I've been trying 'em both out and here are my initial thoughts...
WatiN
The Good
- Fast execution.
- Script creation tools are independent projects; there are 2 that I know of: Wax (Excel based, hosted on CodePlex) and WatiN Test Record (hosted on SourceForge). Neither is as robust as Selenium IDE.
- Very good IE support. Can attach and detach to/from running instances. Can access native window handles etc. (See script example below).
- NuGet packaged, easy to get running in .NET, Visual Studio style environments and keep updated.
The Bad
- Googling WatiN (watin xyz) often causes Google to recommend "watir xyz" instead. Not that much documentation out there.
- What little there is (documentation), it is confusing; for example: at first blush it would appear that there is no native support for CSS selectors. Especially since there are extensions libraries like 'WatiNCssSelectorExtensions' and many blog articles about alternative techniques (such as injecting jQuery/sizzle into the page). On Stack Overflow, I found a comment by Jeroen van Menen which suggests that there is native support. At least the lead-developer spends time on Stack Overflow :)
- No native XPath support.
- No out-of-the-box remote execution/grid based execution.
Script Example (C#). You can't do this with Selenium (not that I know off, at least):
class IEManager
{
IE _ie = null;
object _lock = new object();
IE GetInstance(string UrlFragment)
{
lock (_lock)
{
if (_ie == null)
{
var instances = new IECollection(true); //Find all existing IE instances
var match = instances.FirstOrDefault(ie=>ie.Url.Contains(UrlFragment));
_ie = match ?? new IE();
if (match==null) //we created a new instance, so we should clean it up when done!
_ie.AutoClose = true;
}
}
return _ie;
}
}
Selenium
- Slower than WatiN (especially since a new process has to be created).
- Built-in CSS selectors/XPath support.
- Selenium IDE is good (can't say great, but it’s the best in class!).
- Feels more Java-ish than .NET-ish...but really, it's programming language agnostic; all commands are sent to an out-of-process 'Driver'. The driver is really a 'host' process for the browser instance. All communication must be serialised in/out across process boundaries, which might explain the speed issues relative to WatiN.
- Decoupled processes - "Driver" and "Control" mean more robustness, more complexity, etc., but also easier to create grids/distributed test environments. Would have really liked it if the "distribution" mechanism (i.e. the communication between Driver & Control) were across WebSphere or other existing, robust, message queue manager.
- Support chrome and other browsers out of the box.
Despite everything, I went with WatiN in the end; I mainly intend to write small screen-scraping applications and want to use LINQPad for development. Attaching to a remote IE instance (one that I did not spawn myself) is a big plus. I can fiddle around in an existing instance...then run a bit of script...then fiddle again etc. This is harder to do with Selenium, though I suppose "pauses" could be embedded in the script during which time I could fiddle directly with the browser.