在Express中基于documentId删除Azure Cognitive Search索引文档

我在Azure Cognitive Search中有一个名为document的索引，文档结构示例如下：

{"@search.score": 1,"id": "3412b974-ca9e-4bce-a592-78084a71e4c6","content": null,"documentId": "3d8c0912-e292-483f-80a2-5e63c57b9504","bucket": "test-storage-autonomize","name": "demo-project/75131_Patient.pdf","metadata": "","form": [],"table": [],"medicalEntities": [{"BeginOffset": 1961,"Category": "MEDICAL_CONDITION","EndOffset": 1964,"Id": 56,"Score": 0.808318018913269,"Text": "ARF","Type": "DX_NAME","Attributes": [],"ICD10CMConcepts": [{"Code": "J96.0","Description": "Acute respiratory failure","Score": 0.8527929592132568"},{"Code": "J96.0","Description": "Acute respiratory failure","Score": 0.8527929592132568"},{"Code": "J96.2","Description": "Acute and chronic respiratory failure","Score": 0.8522305774688721"},{"Code": "N17.2","Description": "Acute kidney failure with medullary necrosis","Score": 0.8506965065002441"},{"Code": "N17.2","Description": "Acute renal failure with medullary necrosis","Score": 0.8506684684753418"}],"RxNormConcepts": [],"CPT_Current_Procedural_Terminology": [],"Traits": [{"Name": "SIGN","Score": 0.808318018913269}]}]}

需要基于documentId删除索引中的文档，但documentId不是索引主键（主键是id）。目前在Express框架中使用Azure SDK尝试过两种方式：

• 先通过documentId搜索获取文档再删除，存在额外开销；
• 尝试action.delete方法，但该方法需要传入主键，无法直接用documentId。

可行解决方案 #

1. 使用批量删除的过滤条件（推荐） #

Azure Cognitive Search的批量删除支持通过filter参数匹配非主键字段，一次性删除符合条件的文档，无需先查询再删除。

前提是documentId字段已设置为可过滤（filterable），如果未配置，需要先更新索引的字段定义，将documentId的filterable属性设为true。

Express中使用Azure SDK的示例代码：

const { SearchIndexClient, AzureKeyCredential, IndexDocumentsClient } = require("@azure/search-documents");

async function deleteByDocumentId(serviceEndpoint, apiKey, indexName, targetDocumentId) {
  const client = new SearchIndexClient(serviceEndpoint, new AzureKeyCredential(apiKey));
  const indexDocumentsClient = client.getIndexDocumentsClient(indexName);

  // 构建带过滤条件的删除请求
  const deleteOptions = {
    filter: `documentId eq '${targetDocumentId}'`
  };

  const result = await indexDocumentsClient.deleteDocuments(deleteOptions);
  return result;
}

// 集成到Express路由
app.delete('/documents/:documentId', async (req, res) => {
  try {
    await deleteByDocumentId(
      process.env.AZURE_SEARCH_ENDPOINT,
      process.env.AZURE_SEARCH_API_KEY,
      'document',
      req.params.documentId
    );
    res.status(200).json({ message: '文档删除成功' });
  } catch (error) {
    res.status(500).json({ error: error.message });
  }
});

2. 调整索引主键（如果业务允许） #

如果业务场景支持，可以将索引主键从id修改为documentId，后续即可直接通过action.delete方法传入documentId完成删除。但此操作需要重建索引，且需确保documentId全局唯一，适合索引创建初期或可接受数据迁移的场景。

3. 优化查询后删除的方式（备选） #

如果暂时无法调整索引配置或使用批量过滤删除，可以优化查询步骤，仅返回主键id字段，减少数据传输开销：

async function getPrimaryKeyByDocumentId(serviceEndpoint, apiKey, indexName, targetDocumentId) {
  const client = new SearchIndexClient(serviceEndpoint, new AzureKeyCredential(apiKey));
  const searchClient = client.getSearchClient(indexName);

  const searchResult = await searchClient.search(`documentId eq '${targetDocumentId}'`, {
    select: 'id', // 仅返回主键字段
    top: 1 // 假设documentId唯一，只取第一个结果
  });

  let primaryKey = null;
  for await (const result of searchResult.results) {
    primaryKey = result.document.id;
    break;
  }
  return primaryKey;
}

async function deleteDocumentById(serviceEndpoint, apiKey, indexName, primaryKey) {
  const client = new SearchIndexClient(serviceEndpoint, new AzureKeyCredential(apiKey));
  const indexDocumentsClient = client.getIndexDocumentsClient(indexName);
  
  return await indexDocumentsClient.deleteDocuments([{ id: primaryKey }]);
}

// 路由中调用
app.delete('/documents/:documentId', async (req, res) => {
  try {
    const primaryKey = await getPrimaryKeyByDocumentId(
      process.env.AZURE_SEARCH_ENDPOINT,
      process.env.AZURE_SEARCH_API_KEY,
      'document',
      req.params.documentId
    );
    
    if (!primaryKey) {
      return res.status(404).json({ message: '未找到对应文档' });
    }
    
    await deleteDocumentById(
      process.env.AZURE_SEARCH_ENDPOINT,
      process.env.AZURE_SEARCH_API_KEY,
      'document',
      primaryKey
    );
    res.status(200).json({ message: '文档删除成功' });
  } catch (error) {
    res.status(500).json({ error: error.message });
  }
});

内容的提问来源于stack exchange，提问作者ctaneja