移动测试基础 作为移动测试人员,你应该知道的移动自动化测试协议 Mobile JSON Wire Protocol Specification

easonhan007 · 2014年05月26日 · 最后由 hlm 回复于 2018年07月06日 · 1109 次阅读


  • 移动端测试工具的设计要follow下面的规格,否则哪一天这个规格不小心w3c了,你的工具就成野路子了;
  • 这个规格定义了新的New desired capability keys,这个非常重要,背下来就好;
  • 这个规格定义了新的定位方法,对于工具的使用者来说,这是核心,非常重要;
  • 这个规格对移动化做了更多detail的定义,比如网络状态等;
  • webview的定位规格;

Mobile JSON Wire Protocol Specification 这个标题就不翻译了





This specification is designed to extend the JSON Wire
a W3C working draft for web browser automation. The JSONWP has been greatly
successful for that purpose. The need for automation of native and hybrid
mobile applications can be met by the extension of the JSONWP, which already
has a proven basic automation framework (architecture, interaction model,


The initial details of this specification were worked out at a series of
meetings held in Mozilla's offices in London in August of 2013. The
participants were:

Sessions 会话

跟webdriver一样,POST /session这个URI,然后server给你返回一个sessionId。

如果server没办法start session,那么返回的状态码是500。

Session通过DELETE /session/:id这个URI进行销毁,跟webdriver一样。


Sessions work just like WebDriver: you POST to /session and receive a sessionId
as a response if the server can give you one, at which point you can send
further automation commands. If the server can't start a session, for example
if another session is running and only one session can be handled at a time,
the server must return the appropriate 500 response. Sessions are ended with
a DELETE to /session/:id as per the original WebDriver spec.

The server may but is not required to launch the AUT or a device/simulator in
the process of creating a session. It may but is not required to perform some
kind of cleaning or resetting of the AUT in order to provide a clean test
environment. It may but is not required to stop the running AUT at the session
end. It may but is not required to remove the AUT from the device or otherwise
reset the device state after the session is complete. In general, it is the
responsibility of the user to manage the test environment; it is not a part of
this specification. But a server conforming to this specification may by other
means provide that functionality as a convenience.

Desired Capabilities 不知道怎么翻,反正就是做配置的


  • automationName: 用来指定测试工具,是 appium呢?还是 ios-driver或者是 selendroid
  • platformName: 测试平台。 e.g., Android, iOS
  • platformVersion: 平台版本 e.g., 4.3 (for Android) or 6.1 (for iOS)
  • deviceName: 测试设备名称,要有版本信息的。, e.g., Nexus 4, iPhone 4S, iPhone Simulator, iPad Mini
  • app (可选): AUT的路径或者是uri
  • browserName (可选): 浏览器,其实就是webdriver的session, e.g., Safari, Chrome

New desired capability keys:

  • automationName: specific automation tool, e.g., appium, ios-driver, selendroid
  • platformName: platform to automate, e.g., Android, iOS
  • platformVersion: platform version e.g., 4.3 (for Android) or 6.1 (for iOS)
  • deviceName: specific device names including version information, e.g., Nexus 4, iPhone 4S, iPhone Simulator, iPad Mini
  • app (optional): path or uri to AUT
  • browserName (optional): web browser to automate as a webdriver session, e.g., Safari, Chrome

Locator Strategies 新的定位策略,重点


  • class name:就是个字符串,其实就是SDK里控件的类名。注意,android里要带包名的。, e.g., UIAPickerWheel for iOS or android.widget.Button for Android
    • 这些应该完全匹配由基础自动化框架提供的类名(via google翻译)
  • accessibility id: 就是代表元素可访问的id或者是label的字符串。, e.g., for iOS the accessibility identifier and for Android the content-description
  • xpath: 老熟人了,不罗嗦

The following locator strategies must be supported for non-HTML-based platforms:

  • class name: a string representing the UI element type for a given platform, e.g., UIAPickerWheel for iOS or android.widget.Button for Android
    • These should exactly match the class names given by the underlying automation frameworks
  • accessibility id: a string representing the accessibility id or label attached to a given element, e.g., for iOS the accessibility identifier and for Android the content-description
  • xpath: a valid xpath string applied to the XML document that would be retrieved using the page source command


The following locator strategies may be supported, depending on the automation

  • id: 字符串,代表对象的resource ID
  • -android uiautomator: 字符串,就是 UiAutomator的定位符 (Android only)
    • TODO: 具体描述一下
  • -ios uiautomation: 字符串,也就是 UIAutomation 的定位符 (iOS-only)
  • TODO: 指出server是否需要指明其支持这些策略。


  • id: a string corresponding to a resource ID
  • -android uiautomator: a string corresponding to a recursive element search using the UiAutomator library (Android only)
    • TODO: specify this
  • -ios uiautomation: a string corresponding to a recursive element search using the UIAutomation library (iOS-only)
  • TODO: figure out whether server should report support of these strategies

If automating a mobile browser in WebDriver mode, or a platform that uses HTML
as its element hierarchy, the usual array of WebDriver commands must be
supported instead, with their usual semantics.

Page Source 页面源码,真牵强

所有的平台必须支持GET source命令。返回代表其UI层级的xml(html)。


All platforms must respond to the GET source command with an XML (or HTML in
the case of HTML-based platforms) document representing the UI hierarchy. The
precise structure of the document may differ from platform to platform. Schemas
that must be followed for iOS and Android automation are as follows:

TODO: get together schemas for UIAutomation (iOS), Instruments (Android), and
UiAutomator (Android).

The elements in these documents may be augmented with such attributes as, for
example, ids, in order to support internal behaviors.

Touch Gestures 触摸手势

所有平台都要支持Mozila提出的Multi-Action API(这个谁能告诉我怎么翻),在一些情况下如果无法支持,直接返回500.

All platforms must adopt the Multi-Action API pioneered by Mozilla. In some
cases it will not be possible to support the full range of gestures potentially
described by this API on a given platform. In this case, the platform should
respond with a 500 when it cannot faithfully render the requested gesture.

TODO: show what the gestures API actually looks like in terms of server
endpoints that must be supported.

Device Modes 设备模式


  • GET /session/:sessionid/network_connection
    • 返回 ConnectionType
  • POST /session/:sessionid/network_connection
    • 接受一个 ConnectionType
    • 返回 ConnectionType



Devices have various states of network connectivity. In order to control
those states we have the following endpoints:

  • GET /session/:sessionid/network_connection
    • returns ConnectionType
  • POST /session/:sessionid/network_connection
    • accepts a ConnectionType
    • returns ConnectionType

Setting the network connection in the POST returns the ConnectionType because
the device might not be capable of the network connection type requested.

The remote end MUST reply with the capability "networkConnectionEnabled"

ConnectionType - 连接类型


Value (Alias) | Data | Wifi | Airplane Mode

1 (Airplane Mode) | 0 | 0 | 1
6 (All network on) | 1 | 1 | 0
4 (Data only) | 1 | 0 | 0
2 (Wifi only) | 0 | 1 | 0
0 (None) | 0 | 0 | 0


{ "name": "network_connection", "parameters": { "type": 1 } }


Is a bit mask that should be translated to an integer value when serialized.

Value (Alias) | Data | Wifi | Airplane Mode

1 (Airplane Mode) | 0 | 0 | 1
6 (All network on) | 1 | 1 | 0
4 (Data only) | 1 | 0 | 0
2 (Wifi only) | 0 | 1 | 0
0 (None) | 0 | 0 | 0

Example payload for setting "Airplane Mode":

{ "name": "network_connection", "parameters": { "type": 1 } }

Data is the upper bits since in the future we may want to support setting
certain types of Data the device is capable of. For example 3G, 4G, LTE.

Other Device Features

Mobile devices have a variety of sensors and input methods. These are automated
as follows:

  • The virtual keyboard: use sendKeys
  • acceleromator: TODO @mdas is working on this
  • geolocation: use regular webdriver endpoints
  • rotation (different from orientation): TODO
  • battery level: not in spec, perhaps exposed via executeScript
  • network speed: not in spec, perhaps exposed via executeScript

WebViews and Other Contexts

One common feature of mobile platforms is the ability to embed a chromeless
webbrowser inside of a 'native' application. These are called 'webviews', and,
if possible, a server for a given platform should implement support for
automating the webview using the full, regular, WebDriver API.

This creates a situation where there are two potential contexts for automation
in a given AUT: the native layer and the webview layer. If providing webview
support, the server must have the following endpoints:

  • GET /session/:sessionid/contexts
    • returns an array of strings representing available contexts, e.g. 'WEBVIEW', or 'NATIVE'
  • GET /session/:sessionid/context
    • returns one of:
    • a string representing the current context
    • null, representing the default context ("no context")
  • POST /session/:sessionid/context
    • accepts one of:
    • a string representing an available context
    • null, signifying a return to the default context

The first endpoint must return a possibly-empty array of strings. Each string
must be the arbitrary name of an available context, e.g., one of possibly
multiple webviews. The second must interpret the body of the request as the
name of an available context. If that context is not found, a NoSuchContext
error must be returned. If the context is available, the server must switch
automation to that context, such that all subsequent commands are taken to
apply to that context. If the body of the POST is null, the server must
return to the original context.

If a server receives a request at an endpoint which is valid in some context
but not the currently active context (for example if a user calls
driver.get() in a native context instead of a webview context), the server
must respond with an InvalidContentException.

Waiting for Conditions

The server must respond to the management commands for implicit wait timeouts,
such that when a user sets an implicit wait timeout and tries to find an
element(s), the server keeps trying to find the element(s) until that timeout
expires, rather than responding with the first failure to find the element(s).

TODO: figure out what the serversidewait implementation will be and talk about
Hide details
Change log
2a302804fc1d by Luke Inman-Semerau on May 5, 2014 Diff
using "network_connectivity" enpoint
instead of toggling just airplane mode

passing a 'bitmask' for the types of
network connectivity desired
Go to:

Double click a line to add a comment
Older revisions
7a8d40f61557 by Jonathan Lipps on Apr 2, 2014 Diff
ba61c96e26c8 by Jonathan Lipps on Mar 27, 2014 Diff
5432d87357e0 by Jonathan Lipps on Feb 25, 2014 Diff
All revisions of this file
File info
Size: 9179 bytes, 198 lines
View raw file

共收到 6 条回复 时间 点赞





作者你好,请问您知道用webdriver的capabilities参数,怎么指定火狐的配置启动吗,比如用火狐默认配置路径:C:\Users\Administrator\AppData\Roaming\Mozilla\Firefox\Profiles\w60fyb1o.default 来启动浏览器

需要 登录 后方可回复, 如果你还没有账号请点击这里 注册