QQ交流群: 815453846
各种库的版本号
该项目处于低活跃度的开发中
UiAutomator是Google提供的用来做安卓自动化测试的一个Java库,基于Accessibility服务。功能很强,可以对第三方App进行测试,获取屏幕上任意一个APP的任意一个控件属性,并对其进行任意操作,但有两个缺点:1. 测试脚本只能使用Java语言 2. 测试脚本要打包成jar或者apk包上传到设备上才能运行。
我们希望测试逻辑能够用Python编写,能够在电脑上运行的时候就控制手机。这里要非常感谢 Xiaocong He (@xiaocong),他将这个想法实现了出来(见xiaocong/uiautomator),原理是在手机上运行了一个http rpc服务,将uiautomator中的功能开放出来,然后再将这些http接口封装成Python库。
因为xiaocong/uiautomator
这个库,已经很久不见更新。所以我们直接fork了一个版本,为了方便做区分我们就在后面加了个2 openatx/uiautomator2
除了对原有的库的bug进行了修复,还增加了很多新的Feature。主要有以下部分:
- 设备和开发机可以脱离数据线,通过WiFi互联(基于atx-agent)
- 集成了openstf/minicap达到实时屏幕投频,以及实时截图
- 集成了openstf/minitouch达到精确实时控制设备
- 修复了xiaocong/uiautomator经常性退出的问题
- 代码进行了重构和精简,方便维护
- 实现了一个设备管理平台(也支持iOS) atxserver2
- 扩充了toast获取和展示的功能
这里要先说明下,因为经常有很多人问 openatx/uiautomator2 并不支持iOS测试,需要iOS自动化测试,可以转到这个库 openatx/facebook-wda。
PS: 这个库
https://github.com/NeteaseGame/ATX目前已经不维护了,请尽快更换。
这里有一份快速参考,适合已经入门的人 QUICK REFERENCE GUIDE,欢迎多提意见。
- Android版本 4.4+
- Python 3.6+ (社区反馈3.8.0不支持, 但是3.8.2支持)
如果用python2的pip安装,会安装本库的老版本0.2.3;如果用python3.5的pip安装,会安装本库的老版本0.3.3;两者均已经不会再维护;PYPI上的最近版本是这个:https://pypi.org/project/uiautomator2/
先准备一台(不要两台)开启了开发者选项
的安卓手机,连接上电脑,确保执行adb devices
可以看到连接上的设备。
运行pip3 install -U uiautomator2
安装uiautomator2
运行python3 -m uiautomator2 init
安装包含httprpc服务的apk到手机+atx-agent, minicap, minitouch
(在过去的版本中,这一步是必须执行的,但是从1.3.0之后的版本,当运行python代码u2.connect()
时就会自动推送这些文件了)
命令行运行python
打开python交互窗口。然后将下面的命令输入到窗口中。
import uiautomator2 as u2
d = u2.connect() # connect to device
print(d.info)
这时看到类似下面的输出,就可以正式开始用我们这个库了。因为这个库功能太多,后面还有很多的内容,需要慢慢去看 ....
{'currentPackageName': 'net.oneplus.launcher', 'displayHeight': 1920, 'displayRotation': 0, 'displaySizeDpX': 411, 'displaySizeDpY': 731, 'displayWidth': 1080, 'productName': 'OnePlus5', '
screenOn': True, 'sdkInt': 27, 'naturalOrientation': True}
一般情况下都会成功,不过也可能会有意外。可以加QQ群反馈问题,群里有很多大佬可以帮你解决问题。
Thank you to all our sponsors! ✨🍰✨
霍格沃兹测试开发学社:中国软件测试开发高端教育品牌,产品由国内顶尖软件测试开发技术专家携手打造。为企业与个人提供专业的技能培训与咨询、测试工具与测试平台、测试外包与测试众包服务。领域涵盖App/Web自动化测试、接口自动化测试、性能测试、安全测试、持续交付/DevOps、测试左移、测试右移、精准测试、测试平台开发、测试管理等方向,联系我们
- 设备管理平台,设备多了就会用到 atxserver2
- 专门与adb进行交互的库 adbutils
- atx-agent 运行在设备上的驻守程序,go开发,用于保活设备上相关的服务
- weditor 类似于uiautomatorviewer,专门为本项目开发的辅助编辑器
- Install an app
- Launch an app
- Stop an app
- Stop all running apps
- Push and pull files
- Auto click permission dialogs
- Open Scheme
- Shell commands
- Session
- Retrieve the device info
- Key Events
- Gesture interaction with the device
- Screen-related
- Selector
- Watcher
- Global settings
- Input method
- Toast
- XPath
- Screenrecord
- Image match
常见问题
- 远程投屏
- htmlreport
- 诊断uiautomator2方法
- Plugin
- Hooks
- 失败时弹出提示框
-
Install uiautomator2
# Since uiautomator2 is still under development, you have to add --pre to install the development version pip install --upgrade --pre uiautomator2 # Or you can install directly from github source git clone https://github.com/openatx/uiautomator2 pip install -e uiautomator2
测试是否安装成功
uiautomator2 --help
-
Install weditor (UI Inspector)
因为uiautomator是独占资源,所以当atx运行的时候uiautomatorviewer是不能用的,为了减少atx频繁的启停,我们开发了基于浏览器技术的weditor UI查看器。https://github.com/openatx/weditor
安装方法(备注: 目前最新的稳定版为 0.1.0)
pip install -U weditor
安装好之后,就可以在命令行运行
weditor --help
确认是否安装成功了。Windows系统可以使用命令在桌面创建一个快捷方式
weditor --shortcut
命令行直接输入
weditor
会自动打开浏览器,输入设备的ip或者序列号,点击Connect即可。具体参考文章:浅谈自动化测试工具python-uiautomator2
-
Install daemons to a device (Optional)
电脑连接上一个手机或多个手机, 确保adb已经添加到环境变量中,执行下面的命令会自动安装本库所需要的设备端程序:uiautomator-server 、atx-agent、openstf/minicap、openstf/minitouch
# init 所有的已经连接到电脑的设备 python -m uiautomator2 init # 高阶用法 # init and set atx-agent listen in all address python -m uiautomator2 init --addr :7912
有时候init也会出错,请参考手动Init指南
安装提示
success
即可 -
【可选】AppetizerIO 所见即所得脚本编辑器
AppetizerIO 提供了对uiautomator2的深度集成,可以图形化管理ATX设备,还有所见即所得脚本编辑器
There are two ways to connect to the device.
- Through WiFi
Suppose device IP is 10.0.0.1
and your PC is in the same network.
import uiautomator2 as u2
d = u2.connect('10.0.0.1') # alias for u2.connect_wifi('10.0.0.1')
print(d.info)
- Through USB
Suppose the device serial is 123456f
(seen from adb devices
)
import uiautomator2 as u2
d = u2.connect('123456f') # alias for u2.connect_usb('123456f')
print(d.info)
- Through ADB WiFi
import uiautomator2 as u2
d = u2.connect_adb_wifi("10.0.0.1:5555")
# Equals to
# + Shell: adb connect 10.0.0.1:5555
# + Python: u2.connect_usb("10.0.0.1:5555")
Calling u2.connect()
with no argument, uiautomator2
will obtain device IP from the environment variable ANDROID_DEVICE_IP
or ANDROID_SERIAL
.
If this environment variable is empty, uiautomator will fall back to connect_usb
and you need to make sure that there is only one device connected to the computer.
其中的$device_ip
代表设备的ip地址
如需指定设备需要传入--serial
如 python3 -m uiautomator2 --serial bff1234 <SubCommand>
, SubCommand为子命令(init,或者screenshot等)
1.0.3 Added:
python3 -m uiautomator2
可以简写为uiautomator2
-
screenshot: 截图
$ uiautomator2 screenshot screenshot.jpg
-
current: 获取当前包名和activity
$ uiautomator2 current { "package": "com.android.browser", "activity": "com.uc.browser.InnerUCMobile", "pid": 28478 }
-
uninstall: 卸载
$ uiautomator2 uninstall <package-name> # 卸载一个包 $ uiautomator2 uninstall <package-name-1> <package-name-2> # 卸载多个包 $ uiautomator2 uninstall --all # 全部卸载
-
stop: 停止应用
$ uiautomator2 stop com.example.app # 停止一个app $ uiautomator2 stop --all # 停止所有的app
-
install: 安装apk,apk通过URL给出 (暂时不能用)
-
healthcheck: 健康检查 (暂不能用)
-
doctor: 检查uiautomator2无法使用的原因
$ uiautomator2 doctor I 210519 16:48:45 init:156] uiautomator2 version: 2.14.2.dev1 [D 210519 16:48:45 __main__:105] sdk:29 abi:arm64-v8a CHECK atx-agent GOOD: atx-agent version 0.10.0 CHECK uiautomator-apks GOOD: com.github.uiautomator 2.3.3 CHECK jsonrpc GOOD: d.info success ==> GOOD <==
How long (in seconds) will wait for a new command from the client before assuming the client quit and ending the uiautomator service (Default 3 minutes)
配置accessibility服务的最大空闲时间,超时将自动释放。默认3分钟。
d.set_new_command_timeout(300) # change to 5 minutes, unit seconds
Trace HTTP requests and response to find out how it works.
>>> d.debug = True
>>> d.info
12:32:47.182 $ curl -X POST -d '{"jsonrpc": "2.0", "id": "b80d3a488580be1f3e9cb3e926175310", "method": "deviceInfo", "params": {}}' 'http://127.0.0.1:54179/jsonrpc/0'
12:32:47.225 Response >>>
{"jsonrpc":"2.0","id":"b80d3a488580be1f3e9cb3e926175310","result":{"currentPackageName":"com.android.mms","displayHeight":1920,"displayRotation":0,"displaySizeDpX":360,"displaySizeDpY":640,"displayWidth":1080,"productName"
:"odin","screenOn":true,"sdkInt":25,"naturalOrientation":true}}
<<< END
Set default element wait time, unit seconds
设置元素查找等待时间(默认20s)
d.implicitly_wait(10.0) # 也可以通过d.settings['wait_timeout'] = 10.0 修改
d(text="Settings").click() # if Settings button not show in 10s, UiObjectNotFoundError will raised
print("wait timeout", d.implicitly_wait()) # get default implicit wait
This function will have influence on click
, long_click
, drag_to
, get_text
, set_text
, clear_text
, etc.
This part showcases how to perform app management
We only support installing an APK from a URL
d.app_install('http://some-domain.com/some.apk')
# 默认的这种方法是先通过atx-agent解析apk包的mainActivity,然后调用am start -n $package/$activity启动
d.app_start("com.example.hello_world")
# 使用 monkey -p com.example.hello_world -c android.intent.category.LAUNCHER 1 启动
# 这种方法有个副作用,它自动会将手机的旋转锁定给关掉
d.app_start("com.example.hello_world", use_monkey=True) # start with package name
# 通过指定main activity的方式启动应用,等价于调用am start -n com.example.hello_world/.MainActivity
d.app_start("com.example.hello_world", ".MainActivity")
# equivalent to `am force-stop`, thus you could lose data
d.app_stop("com.example.hello_world")
# equivalent to `pm clear`
d.app_clear('com.example.hello_world')
# stop all
d.app_stop_all()
# stop all app except for com.examples.demo
d.app_stop_all(excludes=['com.examples.demo'])
d.app_info("com.examples.demo")
# expect output
#{
# "mainActivity": "com.github.uiautomator.MainActivity",
# "label": "ATX",
# "versionName": "1.1.7",
# "versionCode": 1001007,
# "size":1760809
#}
# save app icon
img = d.app_icon("com.examples.demo")
img.save("icon.png")
d.app_list_running()
# expect output
# ["com.xxxx.xxxx", "com.github.uiautomator", "xxxx"]
pid = d.app_wait("com.example.android") # 等待应用运行, return pid(int)
if not pid:
print("com.example.android is not running")
else:
print("com.example.android pid is %d" % pid)
d.app_wait("com.example.android", front=True) # 等待应用前台运行
d.app_wait("com.example.android", timeout=20.0) # 最长等待时间20s(默认)
Added in version 1.2.0
-
push a file to the device
# push to a folder d.push("foo.txt", "/sdcard/") # push and rename d.push("foo.txt", "/sdcard/bar.txt") # push fileobj with open("foo.txt", 'rb') as f: d.push(f, "/sdcard/") # push and change file access mode d.push("foo.sh", "/data/local/tmp/", mode=0o755)
-
pull a file from the device
d.pull("/sdcard/tmp.txt", "tmp.txt") # FileNotFoundError will raise if the file is not found on the device d.pull("/sdcard/some-file-not-exists.txt", "tmp.txt")
d.healthcheck()
注意注意 disable_popups
函数,检测发现很不稳定,暂时不要使用,等候通知。
Import in version 0.1.1
d.disable_popups() # automatic skip popups
d.disable_popups(False) # disable automatic skip popups
If this method is not working on your device, You can make a pull request or create an issue to enhance this function. I'll show you how to do it.
- Open
uiautomatorviewer.bat
- Get popup hierarchy
Now you know the button text and current package name. Make a pull request by update function disable_popups
or create an issue if you are not familar with git and python.
You can do it wire adb: adb shell am start -a android.intent.action.VIEW -d "appname://appnamehost"
Also you can do it with python code
d.open_url("https://www.baidu.com")
d.open_url("taobao://taobao.com") # open Taobao app
d.open_url("appname://appnamehost")
This part showcases how to perform common device operations:
-
Run a short-lived shell command with a timeout protection. (Default timeout 60s)
Note: timeout support require
atx-agent >=0.3.3
adb_shell
function is deprecated. Useshell
instead.Simple usage
output, exit_code = d.shell("pwd", timeout=60) # timeout 60s (Default) # output: "/\n", exit_code: 0 # Similar to command: adb shell pwd # Since `shell` function return type is `namedtuple("ShellResponse", ("output", "exit_code"))` # so we can do some tricks output = d.shell("pwd").output exit_code = d.shell("pwd").exit_code
The first argument can be list. for example
output, exit_code = d.shell(["ls", "-l"]) # output: "/....", exit_code: 0
This returns a string for stdout merged with stderr. If the command is a blocking command,
shell
will also block until the command is completed or the timeout kicks in. No partial output will be received during the execution of the command. This API is not suitable for long-running commands. The shell command given runs in a similar environment ofadb shell
, which has a Linux permission level ofadb
orshell
(higher than an app permission). -
Run a long-running shell command
add stream=True will return
requests.models.Response
object. More info see requests streamr = d.shell("logcat", stream=True) # r: requests.models.Response deadline = time.time() + 10 # run maxium 10s try: for line in r.iter_lines(): # r.iter_lines(chunk_size=512, decode_unicode=None, delimiter=None) if time.time() > deadline: break print("Read:", line.decode('utf-8')) finally: r.close() # this method must be called
Command will be terminated when
r.close()
called.
Session represent an app lifecycle. Can be used to start app, detect app crash.
-
Launch and close app
sess = d.session("com.netease.cloudmusic") # start 网易云音乐 sess.close() # 停止网易云音乐 sess.restart() # 冷启动网易云音乐
-
Use python
with
to launch and close appwith d.session("com.netease.cloudmusic") as sess: sess(text="Play").click()
-
Attach to the running app
# launch app if not running, skip launch if already running sess = d.session("com.netease.cloudmusic", attach=True) # raise SessionBrokenError if not running sess = d.session("com.netease.cloudmusic", attach=True, strict=True)
-
Detect app crash
# When app is still running sess(text="Music").click() # operation goes normal # If app crash or quit sess(text="Music").click() # raise SessionBrokenError # other function calls under session will raise SessionBrokenError too
# check if session is ok. # Warning: function name may change in the future sess.running() # True or False
Get basic information
d.info
Below is a possible output:
{
u'displayRotation': 0,
u'displaySizeDpY': 640,
u'displaySizeDpX': 360,
u'currentPackageName': u'com.android.launcher',
u'productName': u'takju',
u'displayWidth': 720,
u'sdkInt': 18,
u'displayHeight': 1184,
u'naturalOrientation': True
}
Get window size
print(d.window_size())
# device upright output example: (1080, 1920)
# device horizontal output example: (1920, 1080)
Get current app info. For some android devices, the output could be empty (see Output example 3)
print(d.app_current())
# Output example 1: {'activity': '.Client', 'package': 'com.netease.example', 'pid': 23710}
# Output example 2: {'activity': '.Client', 'package': 'com.netease.example'}
# Output example 3: {'activity': None, 'package': None}
Wait activity
d.wait_activity(".ApiDemos", timeout=10) # default timeout 10.0 seconds
# Output: true of false
Get device serial number
print(d.serial)
# output example: 74aAEDR428Z9
Get WLAN ip
print(d.wlan_ip)
# output example: 10.0.0.1
Get detailed device info
print(d.device_info)
Below is a possible output:
{'udid': '3578298f-b4:0b:44:e6:1f:90-OD103',
'version': '7.1.1',
'serial': '3578298f',
'brand': 'SMARTISAN',
'model': 'OD103',
'hwaddr': 'b4:0b:44:e6:1f:90',
'port': 7912,
'sdk': 25,
'agentVersion': 'dev',
'display': {'width': 1080, 'height': 1920},
'battery': {'acPowered': False,
'usbPowered': False,
'wirelessPowered': False,
'status': 3,
'health': 0,
'present': True,
'level': 99,
'scale': 100,
'voltage': 4316,
'temperature': 272,
'technology': 'Li-ion'},
'memory': {'total': 3690280, 'around': '4 GB'},
'cpu': {'cores': 8, 'hardware': 'Qualcomm Technologies, Inc MSM8953Pro'},
'presenceChangedAt': '0001-01-01T00:00:00Z',
'usingBeganAt': '0001-01-01T00:00:00Z'}
Get of set clipboard content
设置粘贴板内容或获取内容 (目前已知问题是9.0之后的后台程序无法获取剪贴板的内容)
-
clipboard/set_clipboard
d.set_clipboard('text', 'label') print(d.clipboard)
-
Turn on/off screen
d.screen_on() # turn on the screen d.screen_off() # turn off the screen
-
Get current screen status
d.info.get('screenOn') # require Android >= 4.4
-
Press hard/soft key
d.press("home") # press the home key, with key name d.press("back") # press the back key, with key name d.press(0x07, 0x02) # press keycode 0x07('0') with META ALT(0x02)
-
These key names are currently supported:
- home
- back
- left
- right
- up
- down
- center
- menu
- search
- enter
- delete ( or del)
- recent (recent apps)
- volume_up
- volume_down
- volume_mute
- camera
- power
You can find all key code definitions at Android KeyEvnet
-
Unlock screen
d.unlock() # This is equivalent to # 1. launch activity: com.github.uiautomator.ACTION_IDENTIFY # 2. press the "home" key
-
Click on the screen
d.click(x, y)
-
Double click
d.double_click(x, y) d.double_click(x, y, 0.1) # default duration between two click is 0.1s
-
Long click on the screen
d.long_click(x, y) d.long_click(x, y, 0.5) # long click 0.5s (default)
-
Swipe
d.swipe(sx, sy, ex, ey) d.swipe(sx, sy, ex, ey, 0.5) # swipe for 0.5s(default)
-
SwipeExt 扩展功能
d.swipe_ext("right") # 手指右滑,4选1 "left", "right", "up", "down" d.swipe_ext("right", scale=0.9) # 默认0.9, 滑动距离为屏幕宽度的90% d.swipe_ext("right", box=(0, 0, 100, 100)) # 在 (0,0) -> (100, 100) 这个区域做滑动 # 实践发现上滑或下滑的时候,从中点开始滑动成功率会高一些 d.swipe_ext("up", scale=0.8) # 代码会vkk # 还可以使用Direction作为参数 from uiautomator2 import Direction d.swipe_ext(Direction.FORWARD) # 页面下翻, 等价于 d.swipe_ext("up"), 只是更好理解 d.swipe_ext(Direction.BACKWARD) # 页面上翻 d.swipe_ext(Direction.HORIZ_FORWARD) # 页面水平右翻 d.swipe_ext(Direction.HORIZ_BACKWARD) # 页面水平左翻
-
Drag
d.drag(sx, sy, ex, ey) d.drag(sx, sy, ex, ey, 0.5) # swipe for 0.5s(default)
-
Swipe points
# swipe from point(x0, y0) to point(x1, y1) then to point(x2, y2) # time will speed 0.2s bwtween two points d.swipe_points([(x0, y0), (x1, y1), (x2, y2)], 0.2))
多用于九宫格解锁,提前获取到每个点的相对坐标(这里支持百分比), 更详细的使用参考这个帖子 使用u2实现九宫图案解锁
-
Touch and drap (Beta)
这个接口属于比较底层的原始接口,感觉并不完善,不过凑合能用。注:这个地方并不支持百分比
d.touch.down(10, 10) # 模拟按下 time.sleep(.01) # down 和 move 之间的延迟,自己控制 d.touch.move(15, 15) # 模拟移动 d.touch.up() # 模拟抬起
Note: click, swipe, drag operations support percentage position values. Example:
d.long_click(0.5, 0.5)
means long click center of screen
-
Retrieve/Set device orientation
The possible orientations:
natural
orn
left
orl
right
orr
upsidedown
oru
(can not be set)
# retrieve orientation. the output could be "natural" or "left" or "right" or "upsidedown" orientation = d.orientation # WARNING: not pass testing in my TT-M1 # set orientation and freeze rotation. # notes: setting "upsidedown" requires Android>=4.3. d.set_orientation('l') # or "left" d.set_orientation("l") # or "left" d.set_orientation("r") # or "right" d.set_orientation("n") # or "natural"
-
Freeze/Un-freeze rotation
# freeze rotation d.freeze_rotation() # un-freeze rotation d.freeze_rotation(False)
-
Take screenshot
# take screenshot and save to a file on the computer, require Android>=4.2. d.screenshot("home.jpg") # get PIL.Image formatted images. Naturally, you need pillow installed first image = d.screenshot() # default format="pillow" image.save("home.jpg") # or home.png. Currently, only png and jpg are supported # get opencv formatted images. Naturally, you need numpy and cv2 installed first import cv2 image = d.screenshot(format='opencv') cv2.imwrite('home.jpg', image) # get raw jpeg data imagebin = d.screenshot(format='raw') open("some.jpg", "wb").write(imagebin)
-
Dump UI hierarchy
# get the UI hierarchy dump content (unicoded). xml = d.dump_hierarchy()
-
Open notification or quick settings
d.open_notification() d.open_quick_settings()
Selector is a handy mechanism to identify a specific UI object in the current window.
# Select the object with text 'Clock' and its className is 'android.widget.TextView'
d(text='Clock', className='android.widget.TextView')
Selector supports below parameters. Refer to UiSelector Java doc for detailed information.
text
,textContains
,textMatches
,textStartsWith
className
,classNameMatches
description
,descriptionContains
,descriptionMatches
,descriptionStartsWith
checkable
,checked
,clickable
,longClickable
scrollable
,enabled
,focusable
,focused
,selected
packageName
,packageNameMatches
resourceId
,resourceIdMatches
index
,instance
-
children
# get the children or grandchildren d(className="android.widget.ListView").child(text="Bluetooth")
-
siblings
# get siblings d(text="Google").sibling(className="android.widget.ImageView")
-
children by text or description or instance
# get the child matching the condition className="android.widget.LinearLayout" # and also its children or grandchildren with text "Bluetooth" d(className="android.widget.ListView", resourceId="android:id/list") \ .child_by_text("Bluetooth", className="android.widget.LinearLayout") # get children by allowing scroll search d(className="android.widget.ListView", resourceId="android:id/list") \ .child_by_text( "Bluetooth", allow_scroll_search=True, className="android.widget.LinearLayout" )
-
child_by_description
is to find children whose grandchildren have the specified description, other parameters being similar tochild_by_text
. -
child_by_instance
is to find children with has a child UI element anywhere within its sub hierarchy that is at the instance specified. It is performed on visible views without scrolling.
See below links for detailed information:
- UiScrollable,
getChildByDescription
,getChildByText
,getChildByInstance
- UiCollection,
getChildByDescription
,getChildByText
,getChildByInstance
Above methods support chained invoking, e.g. for below hierarchy
<node index="0" text="" resource-id="android:id/list" class="android.widget.ListView" ...> <node index="0" text="WIRELESS & NETWORKS" resource-id="" class="android.widget.TextView" .../> <node index="1" text="" resource-id="" class="android.widget.LinearLayout" ...> <node index="1" text="" resource-id="" class="android.widget.RelativeLayout" ...> <node index="0" text="Wi‑Fi" resource-id="android:id/title" class="android.widget.TextView" .../> </node> <node index="2" text="ON" resource-id="com.android.settings:id/switchWidget" class="android.widget.Switch" .../> </node> ... </node>
To click the switch widget right to the TextView 'Wi‑Fi', we need to select the switch widgets first. However, according to the UI hierarchy, more than one switch widgets exist and have almost the same properties. Selecting by className will not work. Alternatively, the below selecting strategy would help:
d(className="android.widget.ListView", resourceId="android:id/list") \ .child_by_text("Wi‑Fi", className="android.widget.LinearLayout") \ .child(className="android.widget.Switch") \ .click()
-
-
relative positioning
Also we can use the relative positioning methods to get the view:
left
,right
,top
,bottom
.d(A).left(B)
, selects B on the left side of A.d(A).right(B)
, selects B on the right side of A.d(A).up(B)
, selects B above A.d(A).down(B)
, selects B under A.
So for above cases, we can alternatively select it with:
## select "switch" on the right side of "Wi‑Fi" d(text="Wi‑Fi").right(className="android.widget.Switch").click()
-
Multiple instances
Sometimes the screen may contain multiple views with the same properties, e.g. text, then you will have to use the "instance" property in the selector to pick one of qualifying instances, like below:
d(text="Add new", instance=0) # which means the first instance with text "Add new"
In addition, uiautomator2 provides a list-like API (similar to jQuery):
# get the count of views with text "Add new" on current screen d(text="Add new").count # same as count property len(d(text="Add new")) # get the instance via index d(text="Add new")[0] d(text="Add new")[1] ... # iterator for view in d(text="Add new"): view.info # ...
Notes: when using selectors in a code block that walk through the result list, you must ensure that the UI elements on the screen keep unchanged. Otherwise, when Element-Not-Found error could occur when iterating through the list.
-
Check if the specific UI object exists
d(text="Settings").exists # True if exists, else False d.exists(text="Settings") # alias of above property. # advanced usage d(text="Settings").exists(timeout=3) # wait Settings appear in 3s, same as .wait(3)
-
Retrieve the info of the specific UI object
d(text="Settings").info
Below is a possible output:
{ u'contentDescription': u'', u'checked': False, u'scrollable': False, u'text': u'Settings', u'packageName': u'com.android.launcher', u'selected': False, u'enabled': True, u'bounds': {u'top': 385, u'right': 360, u'bottom': 585, u'left': 200}, u'className': u'android.widget.TextView', u'focused': False, u'focusable': True, u'clickable': True, u'chileCount': 0, u'longClickable': True, u'visibleBounds': {u'top': 385, u'right': 360, u'bottom': 585, u'left': 200}, u'checkable': False }
-
Get/Set/Clear text of an editable field (e.g., EditText widgets)
d(text="Settings").get_text() # get widget text d(text="Settings").set_text("My text...") # set the text d(text="Settings").clear_text() # clear the text
-
Get Widget center point
x, y = d(text="Settings").center() # x, y = d(text="Settings").center(offset=(0, 0)) # left-top x, y
-
Take screenshot of widget
im = d(text="Settings").screenshot() im.save("settings.jpg")
-
Perform click on the specific object
# click on the center of the specific ui object d(text="Settings").click() # wait element to appear for at most 10 seconds and then click d(text="Settings").click(timeout=10) # click with offset(x_offset, y_offset) # click_x = x_offset * width + x_left_top # click_y = y_offset * height + y_left_top d(text="Settings").click(offset=(0.5, 0.5)) # Default center d(text="Settings").click(offset=(0, 0)) # click left-top d(text="Settings").click(offset=(1, 1)) # click right-bottom # click when exists in 10s, default timeout 0s clicked = d(text='Skip').click_exists(timeout=10.0) # click until element gone, return bool is_gone = d(text="Skip").click_gone(maxretry=10, interval=1.0) # maxretry default 10, interval default 1.0
-
Perform long click on the specific UI object
# long click on the center of the specific UI object d(text="Settings").long_click()
-
Drag the UI object towards another point or another UI object
# notes : drag can not be used for Android<4.3. # drag the UI object to a screen point (x, y), in 0.5 second d(text="Settings").drag_to(x, y, duration=0.5) # drag the UI object to (the center position of) another UI object, in 0.25 second d(text="Settings").drag_to(text="Clock", duration=0.25)
-
Swipe from the center of the UI object to its edge
Swipe supports 4 directions:
- left
- right
- top
- bottom
d(text="Settings").swipe("right") d(text="Settings").swipe("left", steps=10) d(text="Settings").swipe("up", steps=20) # 1 steps is about 5ms, so 20 steps is about 0.1s d(text="Settings").swipe("down", steps=20)
-
Two-point gesture from one point to another
d(text="Settings").gesture((sx1, sy1), (sx2, sy2), (ex1, ey1), (ex2, ey2))
-
Two-point gesture on the specific UI object
Supports two gestures:
In
, from edge to centerOut
, from center to edge
# notes : pinch can not be set until Android 4.3. # from edge to center. here is "In" not "in" d(text="Settings").pinch_in(percent=100, steps=10) # from center to edge d(text="Settings").pinch_out()
-
Wait until the specific UI appears or disappears
# wait until the ui object appears d(text="Settings").wait(timeout=3.0) # return bool # wait until the ui object gone d(text="Settings").wait_gone(timeout=1.0)
The default timeout is 20s. see global settings for more details
-
Perform fling on the specific ui object(scrollable)
Possible properties:
horiz
orvert
forward
orbackward
ortoBeginning
ortoEnd
# fling forward(default) vertically(default) d(scrollable=True).fling() # fling forward horizontally d(scrollable=True).fling.horiz.forward() # fling backward vertically d(scrollable=True).fling.vert.backward() # fling to beginning horizontally d(scrollable=True).fling.horiz.toBeginning(max_swipes=1000) # fling to end vertically d(scrollable=True).fling.toEnd()
-
Perform scroll on the specific ui object(scrollable)
Possible properties:
horiz
orvert
forward
orbackward
ortoBeginning
ortoEnd
, orto
# scroll forward(default) vertically(default) d(scrollable=True).scroll(steps=10) # scroll forward horizontally d(scrollable=True).scroll.horiz.forward(steps=100) # scroll backward vertically d(scrollable=True).scroll.vert.backward() # scroll to beginning horizontally d(scrollable=True).scroll.horiz.toBeginning(steps=100, max_swipes=1000) # scroll to end vertically d(scrollable=True).scroll.toEnd() # scroll forward vertically until specific ui object appears d(scrollable=True).scroll.to(text="Security")
目前的这个watch_context是用threading启动的,每2s检查一次 目前还只有click这一种触发操作
with d.watch_context() as ctx:
ctx.when("^立即(下载|更新)").when("取消").click() # 当同时出现 (立即安装 或 立即取消)和 取消 按钮的时候,点击取消
ctx.when("同意").click()
ctx.when("确定").click()
# 上面三行代码是立即执行完的,不会有什么等待
ctx.wait_stable() # 开启弹窗监控,并等待界面稳定(两个弹窗检查周期内没有弹窗代表稳定)
# 使用call函数来触发函数回调
# call 支持两个参数,d和el,不区分参数位置,可以不传参,如果传参变量名不能写错
# eg: 当有元素匹配仲夏之夜,点击返回按钮
ctx.when("仲夏之夜").call(lambda d: d.press("back"))
ctx.when("确定").call(lambda el: el.click())
# 其他操作
# 为了方便也可以使用代码中默认的弹窗监控逻辑
# 下面是目前内置的默认逻辑,可以加群at群主,增加新的逻辑,或者直接提pr
# when("继续使用").click()
# when("移入管控").when("取消").click()
# when("^立即(下载|更新)").when("取消").click()
# when("同意").click()
# when("^(好的|确定)").click()
with d.watch_context(builtin=True) as ctx:
# 在已有的基础上增加
ctx.when("@tb:id/jview_view").when('//*[@content-desc="图片"]').click()
# 其他脚本逻辑
另外一种写法
ctx = d.watch_context()
ctx.when("设置").click()
ctx.wait_stable() # 等待界面不在有弹窗了
ctx.close()
更推荐用WatchContext 写法更简洁一些
You can register watchers to perform some actions when a selector does not find a match.
2.0.0之前使用的是 uiautomator-jar库中提供的[Watcher]((http://developer.android.com/tools/help/uiautomator/UiWatcher.html)方法,但在实践中发现一旦uiautomator所有的watcher配置都是丢失,这肯定是无法接受的。 所以目前采用了后台运行了一个线程的方法(依赖threading库),然后每隔一段时间dump一次hierarchy,匹配到元素之后执行相应的操作。
用法举例
注册监控
# 常用写法,注册匿名监控
d.watcher.when("安装").click()
# 注册名为ANR的监控,当出现ANR和Force Close时,点击Force Close
d.watcher("ANR").when(xpath="ANR").when("Force Close").click()
# 其他回调例子
d.watcher.when("抢红包").press("back")
d.watcher.when("//*[@text = 'Out of memory']").call(lambda d: d.shell('am force-stop com.im.qq'))
# 回调说明
def click_callback(d: u2.Device):
d.xpath("确定").click() # 在回调中调用不会再次触发watcher
d.xpath("继续").click() # 使用d.xpath检查元素的时候,会触发watcher(目前最多触发5次)
监控操作
# 移除ANR的监控
d.watcher.remove("ANR")
# 移除所有的监控
d.watcher.remove()
# 开始后台监控
d.watcher.start()
d.watcher.start(2.0) # 默认监控间隔2.0s
# 强制运行所有监控
d.watcher.run()
# 停止监控
d.watcher.stop()
# 停止并移除所有的监控,常用于初始化
d.watcher.reset()
另外文档还是有很多没有写,推荐直接去看源码watcher.py
d.HTTP_TIMEOUT = 60 # 默认值60s, http默认请求超时时间
# 当设备掉线时,等待设备在线时长,仅当TMQ=true时有效,支持通过环境变量 WAIT_FOR_DEVICE_TIMEOUT 设置
d.WAIT_FOR_DEVICE_TIMEOUT = 70
其他的配置,目前已大部分集中到 d.settings
中,根据后期的需求配置可能会有增减。
print(d.settings)
{'operation_delay': (0, 0),
'operation_delay_methods': ['click', 'swipe'],
'wait_timeout': 20.0,
'xpath_debug': False}
# 配置点击前延时0.5s,点击后延时1s
d.settings['operation_delay'] = (.5, 1)
# 修改延迟生效的方法
# 其中 double_click, long_click 都对应click
d.settings['operation_delay_methods'] = ['click', 'swipe', 'drag', 'press']
d.settings['xpath_debug'] = True # 开启xpath插件的调试日志
d.settings['wait_timeout'] = 20.0 # 默认控件等待时间(原生操作,xpath插件的等待时间)
对于随着版本升级,设置过期的配置时,会提示Deprecated,但是不会抛异常。
>>> d.settings['click_before_delay'] = 1
[W 200514 14:55:59 settings:72] d.settings[click_before_delay] deprecated: Use operation_delay instead
uiautomator恢复方式设置
细心的你可能发现,实际上手机安装了两个APK,一个在前台可见(小黄车)。一个包名为com.github.uiautomator.test
在后台不可见。这两个apk使用同一个证书签名的。
不可见的应用实际上是一个测试包,包含有所有的测试代码,核心的测试服务也是通过其启动的。
但是运行的时候,系统却需要那个小黄车一直在运行(在后台运行也可以)。一旦小黄车应用被杀,后台运行的测试服务也很快的会被杀掉。就算什么也不做,应用应用在后台,也会很快被系统回收掉。(这里希望高手指点一下,如何才能不依赖小黄车应用,感觉理论上是可以的,但是目前我还不会)。
让小黄车在后台运行有两种方式,一种启动应用后,放到后台(默认)。另外通过am startservice
启动一个后台服务也行。
通过 d.settings["uiautomator_runtest_app_background"] = True
可以调整该行为。True代表启动应用,False代表启动服务。
UiAutomator中的超时设置(隐藏方法)
>> d.jsonrpc.getConfigurator()
{'actionAcknowledgmentTimeout': 500,
'keyInjectionDelay': 0,
'scrollAcknowledgmentTimeout': 200,
'waitForIdleTimeout': 0,
'waitForSelectorTimeout': 0}
>> d.jsonrpc.setConfigurator({"waitForIdleTimeout": 100})
{'actionAcknowledgmentTimeout': 500,
'keyInjectionDelay': 0,
'scrollAcknowledgmentTimeout': 200,
'waitForIdleTimeout': 100,
'waitForSelectorTimeout': 0}
为了防止客户端程序响应超时,waitForIdleTimeout
和waitForSelectorTimeout
目前已改为0
Refs: Google uiautomator Configurator
这种方法通常用于不知道控件的情况下的输入。第一步需要切换输入法,然后发送adb广播命令,具体使用方法如下
d.set_fastinput_ime(True) # 切换成FastInputIME输入法
d.send_keys("你好123abcEFG") # adb广播输入
d.clear_text() # 清除输入框所有内容(Require android-uiautomator.apk version >= 1.0.7)
d.set_fastinput_ime(False) # 切换成正常的输入法
d.send_action("search") # 模拟输入法的搜索
send_action 说明
该函数可以使用的参数有 go search send next done previous
什么时候该使用这个函数呢?
有些时候在EditText中输入完内容之后,调用press("search")
or press("enter")
发现并没有什么反应。
这个时候就需要send_action
函数了,这里用到了只有输入法才能用的IME_ACTION_CODE。
send_action
先broadcast命令发送给输入法操作IME_ACTION_CODE
,由输入法完成后续跟EditText的通信。(原理我不太清楚,有了解的,提issue告诉我)
Show Toast
d.toast.show("Hello world")
d.toast.show("Hello world", 1.0) # show for 1.0s, default 1.0s
Get Toast
# [Args]
# 5.0: max wait timeout. Default 10.0
# 10.0: cache time. return cache toast if already toast already show up in recent 10 seconds. Default 10.0 (Maybe change in the furture)
# "default message": return if no toast finally get. Default None
d.toast.get_message(5.0, 10.0, "default message")
# common usage
assert "Short message" in d.toast.get_message(5.0, default="")
# clear cached toast
d.toast.reset()
# Now d.toast.get_message(0) is None
Java uiautoamtor中默认是不支持xpath的,所以这里属于扩展的一个功能。速度不是这么的快。
For example: 其中一个节点的内容
<android.widget.TextView
index="2"
text="05:19"
resource-id="com.netease.cloudmusic:id/qf"
package="com.netease.cloudmusic"
content-desc=""
checkable="false" checked="false" clickable="false" enabled="true" focusable="false" focused="false"
scrollable="false" long-clickable="false" password="false" selected="false" visible-to-user="true"
bounds="[957,1602][1020,1636]" />
xpath定位和使用方法
有些属性的名字有修改需要注意
description -> content-desc
resourceId -> resource-id
常见用法
# wait exists 10s
d.xpath("//android.widget.TextView").wait(10.0)
# find and click
d.xpath("//*[@content-desc='分享']").click()
# check exists
if d.xpath("//android.widget.TextView[contains(@text, 'Se')]").exists:
print("exists")
# get all text-view text, attrib and center point
for elem in d.xpath("//android.widget.TextView").all():
print("Text:", elem.text)
# Dictionary eg:
# {'index': '1', 'text': '999+', 'resource-id': 'com.netease.cloudmusic:id/qb', 'package': 'com.netease.cloudmusic', 'content-desc': '', 'checkable': 'false', 'checked': 'false', 'clickable': 'false', 'enabled': 'true', 'focusable': 'false', 'focused': 'false','scrollable': 'false', 'long-clickable': 'false', 'password': 'false', 'selected': 'false', 'visible-to-user': 'true', 'bounds': '[661,1444][718,1478]'}
print("Attrib:", elem.attrib)
# Coordinate eg: (100, 200)
print("Position:", elem.center())
点击查看其他XPath常见用法
视频录制
这里没有使用手机中自带的screenrecord命令,是通过获取手机图片合成视频的方法,所以需要安装一些其他的依赖,如imageio, imageio-ffmpeg, numpy等 因为有些依赖比较大,推荐使用镜像安装。直接运行下面的命令即可。
pip3 install -U "uiautomator2[image]" -i https://pypi.doubanio.com/simple
使用方法
d.screenrecord('output.mp4')
time.sleep(10)
# or do something else
d.screenrecord.stop() # 停止录制后,output.mp4文件才能打开
录制的时候也可以指定fps(当前是20),这个值是率低于minicap输出图片的速度,感觉已经很好了,不建议你修改。
图像匹配,在使用这个功能之前你需要先把依赖安装上
pip3 install -U "uiautomator2[image]" -i https://pypi.doubanio.com/simple
目前开放两个接口
imdata = "target.png" # 也可以是URL, PIL.Image或OpenCV打开的图像
d.image.match(imdata)
# 匹配待查找的图片,立刻返回一个结果
# 返回一个dict, eg: {"similarity": 0.9, "point": [200, 300]}
d.image.click(imdata, timeout=20.0)
# 在20s的时间内调用match轮询查找图片,当similarity>0.9时,执行点击操作
该功能还在完善中,图片需要手机的原图裁剪后的图才可以。
很多没写在这个地方的,都放到了这里 Common Issues
停止UiAutomator守护服务
因为有atx-agent
的存在,Uiautomator会被一直守护着,如果退出了就会被重新启动起来。但是Uiautomator又是霸道的,一旦它在运行,手机上的辅助功能、电脑上的uiautomatorviewer 就都不能用了,除非关掉该框架本身的uiautomator。下面就说下两种关闭方法
方法1:
直接打开uiautomator app(init成功后,就会安装上的),点击关闭UIAutomator
方法2:
d.uiautomator.stop()
# d.uiautomator.start() # 启动
# d.uiautomator.running() # 是否在运行
ATX与Maxim共存AccessibilityService的方法
优秀文章推荐 (欢迎QQ群里at我反馈)
- termux里如何部署uiautomator2简介 by
成都-测试只会一点点
- 新增接口:UiObject2、Until、By、BySelector
- 引入方式:2.0中,com.android.uiautomator.core.* 引入方式被废弃。改为android.support.test.uiautomator
- 构建系统:Maven 和/或 Ant(1.x);Gradle(2.0)
- 产生的测试包的形式:从zip /jar(1.x) 到 apk(2.0)
- 在本地环境以adb命令运行UIAutomator测试,启动方式的差别:
adb shell uiautomator runtest UiTest.jar -c package.name.ClassName(1.x) adb shell am instrument -e class com.example.app.MyTest com.example.app.test/android.support.test.runner.AndroidJUnitRunner(2.0) - 能否使用Android服务及接口? 1.x
不能;2.0能。 - og输出? 使用System.out.print输出流回显至执行端(1.x); 输出至Logcat(2.0)
- 执行?测试用例无需继承于任何父类,方法名不限,使用注解 Annotation进行(2.0); 需要继承UiAutomatorTestCase,测试方法需要以test开头(1.x)
重大更新
-
1.0.0
移除
d.watchers.watched
(会拖慢自动化的执行速度并且还会降低稳定性)
- uiautomator守护程序 https://github.com/openatx/atx-agent
- uiautomator jsonrpc serverhttps://github.com/openatx/android-uiautomator-server/
- codeskyblue (@codeskyblue)
- Xiaocong He (@xiaocong)
- Yuanyuan Zou (@yuanyuan)
- Qian Jin (@QianJin2013)
- Xu Jingjie (@xiscoxu)
- Xia Mingyuan (@mingyuan-xia)
- Artem Iglikov, Google Inc. (@artikz)
Other contributors
- google/mobly 谷歌内部的测试框架,虽然我不太懂,但是感觉很好用
- https://www.appetizer.io/ 包含一个很好用的IDE,快速编写脚本,也可以插桩采集性能。
- https://github.com/atinfo/awesome-test-automation 所有优秀测试框架的集合,包罗万象
- http://www.sikulix.com/ 基于图像识别的自动化测试框架,非常的老牌
- http://airtest.netease.com/ 本项目的前身,后来被网易广州团队接手并继续优化。实现有一个不错的IDE