• Pablo Aguilar (3/3) Jun 01 2005 ...is missing support for string_views
• Matthew (21/23) Jun 01 2005 Be my guest!

"Pablo Aguilar" <paguilarg hotmail.com> writes:

...is missing support for string_views

Mind if I take a shot at this as well?


Pablo 

"Matthew" <admin.hat stlsoft.dot.org> writes:

"Pablo Aguilar" <paguilarg hotmail.com> wrote in message
news:d7la3v$2jh2$1 digitaldaemon.com...
 ...is missing support for string_views

 Mind if I take a shot at this as well?

Be my guest!

Note: there are two bits of progress on FSC:

1. I updated the implementation after the impl described in IC++ /
http://www.cuj.com/documents/s=8188/cuj0406wilson/, 
in response to feedback from Sean Kelly, who pointed out that I assumed
continguous storage, which is not prescribed by 
the standard for std::basic_string. So the updated impl (as of 1.8.1) uses
iterators and algorithms, rather than 
pointers and memcpy(). (The old functionality is available via specification of
the 
STLSOFT_FAST_STRING_CONCATENATION_ASSUME_CONTIGUOUS_STORAGE pre-processor
symbol.)

So I'll be working on an updated impl using TMP type-detection techniques so as
to work with a wide variety of 
techniques without using the pre-processor, which'll be featuring in XSTL and a
follow-up article in CUJ.

I plan to do this work next week, and so will keep you posted on what's
happening so we don't screw each other up.

2. Craig Henderson and I are looking at combining our techniques on this. Craig
was published in this June's CUJ (mine 
was in June 2004), and uses compile-time selection of functionality for string
classes, c-style strings and characters, 
whereas mine relies on the (nearly ubiquitous) optimisation of the runtime
techniques. Craig's already looked into a 
comparison, and formulated some ideas for combining the techniques, and I'll be
doing the same next week. Again, it 
might behove you to wait a little while, or at least keep in touch, so we all
keep synchronised.


I think overall we're likely to come up with a combination of Craig and my
techniques, followed by application of the 
type-detection stuff to make them work with all kinds of string types. At this
point, we can factor in any work you 
might have done to allow string_view - not a value type, as you know - to work
as well.

Sounds like a lot of fun to be had in next week or two! :-)

Matthew