Thanks for this PR! As I frequently use linear region items I've been suspicious of some performance issue when using a significant number of them...

I'll try and review/test this in the next day or two.

This looks great! I can't think of a reason why the caching should cause any problems here.

However: why not just implement this on GraphicsItem instead? Then potentially many different items will benefit form it.

@campagnola I looked at the code for some other items briefly, and they seemed to have their own caching, albeit implemented in some slightly different way for each. (They also didn't seem to have any performance issues.) Figured I might break something if I try to harmonize that...

@jjuod I'm good w/ implementing it at GraphicsItem, the other objects that inherit GraphicsItem can re-implement those methods...but that may be outside the scope of the PR so if you're reluctant to make that implementation, I'll take this PR as it is.

Ok, moved the caching to GraphicsItem. Works fine with instances of LinearRegionItems. Haven't tested the other items, but it doesn't seem like they could be broken by this in any way.

Merged, thanks for the PR @jjuod !

Turns out we did break something!

this is a screenshot of the basic example app; when you zoom out, notice how the infinite lines do not scale all the way up...

I'm thinking the solution here is not to undo this PR, but to add some more logic to InfiniteLine

Managed to fix this by adding the call to GraphicsItem for viewTransformChanged() for the InfiniteLine class.

    def viewTransformChanged(self):
        """
        Called whenever the transformation matrix of the view has changed.
        (eg, the view range has changed or the view was resized)
        """
        GraphicsItem.viewTransformChanged(self)
        self._invalidateCache()