[phobos] std.process and unbuffered I/O (again)
Andrei Alexandrescu
andrei at erdani.com
Fri Jan 7 08:18:42 PST 2011
I know how to implement the necessary primitives, I just couldn't find
the time to do it. It's not difficult - I looked up
google.com/codesearch for the implementation of I/O routines such as
getc(), ungetc(), fread() and saw how the FILE* maintains its buffer.
We need them anyway. Right now, iterating a stream by character is very
slow because each popFront is a fgetc followed by ungetc. That is
necessary because in readf you need to be able to look at a character
without removing it off the stream.
Andrei
On 1/7/11 9:18 AM, SHOO wrote:
> I am one waiting expectantly for your contribution.
>
> Perhaps I think that UnbufferedTransports meets that purpose.
> I think that it is necessary to define specialized Unformatter(or
> ByLine?) to treat strings.
>
> BTW, In unbuffered stream, AFAIK, the handling of the character string
> is very difficult. For example, you must perform reading by 1 byte for 1
> line input. This is very inefficient reading.
> What do you think about this?
>
> --
> SHOO
>
> (2011/01/04 7:14), Lars Tandle Kyllingstad wrote:
>> Ok, now that bug 3979 is out of the way, there is, as far as we know,
>> only one thing standing in the way of finalising the design of
>> std.process, and that is the lack of unbuffered I/O in Phobos.
>>
>> std.process needs a type akin to File, but without the buffering, for
>> handling pipes. So I wrote a new one, which I called UnbufferedFile.
>> There was a debate about this on this list a while ago, in which Andrei
>> seemed to oppose the idea of having a separate type for that purpose,
>> and instead argued that File should be changed to allow unbuffered I/O.
>> I didn't really get a clear impression of others' opinions. And now the
>> recent proposal for a streaming API has been thrown into the mix.
>> (Should pipe ends be UnbufferedTransports, perhaps? I don't even know
>> how std.stdio.File will relate to the streaming module.)
>>
>> What do you think, folks?
>>
>>
>> -Lars
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos
More information about the phobos
mailing list