如何用elasticlunr搜索嵌套数组？搜索Phoenix返回空数组求助

我有包含嵌套数组的文档，尝试用elasticlunr搜索嵌套数组里的内容，但搜索城市Phoenix时返回空数组，代码示例如下：

const doc1 = 
    {
        "id": 123,
        "firstName": "jon",
        "lastName": "doe",
        "addresses": [
            {
                "line1": "123 any str",
                "city": "Some City",
                "state": "CA"
            }
        ]
    };
const doc2 = {
        "id": 456,
        "firstName": "alan",
        "lastName": "park",
        "addresses": [
            {
                "line1": "1 some str",
                "city": "Phoenix",
                "state": "AZ"
            }
        ]
    };

this.index.addDoc(doc1);
this.index.addDoc(doc2);

let index = elasticlunr(function () {
    this.addField('firstName');
    this.addField('lastName');
    this.addField('addresses');
    this.addField('city');
    this.addField('state');
    this.setRef('id');
});
this.index = index;

searchFreeText() {
    return this.index.search("Phoenix", {
        fields: {
            firstName: { boost: 1, bool: "OR", expand: true },
            lastName: { boost: 1, bool: "OR" },
            addresses: { boost: 1, bool: "OR" },
            line1: { boost: 1, bool: "OR" },
            city: { boost: 1, bool: "OR" },
            state: { boost: 1, bool: "OR" },
        }
    });
}

问题原因 #

elasticlunr默认不会自动解析嵌套对象或数组：

• 直接添加city、state这类字段时，索引会尝试从文档顶层属性取值，但你的文档里这些字段是嵌套在addresses数组的子对象中，所以索引中根本没有这些字段的有效内容
• 添加addresses字段时，elasticlunr会把整个数组转换成字符串（如[object Object]），里面的城市名无法被正确分词索引

解决方案 #

方案1：扁平化嵌套结构后再添加文档 #

把嵌套字段提取到文档顶层，确保索引能直接获取到对应值：

// 定义文档处理函数，扁平化嵌套的地址字段
const flattenDoc = (doc) => {
    return {
        ...doc,
        // 将所有地址的城市合并为字符串（支持多地址场景）
        city: doc.addresses.map(addr => addr.city).join(' '),
        state: doc.addresses.map(addr => addr.state).join(' '),
        line1: doc.addresses.map(addr => addr.line1).join(' ')
    };
};

// 处理原始文档
const processedDoc1 = flattenDoc(doc1);
const processedDoc2 = flattenDoc(doc2);

// 创建索引，只添加顶层字段
let index = elasticlunr(function () {
    this.addField('firstName');
    this.addField('lastName');
    this.addField('city');
    this.addField('state');
    this.addField('line1');
    this.setRef('id');
});

// 添加处理后的文档
index.addDoc(processedDoc1);
index.addDoc(processedDoc2);

// 搜索函数（可给城市字段更高权重提升匹配优先级）
searchFreeText() {
    return this.index.search("Phoenix", {
        fields: {
            firstName: { boost: 1, bool: "OR", expand: true },
            lastName: { boost: 1, bool: "OR" },
            city: { boost: 2, bool: "OR" },
            state: { boost: 1, bool: "OR" },
            line1: { boost: 1, bool: "OR" },
        }
    });
}

方案2：自定义字段提取逻辑 #

利用elasticlunr的字段提取函数，直接从嵌套结构中获取字段值：

// 创建索引时，为嵌套字段自定义提取逻辑
let index = elasticlunr(function () {
    this.addField('firstName');
    this.addField('lastName');
    // 从addresses数组中提取所有城市并合并为字符串
    this.addField('city', doc => doc.addresses.map(addr => addr.city).join(' '));
    this.addField('state', doc => doc.addresses.map(addr => addr.state).join(' '));
    this.addField('line1', doc => doc.addresses.map(addr => addr.line1).join(' '));
    this.setRef('id');
});

// 直接添加原始文档即可
index.addDoc(doc1);
index.addDoc(doc2);

// 搜索函数保持不变，现在能正常搜到Phoenix
searchFreeText() {
    return this.index.search("Phoenix", {
        fields: {
            firstName: { boost: 1, bool: "OR", expand: true },
            lastName: { boost: 1, bool: "OR" },
            city: { boost: 2, bool: "OR" },
            state: { boost: 1, bool: "OR" },
            line1: { boost: 1, bool: "OR" },
        }
    });
}

内容的提问来源于stack exchange，提问作者Maharaj