It occurs to me that the std.stream module ought to be broken down into its component parts... ie into a series of modules like:
	std.stream(.stream?)
	std.stream.file
	std.stream.memory
	std.stream.slice

Thoughts?

C. Sauls
Invironz

Agreed, and maybe add an IStream interface.

There are other problems with the stream module, like using exceptions to flag EOF (every file has an end, it isn't an exceptional circumstance).



"Chris Sauls" <ibisbasenji@yahoo.com> wrote in message news:bqns2v$13i1$1@digitaldaemon.com...
> It occurs to me that the std.stream module ought to be broken down into
> its component parts... ie into a series of modules like:
> std.stream(.stream?)
> std.stream.file
> std.stream.memory
> std.stream.slice
>
> Thoughts?
>
> C. Sauls
> Invironz
>

Ouch!  where ( what file ) is this ?

C


"Julio César Carrascal Urquijo" <adnoctum@phreaker.net> wrote in message news:bqtqnr$1471$1@digitaldaemon.com...
> Agreed, and maybe add an IStream interface.
>
> There are other problems with the stream module, like using exceptions to flag EOF (every file has an end, it isn't an exceptional circumstance).
>
>
>
> "Chris Sauls" <ibisbasenji@yahoo.com> wrote in message news:bqns2v$13i1$1@digitaldaemon.com...
> > It occurs to me that the std.stream module ought to be broken down into
> > its component parts... ie into a series of modules like:
> > std.stream(.stream?)
> > std.stream.file
> > std.stream.memory
> > std.stream.slice
> >
> > Thoughts?
> >
> > C. Sauls
> > Invironz
> >
>
>

"Charles Sanders" <sanders-consulting@comcast.net> wrote in message news:br07nl$1s1r$1@digitaldaemon.com...
> Ouch!  where ( what file ) is this ?
>
> C

In the std.stream module there are lots of sections that look like this:

        try
        {

--cut--cut--

        }
        catch (ReadError e)
        {
            // either this is end of stream, which is okay,
            // or something bad occured while reading
            if (!eof())
                throw e;
        }


And that eof() call it's really expensive (Executes at least four seeks on
the file).

The following is from the d2html.d sample in the DMD distribution.

    // the main part is wrapped into try..catch block because
    // when end of file is reached, an exception is raised;
    // so we can omit any checks for EOF inside this block...
    try
    {

--cut--cut--

    }
    // if end of file is reached and we try to read something
    // with typed read(), a ReadError is thrown; in our case,
    // this means that job is successfully done
    catch (ReadError e)
    {
        // write HTML footer
        dst.writeLine("</code></pre></body></html>");
    }


The tendency is to catch ReadError and discard any real error as end of file.