NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c-part2/13409

Internet Message Format | 1996-08-05 | 3.9 KB

Path: mail2news.demon.co.uk!genesis.demon.co.uk
From: Lawrence Kirby <[e-mail]>
Newsgroups: comp.lang.c
Subject: Re: help with strcmp
Date: Sat, 06 Apr 96 19:32:07 GMT
Organization: none
Message-ID: <[e-mail]>
References: <4jpiek$[e-mail]> <[e-mail]> <4jup7d$[e-mail]> <[e-mail]>
Reply-To: [e-mail]
X-NNTP-Posting-Host: genesis.demon.co.uk
X-Newsreader: Demon Internet Simple News v1.27
X-Mail2News-Path: genesis.demon.co.uk

In article <[e-mail]>
           [e-mail] "Kazimir Kylheku" writes:

>In article <4jup7d$[e-mail]>,
>Peter Seebach <[e-mail]> wrote:
>>Once again: feof() informs you only that the last read failed; it doesn't
>>belong before a read.
>
>Whoa, are you implying that the feof() status is _undefined_ before you do the
>first read or merely that it's an indication of bad style to have the check
>before a read?

No, feof() is well defined before the first read. However the point is that
feof() should be tested (if at all) after the corresponding operation, not
before. That can be no operation if, for example, you don't know whether
there has been any previous I/O performed on the stream.

>I recently posted some code which does check ferror() and feof()
>before doing a read, although not in the same sense that Pascal newbies make
>the mistake. It's the code in which lines from two files are read
>concurrently (in the lame sense of the word, of course!), and as long as one of
>them has no ferror() or feof() condition, the loop has to continue.  The outer
>loop does a bottom test, but inside the loop, the entry into two scanning
>inner-loops is guarded by checking ferror() and feof(), which the first time
>around happens before anything is read.

The point is that here your feof() and ferror() tests do relate to prior
reads so even in your example they come after the corresponding read as far
as the order of execution is concerned. You also perform a test after
each read in the loop i.e. before the data is 'processed'. However now you
bring it up I'll comment on the code!

>#include <stdio.h>
>#include <stdlib.h>
>#include <errno.h>

I don't see any need for stdlib.h or errno.h.

>void concatprint(FILE *f1, FILE *f2)
>
>/*
> * read lines from f1 and f2 simultaneously, and print them side by side
> * f1 and f2 must be valid streams open for reading.
> */
>
>{
>    int c;
>
>    do {
>        if (!feof(f1) && !ferror(f1))
>            while ((c = getc(f1)) != '\n' && c != EOF)
>                putchar(c);
>        if (!feof(f2) && !ferror(f2))
>            while ((c = getc(f2)) != '\n' && c != EOF)
>                putchar(c);
>        putchar('\n');
>    } while ((!feof(f1) && !ferror(f1)) || (!feof(f2) && !ferror(f2)));
>}

If you hit an error on one of the reads it is questionable as to whether
the operation as a whole should continue.

The feof() and ferror() tests in the loop check whether the I/O on a file
has been completed but allows further reading from the other file. There
are other ways to achieve the same effect e.g. you exit the main loop once
either of the files has hit end-of-file and then copy the rest of the other
file as a separate operation. You could also hold the completion indication
separately e.g.

...

{
    int c1 = 0, c2 = 0;

    do {
        if (c1 != EOF) {
            while ((c1 = getc(f1)) != '\n' && c1 != EOF)
                putchar(c1);
        }

        if (c2 != EOF) {
            while ((c2 = getc(f1)) != '\n' && c2 != EOF)
                putchar(c2);
        }

        putchar('\n');
    } while (c1 != EOF || c2 != EOF);
}

avoids the whole issue of feof() by simply not using it!

There is still the issue of whether to continue going on an error. Both of
these functions should test putchar() for failure.

--
-----------------------------------------
Lawrence Kirby | [e-mail]
Wilts, England | [e-mail]
-----------------------------------------