On 2/7/11 9:40 AM, Michel Fortin wrote:
On 2011-02-07 08:24:32 -0500, spir <denis.s...@gmail.com> said:
Does this have anything to do with currently discussed buffered input
ranges? If yes, how does such a design, or any alternative, fit their
proposed interface?
You can build all of this on top of a buffered input range. The buffered
input range is not an alternative for your complex parsing algorithm,
it's just the component managing the buffer.
Having the underlying range manage the buffer (as opposed to having your
parser algorithm doing it) means that the buffer implementation can vary
depending on the type of range. For instance, if you're parsing directly
data in memory, the buffered range can use directly this data in memory
as the buffer, requiring no allocation and no copy; if you're parsing
from a file or a network stream it'll use a more standard buffer.
But how exactly the buffer is implemented does not affect your parsing
algorithm in any way. That's the great thing about it, separation of
concerns: your algorithm will work independently of the buffer
implementation used. All your parser algorithm has to do is say
"shiftFront" and "appendToFront" to control what's available in the buffer.
Very well put, thanks.
Andrei
