At 09:19 PM 7/2/2001 +0100, Graham Barr wrote:
>On Mon, Jul 02, 2001 at 04:12:31PM -0400, Dan Sugalski wrote:
> > At 09:07 PM 7/2/2001 +0100, Graham Barr wrote:
> > >On Mon, Jul 02, 2001 at 03:52:34PM -0400, Dan Sugalski wrote:
> > > > At 08:36 PM 7/2/2001 +0100, Simon Cozens wrote:
> > > > >On Mon, Jul 02, 2001 at 03:00:54PM -0400, Dan Sugalski wrote:
> > > > > > >what about starting offset? that is used now to shorten a
> string from
> > > > > > >the left side.
> > > > > >
> > > > > > D'oh! In. Out goes the unused.
> > > > >
> > > > >Whoa there. Do we still actually want to do this? It's unclear whether
> > > > >or not it's actually a net win.
> > > >
> > > > A valid point. It doesn't seem to be used at all often, but it's always
> > > > tough to equate frequency of appearance to frequency of usage.
> > >
> > >I for one use s/^...// quite often in the knowledge that it is
> optimized to
> > >just move a pointer and not cause a copy of the string.
> >
> > We'll still be doing that.
>
>Then I misread the mails, ignore me.
It wasn't in the original mails--it hadn't occurred to me originally that
the fact we'd have a copying garbage collector would let us drop bytes on
the floor with impunity.
> > (The leftover memory will get cleaned up when we
> > GC) The bigger issue is when we need to move the pointer back because
> we've
> > stuck something back on the front of the string.
>
>Surly the buffer holding the string would just be a malloc'd buffer
>and not GC'd (I was assuming GC would only be for the PMC) In which
>case you need to remember the start of the buffer for free.
Nope. All the string buffers will be GC'd, as will all the variable-length
data buffers. GC's a bigger win in those cases than when managing
fixed-length things. (Though there's a reasonable win there, too)
Dan
--------------------------------------"it's like this"-------------------
Dan Sugalski even samurai
[EMAIL PROTECTED] have teddy bears and even
teddy bears get drunk
