Kevin Kenny wrote:
>  Uhm, Donal, excuse me?  This is a simple forward iteration. Keep an
>  index into the string rep - as opposed to the 'string' internal rep.
>  (That one never shimmers.)  On each trip through the loop, decode
>  one UTF-8 character and advance the index around it.  There's no
>  real slowdown because you don't need to get into quadratic behaviour
>  counting the UTF-8 characters (the way a naively coded [string index]
>  loop at Tcl level would).
>  
>  If you want to cache single character strings (at least for
>  ISO8849 characters), go for it.  But I suspect that in most uses
>  of this, the single characters would be short-lived.

Bah, this whole discussion is getting out of hand (and I'm highly
irritable today because I'm teaching). For the business of [string
foreach], I'd be perfectly happy to see an implementation as you
describe it. Since it's iterating by single characters, the benefit of
buffer sharing would be slight. Whether or not to cache character-objs
would be a question to decide at a later date.

The general question of string rep sharing is complex, and a massively
over-heavy hammer for this particular nut. (Another objection to it I've
just thought of is that it would cause the loss of the NUL-termination
property, with consequent implications for the core of Tcl and many
extensions; quicker and simpler to just copy the string to a new buffer.
This particular point doesn't apply to byte arrays.)

Donal.

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Tcl-Core mailing list
Tcl-Core@[...].net
https://lists.sourceforge.net/lists/listinfo/tcl-core