Re: How to delete charcters in a string?

["David Thompson" <david.thompson1@worldnet.att.net>]

Pascal Obry <p.obry@wanadoo.fr> wrote :
>
> Stephen Leake <stephen.a.leake.1@gsfc.nasa.gov> writes:
>
> > tmoran@acm.org writes:
...
> > >   How would you delete a character from a C string?  Seems to me that's
> > > probably harder that with an Ada string (lack of slice assignment),
> >
> > char *S1 = "ABCDEF";
> >
> > strncpy (S1 + 3, S1 + 4, 3);
> >
> > result in *S1 is "ABDEF".
> >
No it's not.  All C arrays, including strings, are zero-origin,
so _if_ this worked it would produce ABCEF.  But:

1) you can't legally modify (store to) a string literal value.
_Some_ implementations put them in readonly storage
and this will fault.  _Some_ implementations share storage
for identical(ly-suffixed) values, so modifying one changes
the other(s) -- like the classic FORTRAN (II? IV?) bug of
changing say the constant 1 to be 2.  In C++ string literal
values are at least qualified const, which couldn't be done
in Standard C for compatibility with the pre-existing code
base, so _some_ (simple) violations of this rule are caught
by the compiler, but even in C++ there is a legal though
nonpreferred conversion to non-const char pointer.

Note char foo[] = "ABCDEF" is different.  That allocates
space and _initializes_ it by copying the literal value --
at least notionally, for static-duration variables the copy
is often embedded in process creation -- and it can then
be legally modified; but see next.

2) all the <string.h> routines have Undefined Behavior
(in Ada terms, are erroneous) for overlapping buffers,
with the single exception of memmove().  This means
in theory _anything_ is permissible; in practice, the
likely problem is that fetches and/or stores are reordered
and you do get a deterministic value, just the wrong one.

> > No where near as pretty or readable as Ada, but just as short and
> > easy. And I've probably got the "src" and "dest" parameters swapped.
> > Man, I hate C :).
>
Actually (dest,src,len) is the only part you got right.
I decline to guess what if anything this signifies.

> And that is not what you wanted too :)
>
>  result in *S1 is "ABDEFF".
>
That isn't right either.  See above.

--
- David.Thompson 1 now at worldnet.att.net