Re: Reserve_Capacity for Unbounded_String?

["Randy Brukardt" <randy@rrsoftware.com>]

"Maciej Sobczak" <see.my.homepage@gmail.com> wrote in message
news:1185395844.104043.194340@o61g2000hsh.googlegroups.com...
...
> Building a string by appending small chunks to the end seems to be a
> common practice. Optimizing the library for this case is a wise
> implementation strategy.

Yes, but not if performance matters. If performance matters, doing a lot of
small operations is the worst possible way to do it, irrespective of the
implementation of those small operations.

We chose

> > I still don't think it is worth it, but obviously a customer example
could
> > change my mind.
>
> What about a customer who is choosing the competing product? Doesn't
> sound convincing enough?

No, because customers don't chose products because of the performance of one
rarely used library call. They chose products because of price, features,
ease of use, reputation of vendor, and quality of support. And it is also
the case that they rarely tell the truth about why they chose something
(partially because they often don't really have a clear understanding of
that either). I'd rather spend our limited efforts on these factors
(especially ease of use, where I don't think Janus/Ada really measures up
very well) than chasing single customers with single changes, especially
ones that will negatively impact other usage patterns.

After all, my experience with chasing customers has uniformly been bad; they
always have a new reason for not buying once you eliminate the last one.

                                               Randy.