Yeti automates tests written for various test frameworks. Yeti scales from your dev box (where it works by itself) to your CI system (where it launches browsers with Selenium) without changing your existing tests.
- Works with the frameworks you already use.
- Automates tests without any additional software. Selenium not required!
- Built-in code coverage provided by Istanbul.
- Works with IE 6+, Android 4+, Firefox, Safari, Chrome, iOS 4+.
- Server-side AJAX testing with echoecho.
- JUnit XML output makes Yeti play nice with Jenkins.
- Workload can be split across multiple instances of the same browser.
- Optional Selenium/WebDriver browser launching. Works great with Sauce Labs.
You can use any of these test frameworks with Yeti.
- QUnit
- Jasmine
- Mocha with Expect.js assertions
- Dojo Objective Harness
- YUI Test
- Your framework here. Submit a pull request!
npm install -g yeti
Yeti requires Node.js, which provides the npm
command for installation.
You can download Node.js source or pre-built
installers from their website.
Just run Yeti with the HTML files containing your tests.
$ yeti test/*.html
Creating a Hub at http://localhost:9000
Waiting for agents to connect at http://localhost:9000.
When ready, press Enter to begin testing.
Point your browsers at that URL, then come back and press Enter.
[Open some browsers...]
Agent connected: Safari (6.0) / Mac OS
Agent connected: Chrome (22.0.1221.0) / Mac OS
[Come back, press Enter]
✔ Testing started on Safari (6.0) / Mac OS, Chrome (22.0.1221.0) / Mac OS
✔ Agent completed: Safari (6.0) / Mac OS
✔ Agent completed: Chrome (22.0.1221.0) / Mac OS
504 tests passed! (9.1 seconds)
$
Yeti exits automatically when all tests complete. If test failures occur, Yeti will exit with a non-zero status code.
0
indicates testing was completed and all tests passed.1
indicates an error occurred that prevented testing from running to completion.3
indicated testing was completed but some tests failed.
These codes are useful for using Yeti programmatically. The codes 0
and 3
indicate testing completion.
If Yeti ended in code 1
, you may wish to retry the command to workaround transient errors. This is useful
when using Yeti to launch browsers in CI, which sometimes may fail due to temporary Selenium issues.
Yeti can output machine-readable JUnit XML suitable for use in Jenkins with the --junit
option.
$ yeti --junit test/*.html > yeti.xml
Yeti will output XML on stdout and status messages on stderr.
When using Yeti several times in the same Jenkins job, it's useful to label tests with a prefix
to distinguish between different Yeti runs after Jenkins merges the reports together. You can
assign this prefix with the --name
option.
$ yeti --name stable --junit test/stable/*.html > stable.xml
$ yeti --name flaky --junit test/flaky/*.html > flaky.xml
Yeti can generate a code coverage report with the -c
or --coverage
option.
$ yeti -c
Found 1 file to test.
Agent connected: Chrome (34.0.1825.4) / Mac OS from 127.0.0.1
✓ Testing started on Chrome (34.0.1825.4) / Mac OS
✓ Agent completed: Chrome (34.0.1825.4) / Mac OS
Coverage summary:
Statements : 96.02% ( 1110/1156 )
Branches : 88.53% ( 772/872 )
Functions : 97.45% ( 229/235 )
Lines : 96.27% ( 1110/1153 )
✓ 397 tests passed! (7 seconds)
Use the --coverage-report
option to generate reports written to ./coverage
or another directory defined by --coverage-dir
. Values for --coverage-report
include:
lcov
for LCOV and HTML reports.lcovonly
for LCOV without HTML.html
for HTML only.json
for Istanbul JSON.summary
for the default summary shown above.- Any other report supported by Istanbul.
Yeti instruments all JavaScript files automatically with Istanbul. If you'd prefer to instrument
your JavaScript code with Istanbul as a part of a build step, you can pass --no-instrument
to skip all instrumenting. Yeti will still generate code coverage reports if used with
the --coverage
option.
If you'd like to exclude some JavaScript files from instrumenting, you can pass multiple
--instrument-exclude
options defining minimatch patterns to exclude.
$ yeti -c --instrument-exclude "**/vendor/**" --instrument-exclude "**.nocover.js"
To prevent needing to specify these verbose options every time, you can put a
coverageOptions
object in your project's .yeti.json
file containing configuration
that is only used with --coverage
.
$ cd yui3/src/app
$ cat ../../.yeti.json
{
"basedir": ".",
"glob": "**/tests/unit/*.html",
"coverageOptions": {
"instrument": false,
"query": "filter=coverage"
}
}
$ yeti -c # will run with --no-instrument --query "filter=coverage"
Found 1 file to test.
Agent connected: Chrome (34.0.1825.4) / Mac OS from 127.0.0.1
✓ Testing started on Chrome (34.0.1825.4) / Mac OS
✓ Agent completed: Chrome (34.0.1825.4) / Mac OS
Coverage summary:
Statements : 96.02% ( 1110/1156 )
Branches : 88.53% ( 772/872 )
Functions : 97.45% ( 229/235 )
Lines : 96.27% ( 1110/1153 )
✓ 397 tests passed! (7 seconds)
Yeti provides server-side AJAX routes with echoecho. Your test can make relative HTTP requests to test your code against server-side HTTP GET, POST, PUT, DELETE, OPTIONS, GET with delay, JSON or JSONP responses via POST, or any HTTP status code.
Example supported routes:
echo/status/500
returns a 500 response.echo/delay/3
returns a 200 response after 3 seconds.echo/jsonp?callback=foo
returns a JSONP response with the given POST data wrapped in a call tofoo
.
Note these routes are intentionally relative paths. See the echoecho README for more details.
Yeti will move on to the next test if a test takes longer than 5 minutes (300 seconds).
You can adjust this interval with the --timeout
option.
This will run Yeti with a 120 second timeout:
$ yeti --timeout 120 test.html
There isn't a general timeout setting. Yeti actively pings browsers about every 2-5 seconds and disconnects them if they fail to respond to a ping three times.
You can specify query string parameters to add to your test URLs. This can be used to pass information to your tests that control its behavior.
This will append ?filter=coverage
to your tests, which is used by the tests
for the YUI Library to trigger loading instrumented code.
$ yeti --query 'filter=coverage' test/*.html
Yeti will report an uncaught exceptions as Script Errors.
Yeti enforces No-Quirks Mode in your tests because it may impact DOM-related APIs. Add a DOCTYPE to your test document to fix this.
When combined with localtunnel, mobile testing is simple. If you're not dealing with sensitive information, startup your Yeti Hub and then run:
$ localtunnel 9000
Port 9000 is now publicly accessible from http://3z48.localtunnel.com ...
You can then visit that URL on your mobile (or any other) device and have it run new tests.
To save time, start a Yeti Hub.
$ yeti --server
Yeti Hub listening on port 9000.
Point browsers at your local Yeti on port 9000. Now, you're ready to run tests without having to reconnect browsers each time.
Starting Yeti in another terminal will connect to that Hub instead of starting a new one and will begin testing immediately if browsers are already connected.
$ yeti test/*.html
Connected to http://localhost:9000
Agent connected: Chrome (22.0.1221.0) / Mac OS
Agent connected: Safari (6.0) / Mac OS
✔ Testing started on Chrome (22.0.1221.0) / Mac OS, Safari (6.0) / Mac OS
✔ Agent completed: Chrome (22.0.1221.0) / Mac OS
✔ Agent completed: Safari (6.0) / Mac OS
504 tests passed! (11.5 seconds)
$
Your Yeti Hub can be shared with other developers.
First, I'll start a Hub on test.yeti.cx on port 80.
$ yeti --server --port 80
Go ahead and point a few browsers there.
Now, others can connect to it from their computer like so:
$ yeti --hub http://test.yeti.cx/ test/*.html
Connected to http://test.yeti.cx/
Waiting for agents to connect at http://test.yeti.cx/.
When ready, press Enter to begin testing.
Your pwd
and your test file will be served through the Hub. Like magic.
[Hit Enter]
Agent connected: Chrome (22.0.1221.0) / Mac OS
Agent connected: Safari (6.0) / Mac OS
✔ Testing started on Chrome (22.0.1221.0) / Mac OS, Safari (6.0) / Mac OS
✔ Agent completed: Safari (6.0) / Mac OS
✔ Agent completed: Chrome (22.0.1221.0) / Mac OS
504 tests passed! (8.7 seconds)
This makes it really simple to setup an ad-hoc testing lab shared with your team.
You can specify the wd-url
option to connect Yeti to a Selenium 2 Hub using the
WebDriver protocol. Specifying one or more caps
options will cause Yeti to launch
browsers for the given capabilities over WebDriver.
For example, launching 2 Firefox browsers to test multiple files:
yeti --hub http://example.com --wd-url http://localhost:4444 \
--caps "browserName=firefox;version=26;platform=Windows XP" \
--caps "browserName=firefox;version=26;platform=Windows XP" test/*.html
Or launching an iPhone simulator on Sauce Labs:
yeti --caps "platform=OS X 10.8;version=7;device-orientation=portrait;app=safari;device=iPhone Simulator" test.html
Here's a breakdown of all available CLI options.
- query (String) Query string parameters to pass to tests.
- timeout (Number in seconds) Test timeout.
- hub (URL) Location of the Yeti Hub to use.
Set to
false
or specify--no-hub
to override a configuration file. - server Starts a Yeti Hub.
- port (Number) Yeti Hub will listen to this port.
- loglevel (
debug
orinfo
) Print debugging information. - caps (String) Browser capabilities to launch with WebDriver.
Should be used with
wd-url
. - wd-url (URL) WebDriver Hub URL. May contain a username and password.
- help Print usage.
- version Print the Yeti version.
You may use place JSON in a .yeti.json
file to set project or user specific configuration.
Yeti will look for .yeti.json
in these places:
- Recursively starting in the directory you start Yeti
- In your home folder
Here is an example .yeti.json
for the YUI project, which is placed in the repository root:
{
"hub": "http://test.yeti.cx/",
"basedir": ".",
"glob": "**/tests/unit/*.html"
}
Here is the breakdown of these settings:
- The hub option defines a Yeti Hub URL to use.
- The basedir option indicates that the directory where
.yeti.json
lives is permitted to serve files to the Yeti Hub. - The glob option defines a pattern to search for test files.
These settings let YUI developers simply run yeti
inside of the project directory
to run tests. Since all tests in the project match the glob pattern, the yeti
command works for specific components as well as for the entire project.
This configuration can be overridden on the command line. For example, to ignore the
hub setting, you can run Yeti with --no-hub
.
You can require("yeti")
inside your application to script Yeti for your own use.
For API documentation:
- Run
make html
to build HTML documentation to./build_docs
. - Review code comments inside
lib/yeti.js
,lib/client.js
andlib/hub/index.js
.
Yeti follows Semantic Versioning but is currently at a 0.x.y release. The public API is not stable. There will be changes.
Yeti typically automates test frameworks, but you can integrate any client-side test or performance framework into Yeti. Combined with the Yeti API, you can easily build your own automation tools. YUI uses Yeti in this way to automate performance benchmarks.
Normally Yeti will scan pages in order to find test frameworks. When serving a page to Yeti, you can set
window.stopYetiScan
to true to signal that your page will explicitly submit results to Yeti.
When your framework has results and is ready to move to the next page, you can call window.sendYetiResults
with an object containing data to report. This data will be passed through verbatim to the Node.js Yeti API
for further processing in your tool.
Yeti should work on all platforms supported by Node.js. It's tested on Linux and OS X.
You must start Yeti's client in the directory you'll be serving tests from. For security reasons, Yeti will reject requests that try to access files outside of the directory you start Yeti in.
You can install the latest development snapshot of Yeti easily:
npm install -g http://latest.yeti.cx
This will install Yeti as it exists on the yui/yeti GitHub repository. You can check the stability of the Yeti snapshot by checking yui/yeti on Travis.
Do you want to add new features or fix bugs in Yeti itself? We made it easy for you to hack on Yeti.
After running npm install
, replace the make
commands below with
.\jake.bat
to use the experimental Jake tasks that are Windows ready.
Clone Yeti.
git clone https://github.com/yui/yeti.git
cd yeti
Install Yeti's dependencies.
npm install
Yeti's automated tests require PhantomJS.
You can download PhantomJS source or pre-built
binaries from their website. Make sure the phantomjs
binary is installed in your PATH.
make test
make coverage
The latter command uses JSCoverage for Node.js,
which will be built and installed to ./tools/jscoverage
.
make lint
You may also run the linter on individual files with ./go lint
:
./go lint test/blizzard.js
Yeti uses JSHint to analyze code for problems. See .jshintrc
for options used by Yeti.
Requires Google Chrome Canary and OS X.
Profile the Yeti Hub:
./go profile --server
Using ./go profile
without --server
to profile the Yeti client
requires an interactive terminal, which does not yet work.
Yeti uses Selleck to generate its website. Selleck files are located in doc/
.
make html
Documentation will be built to build_docs/
.
Yeti uses YUIDocJS to generate API documentation from inline JSDoc comment blocks.
make html-api
Documentation will be built to build_docs/api/everything/
.
See CONTRIBUTING.md.
Open a ticket on YUILibrary.com's Yeti Issue Tracker to report bugs or feature requests.
Yeti is free to use under YUI's BSD license. See the LICENSE file or the YUI license page for license text and copyright information.