Please wait. This can take some minutes ...
Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance.
Project price only 1 $
You can buy this project and download/modify it how often you want.
org.graylog.plugins.views.search.elasticsearch.searchtypes.pivot.ESPivot Maven / Gradle / Ivy
/**
* This file is part of Graylog.
*
* Graylog is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* Graylog is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with Graylog. If not, see .
*/
package org.graylog.plugins.views.search.elasticsearch.searchtypes.pivot;
import com.google.common.base.Preconditions;
import com.google.common.collect.ImmutableList;
import io.searchbox.core.SearchResult;
import io.searchbox.core.search.aggregation.Aggregation;
import io.searchbox.core.search.aggregation.MetricAggregation;
import one.util.streamex.EntryStream;
import org.elasticsearch.search.aggregations.AggregationBuilder;
import org.elasticsearch.search.aggregations.AggregationBuilders;
import org.elasticsearch.search.aggregations.metrics.max.MaxAggregationBuilder;
import org.elasticsearch.search.aggregations.metrics.min.MinAggregationBuilder;
import org.elasticsearch.search.builder.SearchSourceBuilder;
import org.graylog.plugins.views.search.Query;
import org.graylog.plugins.views.search.SearchJob;
import org.graylog.plugins.views.search.SearchType;
import org.graylog.plugins.views.search.elasticsearch.ESGeneratedQueryContext;
import org.graylog.plugins.views.search.elasticsearch.searchtypes.ESSearchTypeHandler;
import org.graylog.plugins.views.search.searchtypes.pivot.BucketSpec;
import org.graylog.plugins.views.search.searchtypes.pivot.Pivot;
import org.graylog.plugins.views.search.searchtypes.pivot.PivotResult;
import org.graylog.plugins.views.search.searchtypes.pivot.PivotSpec;
import org.graylog.plugins.views.search.searchtypes.pivot.SeriesSpec;
import org.graylog2.plugin.indexer.searches.timeranges.AbsoluteRange;
import org.graylog2.plugin.indexer.searches.timeranges.InvalidRangeParametersException;
import org.graylog2.plugin.indexer.searches.timeranges.RelativeRange;
import org.graylog2.plugin.indexer.searches.timeranges.TimeRange;
import org.joda.time.DateTime;
import org.joda.time.DateTimeZone;
import org.jooq.lambda.tuple.Tuple;
import org.jooq.lambda.tuple.Tuple2;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import javax.inject.Inject;
import java.util.ArrayDeque;
import java.util.IdentityHashMap;
import java.util.Iterator;
import java.util.List;
import java.util.Map;
import java.util.Optional;
import java.util.stream.Stream;
public class ESPivot implements ESSearchTypeHandler {
private static final Logger LOG = LoggerFactory.getLogger(ESPivot.class);
private final Map> bucketHandlers;
private final Map> seriesHandlers;
private static final TimeRange ALL_MESSAGES_TIMERANGE = allMessagesTimeRange();
private static TimeRange allMessagesTimeRange() {
try {
return RelativeRange.create(0);
} catch (InvalidRangeParametersException e){
LOG.error("Unable to instantiate all messages timerange: ", e);
}
return null;
}
@Inject
public ESPivot(Map> bucketHandlers,
Map> seriesHandlers) {
this.bucketHandlers = bucketHandlers;
this.seriesHandlers = seriesHandlers;
}
@Override
public void doGenerateQueryPart(SearchJob job, Query query, Pivot pivot, ESGeneratedQueryContext queryContext) {
LOG.debug("Generating aggregation for {}", pivot);
final SearchSourceBuilder searchSourceBuilder = queryContext.searchSourceBuilder(pivot);
final Map contextMap = queryContext.contextMap();
final AggTypes aggTypes = new AggTypes();
contextMap.put(pivot.id(), aggTypes);
// holds the initial level aggregation to be added to the query
AggregationBuilder topLevelAggregation = null;
// holds the last complete bucket aggregation into which subsequent buckets get added
AggregationBuilder previousAggregation = null;
// add global rollup series if those were requested
if (pivot.rollup()) {
seriesStream(pivot, queryContext, "global rollup")
.forEach(previousAggregation != null ? previousAggregation::subAggregation : searchSourceBuilder::aggregation);
}
final Iterator rowBuckets = pivot.rowGroups().iterator();
while (rowBuckets.hasNext()) {
final BucketSpec bucketSpec = rowBuckets.next();
final String name = queryContext.nextName();
LOG.debug("Creating row group aggregation '{}' as {}", bucketSpec.type(), name);
final ESPivotBucketSpecHandler handler = bucketHandlers.get(bucketSpec.type());
if (handler == null) {
throw new IllegalArgumentException("Unknown row_group type " + bucketSpec.type());
}
final Optional generatedAggregation = handler.createAggregation(name, pivot, bucketSpec, this, queryContext, query);
if (generatedAggregation.isPresent()) {
final AggregationBuilder aggregationBuilder = generatedAggregation.get();
if (topLevelAggregation == null) {
topLevelAggregation = aggregationBuilder;
}
// always insert the series for the final row group, or for each one if explicit rollup was requested
if (!rowBuckets.hasNext() || pivot.rollup()) {
seriesStream(pivot, queryContext, !rowBuckets.hasNext() ? "leaf row" : "row rollup")
.forEach(aggregationBuilder::subAggregation);
}
if (previousAggregation != null) {
previousAggregation.subAggregation(aggregationBuilder);
} else {
searchSourceBuilder.aggregation(aggregationBuilder);
}
previousAggregation = aggregationBuilder;
}
}
final Iterator colBuckets = pivot.columnGroups().iterator();
while (colBuckets.hasNext()) {
final BucketSpec bucketSpec = colBuckets.next();
final String name = queryContext.nextName();
LOG.debug("Creating column group aggregation '{}' as {}", bucketSpec.type(), name);
final ESPivotBucketSpecHandler handler = bucketHandlers.get(bucketSpec.type());
if (handler == null) {
throw new IllegalArgumentException("Unknown column_group type " + bucketSpec.type());
}
final Optional generatedAggregation = handler.createAggregation(name, pivot, bucketSpec, this, queryContext, query);
if (generatedAggregation.isPresent()) {
final AggregationBuilder aggregationBuilder = generatedAggregation.get();
// always insert the series for the final row group, or for each one if explicit rollup was requested
if (!colBuckets.hasNext() || pivot.rollup()) {
seriesStream(pivot, queryContext, !colBuckets.hasNext() ? "leaf column" : "column rollup")
.forEach(aggregationBuilder::subAggregation);
}
if (previousAggregation != null) {
previousAggregation.subAggregation(aggregationBuilder);
} else {
searchSourceBuilder.aggregation(aggregationBuilder);
}
previousAggregation = aggregationBuilder;
}
}
final MinAggregationBuilder startTimestamp = AggregationBuilders.min("timestamp-min").field("timestamp");
final MaxAggregationBuilder endTimestamp = AggregationBuilders.max("timestamp-max").field("timestamp");
searchSourceBuilder.aggregation(startTimestamp);
searchSourceBuilder.aggregation(endTimestamp);
if (topLevelAggregation == null) {
LOG.debug("No aggregations generated for {}", pivot);
}
}
private Stream seriesStream(Pivot pivot, ESGeneratedQueryContext queryContext, String reason) {
return EntryStream.of(pivot.series())
.mapKeyValue((integer, seriesSpec) -> {
final String seriesName = queryContext.seriesName(seriesSpec, pivot);
LOG.debug("Adding {} series '{}' with name '{}'", reason, seriesSpec.type(), seriesName);
final ESPivotSeriesSpecHandler esPivotSeriesSpecHandler = seriesHandlers.get(seriesSpec.type());
if (esPivotSeriesSpecHandler == null) {
throw new IllegalArgumentException("No series handler registered for: " + seriesSpec.type());
}
return esPivotSeriesSpecHandler.createAggregation(seriesName, pivot, seriesSpec, this, queryContext);
})
.filter(Optional::isPresent)
.map(Optional::get);
}
private boolean isAllMessagesTimeRange(TimeRange timeRange) {
return ALL_MESSAGES_TIMERANGE.equals(timeRange);
}
private AbsoluteRange extractEffectiveTimeRange(SearchResult queryResult, Query query, Pivot pivot) {
final Double from = queryResult.getAggregations().getMinAggregation("timestamp-min").getMin();
final Double to = queryResult.getAggregations().getMaxAggregation("timestamp-max").getMax();
final TimeRange pivotRange = query.effectiveTimeRange(pivot);
return AbsoluteRange.create(
isAllMessagesTimeRange(pivotRange) && from != null
? new DateTime(from.longValue(), DateTimeZone.UTC)
: query.effectiveTimeRange(pivot).getFrom(),
isAllMessagesTimeRange(pivotRange) && to != null
? new DateTime(to.longValue(), DateTimeZone.UTC)
: query.effectiveTimeRange(pivot).getTo()
);
}
@Override
public SearchType.Result doExtractResult(SearchJob job, Query query, Pivot pivot, SearchResult queryResult, MetricAggregation aggregations, ESGeneratedQueryContext queryContext) {
final AbsoluteRange effectiveTimerange = extractEffectiveTimeRange(queryResult, query, pivot);
final PivotResult.Builder resultBuilder = PivotResult.builder()
.id(pivot.id())
.effectiveTimerange(effectiveTimerange)
.total(extractDocumentCount(queryResult, pivot, queryContext));
// pivot results are a table where cells can contain multiple "values" and not only scalars:
// each combination of row and column groups can contain all series (if rollup is true)
// if rollup is false, only the "leaf" components contain the series
// in the elasticsearch result, rows and columns are simply nested aggregations (first aggregations from rows, then from columns)
// with metric aggregations on the corresponding levels.
// first we iterate over all row groups (whose values generate a "key array", corresponding to the nesting level)
// once we exhaust the row groups, we descend into the columns, which get added as values to their corresponding rows
// on each nesting level and combination we have to check for series which we also add as values to the containing row
processRows(resultBuilder, queryResult, queryContext, pivot, pivot.rowGroups(), new ArrayDeque<>(), aggregations);
return pivot.name().map(resultBuilder::name).orElse(resultBuilder).build();
}
private long extractDocumentCount(SearchResult queryResult, Pivot pivot, ESGeneratedQueryContext queryContext) {
return queryResult.getTotal();
}
/*
results from elasticsearch are nested so we need to recurse into the aggregation tree, but our result is a table, thus we need
to keep track of the current row keys manually
*/
private void processRows(PivotResult.Builder resultBuilder,
SearchResult searchResult,
ESGeneratedQueryContext queryContext,
Pivot pivot,
List remainingRows,
ArrayDeque rowKeys,
MetricAggregation aggregation) {
if (remainingRows.isEmpty()) {
// this is the last row group, so we need to fork into the columns if they exist.
// being here also means that `rowKeys` contains the maximum number of parts, one for each combination of row bucket keys
// we will always add the series for this bucket, because that's the entire point of row groups
final PivotResult.Row.Builder rowBuilder = PivotResult.Row.builder().key(ImmutableList.copyOf(rowKeys));
// do the same for columns as we did for the rows
processColumns(rowBuilder, searchResult, queryContext, pivot, pivot.columnGroups(), new ArrayDeque<>(), aggregation);
// also add the series for the entire row
// columnKeys is empty, because this is a rollup per row bucket, thus for all columns in that bucket (IOW it's not a leaf!)
if (pivot.rollup()) {
processSeries(rowBuilder, searchResult, queryContext, pivot, new ArrayDeque<>(), aggregation, true, "row-leaf");
}
resultBuilder.addRow(rowBuilder.source("leaf").build());
} else {
// this is not a leaf for the rows, so we add its key to the rowKeys and descend into the aggregation tree
// afterwards we'll check if we need to add rollup for intermediate buckets. not all clients need them so they can request
// to not calculate them
final BucketSpec currentBucket = remainingRows.get(0);
// this handler should never be missing, because we used it above to generate the query
// if it is missing for some weird reason, it's ok to fail hard here
final ESPivotBucketSpecHandler handler = bucketHandlers.get(currentBucket.type());
final Aggregation aggregationResult = handler.extractAggregationFromResult(pivot, currentBucket, aggregation, queryContext);
final Stream bucketStream = handler.handleResult(pivot, currentBucket, searchResult, aggregationResult, this, queryContext);
// for each bucket, recurse and eventually collect all the row keys. once we reach a leaf, we'll end up in the other if branch above
bucketStream.forEach(bucket -> {
// push the bucket's key and use its aggregation as the new source for sub-aggregations
rowKeys.addLast(bucket.key());
processRows(resultBuilder, searchResult, queryContext, pivot, tail(remainingRows), rowKeys, bucket.aggregation());
rowKeys.removeLast();
});
// also add the series for this row key if the client wants rollups
if (pivot.rollup()) {
final PivotResult.Row.Builder rowBuilder = PivotResult.Row.builder().key(ImmutableList.copyOf(rowKeys));
// columnKeys is empty, because this is a rollup per row bucket, thus for all columns in that bucket (IOW it's not a leaf!)
processSeries(rowBuilder, searchResult, queryContext, pivot, new ArrayDeque<>(), aggregation, true, "row-inner");
resultBuilder.addRow(rowBuilder.source("non-leaf").build());
}
}
}
private void processColumns(PivotResult.Row.Builder rowBuilder,
SearchResult searchResult,
ESGeneratedQueryContext queryContext,
Pivot pivot,
List remainingColumns,
ArrayDeque columnKeys,
MetricAggregation aggregation) {
if (remainingColumns.isEmpty()) {
// this is the leaf cell of the pivot table, simply add all the series for the complete column key array
// in the rollup: false case, this is the only set of series that is going to be added to the result
// if we simply don't have any column groups, then don't bother adding the series, this is the special case that
// only row grouping was requested, and the rollup for rows is automatically added anyway. otherwise we'll end up
// with duplicate data entries
if (!columnKeys.isEmpty()) {
processSeries(rowBuilder, searchResult, queryContext, pivot, columnKeys, aggregation, false, "col-leaf");
}
} else {
// for a non-leaf column group, we need to recurse further into the aggregation tree
// and if rollup was requested we'll add intermediate series according to the column keys
final BucketSpec currentBucket = remainingColumns.get(0);
// this handler should never be missing, because we used it above to generate the query
// if it is missing for some weird reason, it's ok to fail hard here
final ESPivotBucketSpecHandler handler = bucketHandlers.get(currentBucket.type());
final Aggregation aggregationResult = handler.extractAggregationFromResult(pivot, currentBucket, aggregation, queryContext);
final Stream bucketStream = handler.handleResult(pivot, currentBucket, searchResult, aggregationResult, this, queryContext);
// for each bucket, recurse and eventually collect all the column keys. once we reach a leaf, we'll end up in the other if branch above
bucketStream.forEach(bucket -> {
// push the bucket's key and use its aggregation as the new source for sub-aggregations
columnKeys.addLast(bucket.key());
processColumns(rowBuilder, searchResult, queryContext, pivot, tail(remainingColumns), columnKeys, bucket.aggregation());
columnKeys.removeLast();
});
// also add the series for the base column key if the client wants rollups, the complete column key is processed in the leaf branch
// don't add the empty column key rollup, because that's not the correct bucket here, it's being done in the row-leaf code
if (pivot.rollup() && !columnKeys.isEmpty()) {
// columnKeys is not empty, because this is a rollup per column in a row
processSeries(rowBuilder, searchResult, queryContext, pivot, columnKeys, aggregation, true, "col-inner");
}
}
}
private void processSeries(PivotResult.Row.Builder rowBuilder,
SearchResult searchResult,
ESGeneratedQueryContext queryContext,
Pivot pivot,
ArrayDeque columnKeys,
MetricAggregation aggregation,
boolean rollup,
String source) {
pivot.series().forEach(seriesSpec -> {
final ESPivotSeriesSpecHandler seriesHandler = seriesHandlers.get(seriesSpec.type());
final Aggregation series = seriesHandler.extractAggregationFromResult(pivot, seriesSpec, aggregation, queryContext);
seriesHandler.handleResult(pivot, seriesSpec, searchResult, series, this, queryContext)
.map(value -> {
columnKeys.addLast(value.id());
final PivotResult.Value v = PivotResult.Value.create(columnKeys, value.value(), rollup, source);
columnKeys.removeLast();
return v;
})
.forEach(rowBuilder::addValue);
});
}
private static List tail(List list) {
Preconditions.checkArgument(!list.isEmpty(), "List must not be empty!");
return list.subList(1, list.size());
}
/**
* This solely exists to hide the nasty type signature of the aggregation type map from the rest of the code.
* It's just ugly and in the way.
*/
public static class AggTypes {
final IdentityHashMap>> aggTypeMap = new IdentityHashMap<>();
public void record(PivotSpec pivotSpec, String name, Class aggClass) {
aggTypeMap.put(pivotSpec, Tuple.tuple(name, aggClass));
}
public Aggregation getSubAggregation(PivotSpec pivotSpec, MetricAggregation currentAggregationOrBucket) {
final Tuple2> tuple2 = getTypes(pivotSpec);
return currentAggregationOrBucket.getAggregation(tuple2.v1, tuple2.v2);
}
public Tuple2> getTypes(PivotSpec pivotSpec) {
return aggTypeMap.get(pivotSpec);
}
}
}