Assign the correct XHR method to NewXHR rather than re-testing each time it's invoked

[barneycarroll]

NewXHR's initial assignment is now to a self executing function which invokes the same tests, but assigns the result to a variable instead of immediately returning it. Instead it, returns a function which calls that variable. This way the test runs immediately and doesn't need to be executed multiple times.

[ProgerXP]

This won't work: NewXHR not just determines available method but creates a new XHR object. Your version, however, creates a singleton XHR object and returns it on each NewXHR call. In other words the instance is created once and then shared among the calls. This will break things in more ways than one.

Moreover, I don't see a point in optimizing this function in any way because in modern browsers (actually, starting with IE 6, i.e. for the last decade) its execution is already as quick as possible running a single instruction: return new XMLHttpRequest;

I'm closing this pull unless you have more ideas.

[barneycarroll]

Woops. Yeah, talk about premature optimisation ;)

I fixed the fallacious constructor invocation (with a commit --amend — @9d2ad3bed583cde1a0f9acd65ea6725a25c3a196 — don't understand why that hasn't pulled through to this pull request) … still saves iteration in old old old IE but yeah, you're right: kind of the point was to avoid repeatedly invoking a try/catch (I thought there was a notable performance hit there; turns out I was wrong).

Still, new version avoids repeated query… Good practice but not that significant an improvement I suppose.

[ProgerXP]

Still, new version avoids repeated query… Good practice but not that significant an improvement I suppose.

You're right. It makes the code more complex and in this particular case it's optimizing for IE 6 - so I will leave current version.