On Sun, 26 Nov 2000, Baruch Even wrote:
> In insetgraphics draw() method I've now added the following construct in
> order to make sure that the x value is updated upon exit.
>
> The construct:
>
> | class set_var_on_exit {
> | public:
> | set_var_on_exit(float & var, float new_val)
> | : var_(var), new_val_(new_val) {};
> |
> | ~set_var_on_exit() { var_ = new_val_; };
> | private:
> | float & var_;
> | float new_val_;
> | };
This would be more useful as a template.
> and in the method I use:
>
> | int lwidth = some_width_calculation_at_the_beginning;
> | set_var_on_exit set_x(x, x+lwidth);
>
> I do this in order to guarantee that no matter what the exit point it will
> be, the x variable will be adjusted correctly, without any extra work,
> both for me now and for a future maintainer.
Any particular reason you can't just make the assignment in the first
place? If you use the value of x throughout the function but know what
it's return value will be early on then you could instead use a temporary
variable to hold x's original value and assign the new value to x.
> In optimized code (-O2) it amounts to nothing more than the calculation at
> the start and assignment on exit. And for maintenance, this is most
> probably easier and less bug prone.
I agree it's a simple and neat trick but it's use is also very limited.
Allan. (ARRae)
