On 8/9/2011 3:14 PM, Graham Fawcett wrote:
Interesting. A possible use-case would be reading from a set of files,
say, a directory full of log files which need to be analyzed
line-by-line. So a reader that accepted a range of files might be
handy. (I guess that the file concatenation could be handled at the OS
level, with "cat" and a pipe, but a range-of-files would be more
convenient.)
On a tangent, the first thing that came to mind when you mentioned
this was Tim Bray's WideFinder benchmark:
http://wikis.sun.com/display/WideFinder/Wide+Finder+Home
ParallelByLine would make for a great WideFinder implementation.
Best,
Graham
I made a pull request
(https://github.com/D-Programming-Language/phobos/pull/179) for another
overload of asyncBuf in std.parallelism that will make parallel byLine
or byChunk trivial to implement (reading lines in parallel is actually
the documentation example). I'm not sure whether it's worth adding an
actual ParallelByLine to std.stdio, given how trivial the new asyncBuf
overload makes it. At any rate, I'd rather get the lower level
infrastructure finalized and into Phobos before I build the higher level
stuff on top of it.
