[m-dev.] a chunk type?

[Paul Bone]

On Thu, Nov 01, 2007 at 05:07:01PM +1100, Peter Wang wrote:
> On 2007-11-01, Paul Bone <pbone at csse.unimelb.edu.au> wrote:
> > 
> >  On 01/11/2007, at 4:17 PM, Peter Wang wrote:
> > >
> > > Actually I was thinking of
> > >
> > >     :- type chunk ---> chunk(string).
> > >
> > 
> >  Although the system libraries (fread/fwrite) will do buffering I can see 
> >  this being useful when somebody wants to use records of particular lenghts, 
> >  or possibly for some other use cases:
> >  	+ Reading delimetered text (CSV),
> >  	+ Reading variable width records.  say, read a header of a known size,
> >  	  interogate it for the size of the body, then read that many bytes into
> >  	  a chunk.
> 
> Well, those imply that the underlying stream to respect the chunk size.
> My intention was that the underlying stream is free NOT to do that.
> e.g. if you are reading from a network socket, it would be best to
> return data whatever it's got handy, rather than blocking indefinitely.
> 
> The problem I'm trying to solve is to efficiently read data out of a
> network socket to feed an XML parser.  If the parser says there's not
> enough data, then we just go back and get more.
> 
> Other uses might include efficient message digest or encryption
> routines, etc.  Or just copying one stream into another.
> 

Ah.  Those seem like good uses.  The same type could be used for any of
these uses.

I see no reason why not to introduce this type.



--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at csse.unimelb.edu.au
Administrative Queries: owner-mercury-developers at csse.unimelb.edu.au
Subscriptions:          mercury-developers-request at csse.unimelb.edu.au
--------------------------------------------------------------------------