On Saturday, 9 February 2013 at 07:15:57 UTC, deadalnix wrote:
On Friday, 8 February 2013 at 20:54:32 UTC, Jonathan M Davis
wrote:
On Friday, February 08, 2013 23:50:13 Dmitry Olshansky wrote:
Complication - yes, slight performance cost is what I doubt
it in RA
case. Seems like a compiler/optimizer issue.
The main problem isn't RA ranges. It's quite probable that the
optimizer takes
care of any minor additional overhead that's incurred there.
The issue is
really pure forward ranges, because you're forced to count the
code units as
they're processed (or even save the range as you go along,
given how expensive
it would be to go find the index even if you have it). But we
really have no
way of knowing how bad it is without hard data. It would
certainly be trivial
to end up doing other stuff in the implementation which cost
far more.
For pure InputRanges, that is pretty bad as the decoding of UTF
chars basically have to be done 2 times, and each codepoint
popped individually, or the lexer have to embed its own
homebrew version of std.utf .
Wow that is complete bullshit xD I completely messed up what I
wanted to say.
So, std.utf.decodeFront pop or not the utf character. And in case
it does not, you ends up having to do extra hanky panky (and
duplicate logic in std.utf to know if it does pop or not, which
is likely to be very bug prone).
