如何为指定Kotlin抽象函数编写Flow风格的扩展函数？

现有Kotlin抽象函数定义如下：

public abstract fun <T : TargetValue> getAverage(
    metric: StatisticsMetric, 
    interval: DateInterval, 
    timeUnit: TimeUnit,
    mode: GroupingMode, 
    result: (AsyncResult<kotlin.collections.Map<Info, T>, Error>) -> kotlin.Unit
): kotlin.Unit

其中TargetValueV2继承自TargetValue，需要编写扩展函数，让该函数支持链式调用.asFlow()，返回Flow类型结果，目标调用方式如下：

val flowResult = sdk.getAverage<TargetValueV2>(
    metric = TIME,
    interval = DateInterval(startDay, currentDate),
    timeUnit = DAY,
    mode = private
).asFlow()

解决方案 #

1. 封装请求参数的辅助类 #

首先定义一个辅助类，用来承接getAverage的调用参数，后续扩展函数基于这个类实现Flow转换：

// 替换YourSdkType为定义getAverage的抽象类/接口的实际类型
class GetAverageRequest<T : TargetValue>(
    internal val sdk: YourSdkType,
    internal val metric: StatisticsMetric,
    internal val interval: DateInterval,
    internal val timeUnit: TimeUnit,
    internal val mode: GroupingMode
)

2. 为SDK添加省略回调的扩展函数 #

编写扩展函数，让getAverage可以不传result参数，转而返回上面定义的辅助类实例：

fun <T : TargetValue> YourSdkType.getAverage(
    metric: StatisticsMetric,
    interval: DateInterval,
    timeUnit: TimeUnit,
    mode: GroupingMode
): GetAverageRequest<T> {
    return GetAverageRequest(
        sdk = this,
        metric = metric,
        interval = interval,
        timeUnit = timeUnit,
        mode = mode
    )
}

3. 实现.asFlow()扩展函数 #

基于GetAverageRequest编写asFlow扩展，利用callbackFlow将异步回调转换为Flow（需要依赖kotlinx-coroutines-core库）：

import kotlinx.coroutines.flow.Flow
import kotlinx.coroutines.flow.callbackFlow

fun <T : TargetValue> GetAverageRequest<T>.asFlow(): Flow<AsyncResult<Map<Info, T>, Error>> {
    return callbackFlow {
        sdk.getAverage(
            metric = metric,
            interval = interval,
            timeUnit = timeUnit,
            mode = mode,
            result = { asyncResult ->
                trySend(asyncResult)
                close()
            }
        )
        awaitClose()
    }
}

4. 类型推导优化（可选） #

如果希望调用时省略显式类型参数，可以使用reified类型参数优化扩展函数：

inline fun <reified T : TargetValue> YourSdkType.getAverage(
    metric: StatisticsMetric,
    interval: DateInterval,
    timeUnit: TimeUnit,
    mode: GroupingMode
): GetAverageRequest<T> {
    return GetAverageRequest(
        sdk = this,
        metric = metric,
        interval = interval,
        timeUnit = timeUnit,
        mode = mode
    )
}

此时调用可以简化为：

val flowResult = sdk.getAverage(
    metric = TIME,
    interval = DateInterval(startDay, currentDate),
    timeUnit = DAY,
    mode = private
).asFlow()

编译器会根据上下文自动推导T为TargetValueV2。

完整调用示例 #

// 显式指定类型
val flowResult: Flow<AsyncResult<Map<Info, TargetValueV2>, Error>> = sdk.getAverage<TargetValueV2>(
    metric = TIME,
    interval = DateInterval(startDay, currentDate),
    timeUnit = DAY,
    mode = private
).asFlow()

// 自动推导类型（需reified扩展函数支持）
val flowResult = sdk.getAverage(
    metric = TIME,
    interval = DateInterval(startDay, currentDate),
    timeUnit = DAY,
    mode = private
).asFlow()

内容的提问来源于stack exchange，提问作者yozhik