Can I use Selenium IDE on a Drupal system?
I found http://drupal.org/project/selenium but that involves downloading Core and not using my current machine.
Does anybody know of a way to use the IDE with Drupal, or if not what do you suggest I do?
You can, in the same way that you can use selenium on just about any web page.
From what I can see (though I haven’t' used it) the selenium module seems to be more about using Drupal to store and run selenium tests, which is useful, but will not explicitly test your Drupal site. (edit: this module dosn't look like it is currently maintained)
Drupal comes with it's own testing framework which uses simpletest. This can do a lot of the testing that you will need. It can handle form submissions and check for text in a page. There are two cases that I personally have found that it is not very useful for:
Data integrity, where you would like to check that the content of a particular page is correct where you know what to expect on your system
Two is checking the site layout, where you wish to check your live site that certain elements appear as they should do.
Selenium can be used for both of these, in the simplest case you can record tests against your site and play them back to check it. You can check for an elements existence and style. You can check that elements have the text you would expect. Even a few of these can be useful as a tool for checking the configuration of your site.
It is also possible to integrate selenium with simplest so that while running your simpletests through Drupal it will run selenium tests, however this can get very complicated and confusing.
There's a nice new module that takes Selenium exports (in PHP) and converts them to Drupal SimpleTest suites. http://drupal.org/project/simpletest_selenium
This is nice because you can then do continuous integration testing by using Drush to script the running of your tests.
The question is what do you want to achieve?
testing a Drupal site?
saving / playing back configuration settings (automation through the browser)?
The Firefox & Selenium IDE (FF extension) combination works great with Drupal for functional testing.
It also works great if you want to capture some Drupal configuration settings (that you'd like to play back e.g. in another environment) but there are better ways - Patterns and Strongarm modules - to achieve that.
A couple of tips & tricks:
while capturing clicks be careful with timings (between clicks), sometimes you need to add "pause" lines to your test case,
for HTML form checkboxes Selenium often records the "click" action instead of the much safer "check" or "uncheck" action (this is important if you run your test case several times against the same page / form),
if your page reloads when you click a certain link, use "clickAndWait" instead of the simple "click" action,
use "verifyTextPresent" often, especially if you're running your tests "blindly", using Selenium RC.
I use selenium IDE all the time. It's an awesome tool for productivity and communication.
But Drupal's Simpletest framework has a built in browser for user testing. So you probably won't that module.
See:
http://www.lastcraft.com/web_tester_documentation.php
http://drupal.org/node/265762
http://www.lullabot.com/articles/drupal-module-developer-guide-simpletest
You can test Drupal 7 with SeLite, which extends Selenium IDE. It allows your tests to access (read and write to) a test DB (isolated from the DB of the tested application).
It's especially good for Drupal 7, which can have its data in (separate) SQLite DB. So if your test environment can use SQLite, your test data lifecycle will be very easy. However, you can benefit from SeLite even if your test Drupal instance uses other type of DB.
See https://code.google.com/p/selite/wiki/ProjectHome and https://code.google.com/p/selite/wiki/DrupalTutorial.
Related
I have a question regarding the version control integration in Rstudio.
What's the advantage of using it instead of an external version control manager like SourceTree?
From what I understand the main advantage is to avoid using another tool and saving the switch from a window to another.
But the problem is we often need to use other languages in a data mining projects (SQL, Python, ...) so in almost every scenario we also need a dedicated version control manager.
It's not an either/or. Think of RStudio's version control as a way to accomplish your most common source control tasks without leaving the IDE. For instance, press Ctrl + Alt + D to see a diff of the current file. How long would that take using an external tool?
Likewise, you can easily commit a change in just a few keystrokes (try Ctrl + Alt + M), and you're never more than a click away from seeing which branch you're on.
Most IDEs have some basic form of version control built-in for this reason. It isn't intended to completely replace any VCS tools you're using; it's intended to keep you from having to switch context when you're doing common tasks. You will still need to use an external VCS for more complicated operations.
Another thing to keep in mind is that RStudio can manage your whole data science project, not just its R code. You can use its VCS integration to manage and commit changes to files of any kind. We sometimes use it as a VCS interface to projects that contain no R code at all!
Some people prefer to do everything VCS-related in a single interface. That's great, and hopefully the RStudio VCS integration stays out of your way if that's your workflow. But most IDE users like being able to do day-to-day VCS operations without breaking context, and that's what the VCS integration in RStudio is for.
I'm writing tests using webdriver in Java.
Is there a way to run browsers (ff / ie) in headless mode in a windows box?
I cant use a linux box and HtmlUnitDriver is not helping as our webapp has many ajax controls.
I have started a small project to supply a driver for WebDriver using SimpleBrowser. You may want to check it out: https://github.com/Teun/SimpleBrowser.WebDriver
It's still early days though. I would appreciate help in the form of contributions and testing.
At this point there is no headless version that I am aware of the Firefox or IE. Since Firefox is open source it's possible that you could get the source code for that browser and then remove the UI portion but there are a few problems with it that I can see. First you have to build a custom app to setup your own load testing engine. Second you are locked to one web render, though I think chrome is open source as well but it doesn't play as nicely as Firefox or IE (using the system .dlls vs. a custom executable). Third without a lot of testing I am sure there will be stability issues and potentially other unforseen issues with this option. That being said there is a .net headless browser that's fairly simple and would probably be decent for load testing called simple browser. However it depends on what you are wanting to load. If you just looking for bandwidth and server load then you should be find using multiple instances of the browser. I would be curious to know if using the headless browser in this fashion if you instantiated different WebDriver if it would play nicely. i.e.
WebDriver driver1 = new SimpleBrowser();
WebDriver driver2 = new SimpleBrowser();
.
.
.
.
.
WebDriver driver[n] = new SimpleBrowser();
and then you could run [n] parallel tests. Also of note in the above code SimpleBrowser() doesn't hook directly to the WebDriver interface you would have to Extend the interface to reference that .dll
Start the test program as a Windows service and the browsers will run just fine, even on a machine with no monitors. Technically speaking, they're running in a "session" that is disconnected, but it's essentially what you mean by "headless".
HtmlUnitDriver does support ajax, and can run in both java and c# (if ported via IKVM).
There is, however, no way to run IE and FF in headless mode. I can't tell you how many times I've wished for that...
We are a MSFT shop with a far-reaching MSDN license.
After many years of doing things wrong, we finally have to start doing automated testing.
My group is the Guinea pigs at this. We need to create what was not there before. We looked at the multitude of options out there. Some people get by just fine with open-source alternatives such as CC.Net, Bamboo, MbUnit, etc. We want to give MsTest, CodedUI, Team Build a good try ... might as well because of MSDN licensing and MSFT focus.
The plus and minus of doing things the MSFT way is that MSFT makes monolithic things. You have got to install various tools that play with each other nicely, but with outsiders - not necessarily. The plus is that when things are done correctly, it should all function rather smoothly. There is the option of gated check-ins, of using TFS to store the reports, etc.
Frankly, I am confused by all of the options. Our traditional build system was hacked together with a bunch of perl, batch scripts, executables, but now the build team switched to Team Build, which ought to be cleaner, but for the most part it is just a wrapper to the same old perl crap.
I am inclined to hack things together for testing too, because I can at least see what the pieces are. So, I envision the poor man's version as:
* A dedicated fast computer to run tests
* Some script to copy build files (test code as well as product code) over to that computer.
* A batch/perl script which would run mstest.exe from command line and execute a few test batches on some by-category filter within some test dlls (the product is so huge, that we do want to organize tests by various categories).
* Some script which will invoke the latter script remotely from the build server using psexec.exe (http://technet.microsoft.com/en-us/sysinternals/bb897553), as well as grabbing the xml output from a shared drive, and then sending out an email with results to those who are interested.
This can probably work, but then I have to worry about how well error handling can work with so many potential points of failure. It would be nice to configure things the "right way", taking advantage of whatever MSFT has cooked up. I am just not sure where to turn for a good guide. Have you done something like this?
Eventually we will want to have a farm of test computers, if we are to run out of the allotted time. Something else of concern is - for coded ui tests to succeed, I think a user has to be logged in, so I am not sure if psexec will be of much help here.
Can you share your positive/negative experience, point me to a good guide perhaps? Thanks!
Here are some tips off the top of my head if you want to get started with testing using the MS tools:
If you have an MSDN subscription, install a Test Rig by installing the Test Controller on your network and the the Test Agent service on each of the machines that will be collecting diagnostic data. See the following link for reference: http://msdn.microsoft.com/en-us/library/dd293551.aspx.
Add a Test Project to your solution. See the first part of the following blog post: http://blogs.microsoft.co.il/blogs/eranruso/archive/2010/03/27/visual-studio-2010-coded-ui-test-user-guide-create-a-simple-coded-ui-test.aspx.
Automated test options can be configured through the .testsettings file(s) that are added automatically when you add a Test project (you can also manually add these files to your solution).
Install Team Foundation Server (2010 recommended) in order to take advantage automating your tests with a daily build. You will also need TFS 2010 if you want to use the VS2010 Test Manager tool to define test environments and plan manual tests (these can be fully automated with CodedUI). Customize your new automated build to setup / deploy your application after build and set the build to run tests. Deployment will likely not be necessary for unit tests, but they will be for Web Performance and CodedUI test types.
If you have VS Ultimate or Test Professional licenses, you can also go further and set up virtual test labs using "Lab Management" features.
Ok.. I have looked through this site and just can't seem to find the answer I am looking for.
We have a multi server setup for our web sites, typically (DEV/QA/PRODUCTION). Our sites are .net which require some sort of build before being deployed. We are using SVN for source control and are looking for a tool/website/something that will allow our project managers to push button deploy changes to the different environments. It seems there is no silver bullet for this, am I correct in this assumption?
I like the functionality of Springloops but can't find any information regarding its use with .net. Sparing details, it would be nice to see a diagram of sorts on the whole end to end process. What I mean by that is, if multiple different tools are the only answer (no silver bullet), then a diagram that shows where the tools sit in relation to the whole process.
Look at CruiseControl.Net. It'll let you automate your builds and if there are errors it can notify certain people and even flash a big red light in the office.
The other thing it can do is automatically deploy to a chosen region such as Dev/Test/Stage/Prod when there are no errors.
Though you may want to make Prod a manual press of a button. :)
The new on click publishing features in Visual Studio 2010 are pretty easy to set up ans use: http://vishaljoshi.blogspot.com/2009/05/web-1-click-publish-with-vs-2010.html
Supports configuring IIS and deploying SQL packages too.
There is a msdeploy.exe file included in the that you could setup to run with .bat files or some configurable script runner.
Try ClickOnce deployment. Once a update is deployed on the server the client applications are automatically get updated on every client on their next run.
I wrote an automated test using dijit robot - but in order to be able to use relative paths within our web application, I created an OSGi service for our tests and put the test code in a velocity template. When I try to run the tests, nothing happens. If I use the same script in an html file and access it directly from windows explorer (not via localhost), it works fine. I find that there are many cases that will make the dijit robot tests just not run - has anyone ran into this and found out all the little gotchas to make dijit tests run?
Check out dijit.initRobot(), that might take care of some things for you.
One thing that was screwing up a lot of my tests is described in this blog post - basically the robot was not initializing because I was obscuring a special div that the robot clicks to initialize.
However I have realized that there are still quite a few problems with the doh robot - it just seems very fragile. Often I will have a working test, then add 1 robot command and the test will break. When I remove the line and try it again...well the robot wont run even though it is the exact same code as before.
I've found the best thing to do when writing robot code is to just clear the cache every time and cross your fingers. Good luck.
Problem can be if you trying it with openjdk, run it on oracle java version