;
}).collect(Collectors.toList());
});
}

@Override
Expand Down
Loading