🚀 go-pugleaf

Jeff,

> Great discussions, thanks all for the input! 

Thanks for the summary!

> It seems that the decision points to using a WideString (or SAXString...)
> instead of some type of buffer within the characters function:
> 
> For Dynarray/Buffer
> ==============
> Jeff Rafter
> Binh Ly
> 
> For WideString/SAXString
> ==================
> Robert Loof
> Mark Edington
> Vincent Parrett
> Keith Wood
> R�mon Sinnema

Hmm, I actually prefer characters(ch: PWideChar; length: Integer), with
characters(WideString) as second choice. Both would be even better,
using overloading and a property (as discussed earlier). I am
experimenting with a parser at the moment, and will let you know as soon
as I have some results (like everybody else, I am also trying to make a
living <g>). I plan to release a high performance and fully compliant
parser in the near future, and that will support whatever the group
decides (in addition to its native API).

> Ease of use (based on EVERYONE'S statements) leaves us the clear winner of
> WideString/SAXString.  Speed/Memory will depend on implementation (though if
> parsers are maintaining back end buffers, as I am, for reuse and not
> reallocating for each read, then the buffer model wins out hands down).
> When someone steps up with a Gig of data we'll see what happens-- but I
> think improving entry level for new XML'ers wins out in this case which
> ultimately points to WideString/SAXString.

You are probably right here, except that WideStrings have their own
problems if you primarily use 8-bit encodings. What is your opinion on
having two sets of interfaces, widestring and string based? It is up to
component vendors to support them, so in the end the market will decide
on the viability of this approach.

> Does everyone agree with this roundup?  Does anyone want to argue the
> PWideChar/PSAXChar model (Would this provide a better way to typecast my
> buffer as a WideString based on my internal start and length without
> re-allocating)?  Does anyone want to argue for the mixed function types with
> a property?

See above.

All the best,

Robert