Finding the performance problem
Background: I was pairing with another dev to do Watin tests for a page, we have code similar to this:
[Test]
public void CanAddNoteToPolicy()
{
ie.Link("AddNote").Click();
ie.TextField("NoteText").TypeText("This is the comment text");
ie.Button("SaveNewNote").Click();
Thread.Sleep(500);//wait for the ajax call to return
AssertPageContainsText("This is the comment text");
}
The test failed, and when we investigated why, we found that it was because the call took ~1.3 seconds to complete. This is on a local machine, so I was naturally worried about it. I increased the time and the test passed, but it was still worrying that it took so long. It basically was filling an entity from the user input and saving it to the database.
I went over the code, and couldn't really find something wrong with it, but it consistently took way more time than it should. I was about to open up a profiler when I remembered...
/// <summary>
/// Notify the web service that we created a new note
/// If completes, the web service accepted the new note and will
/// process it in due time.
/// If failed, the web service didn't accept the note, probably need
/// to fail the transaction in this case.
/// </summary>
public virtual void BeginNoteInsertion(Note note)
{
//Oren: we don't have the web service yet, so we just simulate the
//duration of avg call at the moment
Thread.Sleep(1000);
}
That was some mighty head smack, I tell you that.
Comments
You just lost my respect for writing that Thread.Sleep(1000) then regained it when I remembered you'd just admitted it to the world.
The reasoning behind it is sound (I think), I want to simulate a real latency for now, since I would be replacing this with a web service call, and I don't want all the tests that makes assumptions about the length of the call to break when I do.
I should also mention that it is in a class called: FakeInserter
Sounds like a great use case for that expiring hack exception you wrote about!