elasticsearch java原生打分插件开发

能有影响elasticsearch score的方法有很多，官方推荐的是使用内置的painless脚本语言结合function_score来重新定义score。由于本人开发的项目其算法是由java语言开发的，于是决定尝试原生脚本开发。 elasticsearch脚本由plugin-descriptor.properties文件以及运行jar包组成，plugin-descriptor.properties主要用来定义版本信息、对应es的版本信息等属性。

官方的例子

public class ExpertScriptPlugin extends Plugin implements ScriptPlugin {
    @Override
    public ScriptEngineService getScriptEngineService(Settings settings) {
        return new MyExpertScriptEngine();
    }
    /** An example {@link ScriptEngineService} that uses Lucene segment details to implement pure document frequency scoring. */
    // tag::expert_engine
    private static class MyExpertScriptEngine implements ScriptEngineService {
        @Override
        public String getType() {
            return "expert_scripts";
        }
        @Override
        public Function<Map<String,Object>,SearchScript> compile(String scriptName, String scriptSource, Map<String, String> params) {
            // we use the script "source" as the script identifier
            if ("pure_df".equals(scriptSource)) {
                return p -> new SearchScript() {
                    final String field;
                    final String term;
                    {
                        if (p.containsKey("field") == false) {
                            throw new IllegalArgumentException("Missing parameter [field]");
                        }
                        if (p.containsKey("term") == false) {
                            throw new IllegalArgumentException("Missing parameter [term]");
                        }
                        field = p.get("field").toString();
                        term = p.get("term").toString();
                    }
                    @Override
                    public LeafSearchScript getLeafSearchScript(LeafReaderContext context) throws IOException {
                        PostingsEnum postings = context.reader().postings(new Term(field, term));
                        if (postings == null) {
                            // the field and/or term don't exist in this segment, so always return 0
                            return () -> 0.0d;
                        }
                        return new LeafSearchScript() {
                            int currentDocid = -1;
                            @Override
                            public void setDocument(int docid) {
                                // advance has undefined behavior calling with a docid <= its current docid
                                if (postings.docID() < docid) {
                                    try {
                                        postings.advance(docid);
                                    } catch (IOException e) {
                                        throw new UncheckedIOException(e);
                                    }
                                }
                                currentDocid = docid;
                            }
                            @Override
                            public double runAsDouble() {
                                if (postings.docID() != currentDocid) {
                                    // advance moved past the current doc, so this doc has no occurrences of the term
                                    return 0.0d;
                                }
                                try {
                                    return postings.freq();
                                } catch (IOException e) {
                                    throw new UncheckedIOException(e);
                                }
                            }
                        };
                    }
                    @Override
                    public boolean needsScores() {
                        return false;
                    }
                };
            }
            throw new IllegalArgumentException("Unknown script name " + scriptSource);
        }

        @Override
        @SuppressWarnings("unchecked")
        public SearchScript search(CompiledScript compiledScript, SearchLookup lookup, @Nullable Map<String, Object> params) {
          Function<Map<String,Object>,SearchScript> scriptFactory = (Function<Map<String,Object>,SearchScript>) compiledScript.compiled();
          return scriptFactory.apply(params);
        }

        @Override
        public ExecutableScript executable(CompiledScript compiledScript, @Nullable Map<String, Object> params) {
            throw new UnsupportedOperationException();
        }

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

        @Override
        public void close() {}
    }
}

代码解读： 本例在elasticsearch源码中，https://github.com/elastic/elasticsearch/tree/master/plugins/examples/script-expert-scoring

MyExpertScriptEngine类是其中最重要的类，用于实现脚本参数定义，编译，以及打分机制的实现。其中compile方法返回我们定义好打分逻辑的java function。search方法用于我们在搜索过程中实施定义好的打分逻辑。 怎奈笔者对于函数式编程知道的不多（后续需要补课），其实评分逻辑也可以在search方法中实现，于是有了下面的一段代码。

public class fieldaddScriptPlugin extends Plugin implements ScriptPlugin {
    @Override
    public ScriptEngineService getScriptEngineService(Settings settings) {
        return new MyExpertScriptEngine();
    }
    private static class MyExpertScriptEngine implements ScriptEngineService {
        @Override
        public String getType() {
            return "expert_scripts";
        }

        @Override
        public Object compile(String scriptName, String scriptSource, Map<String, String> params) {
            if ("example_add".equals(scriptSource)) {
                return scriptSource;
            }
            throw new IllegalArgumentException("Unknown script name " + scriptSource);
        }

        @Override
        @SuppressWarnings("unchecked")
        public SearchScript search(CompiledScript compiledScript, SearchLookup lookup, @Nullable Map<String, Object> vars) {

            /**
             * 校验输入参数，DSL中params 参数列表
             */
            final long inc;
            final String fieldname;
            if (vars == null || vars.containsKey("inc") == false) {
                inc = 0;
            } else {
                inc = ((Number) vars.get("inc")).longValue();
            }

            if (vars == null || vars.containsKey("fieldname") == false) {
                throw new IllegalArgumentException("Missing parameter [fieldname]");
            } else {
                fieldname = (String) vars.get("fieldname");
            }

            return new SearchScript() {
                @Override
                public LeafSearchScript getLeafSearchScript(LeafReaderContext context) throws IOException {
                    final LeafSearchLookup leafLookup = lookup.getLeafSearchLookup(context);

                    return new LeafSearchScript() {
                        @Override
                        public void setDocument(int doc) {
                            if (leafLookup != null) {
                                leafLookup.setDocument(doc);
                            }
                        }

                        @Override
                        public double runAsDouble() {
                            long values = 0;
                            /**
                             * 获取document中字段内容
                             */
                            for (Object v : (List<?>) leafLookup.doc().get(fieldname)) {
                                values = ((Number) v).longValue() + values;
                            }
                            return values + inc;
                        }
                    };
                }

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

     这段代码的逻辑是把给定的字段(字段类型long)的每个元素相加后再加上给定的增量参数最后形成score分值。为了实现上述逻辑需要实现参数获取、根据给定的字段名获取内容列表量的关键件。下面结合代码说说这两个步骤如何实现的。

search方法中Map<String, Object> vars参数对应DSL中"params"参数，用于接受实际给定的运行时参数。SearchLookup lookup参数由系统传入，通过lookup.getLeafSearchLookup(context)获取LeafSearchLookup通过该对象可以获取给定字段的值。

对于elasticsearch 2.x以前的版本可以通过NativeScriptFactory实现原生脚本。

public class MyNativeScriptPlugin extends Plugin implements ScriptPlugin {
    private final static Logger LOGGER = LogManager.getLogger(MyFirstPlugin.class);

    public MyNativeScriptPlugin() {
        super();
        LOGGER.warn("This is MyNativeScriptPlugin");
    }

    @Override
    public List<NativeScriptFactory> getNativeScripts() {
        return Collections.singletonList(new MyNativeScriptFactory());
    }

    public static class MyNativeScriptFactory implements NativeScriptFactory {
        @Override
        public ExecutableScript newScript(@Nullable Map<String, Object> params) {

//            return new MyNativeScript();
            return new AbstractDoubleSearchScript(){

                @Override
                public double runAsDouble() {
                    int b=0;
                    if(params.get("add")!=null){
                        b= (int) params.get("add");
                    }

                    String s =  source().get("last").toString();
                    double a = s.length()+b;
                    return a;                }
            };
        }

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

        @Override
        public String getName() {
            return "my_script";
        }
    }
}

工程组织 elasticsearch工程使用gradle进行依赖管理和生命周期管理，为此es项目自己也开发了esplugin的gradle插件，但不兼容gradle4.2以上的版本。参考github中的成熟插件，使用maven组织工程。

主要涉及两个文件 pom.xml plugin.xml 工程利用maven-assembly-plugin打包jar。

本例github地址：https://github.com/jiashiwen/elasticsearchpluginsample 欢迎点赞或拍砖