I don't know if "feature requests" is the right place for this, but could you make plots calculate only a finite number of points? Meaning, if there are several graphs, each one deriving from previous ones and containing some exotic formulae, the graphs would take quite some time to calculate each instance plot. Resizing/etc will do the same.
So, is it possible to calculate the plot for a predefined number of points only (100, for example)? That should be valid for any kind of zoom/resize/etc. Also, while zooming/resizing/etc make it so that it doesn't calculate, just do an estimate resize of it and display the already calculated points, not the continuously calculated ones; it's faster. If the user wants more points, so be it, let the user define them with a range(). And, while we're at it, an auto-fit would be nice, too: you have the range ( => x for the graph) and the max/min values (the formula needs to be calculated) => the graph is auto-adjusting at +/- 5% or so for y.

As a side note, I noticed that formulae that span from others require more and more time, as if it were progressive. I may be wrong, but I have the impression that if the user has, let's say:
f(x):=x^2
g(x):=f(x)^2
the program calculates g(x) by substituting actual f(x), not the values already calculated up to that point. So g(x) isn't calculated with ^2, but instead with ^2. So, if for f(x) it takes 1 sec to calculate, then, if the user hits after g(x), since it's the same function as f(x) it should take 1 sec, too, but it takes more. If I have h(x)=g(x)^2, then it would take even more.
So, theoretically, wouldn't it be easier if after the user hits after f(x), the values are calculated, then stored in memory (maybe a small program-swapfile? for great numbers/results to avoid memory overloading)? Of course, for a predefined range like in the case above, with the graphs or, with a custom range() defined by the user, if he/she/it so wants. Then, when reaching g(x), the values are already there, ready to be used, not re-calculated. Maybe this could be applied for the graphs, too.
I know there is eval(), but it is limited and it cannot be applied everywhere. Anyway, it's just an idea.


Regards,
Vlad.