Add DerivedFieldMapper and support parsing it in mappings (#12569)

Adds a DerivedFieldMapper to support the Derived Fields feature enhancement as well as updating the mapper parsing logic to recognize and currently parse derived fields in the mappings.

---------

Signed-off-by: Mohammad Qureshi <qreshi@amazon.com>
Signed-off-by: Rishabh Maurya <rishabhmaurya05@gmail.com>
Co-authored-by: Rishabh Maurya <rishabhmaurya05@gmail.com>

Author: qreshi

CHANGELOG.md

@@ -109,6 +109,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - Allow setting KEYSTORE_PASSWORD through env variable ([#12865](https://github.com/opensearch-project/OpenSearch/pull/12865))
 - [Concurrent Segment Search] Perform buildAggregation concurrently and support Composite Aggregations ([#12697](https://github.com/opensearch-project/OpenSearch/pull/12697))
 - [Concurrent Segment Search] Disable concurrent segment search for system indices and throttled requests ([#12954](https://github.com/opensearch-project/OpenSearch/pull/12954))
+- Derived fields support to derive field values at query time without indexing ([#12569](https://github.com/opensearch-project/OpenSearch/pull/12569))
 - Detect breaking changes on pull requests ([#9044](https://github.com/opensearch-project/OpenSearch/pull/9044))
 - Add cluster primary balance contraint for rebalancing with buffer ([#12656](https://github.com/opensearch-project/OpenSearch/pull/12656))
 

server/src/main/java/org/opensearch/index/mapper/DerivedFieldMapper.java

@@ -0,0 +1,129 @@
+/*
+ * SPDX-License-Identifier: Apache-2.0
+ *
+ * The OpenSearch Contributors require contributions made to
+ * this file be licensed under the Apache-2.0 license or a
+ * compatible open source license.
+ */
+
+package org.opensearch.index.mapper;
+
+import org.apache.lucene.index.IndexableField;
+import org.opensearch.core.xcontent.XContentBuilder;
+import org.opensearch.script.Script;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.List;
+import java.util.function.Function;
+
+/**
+ * A field mapper for derived fields
+ *
+ * @opensearch.internal
+ */
+public class DerivedFieldMapper extends ParametrizedFieldMapper {
+
+    public static final String CONTENT_TYPE = "derived";
+
+    private static DerivedFieldMapper toType(FieldMapper in) {
+        return (DerivedFieldMapper) in;
+    }
+
+    /**
+     * Builder for this field mapper
+     *
+     * @opensearch.internal
+     */
+    public static class Builder extends ParametrizedFieldMapper.Builder {
+        // TODO: The type of parameter may change here if the actual underlying FieldType object is needed
+        private final Parameter<String> type = Parameter.stringParam("type", false, m -> toType(m).type, "text");
+
+        private final Parameter<Script> script = new Parameter<>(
+            "script",
+            false,
+            () -> null,
+            (n, c, o) -> o == null ? null : Script.parse(o),
+            m -> toType(m).script
+        ).setSerializerCheck((id, ic, value) -> value != null);
+
+        public Builder(String name) {
+            super(name);
+        }
+
+        @Override
+        protected List<Parameter<?>> getParameters() {
+            return Arrays.asList(type, script);
+        }
+
+        @Override
+        public DerivedFieldMapper build(BuilderContext context) {
+            FieldMapper fieldMapper = DerivedFieldSupportedTypes.getFieldMapperFromType(type.getValue(), name, context);
+            Function<Object, IndexableField> fieldFunction = DerivedFieldSupportedTypes.getIndexableFieldGeneratorType(
+                type.getValue(),
+                name
+            );
+            DerivedFieldType ft = new DerivedFieldType(
+                buildFullName(context),
+                type.getValue(),
+                script.getValue(),
+                fieldMapper,
+                fieldFunction
+            );
+            return new DerivedFieldMapper(name, ft, multiFieldsBuilder.build(this, context), copyTo.build(), this);
+        }
+    }
+
+    public static final TypeParser PARSER = new TypeParser((n, c) -> new Builder(n));
+    private final String type;
+    private final Script script;
+
+    protected DerivedFieldMapper(
+        String simpleName,
+        MappedFieldType mappedFieldType,
+        MultiFields multiFields,
+        CopyTo copyTo,
+        Builder builder
+    ) {
+        super(simpleName, mappedFieldType, multiFields, copyTo);
+        this.type = builder.type.getValue();
+        this.script = builder.script.getValue();
+    }
+
+    @Override
+    public DerivedFieldType fieldType() {
+        return (DerivedFieldType) super.fieldType();
+    }
+
+    @Override
+    protected void parseCreateField(ParseContext context) throws IOException {
+        // Leaving this empty as the parsing should be handled via the Builder when root object is parsed.
+        // The context would not contain anything in this case since the DerivedFieldMapper is not indexed or stored.
+        throw new UnsupportedOperationException("should not be invoked");
+    }
+
+    @Override
+    public ParametrizedFieldMapper.Builder getMergeBuilder() {
+        return new Builder(simpleName()).init(this);
+    }
+
+    @Override
+    protected String contentType() {
+        return CONTENT_TYPE;
+    }
+
+    @Override
+    protected void doXContentBody(XContentBuilder builder, boolean includeDefaults, Params params) throws IOException {
+        getMergeBuilder().toXContent(builder, includeDefaults);
+        multiFields.toXContent(builder, params);
+        copyTo.toXContent(builder, params);
+    }
+
+    public String getType() {
+        return type;
+    }
+
+    public Script getScript() {
+        return script;
+    }
+}

server/src/main/java/org/opensearch/index/mapper/ObjectMapper.java

@@ -291,6 +291,15 @@ protected static boolean parseObjectOrDocumentTypeProperties(
             } else if (fieldName.equals("enabled")) {
                 builder.enabled(XContentMapValues.nodeBooleanValue(fieldNode, fieldName + ".enabled"));
                 return true;
+            } else if (fieldName.equals("derived")) {
+                if (fieldNode instanceof Collection && ((Collection) fieldNode).isEmpty()) {
+                    // nothing to do here, empty (to support "derived: []" case)
+                } else if (fieldNode instanceof Map) {
+                    parseDerived(builder, (Map<String, Object>) fieldNode, parserContext);
+                } else {
+                    throw new OpenSearchParseException("derived must be a map type");
+                }
+                return true;
             } else if (fieldName.equals("properties")) {
                 if (fieldNode instanceof Collection && ((Collection) fieldNode).isEmpty()) {
                     // nothing to do here, empty (to support "properties: []" case)
@@ -349,6 +358,55 @@ protected static void parseNested(
             }
         }
 
+        protected static void parseDerived(ObjectMapper.Builder objBuilder, Map<String, Object> derivedNode, ParserContext parserContext) {
+            Iterator<Map.Entry<String, Object>> iterator = derivedNode.entrySet().iterator();
+            while (iterator.hasNext()) {
+                Map.Entry<String, Object> entry = iterator.next();
+                String fieldName = entry.getKey();
+                // Should accept empty arrays, as a work around for when the
+                // user can't provide an empty Map. (PHP for example)
+                boolean isEmptyList = entry.getValue() instanceof List && ((List<?>) entry.getValue()).isEmpty();
+
+                if (entry.getValue() instanceof Map) {
+                    @SuppressWarnings("unchecked")
+                    Map<String, Object> node = (Map<String, Object>) entry.getValue();
+
+                    // Derived fields are a bit unique in that the 'type' attribute does not map to the TypeParser
+                    // like it would for traditional fields in properties.
+                    // So in this case, the DerivedFieldMapper's TypeParser will explicitly be used
+                    Mapper.TypeParser typeParser = parserContext.typeParser(DerivedFieldMapper.CONTENT_TYPE);
+                    String[] fieldNameParts = fieldName.split("\\.");
+                    // field name is just ".", which is invalid
+                    if (fieldNameParts.length < 1) {
+                        throw new MapperParsingException("Invalid field name " + fieldName);
+                    }
+                    String realFieldName = fieldNameParts[fieldNameParts.length - 1];
+                    Mapper.Builder<?> fieldBuilder = typeParser.parse(realFieldName, node, parserContext);
+                    for (int i = fieldNameParts.length - 2; i >= 0; --i) {
+                        ObjectMapper.Builder<?> intermediate = new ObjectMapper.Builder<>(fieldNameParts[i]);
+                        intermediate.add(fieldBuilder);
+                        fieldBuilder = intermediate;
+                    }
+                    objBuilder.add(fieldBuilder);
+                    node.remove("type");
+                    DocumentMapperParser.checkNoRemainingFields(fieldName, node, parserContext.indexVersionCreated());
+                    iterator.remove();
+                } else if (isEmptyList) {
+                    iterator.remove();
+                } else {
+                    throw new MapperParsingException(
+                        "Expected map for property [derived_fields] on field [" + fieldName + "] but got a " + fieldName.getClass()
+                    );
+                }
+            }
+
+            DocumentMapperParser.checkNoRemainingFields(
+                derivedNode,
+                parserContext.indexVersionCreated(),
+                "DocType mapping definition has unsupported parameters: "
+            );
+        }
+
         protected static void parseProperties(ObjectMapper.Builder objBuilder, Map<String, Object> propsNode, ParserContext parserContext) {
             Iterator<Map.Entry<String, Object>> iterator = propsNode.entrySet().iterator();
             while (iterator.hasNext()) {
@@ -663,7 +721,21 @@ public void toXContent(XContentBuilder builder, Params params, ToXContent custom
         doXContent(builder, params);
 
         // sort the mappers so we get consistent serialization format
-        Mapper[] sortedMappers = mappers.values().stream().toArray(size -> new Mapper[size]);
+        Mapper[] derivedSortedMappers = mappers.values()
+            .stream()
+            .filter(m -> m instanceof DerivedFieldMapper)
+            .toArray(size -> new Mapper[size]);
+        Arrays.sort(derivedSortedMappers, new Comparator<Mapper>() {
+            @Override
+            public int compare(Mapper o1, Mapper o2) {
+                return o1.name().compareTo(o2.name());
+            }
+        });
+
+        Mapper[] sortedMappers = mappers.values()
+            .stream()
+            .filter(m -> !(m instanceof DerivedFieldMapper))
+            .toArray(size -> new Mapper[size]);
         Arrays.sort(sortedMappers, new Comparator<Mapper>() {
             @Override
             public int compare(Mapper o1, Mapper o2) {
@@ -672,6 +744,17 @@ public int compare(Mapper o1, Mapper o2) {
         });
 
         int count = 0;
+        for (Mapper mapper : derivedSortedMappers) {
+            if (count++ == 0) {
+                builder.startObject("derived");
+            }
+            mapper.toXContent(builder, params);
+        }
+        if (count > 0) {
+            builder.endObject();
+        }
+
+        count = 0;
         for (Mapper mapper : sortedMappers) {
             if (!(mapper instanceof MetadataFieldMapper)) {
                 if (count++ == 0) {

server/src/main/java/org/opensearch/index/mapper/ParametrizedFieldMapper.java

@@ -670,7 +670,11 @@ public final void parse(String name, ParserContext parserContext, Map<String, Ob
                     deprecatedParamsMap.put(deprecatedName, param);
                 }
             }
-            String type = (String) fieldNode.remove("type");
+            String type = (String) fieldNode.get("type");
+            if (paramsMap.get("type") == null) {
+                fieldNode.remove("type");
+            }
+
             for (Iterator<Map.Entry<String, Object>> iterator = fieldNode.entrySet().iterator(); iterator.hasNext();) {
                 Map.Entry<String, Object> entry = iterator.next();
                 final String propName = entry.getKey();

server/src/main/java/org/opensearch/indices/IndicesModule.java

@@ -49,6 +49,7 @@
 import org.opensearch.index.mapper.ConstantKeywordFieldMapper;
 import org.opensearch.index.mapper.DataStreamFieldMapper;
 import org.opensearch.index.mapper.DateFieldMapper;
+import org.opensearch.index.mapper.DerivedFieldMapper;
 import org.opensearch.index.mapper.DocCountFieldMapper;
 import org.opensearch.index.mapper.FieldAliasMapper;
 import org.opensearch.index.mapper.FieldNamesFieldMapper;
@@ -170,6 +171,7 @@ public static Map<String, Mapper.TypeParser> getMappers(List<MapperPlugin> mappe
         mappers.put(GeoPointFieldMapper.CONTENT_TYPE, new GeoPointFieldMapper.TypeParser());
         mappers.put(FlatObjectFieldMapper.CONTENT_TYPE, FlatObjectFieldMapper.PARSER);
         mappers.put(ConstantKeywordFieldMapper.CONTENT_TYPE, new ConstantKeywordFieldMapper.TypeParser());
+        mappers.put(DerivedFieldMapper.CONTENT_TYPE, DerivedFieldMapper.PARSER);
 
         for (MapperPlugin mapperPlugin : mapperPlugins) {
             for (Map.Entry<String, Mapper.TypeParser> entry : mapperPlugin.getMappers().entrySet()) {