Skip to content

Commit

Permalink
[7.x][ML] Prevent fetching multi-field from source (#48770) (#48797)
Browse files Browse the repository at this point in the history
Aggregatable mutli-fields are at the moment wrongly mapped
as normal doc_value fields and thus they support fetching from
source. However, they do not exist in the source. This results
to failure to extract such fields.

This commit fixes this bug. While a fix could be worked out
on top of the existing code, it is evident the extraction logic
has become difficult to understand and maintain. As we also
want to deduplicate multi-fields for data frame analytics,
it seemed appropriate to refactor the code to simplify and
better handle the extraction of multi-fields.

Relates #48756

Backport of #48770
  • Loading branch information
dimitris-athanasiou committed Nov 1, 2019
1 parent e1e9b23 commit 1f662e0
Show file tree
Hide file tree
Showing 30 changed files with 1,214 additions and 675 deletions.
Original file line number Diff line number Diff line change
Expand Up @@ -55,7 +55,7 @@ protected boolean preserveTemplatesUponCompletion() {
return true;
}

private void setupDataAccessRole(String index) throws IOException {
private static void setupDataAccessRole(String index) throws IOException {
Request request = new Request("PUT", "/_security/role/test_data_access");
request.setJsonEntity("{"
+ " \"indices\" : ["
Expand Down Expand Up @@ -283,10 +283,12 @@ public void testLookbackOnlyWithSourceDisabled() throws Exception {
new LookbackOnlyTestHelper("test-lookback-only-with-source-disabled", "airline-data-disabled-source").execute();
}

@AwaitsFix(bugUrl = "This test uses painless which is not available in the integTest phase")
public void testLookbackOnlyWithScriptFields() throws Exception {
new LookbackOnlyTestHelper("test-lookback-only-with-script-fields", "airline-data-disabled-source")
.setAddScriptedFields(true).execute();
new LookbackOnlyTestHelper("test-lookback-only-with-script-fields", "airline-data")
.setScriptedFields(
"{\"scripted_airline\":{\"script\":{\"lang\":\"painless\",\"source\":\"doc['airline.keyword'].value\"}}}")
.setAirlineVariant("scripted_airline")
.execute();
}

public void testLookbackOnlyWithNestedFields() throws Exception {
Expand Down Expand Up @@ -1088,7 +1090,7 @@ private class LookbackOnlyTestHelper {
private String jobId;
private String airlineVariant;
private String dataIndex;
private boolean addScriptedFields;
private String scriptedFields;
private boolean shouldSucceedInput;
private boolean shouldSucceedProcessing;

Expand All @@ -1100,8 +1102,8 @@ private class LookbackOnlyTestHelper {
this.airlineVariant = "airline";
}

public LookbackOnlyTestHelper setAddScriptedFields(boolean value) {
addScriptedFields = value;
public LookbackOnlyTestHelper setScriptedFields(String scriptFields) {
this.scriptedFields = scriptFields;
return this;
}

Expand All @@ -1124,10 +1126,7 @@ public LookbackOnlyTestHelper setShouldSucceedProcessing(boolean value) {
public void execute() throws Exception {
createJob(jobId, airlineVariant);
String datafeedId = "datafeed-" + jobId;
new DatafeedBuilder(datafeedId, jobId, dataIndex)
.setScriptedFields(addScriptedFields ?
"{\"airline\":{\"script\":{\"lang\":\"painless\",\"inline\":\"doc['airline'].value\"}}}" : null)
.build();
new DatafeedBuilder(datafeedId, jobId, dataIndex).setScriptedFields(scriptedFields).build();
openJob(client(), jobId);

startDatafeedAndWaitUntilStopped(datafeedId);
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -130,7 +130,7 @@ private SearchRequestBuilder buildSearchRequest(long start) {
context.query, context.extractedFields.timeField(), start, context.end));

for (ExtractedField docValueField : context.extractedFields.getDocValueFields()) {
searchRequestBuilder.addDocValueField(docValueField.getName(), docValueField.getDocValueFormat());
searchRequestBuilder.addDocValueField(docValueField.getSearchField(), docValueField.getDocValueFormat());
}
String[] sourceFields = context.extractedFields.getSourceFields();
if (sourceFields.length == 0) {
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@ class SearchHitToJsonProcessor implements Releasable {
public void process(SearchHit hit) throws IOException {
jsonBuilder.startObject();
for (ExtractedField field : fields.getAllFields()) {
writeKeyValue(field.getAlias(), field.value(hit));
writeKeyValue(field.getName(), field.value(hit));
}
jsonBuilder.endObject();
}
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,6 @@

import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collections;
import java.util.List;
import java.util.Objects;
import java.util.Set;
Expand Down Expand Up @@ -42,13 +41,13 @@ public String timeField() {
public Long timeFieldValue(SearchHit hit) {
Object[] value = timeField.value(hit);
if (value.length != 1) {
throw new RuntimeException("Time field [" + timeField.getAlias() + "] expected a single value; actual was: "
throw new RuntimeException("Time field [" + timeField.getName() + "] expected a single value; actual was: "
+ Arrays.toString(value));
}
if (value[0] instanceof Long) {
return (Long) value[0];
}
throw new RuntimeException("Time field [" + timeField.getAlias() + "] expected a long value; actual was: " + value[0]);
throw new RuntimeException("Time field [" + timeField.getName() + "] expected a long value; actual was: " + value[0]);
}

public static TimeBasedExtractedFields build(Job job, DatafeedConfig datafeed, FieldCapabilitiesResponse fieldsCapabilities) {
Expand All @@ -58,20 +57,18 @@ public static TimeBasedExtractedFields build(Job job, DatafeedConfig datafeed, F
if (scriptFields.contains(timeField) == false && extractionMethodDetector.isAggregatable(timeField) == false) {
throw new IllegalArgumentException("cannot retrieve time field [" + timeField + "] because it is not aggregatable");
}
ExtractedField timeExtractedField = extractedTimeField(timeField, scriptFields, fieldsCapabilities);
ExtractedField timeExtractedField = extractedTimeField(timeField, scriptFields);
List<String> remainingFields = job.allInputFields().stream().filter(f -> !f.equals(timeField)).collect(Collectors.toList());
List<ExtractedField> allExtractedFields = new ArrayList<>(remainingFields.size() + 1);
allExtractedFields.add(timeExtractedField);
remainingFields.stream().forEach(field -> allExtractedFields.add(extractionMethodDetector.detect(field)));

return new TimeBasedExtractedFields(timeExtractedField, allExtractedFields);
}

private static ExtractedField extractedTimeField(String timeField, Set<String> scriptFields,
FieldCapabilitiesResponse fieldCapabilities) {
if (scriptFields.contains(timeField)) {
return ExtractedField.newTimeField(timeField, Collections.emptySet(), ExtractedField.ExtractionMethod.SCRIPT_FIELD);
}
return ExtractedField.newTimeField(timeField, fieldCapabilities.getField(timeField).keySet(),
ExtractedField.ExtractionMethod.DOC_VALUE);
private static ExtractedField extractedTimeField(String timeField, Set<String> scriptFields) {
ExtractedField.Method method = scriptFields.contains(timeField) ? ExtractedField.Method.SCRIPT_FIELD
: ExtractedField.Method.DOC_VALUE;
return ExtractedFields.newTimeField(timeField, method);
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -24,8 +24,8 @@
import org.elasticsearch.search.sort.SortOrder;
import org.elasticsearch.xpack.core.ClientHelper;
import org.elasticsearch.xpack.core.ml.dataframe.analyses.DataFrameAnalysis;
import org.elasticsearch.xpack.ml.extractor.ExtractedField;
import org.elasticsearch.xpack.ml.dataframe.DataFrameAnalyticsIndex;
import org.elasticsearch.xpack.ml.extractor.ExtractedField;

import java.io.IOException;
import java.util.ArrayList;
Expand Down Expand Up @@ -138,7 +138,7 @@ private SearchRequestBuilder buildSearchRequest() {
setFetchSource(searchRequestBuilder);

for (ExtractedField docValueField : context.extractedFields.getDocValueFields()) {
searchRequestBuilder.addDocValueField(docValueField.getName(), docValueField.getDocValueFormat());
searchRequestBuilder.addDocValueField(docValueField.getSearchField(), docValueField.getDocValueFormat());
}

return searchRequestBuilder;
Expand Down Expand Up @@ -231,7 +231,7 @@ private void clearScroll(String scrollId) {
}

public List<String> getFieldNames() {
return context.extractedFields.getAllFields().stream().map(ExtractedField::getAlias).collect(Collectors.toList());
return context.extractedFields.getAllFields().stream().map(ExtractedField::getName).collect(Collectors.toList());
}

public DataSummary collectDataSummary() {
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -11,11 +11,9 @@
import org.elasticsearch.action.fieldcaps.FieldCapabilities;
import org.elasticsearch.action.fieldcaps.FieldCapabilitiesResponse;
import org.elasticsearch.common.Strings;
import org.elasticsearch.common.document.DocumentField;
import org.elasticsearch.common.regex.Regex;
import org.elasticsearch.index.IndexSettings;
import org.elasticsearch.index.mapper.BooleanFieldMapper;
import org.elasticsearch.search.SearchHit;
import org.elasticsearch.search.fetch.subphase.FetchSourceContext;
import org.elasticsearch.xpack.core.ml.dataframe.DataFrameAnalyticsConfig;
import org.elasticsearch.xpack.core.ml.dataframe.DataFrameAnalyticsDest;
Expand All @@ -24,9 +22,9 @@
import org.elasticsearch.xpack.core.ml.job.messages.Messages;
import org.elasticsearch.xpack.core.ml.utils.ExceptionsHelper;
import org.elasticsearch.xpack.core.ml.utils.NameResolver;
import org.elasticsearch.xpack.ml.dataframe.DataFrameAnalyticsIndex;
import org.elasticsearch.xpack.ml.extractor.ExtractedField;
import org.elasticsearch.xpack.ml.extractor.ExtractedFields;
import org.elasticsearch.xpack.ml.dataframe.DataFrameAnalyticsIndex;

import java.util.ArrayList;
import java.util.Arrays;
Expand Down Expand Up @@ -264,13 +262,13 @@ private ExtractedFields fetchBooleanFieldsAsIntegers(ExtractedFields extractedFi
List<ExtractedField> adjusted = new ArrayList<>(extractedFields.getAllFields().size());
for (ExtractedField field : extractedFields.getAllFields()) {
if (isBoolean(field.getTypes())) {
if (config.getAnalysis().getAllowedCategoricalTypes(field.getAlias()).contains(BooleanFieldMapper.CONTENT_TYPE)) {
if (config.getAnalysis().getAllowedCategoricalTypes(field.getName()).contains(BooleanFieldMapper.CONTENT_TYPE)) {
// We convert boolean field to string if it is a categorical dependent variable
adjusted.add(new BooleanMapper<>(field, Boolean.TRUE.toString(), Boolean.FALSE.toString()));
adjusted.add(ExtractedFields.applyBooleanMapping(field, Boolean.TRUE.toString(), Boolean.FALSE.toString()));
} else {
// We convert boolean fields to integers with values 0, 1 as this is the preferred
// way to consume such features in the analytics process.
adjusted.add(new BooleanMapper<>(field, 1, 0));
adjusted.add(ExtractedFields.applyBooleanMapping(field, 1, 0));
}
} else {
adjusted.add(field);
Expand All @@ -282,33 +280,4 @@ private ExtractedFields fetchBooleanFieldsAsIntegers(ExtractedFields extractedFi
private static boolean isBoolean(Set<String> types) {
return types.size() == 1 && types.contains(BooleanFieldMapper.CONTENT_TYPE);
}

/**
* {@link BooleanMapper} makes boolean field behave as a field of different type.
*/
private static final class BooleanMapper<T> extends ExtractedField {

private final T trueValue;
private final T falseValue;

BooleanMapper(ExtractedField field, T trueValue, T falseValue) {
super(field.getAlias(), field.getName(), Collections.singleton(BooleanFieldMapper.CONTENT_TYPE), ExtractionMethod.DOC_VALUE);
this.trueValue = trueValue;
this.falseValue = falseValue;
}

@Override
public Object[] value(SearchHit hit) {
DocumentField keyValue = hit.field(name);
if (keyValue != null) {
return keyValue.getValues().stream().map(v -> Boolean.TRUE.equals(v) ? trueValue : falseValue).toArray();
}
return new Object[0];
}

@Override
public boolean supportsFromSource() {
return false;
}
}
}
Original file line number Diff line number Diff line change
@@ -0,0 +1,49 @@
/*
* 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.extractor;

import org.elasticsearch.common.document.DocumentField;
import org.elasticsearch.search.SearchHit;

import java.util.List;
import java.util.Objects;
import java.util.Set;

abstract class AbstractField implements ExtractedField {

private final String name;

private final Set<String> types;

AbstractField(String name, Set<String> types) {
this.name = Objects.requireNonNull(name);
this.types = Objects.requireNonNull(types);
}

@Override
public String getName() {
return name;
}

@Override
public String getSearchField() {
return name;
}

@Override
public Set<String> getTypes() {
return types;
}

protected Object[] getFieldValue(SearchHit hit) {
DocumentField keyValue = hit.field(getSearchField());
if (keyValue != null) {
List<Object> values = keyValue.getValues();
return values.toArray(new Object[0]);
}
return new Object[0];
}
}
Original file line number Diff line number Diff line change
@@ -0,0 +1,48 @@
/*
* 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.extractor;

import org.elasticsearch.common.Nullable;
import org.elasticsearch.search.SearchHit;

import java.util.Set;

public class DocValueField extends AbstractField {

public DocValueField(String name, Set<String> types) {
super(name, types);
}

@Override
public Method getMethod() {
return Method.DOC_VALUE;
}

@Override
public Object[] value(SearchHit hit) {
return getFieldValue(hit);
}

@Override
public boolean supportsFromSource() {
return true;
}

@Override
public ExtractedField newFromSource() {
return new SourceField(getSearchField(), getTypes());
}

@Override
public boolean isMultiField() {
return false;
}

@Nullable
public String getDocValueFormat() {
return null;
}
}
Loading

0 comments on commit 1f662e0

Please sign in to comment.