Front page | perl.perl5.porters |
Postings from April 2000
Ilya Zakharevich wrote:
> John Holdsworth writes:
> > Included is a bug fix to the decode of unpack 'N/Z*' which
> > was not adding a byte to the encoded length as pack does.
>
Hello IIlya, thanks for your comments.
>
> Not clear. unpack 'N/Z*' with N unpacking to 14 should be the same as
> Z14. Is it?
>
It seems not. There is some rather strange code in pp_pack()
case 'A':
case 'Z':
case 'a':
fromstr = NEXTFROM;
aptr = SvPV(fromstr, fromlen);
if (pat[-1] == '*') {
len = fromlen;
if (datumtype == 'Z')
++len;
}
which bumps the length. I copied this into unpack so unpack(pack("N/Z*")) worked.
>
> > There is also option to use 'N/a*4' to round the length of encoded
> > strings to four bytes as required by SUN XDR RFC 1832.
>
> I think a more general command is needed. Say, 'x!8' which would add
> 0-bytes until the next position which is multiple of 8. Then you
> would write 'N/a*x!4'.
I would agree with this. I was trying to make the absolute
minum change possible to what is very much core code.
> > More contentiously perhaps the code also supports a new length
> > character '?' which steals the length from the next argument.
> > This allows multiple arrays to be encoded and decoded viz.:
> >
> > my @arr = (1,2,3,4);
> > my $packed = pack "N/N?N/N?", scalar @arr, @arr, scalar @arr, @arr;
>
> Again, something more general is needed. Something like
>
> my $packed = pack "(N//N)(N//N)", scalar @arr, @arr, scalar @arr, @arr;
I'm not so sure this would be more intuitive. '?' is in effect a new way to specify
the length of operands after '123' and '*'. It steals an argument for a length in
pack and leaves it on the stack on unpack. This is very useful for processing
SUN/RPC calls along with "@?" which pushes the current position in the string
onto the stack so you know how much of the string has been processed.
> or similar.
>
> Ilya
Cheers,
john.
Thread Previous
|
Thread Next