- A+
所属分类：.NET技术
前言
上传大文件时，原始HTTP文件上传功能可能会影响使用体验，此时使用分片上传功能可以有效避免原始上传的弊端。由于分片上传不是HTTP标准的一部分，所以只能自行开发相互配合的服务端和客户端。文件分片上传在许多情况时都拥有很多好处，除非已知需要上传的文件一定非常小。分片上传可以对上传的文件进行快速分片校验，避免大文件上传时长时间等待校验，当然完整校验可以在秒传时使用，有这种需求的情况就只能老实等待校验了。
Blazr WASM提供了在 .NET环境中使用浏览器功能的能力，充分利用C#和 .NET能够大幅简化分片上传功能的开发。本次示例使用HTTP标准上传作为分片上传的底层基础，并提供分片校验功能保障上传数据的完整性。
新书宣传
有关新书的更多介绍欢迎查看《C#与.NET6 开发从入门到实践》上市，作者亲自来打广告了！
正文
本示例的Blazor代码位于默认ASP.NET Core托管的Blazor WASM应用模板的Index页面。
在Shared项目添加公共数据模型
/// <summary> /// 文件分片上传输入模型 /// </summary> public class FileChunkUploadInput { /// <summary> /// 上传任务代码 /// </summary> public string? UploadTaskCode { get; set; } /// <summary> /// 上传请求类型 /// </summary> public string UploadType { get; set; } = null!; /// <summary> /// 文件名 /// </summary> public string FileName { get; set; } = null!; /// <summary> /// 文件大小 /// </summary> public long? FileSize { get; set; } /// <summary> /// 支持的Hash算法，优选算法请靠前 /// </summary> public List<string>? AllowedHashAlgorithm { get; set; } /// <summary> /// 使用的Hash算法 /// </summary> public string? HashAlgorithm { get; set; } /// <summary> /// Hash值 /// </summary> public string? HashValue { get; set; } /// <summary> /// 文件分片数量 /// </summary> public int FileChunkCount { get; set; } /// <summary> /// 文件片段大小 /// </summary> public int? FileChunkSize { get; set; } /// <summary> /// 文件片段偏移量（相对于整个文件） /// </summary> public long? FileChunkOffset { get; set; } /// <summary> /// 文件片段索引 /// </summary> public int? FileChunkIndex { get; set; } /// <summary> /// 取消上传的原因 /// </summary> public string? CancelReason { get; set; } } /// <summary> /// 文件分片上传开始结果 /// </summary> public class FileChunkUploadStartReault { /// <summary> /// 上传任务代码 /// </summary> public string UploadTaskCode { get; set; } = null!; /// <summary> /// 选中的Hash算法 /// </summary> public string SelectedHashAlgorithm { get; set; } = null!; } /// <summary> /// Hash助手 /// </summary> public static class HashHelper { /// <summary> /// 把Hash的字节数组转换为16进制字符串表示 /// </summary> /// <param name="bytes">原始Hash值</param> /// <returns>Hash值的16进制文本表示（大写）</returns> public static string ToHexString(this byte[] bytes) { StringBuilder sb = new(bytes.Length * 2); foreach (var @byte in bytes) { sb.Append(@byte.ToString("X2")); } return sb.ToString(); } } 服务端控制器
[ApiController] [Route("[controller]")] public class UploadController : ControllerBase { /// <summary> /// 支持的Hash算法，优选算法请靠前 /// </summary> private static string[] supportedHashAlgorithm = new[] { "MD5", "SHA1", "SHA256" }; /// <summary> /// 文件写入锁的线程安全字典，每个上传任务对应一把锁 /// </summary> private static readonly ConcurrentDictionary<string, AsyncLock> fileWriteLockerDict = new(); private readonly ILogger<UploadController> _logger; private readonly IWebHostEnvironment _env; public UploadController(ILogger<UploadController> logger, IWebHostEnvironment env) { _logger = logger; _env = env; } /// <summary> /// 分片上传动作 /// </summary> /// <param name="input">上传表单</param> /// <param name="fileChunkData">文件片段数据</param> /// <param name="requestAborted">请求取消令牌</param> /// <returns>片段上传结果</returns> [HttpPost, RequestSizeLimit(1024 * 1024 * 11)] [ProducesResponseType(StatusCodes.Status200OK)] [ProducesResponseType(StatusCodes.Status400BadRequest)] [ProducesDefaultResponseType] public async Task<IActionResult> Upload( [FromForm]FileChunkUploadInput input, [FromForm]IFormFile? fileChunkData, CancellationToken requestAborted) { switch (input.UploadType) { // 请求开始一个新的上传任务，协商上传参数 case "startUpload": { //var trustedFileNameForDisplay = // WebUtility.HtmlEncode(fileChunkData?.FileName ?? input.FileName); // 选择双方都支持的优选Hash算法 var selectedHashAlgorithm = supportedHashAlgorithm .Intersect(input.AllowedHashAlgorithm ?? Enumerable.Empty<string>()) .FirstOrDefault(); // 验证必要的表单数据 if (selectedHashAlgorithm is null or "") { ModelState.AddModelError<FileChunkUploadInput>(x => x.AllowedHashAlgorithm, "can not select hash algorithm"); } if (input.FileSize is null) { ModelState.AddModelError<FileChunkUploadInput>(x => x.FileSize, "must have value for start、upload and complete"); } if (ModelState.ErrorCount > 0) { return ValidationProblem(ModelState); } // 使用随机文件名提高安全性，并把文件名作为任务代码使用 var trustedFileNameForFileStorage = Path.GetRandomFileName(); var savePath = Path.Combine( _env.ContentRootPath, _env.EnvironmentName, "unsafe_uploads", trustedFileNameForFileStorage); var savePathWithFile = Path.Combine( savePath, $"{input.FileName}.tmp"); if (!Directory.Exists(savePath)) { Directory.CreateDirectory(savePath); } // 根据表单创建对应大小的文件 await using (var fs = new FileStream(savePathWithFile, FileMode.Create)) { fs.SetLength(input.FileSize!.Value); await fs.FlushAsync(); } // 设置锁 fileWriteLockerDict.TryAdd(trustedFileNameForFileStorage, new()); // 返回协商结果 return Ok(new FileChunkUploadStartReault { UploadTaskCode = trustedFileNameForFileStorage, SelectedHashAlgorithm = selectedHashAlgorithm! }); } // 上传文件片段 case "uploadChunk": // 验证表单 if (!fileWriteLockerDict.TryGetValue(input.UploadTaskCode!, out var _)) { ModelState.AddModelError<FileChunkUploadInput>(x => x.UploadTaskCode, $"file upload task with code {input.UploadTaskCode} is not exists"); return ValidationProblem(ModelState); } // 使用内存池缓冲数据，注意使用using释放内存 using (var pooledMemory = MemoryPool<byte>.Shared.Rent((int)fileChunkData!.Length)) { // 使用切片语法获取精准大小的内存缓冲区装载上传的数据 var buffer = pooledMemory.Memory[..(int)fileChunkData!.Length]; var readBytes = await fileChunkData.OpenReadStream().ReadAsync(buffer, requestAborted); var readBuffer = buffer[..readBytes]; Debug.Assert(readBytes == fileChunkData!.Length); // 校验Hash var hash = input.HashAlgorithm switch { "SHA1" => SHA1.HashData(readBuffer.Span), "SHA256" => SHA256.HashData(readBuffer.Span), "MD5" => MD5.HashData(readBuffer.Span), _ => Array.Empty<byte>() }; if (hash.ToHexString() != input.HashValue) { ModelState.AddModelError<FileChunkUploadInput>(x => x.HashValue, "hash does not match"); return ValidationProblem(ModelState); } var savePath = Path.Combine( _env.ContentRootPath, _env.EnvironmentName, "unsafe_uploads", input.UploadTaskCode!); var savePathWithFile = Path.Combine( savePath, $"{input.FileName}.tmp"); // 使用锁写入数据，文件流不支持写共享，必须串行化 if(fileWriteLockerDict.TryGetValue(input.UploadTaskCode!, out var locker)) { using (await locker.LockAsync()) { await using (var fs = new FileStream(savePathWithFile, FileMode.Open, FileAccess.Write)) { // 定位文件流 fs.Seek(input.FileChunkOffset!.Value, SeekOrigin.Begin); await fs.WriteAsync(readBuffer, requestAborted); await fs.FlushAsync(); } } } } return Ok(); // 取消上传 case "cancelUpload": // 验证表单 if (!fileWriteLockerDict.TryGetValue(input.UploadTaskCode!, out var _)) { ModelState.AddModelError<FileChunkUploadInput>(x => x.UploadTaskCode, $"file upload task with code {input.UploadTaskCode} is not exists"); return ValidationProblem(ModelState); } { var deletePath = Path.Combine( _env.ContentRootPath, _env.EnvironmentName, "unsafe_uploads", input.UploadTaskCode!); // 删除文件，清除锁 if (fileWriteLockerDict.TryGetValue(input.UploadTaskCode!, out var locker)) { using (await locker.LockAsync()) { if (Directory.Exists(deletePath)) { var dir = new DirectoryInfo(deletePath); dir.Delete(true); } fileWriteLockerDict.TryRemove(input.UploadTaskCode!, out _); } } } return Ok(); // 完成上传 case "completeUpload": // 验证表单 if (!fileWriteLockerDict.TryGetValue(input.UploadTaskCode!, out var _)) { ModelState.AddModelError<FileChunkUploadInput>(x => x.UploadTaskCode, $"file upload task with code {input.UploadTaskCode} is not exists"); return ValidationProblem(ModelState); } { var savePath = Path.Combine( _env.ContentRootPath, _env.EnvironmentName, "unsafe_uploads", input.UploadTaskCode!); // 去除文件的临时扩展名，清除锁 var savePathWithFile = Path.Combine(savePath, $"{input.FileName}.tmp"); var fi = new FileInfo(savePathWithFile); fi.MoveTo(Path.Combine(savePath, input.FileName)); fileWriteLockerDict.TryRemove(input.UploadTaskCode!, out _); } return Ok(); default: return BadRequest(); } } } 服务端使用三段式上传模式，开始上传，上传数据，完成（取消）上传。开始上传负责协商Hash算法和分配任务代码；上传数据负责具体的传输，并通过表单提供附加信息方便服务端操作。完成上传负责善后和资源清理。其中文件写入的异步锁使用Nito.AsyncEx代替不支持在异步中使用的lock语句。
页面代码（Index.razor）,在结尾追加
<p>支持随时取消的多文件并行分片上传，示例同时上传2个文件，每个文件同时上传2个分片，合计同时上传4个分片</p> <InputFile OnChange="UploadFile" multiple></InputFile> <button @onclick="async (MouseEventArgs e) => uploadCancelSource?.Cancel()">取消上传</button> @code{ [Inject] private HttpClient _http { get; init; } = null!; [Inject] private ILogger<Index> _logger { get; init; } = null!; private CancellationTokenSource? uploadCancelSource; /// <summary> /// 上传文件 /// </summary> /// <param name="args">上传文件的事件参数</param> /// <returns></returns> private async Task UploadFile(InputFileChangeEventArgs args) { // 设置文件并发选项 var parallelCts = new CancellationTokenSource(); uploadCancelSource = parallelCts; var parallelOption = new ParallelOptions { MaxDegreeOfParallelism = 2, CancellationToken = parallelCts.Token }; // 并发上传所有文件 await Parallel.ForEachAsync( args.GetMultipleFiles(int.MaxValue), parallelOption, async (file, cancellation) => { // 这里的取消令牌是并发方法创建的，和并发选项里的令牌不是一个 if (cancellation.IsCancellationRequested) { parallelCts.Cancel(); return; } // 使用链接令牌确保外部取消能传递到内部 var chunkUploadResult = await UploadChunkedFile( file, CancellationTokenSource.CreateLinkedTokenSource( parallelCts.Token, cancellation ).Token ); // 如果上传不成功则取消后续上传 if (chunkUploadResult != FileUploadResult.Success) { parallelCts.Cancel(); return; } } ); } /// <summary> /// 分片上传文件 /// </summary> /// <param name="file">要上传的文件</param> /// <param name="cancellation">取消令牌</param> /// <returns>上传结果</returns> private async Task<FileUploadResult> UploadChunkedFile(IBrowserFile file, CancellationToken cancellation = default) { if (cancellation.IsCancellationRequested) return FileUploadResult.Canceled; _logger.LogInformation("开始上传文件：{0}", file.Name); // 计算分片大小，文件小于10MB分片1MB，大于100MB分片10MB，在其间则使用不超过10片时的所需大小 var coefficient = file.Size switch { <= 1024 * 1024 * 10 => 1, > 1024 * 1024 * 10 and <= 1024 * 1024 *100 => (int)Math.Ceiling(file.Size / (1024.0 * 1024) / 10), _ => 10 }; // 初始化分片参数，准备字符串格式的数据供表单使用 var bufferSize = 1024 * 1024 * coefficient; // MB var stringBufferSize = bufferSize.ToString(); var chunkCount = (int)Math.Ceiling(file.Size / (double)bufferSize); var stringChunkCount = chunkCount.ToString(); var stringFileSize = file.Size.ToString(); // 发起分片上传，协商Hash算法，获取任务代码 var uploadStartContent = new List<KeyValuePair<string, string>> { new("uploadType", "startUpload"), new("fileName", file.Name), new("fileSize", stringFileSize), new("allowedHashAlgorithm", "SHA1"), new("allowedHashAlgorithm", "SHA256"), new("fileChunkCount", stringChunkCount), new("fileChunkSize", stringBufferSize), }; var uploadStartForm = new FormUrlEncodedContent(uploadStartContent); HttpResponseMessage? uploadStartResponse = null; try { uploadStartResponse = await _http.PostAsync("/upload", uploadStartForm, cancellation); } catch(TaskCanceledException e) { _logger.LogWarning(e, "外部取消上传，已停止文件：{0} 的上传", file.Name); return FileUploadResult.Canceled; } catch(Exception e) { _logger.LogError(e, "文件：{0} 的上传参数协商失败", file.Name); return FileUploadResult.Fail; } // 如果服务器响应失败，结束上传 if (uploadStartResponse?.IsSuccessStatusCode is null or false) { _logger.LogError("文件：{0} 的上传参数协商失败", file.Name); return FileUploadResult.Fail; } // 解析协商的参数 var uploadStartReault = await uploadStartResponse.Content.ReadFromJsonAsync<FileChunkUploadStartReault>(); var uploadTaskCode = uploadStartReault!.UploadTaskCode; var selectedHashAlgorithm = uploadStartReault!.SelectedHashAlgorithm; _logger.LogInformation("文件：{0} 的上传参数协商成功", file.Name); // 设置分片并发选项 var parallelOption = new ParallelOptions { MaxDegreeOfParallelism = 2, }; var fileUploadCancelSource = new CancellationTokenSource(); var sliceEnumeratorCancelSource = CancellationTokenSource .CreateLinkedTokenSource( cancellation, fileUploadCancelSource.Token ); // 各个分片的上传结果 var sliceUploadResults = new FileUploadResult?[chunkCount]; // 并发上传各个分片，并发循环本身不能用并发选项的取消令牌取消，可能会导致内存泄漏，应该通过切片循环的取消使并发循环因没有可用元素自然结束 await Parallel.ForEachAsync( SliceFileAsync( file, bufferSize, sliceEnumeratorCancelSource.Token ), parallelOption, async (fileSlice, sliceUploadCancel) => { // 解构参数 var (memory, sliceIndex, readBytes, fileOffset) = fileSlice; // 使用using确保结束后把租用的内存归还给内存池 using (memory) { var stringSliceIndex = sliceIndex.ToString(); // 主动取消上传，发送取消请求，通知服务端清理资源 if (sliceUploadCancel.IsCancellationRequested) { _logger.LogWarning("外部取消上传，已停止文件：{0} 的上传", file.Name); fileUploadCancelSource.Cancel(); sliceUploadResults[sliceIndex] = FileUploadResult.Canceled; var uploadCancelContent = new Dictionary<string, string>() { {"uploadType", "cancelUpload"}, {"uploadTaskCode", uploadTaskCode!}, {"fileName", file.Name}, {"hashAlgorithm", selectedHashAlgorithm}, {"fileChunkCount", stringChunkCount}, {"fileChunkIndex", stringSliceIndex}, {"cancelReason", "调用方要求取消上传。"}, }; var uploadCancelForm = new FormUrlEncodedContent(uploadCancelContent); var uploadCancelResponse = await _http.PostAsync("/upload", uploadCancelForm); return; } // 当前上传分片索引应当小于预计的分片数 Debug.Assert(sliceIndex < chunkCount); // 获取准确大小的缓冲区，从内存池租用时得到的容量可能大于申请的大小，使用C#的新集合切片语法 var readBuffer = memory.Memory[..readBytes]; var sw = Stopwatch.StartNew(); // 根据协商的算法计算Hash，wasm环境不支持MD5和全部非对称加密算法 var hash = selectedHashAlgorithm switch { "SHA1" => SHA1.HashData(readBuffer.Span), "SHA256" => SHA256.HashData(readBuffer.Span), _ => Array.Empty<byte>() }; sw.Stop(); _logger.LogInformation("文件：{0} 的片段 {1}({2} Bytes) 计算Hash用时 {3}", file.Name, sliceIndex, readBytes, sw.Elapsed); var stringReadBytes = readBytes.ToString(); var stringFileOffset = fileOffset.ToString(); // 上传当前分片 MultipartFormDataContent uploadFileForm = new(); uploadFileForm.Add(new StringContent(uploadTaskCode!), "uploadTaskCode"); uploadFileForm.Add(new StringContent("uploadChunk"), "uploadType"); uploadFileForm.Add(new StringContent(file.Name), "fileName"); uploadFileForm.Add(new StringContent(stringFileSize), "fileSize"); uploadFileForm.Add(new StringContent(selectedHashAlgorithm!), "hashAlgorithm"); uploadFileForm.Add(new StringContent(hash.ToHexString()), "hashValue"); uploadFileForm.Add(new StringContent(stringChunkCount), "fileChunkCount"); uploadFileForm.Add(new StringContent(stringReadBytes), "fileChunkSize"); uploadFileForm.Add(new StringContent(stringFileOffset), "fileChunkOffset"); uploadFileForm.Add(new StringContent(stringSliceIndex), "fileChunkIndex"); // 如果是未知的文件类型，设置为普通二进制流的MIME类型 var fileChunk = new ReadOnlyMemoryContent(readBuffer); fileChunk.Headers.ContentType = new MediaTypeHeaderValue(string.IsNullOrEmpty(file.ContentType) ? "application/octet-stream" : file.ContentType); uploadFileForm.Add(fileChunk, "fileChunkData", file.Name); HttpResponseMessage? uploadResponse = null; try { var uploadTaskCancel = CancellationTokenSource .CreateLinkedTokenSource( sliceUploadCancel, sliceEnumeratorCancelSource.Token ); _logger.LogInformation("文件：{0} 的片段 {1}({2} Bytes) 开始上传", file.Name, sliceIndex, readBytes); sw.Restart(); uploadResponse = await _http.PostAsync("/upload", uploadFileForm, uploadTaskCancel.Token); } catch (TaskCanceledException e) { _logger.LogWarning(e, "外部取消上传，已停止文件：{0} 的上传", file.Name); fileUploadCancelSource.Cancel(); sliceUploadResults[sliceIndex] = FileUploadResult.Canceled; var uploadCancelContent = new Dictionary<string, string>() { {"uploadType", "cancelUpload"}, {"uploadTaskCode", uploadTaskCode!}, {"fileName", file.Name}, {"hashAlgorithm", selectedHashAlgorithm}, {"fileChunkCount", stringChunkCount}, {"fileChunkIndex", stringSliceIndex}, {"cancelReason", "调用方要求取消上传。"}, }; var uploadCancelForm = new FormUrlEncodedContent(uploadCancelContent); var uploadCancelResponse = await _http.PostAsync("/upload", uploadCancelForm); return; } catch (Exception e) { _logger.LogError(e, "上传发生错误，已停止文件：{0} 的上传", file.Name); fileUploadCancelSource.Cancel(); sliceUploadResults[sliceIndex] = FileUploadResult.Fail; var uploadCancelContent = new Dictionary<string, string>() { {"uploadType", "cancelUpload"}, {"uploadTaskCode", uploadTaskCode!}, {"fileName", file.Name}, {"hashAlgorithm", selectedHashAlgorithm}, {"fileChunkCount", stringChunkCount}, {"fileChunkSize", stringReadBytes}, {"fileChunkOffset", stringFileOffset}, {"fileChunkIndex", stringSliceIndex}, {"cancelReason", "上传过程中发生错误。"}, }; var uploadCancelForm = new FormUrlEncodedContent(uploadCancelContent); var uploadCancelResponse = await _http.PostAsync("/upload", uploadCancelForm); return; } finally { sw.Stop(); } // 上传发生错误，发送取消请求，通知服务端清理资源 if (uploadResponse?.IsSuccessStatusCode is null or false) { _logger.LogError("上传发生错误，已停止文件：{0} 的上传", file.Name); fileUploadCancelSource.Cancel(); sliceUploadResults[sliceIndex] = FileUploadResult.Fail; var uploadCancelContent = new Dictionary<string, string>() { {"uploadType", "cancelUpload"}, {"uploadTaskCode", uploadTaskCode!}, {"fileName", file.Name}, {"hashAlgorithm", selectedHashAlgorithm}, {"fileChunkCount", stringChunkCount}, {"fileChunkSize", stringReadBytes}, {"fileChunkOffset", stringFileOffset}, {"fileChunkIndex", stringSliceIndex}, {"cancelReason", "上传过程中发生错误。"}, }; var uploadCancelForm = new FormUrlEncodedContent(uploadCancelContent); var uploadCancelResponse = await _http.PostAsync("/upload", uploadCancelForm); return; } _logger.LogInformation("文件：{0} 的片段 {1}({2} Bytes) 上传成功，用时 {3}", file.Name, sliceIndex, readBytes, sw.Elapsed); sliceUploadResults[sliceIndex] = FileUploadResult.Success; } } ); // 如果所有分片都上传成功，则发送完成请求完成上传 if (sliceUploadResults.All(success => success is FileUploadResult.Success)) { var uploadCompleteContent = new Dictionary<string, string>() { {"uploadType", "completeUpload"}, {"uploadTaskCode", uploadTaskCode!}, {"fileName", file.Name}, {"fileSize", stringFileSize}, {"hashAlgorithm", selectedHashAlgorithm}, {"fileChunkCount", stringChunkCount}, {"fileChunkSize", stringBufferSize}, }; var uploadCompleteForm = new FormUrlEncodedContent(uploadCompleteContent); var uploadCompleteResponse = await _http.PostAsync("/upload", uploadCompleteForm); if (uploadCompleteResponse.IsSuccessStatusCode) { _logger.LogInformation("文件：{0} 上传成功，共 {1} 个片段", file.Name, chunkCount); return FileUploadResult.Success; } else { _logger.LogError("上传发生错误，已停止文件：{0} 的上传", file.Name); var uploadCancelContent = new Dictionary<string, string>() { {"uploadType", "cancelUpload"}, {"uploadTaskCode", uploadTaskCode!}, {"fileName", file.Name}, {"hashAlgorithm", selectedHashAlgorithm}, {"fileChunkCount", stringChunkCount}, {"cancelReason", "上传过程中发生错误。"}, }; var uploadCancelForm = new FormUrlEncodedContent(uploadCancelContent); var uploadCancelResponse = await _http.PostAsync("/upload", uploadCancelForm); return FileUploadResult.Fail; } } else if (sliceUploadResults.Any(success => success is FileUploadResult.Fail)) { return FileUploadResult.Fail; } else { return FileUploadResult.Canceled; } } /// <summary> /// 异步切分要上传的文件 /// <br/>如果想中途结束切分，不要在调用此方法的foreach块中使用break，请使用取消令牌，否则会出现内存泄漏 /// </summary> /// <param name="file">要分片的文件</param> /// <param name="sliceSize">分片大小</param> /// <param name="cancellation">取消令牌</param> /// <returns>已切分的文件片段数据，用完切记释放其中的内存缓冲</returns> private static async IAsyncEnumerable<(IMemoryOwner<byte> memory, int sliceIndex, int readBytes, long fileOffset)> SliceFileAsync( IBrowserFile file, int sliceSize, [EnumeratorCancellation] CancellationToken cancellation = default) { if (cancellation.IsCancellationRequested) yield break; int fileSliceIndex; long fileOffset; IMemoryOwner<byte> memory; await using var fileStream = file.OpenReadStream(long.MaxValue); for (fileSliceIndex = 0, fileOffset = 0, memory = MemoryPool<byte>.Shared.Rent(sliceSize); (await fileStream.ReadAsync(memory.Memory[..sliceSize], cancellation)) is int readBytes and > 0; fileSliceIndex++, fileOffset += readBytes, memory = MemoryPool<byte>.Shared.Rent(sliceSize) ) { if(cancellation.IsCancellationRequested) { // 如果取消切分，缓冲不会返回到外部，只能在内部释放 memory.Dispose(); yield break; } yield return (memory, fileSliceIndex, readBytes, fileOffset); } // 切分结束后会多出一个没用的缓冲，只能在内部释放 memory.Dispose(); } /// <summary> /// 上传结果 /// </summary> public enum FileUploadResult { /// <summary> /// 失败 /// </summary> Fail = -2, /// <summary> /// 取消 /// </summary> Canceled = -1, /// <summary> /// 没有结果，未知结果 /// </summary> None = 0, /// <summary> /// 成功 /// </summary> Success = 1 } } 示例使用Parallel.ForEachAsync方法并行启动多个文件和每个文件的多个片段的上传，并发量由方法的参数控制。UploadChunkedFile方法负责单个文件的上传，其中的IBrowserFile类型是.NET 6新增的文件选择框选中项的包装，可以使用其中的OpenReadStream方法流式读取文件数据，确保大文件上传不会在内存中缓冲所有数据导致内存占用问题。
UploadChunkedFile方法内部使用自适应分片大小算法，规则为片段最小1MB，最大10MB，尽可能平均分为10份。得出片段大小后向服务端请求开始上传文件，服务端成功返回后开始文件切分、校验和上传。
SliceFileAsync负责切分文件并流式返回每个片段，切分方法是惰性的，所以不用担心占用大量内存，但是这个方法只能使用取消令牌中断切分，如果在调用该方法的await foreach块中使用break中断会产生内存泄漏。切分完成后会返回包含片段数据的内存缓冲和其他附加信息。OpenReadStream需要使用参数控制允许读取的最大字节数（默认512KB），因为这里是分片上传，直接设置为long.MaxValue即可。for循环头使用逗号表达式定义多个循环操作，使循环体的代码清晰简洁。
UploadChunkedFile方法使用Parallel.ForEachAsync并行启动多个片段的校验和上传，WASM中不支持MD5和所有非对称加密算法，需要注意。完成文件的并行上传或发生错误后会检查所有片段的上传情况，如果所有片段都上传成功，就发送完成上传请求通知服务端收尾善后，否则删除临时文件。
结语
这应该是一个比较清晰易懂的分片上传示例。示例使用Blazor 和C#以非常流畅的异步代码实现了并发分片上传。但是本示例依然有许多可优化的点，例如实现断点续传，服务端如果没有收到结束请求时的兜底处理等，这些就留给朋友们思考了。
又是很久没有写文章了，一直没有找到什么好选题，难得找到一个，经过将近1周的研究开发终于搞定了。
QQ群
读者交流QQ群：540719365
欢迎读者和广大朋友一起交流，如发现本书错误也欢迎通过博客园、QQ群等方式告知我。
![[C#] Bgr24彩色位图转为灰度的Bgr24位图的跨平台SIMD硬件加速向量算法](https://www.ztsky.cn/wp-content/themes/ztsky/img/random/13.jpg)
