Netflix Java GraphQL Client：如何为CustomGraphQLClient动态添加请求头

问题背景 #

之前使用DefaultGraphQLClient时，我们可以在请求到来时动态添加请求头——因为部分特定请求头只有在客户端发送请求时才能确定。现在迁移到CustomGraphQLClient后，发现RequestExecutor只能在创建CustomGraphQLClient时传入，所有请求头必须预先定义，无法实现旧版的动态添加逻辑。

旧版实现代码：

@Bean(name = [CORE_PRODUCT_GRAPHQL_CLIENT])
fun coreGraphQLClient(): GraphQLClient {
      return DefaultGraphQLClient(url)
}

fun executeQuery(query: String, graphQLHeaders: GraphqlHeaderMaps? = null): GraphQLResponse {
    return graphQLClient.executeQuery(query, HashMap<String, List<String>>()) { url, headers, body ->
        val requestHeaders = HttpHeaders()
        headers.forEach(requestHeaders::put)

        graphQLHeaders?.forEach(requestHeaders::put)

        val exchange: ResponseEntity<String> = restTemplate.exchange(
            url, HttpMethod.POST, HttpEntity<Any?>(body, requestHeaders),
            String::class.java
        )
        HttpResponse(exchange.statusCodeValue, exchange.body)
    }
}

解决方案 #

方法1：自定义支持动态头的RequestExecutor #

通过封装RequestExecutor，让它能在执行阶段接收额外请求头，同时保证线程安全：

class DynamicRequestExecutor(private val delegate: RequestExecutor) : RequestExecutor {
    // 用ThreadLocal存储当前请求的额外头，避免多线程冲突
    private val extraHeaders = ThreadLocal<Map<String, List<String>>>()

    fun setExtraHeaders(headers: Map<String, List<String>>?) {
        extraHeaders.set(headers)
    }

    override fun execute(request: GraphQLRequest): GraphQLResponse {
        // 合并原有请求头与额外头
        val mergedHeaders = request.headers.toMutableMap()
        extraHeaders.get()?.let { mergedHeaders.putAll(it) }
        val newRequest = request.copy(headers = mergedHeaders)
        
        // 执行后清理ThreadLocal，防止内存泄漏
        try {
            return delegate.execute(newRequest)
        } finally {
            extraHeaders.remove()
        }
    }
}

创建CustomGraphQLClient时注入这个动态执行器：

@Bean(name = [CORE_PRODUCT_GRAPHQL_CLIENT])
fun coreGraphQLClient(): GraphQLClient {
    val defaultExecutor = DefaultRequestExecutor(restTemplate)
    val dynamicExecutor = DynamicRequestExecutor(defaultExecutor)
    return CustomGraphQLClient(url, dynamicExecutor)
}

执行查询时传入动态头：

fun executeQuery(query: String, graphQLHeaders: GraphqlHeaderMaps? = null): GraphQLResponse {
    val client = coreGraphQLClient() as CustomGraphQLClient
    val executor = client.requestExecutor as DynamicRequestExecutor
    
    executor.setExtraHeaders(graphQLHeaders)
    return client.executeQuery(query)
}

方法2：直接通过executeQuery的headers参数传入 #

如果不需要在RequestExecutor中做统一处理，可直接在调用executeQuery时合并默认头与动态头：

fun executeQuery(query: String, graphQLHeaders: GraphqlHeaderMaps? = null): GraphQLResponse {
    val requestHeaders = HashMap<String, List<String>>()
    // 先添加预先定义的默认头
    // requestHeaders.put("default-header", listOf("value"))
    
    // 再合并动态传入的请求头
    graphQLHeaders?.let { requestHeaders.putAll(it) }
    
    return coreGraphQLClient().executeQuery(query, requestHeaders)
}

注意事项 #

• 使用ThreadLocal方案时，必须在请求执行完成后清理ThreadLocal，避免线程复用导致的内存泄漏。
• 多线程环境下，ThreadLocal是线程安全的，但如果使用线程池，要确保线程复用前已清理ThreadLocal中的值。

内容的提问来源于stack exchange，提问作者Panda World