Alamofire之form多表单上传

作者: lb_ | 来源:发表于2019-08-26 16:02 被阅读0次

使用 AFNetworking 进行文件上传的使用方法大家想必已经很熟悉. 那么作为其 Swift 版本, 我们当然也不能少.

先来个例子.

upload(multipartFormData: { (multipartFormData) in
    
    multipartFormData.append("lb".data(using: .utf8)!, withName: "username")
    multipartFormData.append("123456".data(using: .utf8)!, withName: "password")
    multipartFormData.append("wdms82jnds".data(using: .utf8)!, withName: "token")
    
}, to: "http://www.baidu.com") { (result) in
    print(result)
}

打开 Charles或者其他抓包工具, 运行项目查看我们本次请求.

其具体结构很清晰

alamofire.bounday.91b32560f55a049f 分隔符
Content-Dispositon:form-data;name="name" (key)
/r/n
value

接下来分析下其内部逻辑是如何处理的.
点击进入 upload 具体实现方法, 中间过渡方法省略.

open func upload(
    multipartFormData: @escaping (MultipartFormData) -> Void,
    usingThreshold encodingMemoryThreshold: UInt64 = SessionManager.multipartFormDataEncodingMemoryThreshold,
    with urlRequest: URLRequestConvertible,
    queue: DispatchQueue? = nil,
    encodingCompletion: ((MultipartFormDataEncodingResult) -> Void)?)
{
    DispatchQueue.global(qos: .utility).async {
        let formData = MultipartFormData()
        multipartFormData(formData)

        var tempFileURL: URL?
        
        var urlRequestWithContentType = try urlRequest.asURLRequest()
        urlRequestWithContentType.setValue(formData.contentType, forHTTPHeaderField: "Content-Type")

        let isBackgroundSession = self.session.configuration.identifier != nil

        if formData.contentLength < encodingMemoryThreshold && !isBackgroundSession {
            let data = try formData.encode()

            let encodingResult = MultipartFormDataEncodingResult.success(
                request: self.upload(data, with: urlRequestWithContentType),
                streamingFromDisk: false,
                streamFileURL: nil
            )

            (queue ?? DispatchQueue.main).async { encodingCompletion?(encodingResult) }
        } else {
            let fileManager = FileManager.default
            let tempDirectoryURL = URL(fileURLWithPath: NSTemporaryDirectory())
            let directoryURL = tempDirectoryURL.appendingPathComponent("org.alamofire.manager/multipart.form.data")
            let fileName = UUID().uuidString
            let fileURL = directoryURL.appendingPathComponent(fileName)

            tempFileURL = fileURL

            var directoryError: Error?

            // Create directory inside serial queue to ensure two threads don't do this in parallel
            self.queue.sync {
               try fileManager.createDirectory(at: directoryURL, withIntermediateDirectories: true, attributes: nil)
            }

            if let directoryError = directoryError { throw directoryError }

            try formData.writeEncodedData(to: fileURL)

            let upload = self.upload(fileURL, with: urlRequestWithContentType)

            //...
            (queue ?? DispatchQueue.main).async {
                let encodingResult = MultipartFormDataEncodingResult.success(
                    request: upload,
                    streamingFromDisk: true,
                    streamFileURL: fileURL
                )

                encodingCompletion?(encodingResult)
            }
        }
        
    }
}

提示:

(中间有一些 try catch 异常处理因篇幅原因省略掉了. 不影响)

先看 if else 逻辑控制语句. 搞清楚后把其种之一折起来, 有助于在查看源码时理清思路.

该方法具体操作如下:

1️⃣: 在全局队列开启异步执行下面任务.
2️⃣: 初始化一个 MultipartFormData 对象并调用闭包参数。调用用户在外界闭包中所做的处理, 也就是收取用户拼接的数据

let formData = MultipartFormData()
multipartFormData(formData)

执行用户传递的 multipartFormData 闭包, 我们在这个闭包中调用了 MultipartFormData 的 append 方法. 存到了这个临时变量里.

3️⃣: 设置请求头格式
4️⃣: 根据数据长度来区分处理, 长度为该方法的 usingThreshold 参数. 用户不传时为默认阈值.

public static let multipartFormDataEncodingMemoryThreshold: UInt64 = 10_000_000

接下来:

当满足长度阈值和 SessionManager 的会话环境为默认时, 进行 formData 编码, 上传.
否则,则将数据写入文件, 上传.

那么我们分别探讨.

Alamofire form表单数据编码

由于 upload 方法中, 调用了用户所添加的参数, 也就是调用 multipartFormData.append , 那我们点进去 append 方法

public func append(_ data: Data, withName name: String) {
    let headers = contentHeaders(withName: name)
    let stream = InputStream(data: data)
    let length = UInt64(data.count)

    append(stream, withLength: length, headers: headers)
}

先处理头进入

private func contentHeaders(withName name: String, fileName: String? = nil, mimeType: String? = nil) -> [String: String] {
    var disposition = "form-data; name=\"\(name)\""
    if let fileName = fileName { disposition += "; filename=\"\(fileName)\"" }

    var headers = ["Content-Disposition": disposition]
    if let mimeType = mimeType { headers["Content-Type"] = mimeType }

    return headers
}

对照我们抓包结果来看, 其实就是
Content-Disposition: form-data; name=\"\(name)\"
然后进行 append

append(stream, withLength: length, headers: headers)

这里就是对 bodyParts 这个数组中添加 BodyPart 对象进去. 那么我们回到判断条件满足阈值判定满足时, 首先调用 encode 方法.

let data = try formData.encode()

直接点击进入 encode 方法

public func encode() throws -> Data {
    if let bodyPartError = bodyPartError {
        throw bodyPartError
    }

    var encoded = Data()

    bodyParts.first?.hasInitialBoundary = true
    bodyParts.last?.hasFinalBoundary = true

    for bodyPart in bodyParts {
        let encodedData = try encode(bodyPart)
        encoded.append(encodedData)
    }

    return encoded
}

对头尾进行标识. 然后遍历调用 encode .

private func encode(_ bodyPart: BodyPart) throws -> Data {
    var encoded = Data()

    let initialData = bodyPart.hasInitialBoundary ? initialBoundaryData() : encapsulatedBoundaryData()
    encoded.append(initialData)

    let headerData = encodeHeaders(for: bodyPart)
    encoded.append(headerData)

    let bodyStreamData = try encodeBodyStream(for: bodyPart)
    encoded.append(bodyStreamData)

    if bodyPart.hasFinalBoundary {
        encoded.append(finalBoundaryData())
    }

    return encoded
}

针对数组中的头尾和中间三种情况,分别处理, 返回.

是首元素，则调用 initialBoundaryData() 函数。
是中间元素，则调用 encapsulatedBoundaryData() 函数。
是尾元素，则调用 finalBoundaryData() 函数。

通过type区分最终调用如下函数:

static func boundaryData(forBoundaryType boundaryType: BoundaryType, boundary: String) -> Data {
    let boundaryText: String

    switch boundaryType {
    case .initial:
        boundaryText = "--\(boundary)\(EncodingCharacters.crlf)"
    case .encapsulated:
        boundaryText = "\(EncodingCharacters.crlf)--\(boundary)\(EncodingCharacters.crlf)"
    case .final:
        boundaryText = "\(EncodingCharacters.crlf)--\(boundary)--\(EncodingCharacters.crlf)"
    }

    return boundaryText.data(using: String.Encoding.utf8, allowLossyConversion: false)!
}

其中 boundary 为分隔符. 也就是

static func randomBoundary() -> String {
    return String(format: "alamofire.boundary.%08x%08x", arc4random(), arc4random())
}

EncodingCharacters.crlf 为换行

struct EncodingCharacters {
    static let crlf = "\r\n"
}

还有值得一提的一点是在循环数组进行 encode 方法中还调用了一个
encodeBodyStream 方法

private func encodeBodyStream(for bodyPart: BodyPart) throws -> Data {
    let inputStream = bodyPart.bodyStream
    inputStream.open()
    defer { inputStream.close() }

    var encoded = Data()

    while inputStream.hasBytesAvailable {
        var buffer = [UInt8](repeating: 0, count: streamBufferSize)
        let bytesRead = inputStream.read(&buffer, maxLength: streamBufferSize)

        if let error = inputStream.streamError {
            throw AFError.multipartEncodingFailed(reason: .inputStreamReadFailed(error: error))
        }

        if bytesRead > 0 {
            encoded.append(buffer, count: bytesRead)
        } else {
            break
        }
    }

    return encoded
}

使用数据流stream 的方式读取数据, 利用其优化措施, 防止内存读取存储暴增问题.

Alamofire 表单上传 - 写入文件方式

也就是我们刚刚 if else 时, else 的情况
upload 方法中 else 部分:

let fileManager = FileManager.default
let tempDirectoryURL = URL(fileURLWithPath: NSTemporaryDirectory())
let directoryURL = tempDirectoryURL.appendingPathComponent("org.alamofire.manager/multipart.form.data")
let fileName = UUID().uuidString
let fileURL = directoryURL.appendingPathComponent(fileName)

tempFileURL = fileURL

var directoryError: Error?

// Create directory inside serial queue to ensure two threads don't do this in parallel
self.queue.sync {
    do {
        try fileManager.createDirectory(at: directoryURL, withIntermediateDirectories: true, attributes: nil)
    } catch {
        directoryError = error
    }
}

if let directoryError = directoryError { throw directoryError }

try formData.writeEncodedData(to: fileURL)

let upload = self.upload(fileURL, with: urlRequestWithContentType)

前面的文件操作就不多赘述, 直接来到

try formData.writeEncodedData(to: fileURL)

点击进入方法

public func writeEncodedData(to fileURL: URL) throws {
    if let bodyPartError = bodyPartError {
        throw bodyPartError
    }

    if FileManager.default.fileExists(atPath: fileURL.path) {
        throw AFError.multipartEncodingFailed(reason: .outputStreamFileAlreadyExists(at: fileURL))
    } else if !fileURL.isFileURL {
        throw AFError.multipartEncodingFailed(reason: .outputStreamURLInvalid(url: fileURL))
    }

    guard let outputStream = OutputStream(url: fileURL, append: false) else {
        throw AFError.multipartEncodingFailed(reason: .outputStreamCreationFailed(for: fileURL))
    }

    outputStream.open()
    defer { outputStream.close() }

    self.bodyParts.first?.hasInitialBoundary = true
    self.bodyParts.last?.hasFinalBoundary = true

    for bodyPart in self.bodyParts {
        try write(bodyPart, to: outputStream)
    }
}

老样子前面文件名重复, 文件路径无效等异常处理直接过.
我们看到同样是开流的方式. outputStream.open / .close. 然后标识头尾. 最后循环调用 .write.

private func write(_ bodyPart: BodyPart, to outputStream: OutputStream) throws {
    try writeInitialBoundaryData(for: bodyPart, to: outputStream)
    try writeHeaderData(for: bodyPart, to: outputStream)
    try writeBodyStream(for: bodyPart, to: outputStream)
    try writeFinalBoundaryData(for: bodyPart, to: outputStream)
}

然后分别调用四个过渡方法, 最终来到

private func write(_ buffer: inout [UInt8], to outputStream: OutputStream) throws {
    var bytesToWrite = buffer.count

    while bytesToWrite > 0, outputStream.hasSpaceAvailable {
        let bytesWritten = outputStream.write(buffer, maxLength: bytesToWrite)

        if let error = outputStream.streamError {
            throw AFError.multipartEncodingFailed(reason: .outputStreamWriteFailed(error: error))
        }

        bytesToWrite -= bytesWritten

        if bytesToWrite > 0 {
            buffer = Array(buffer[bytesWritten..<buffer.count])
        }
    }
}

循环写入到流中, 最后使用文件 URL 上传.

private func upload(_ uploadable: UploadRequest.Uploadable) -> UploadRequest {
    do {
        let task = try uploadable.task(session: session, adapter: adapter, queue: queue)
        let upload = UploadRequest(session: session, requestTask: .upload(uploadable, task))

        if case let .stream(inputStream, _) = uploadable {
            upload.delegate.taskNeedNewBodyStream = { _, _ in inputStream }
        }

        delegate[task] = upload

        if startRequestsImmediately { upload.resume() }

        return upload
    } catch {
        return upload(uploadable, failedWith: error)
    }
}

这里就回到了我们熟悉的 request 的流程 , 启动 -> 响应 -> 回调. 不熟悉的可以去阅读一下
Alamofire之Request(二)和队列执行顺序分析
 Alamofire之Request(一)
这两篇文章. 本文就不重复阐述了.
.

Alamofire之form多表单上传

Alamofire form表单数据编码

Alamofire 表单上传 - 写入文件方式

相关文章

网友评论

延伸阅读

深度阅读

栏目导航

热点阅读