On Fri, 18 May 2012 13:27:22 -0400, kenji hara <k.hara...@gmail.com> wrote:
2012/5/19 Steven Schveighoffer <schvei...@yahoo.com>:
On Fri, 18 May 2012 10:39:55 -0400, kenji hara <k.hara...@gmail.com>
wrote:
I'm designing experimental IO primitives:
https://github.com/9rnsr/dio
I'm having trouble following the code, is there a place with the
generated
docs? I'm looking for an overview to understand where to look.
I have created gh-pages:
http://9rnsr.github.com/dio/d/io_core.html
OK, *now* I understand what you mean by non-blocking. There are some I/O
packages that use asynchronous i/o which return even before any data is
given to the buffer. I thought this is what you were talking about.
I'm fully on board with synchronous but non-blocking. That's what I
assumed we would be doing, and it's well supported by low-level OS
routines on all OSes.
In my implementation for a buffer, I have two calls:
read(buf[]) -> read until buf.length bytes are read or EOF
readPartial(buf[]) -> read from 1 to buf.length bytes, but performs at
most 1 low-level read. Returns 0 bytes on EOF.
readPartial will block if no data is yet available, but obviously can be
made to not block if the underlying OS handle is marked as non-blocking (I
need to add some extra structure to account for this).
Typically, this is the normal mechanism that I use for reading data that
is not always available. First, I select on a socket until data is
available, then use synchronous read to get whatever data exists.
continuing reading...
-Steve
