strcpy

[Mike Mimic]

Hi!

Is it possible to write this (and that it will work):

char *buffer = ... // some array of chars terminated by \0 (allocated

// with new and filled somewhere else)
int t = ... // some value - how much to delete at the begining

char *rest = new char[strlen(buffer) - t + 1];
strcpy(rest, buffer + t);
delete[] buffer;
buffer = rest;

in this way:

char *buffer = ... // some array of chars terminated by \0 (allocated

// with new and filled somewhere else)
int t = ... // some value - how much to delete at the begining

strcpy(buffer, buffer + t);

(This is only a code simmilar to that in my program but the idea is
here, t is always smaller than length.)


Mike

[John Harrison]

Re: strcpy


"Mike Mimic" <ppagee@yahoo.c om> wrote in message
news:c87itd$2cp $1@planja.arnes .si...[color=blue]
> Hi!
>
> Is it possible to write this (and that it will work):
>
> char *buffer = ... // some array of chars terminated by \0 (allocated
>
> // with new and filled somewhere else)
> int t = ... // some value - how much to delete at the begining
>
> char *rest = new char[strlen(buffer) - t + 1];
> strcpy(rest, buffer + t);
> delete[] buffer;
> buffer = rest;
>
> in this way:
>
> char *buffer = ... // some array of chars terminated by \0 (allocated
>
> // with new and filled somewhere else)
> int t = ... // some value - how much to delete at the begining
>
> strcpy(buffer, buffer + t);
>[/color]

No, the behaviour of strcpy is undefined if the source and destination
strings overlap.

You can use memmove instead which is guaranteed to work even for overlapping
memory blocks

memmove(buffer, buffer + t, strlen(buffer + t) + 1);

john

[Simon Saunders]

Re: strcpy

On Sun, 16 May 2004 13:22:25 +0200, Mike Mimic wrote:
[color=blue]
> Hi!
>
> Is it possible to write this (and that it will work):
>
> char *buffer = ... // some array of chars terminated by \0 (allocated
>
> // with new and filled somewhere else)
> int t = ... // some value - how much to delete at the begining
>
> char *rest = new char[strlen(buffer) - t + 1];
> strcpy(rest, buffer + t);
> delete[] buffer;
> buffer = rest;
>
> in this way:
>
> strcpy(buffer, buffer + t);
>[/color]

strcpy isn't required to work properly when the source and destination
overlap; for that case you need memmove.

[Julie]

Re: strcpy

John Harrison wrote:[color=blue]
>
> "Mike Mimic" <ppagee@yahoo.c om> wrote in message
> news:c87itd$2cp $1@planja.arnes .si...[color=green]
> > Hi!
> >
> > Is it possible to write this (and that it will work):
> >
> > char *buffer = ... // some array of chars terminated by \0 (allocated
> >
> > // with new and filled somewhere else)
> > int t = ... // some value - how much to delete at the begining
> >
> > char *rest = new char[strlen(buffer) - t + 1];
> > strcpy(rest, buffer + t);
> > delete[] buffer;
> > buffer = rest;
> >
> > in this way:
> >
> > char *buffer = ... // some array of chars terminated by \0 (allocated
> >
> > // with new and filled somewhere else)
> > int t = ... // some value - how much to delete at the begining
> >
> > strcpy(buffer, buffer + t);
> >[/color]
>
> No, the behaviour of strcpy is undefined if the source and destination
> strings overlap.
>
> You can use memmove instead which is guaranteed to work even for overlapping
> memory blocks
>
> memmove(buffer, buffer + t, strlen(buffer + t) + 1);
>
> john[/color]

Should be:

memmove(buffer, buffer + t, (strlen(buffer+ t)+1)*sizeof(*b uffer));

[John Harrison]

Re: strcpy


"Julie" <julie@nospam.c om> wrote in message
news:40A8D070.8 C571F42@nospam. com...[color=blue]
> John Harrison wrote:[color=green]
> >
> > "Mike Mimic" <ppagee@yahoo.c om> wrote in message
> > news:c87itd$2cp $1@planja.arnes .si...[color=darkred]
> > > Hi!
> > >
> > > Is it possible to write this (and that it will work):
> > >
> > > char *buffer = ... // some array of chars terminated by \0 (allocated
> > >
> > > // with new and filled somewhere else)
> > > int t = ... // some value - how much to delete at the begining
> > >
> > > char *rest = new char[strlen(buffer) - t + 1];
> > > strcpy(rest, buffer + t);
> > > delete[] buffer;
> > > buffer = rest;
> > >
> > > in this way:
> > >
> > > char *buffer = ... // some array of chars terminated by \0 (allocated
> > >
> > > // with new and filled somewhere else)
> > > int t = ... // some value - how much to delete at the begining
> > >
> > > strcpy(buffer, buffer + t);
> > >[/color]
> >
> > No, the behaviour of strcpy is undefined if the source and destination
> > strings overlap.
> >
> > You can use memmove instead which is guaranteed to work even for[/color][/color]
overlapping[color=blue][color=green]
> > memory blocks
> >
> > memmove(buffer, buffer + t, strlen(buffer + t) + 1);
> >
> > john[/color]
>
> Should be:
>
> memmove(buffer, buffer + t, (strlen(buffer+ t)+1)*sizeof(*b uffer));[/color]

Why? sizeof(*buffer) is guaranteed to be one. Or are you making a stylistic
comment?

john

[Julie]

Re: strcpy

John Harrison wrote:[color=blue]
>
> "Julie" <julie@nospam.c om> wrote in message
> news:40A8D070.8 C571F42@nospam. com...[color=green]
> > John Harrison wrote:[color=darkred]
> > >
> > > "Mike Mimic" <ppagee@yahoo.c om> wrote in message
> > > news:c87itd$2cp $1@planja.arnes .si...
> > > > Hi!
> > > >
> > > > Is it possible to write this (and that it will work):
> > > >
> > > > char *buffer = ... // some array of chars terminated by \0 (allocated
> > > >
> > > > // with new and filled somewhere else)
> > > > int t = ... // some value - how much to delete at the begining
> > > >
> > > > char *rest = new char[strlen(buffer) - t + 1];
> > > > strcpy(rest, buffer + t);
> > > > delete[] buffer;
> > > > buffer = rest;
> > > >
> > > > in this way:
> > > >
> > > > char *buffer = ... // some array of chars terminated by \0 (allocated
> > > >
> > > > // with new and filled somewhere else)
> > > > int t = ... // some value - how much to delete at the begining
> > > >
> > > > strcpy(buffer, buffer + t);
> > > >
> > >
> > > No, the behaviour of strcpy is undefined if the source and destination
> > > strings overlap.
> > >
> > > You can use memmove instead which is guaranteed to work even for[/color][/color]
> overlapping[color=green][color=darkred]
> > > memory blocks
> > >
> > > memmove(buffer, buffer + t, strlen(buffer + t) + 1);
> > >
> > > john[/color]
> >
> > Should be:
> >
> > memmove(buffer, buffer + t, (strlen(buffer+ t)+1)*sizeof(*b uffer));[/color]
>
> Why? sizeof(*buffer) is guaranteed to be one. Or are you making a stylistic
> comment?
>
> john[/color]

No, stylistic comment, but:

- strlen returns a character count, memmove takes a byte count, multiplying
simply converts from char count to bytes.

- More importantly, if the user changes from ASCII to Unicode (or other mbcs),
then their code is much more portable/convertible.

In this day and age, it is no longer safe to make assumptions that
sizeof('charact er type') == 1

[Peter Koch Larsen]

Re: strcpy


"Julie" <julie@nospam.c om> skrev i en meddelelse
news:40A8FE66.B E4A1902@nospam. com...[color=blue]
> John Harrison wrote:[color=green]
> >
> > "Julie" <julie@nospam.c om> wrote in message
> > news:40A8D070.8 C571F42@nospam. com...[color=darkred]
> > > John Harrison wrote:
> > > >
> > > > "Mike Mimic" <ppagee@yahoo.c om> wrote in message
> > > > news:c87itd$2cp $1@planja.arnes .si...
> > > > > Hi!
> > > > >
> > > > > Is it possible to write this (and that it will work):
> > > > >
> > > > > char *buffer = ... // some array of chars terminated by \0[/color][/color][/color]
(allocated[color=blue][color=green][color=darkred]
> > > > >
> > > > > // with new and filled somewhere else)
> > > > > int t = ... // some value - how much to delete at the begining
> > > > >
> > > > > char *rest = new char[strlen(buffer) - t + 1];
> > > > > strcpy(rest, buffer + t);
> > > > > delete[] buffer;
> > > > > buffer = rest;
> > > > >
> > > > > in this way:
> > > > >
> > > > > char *buffer = ... // some array of chars terminated by \0[/color][/color][/color]
(allocated[color=blue][color=green][color=darkred]
> > > > >
> > > > > // with new and filled somewhere else)
> > > > > int t = ... // some value - how much to delete at the begining
> > > > >
> > > > > strcpy(buffer, buffer + t);
> > > > >
> > > >
> > > > No, the behaviour of strcpy is undefined if the source and[/color][/color][/color]
destination[color=blue][color=green][color=darkred]
> > > > strings overlap.
> > > >
> > > > You can use memmove instead which is guaranteed to work even for[/color]
> > overlapping[color=darkred]
> > > > memory blocks
> > > >
> > > > memmove(buffer, buffer + t, strlen(buffer + t) + 1);
> > > >
> > > > john
> > >
> > > Should be:
> > >
> > > memmove(buffer, buffer + t, (strlen(buffer+ t)+1)*sizeof(*b uffer));[/color]
> >
> > Why? sizeof(*buffer) is guaranteed to be one. Or are you making a[/color][/color]
stylistic[color=blue][color=green]
> > comment?
> >
> > john[/color]
>
> No, stylistic comment, but:
>
> - strlen returns a character count, memmove takes a byte count,[/color]
multiplying[color=blue]
> simply converts from char count to bytes.[/color]

And - as John said - they are exactly the same.[color=blue]
>
> - More importantly, if the user changes from ASCII to Unicode (or other[/color]
mbcs),[color=blue]
> then their code is much more portable/convertible.[/color]

Well... that could be argued. The code probably deserves a major review if
not using plain chars.
[color=blue]
>
> In this day and age, it is no longer safe to make assumptions that
> sizeof('charact er type') == 1[/color]

/Peter

[Julie]

Re: strcpy

Peter Koch Larsen wrote:[color=blue][color=green][color=darkred]
> > > > > You can use memmove instead which is guaranteed to work even for
> > > overlapping
> > > > > memory blocks
> > > > >
> > > > > memmove(buffer, buffer + t, strlen(buffer + t) + 1);
> > > > >
> > > > > john
> > > >
> > > > Should be:
> > > >
> > > > memmove(buffer, buffer + t, (strlen(buffer+ t)+1)*sizeof(*b uffer));
> > >
> > > Why? sizeof(*buffer) is guaranteed to be one. Or are you making a[/color][/color]
> stylistic[color=green][color=darkred]
> > > comment?
> > >
> > > john[/color]
> >
> > No, stylistic comment, but:
> >
> > - strlen returns a character count, memmove takes a byte count,[/color]
> multiplying[color=green]
> > simply converts from char count to bytes.[/color]
>
> And - as John said - they are exactly the same.[color=green]
> >
> > - More importantly, if the user changes from ASCII to Unicode (or other[/color]
> mbcs),[color=green]
> > then their code is much more portable/convertible.[/color]
>
> Well... that could be argued. The code probably deserves a major review if
> not using plain chars.[/color]

No disagreement for changes to existing code -- it should be reviewed in
detail.

However, any new code should be built to be as portable to non _char_ character
types. Duty now for the future.

[color=blue][color=green]
> >
> > In this day and age, it is no longer safe to make assumptions that
> > sizeof('charact er type') == 1[/color]
>
> /Peter[/color]

[John Harrison]

Re: strcpy

>[color=blue]
> - More importantly, if the user changes from ASCII to Unicode (or other[/color]
mbcs),[color=blue]
> then their code is much more portable/convertible.
>
> In this day and age, it is no longer safe to make assumptions that
> sizeof('charact er type') == 1[/color]

If the change is made from ASCII to Unicode then the call to strlen would
also need to change to wcslen. I'd hope sizeof would be added at the same
time.

I'm not disagreeing with you, but I don't think I'd be bothered with sizeof,
if I wrote the code.

john

[Peter Koch Larsen]

Re: strcpy


"Julie" <julie@nospam.c om> skrev i en meddelelse
news:40A907C4.7 4D836AD@nospam. com...[color=blue]
> Peter Koch Larsen wrote:[/color]
[snip9[color=blue][color=green][color=darkred]
> > > > > memmove(buffer, buffer + t, (strlen(buffer+ t)+1)*sizeof(*b uffer));
> > > >
> > > > Why? sizeof(*buffer) is guaranteed to be one. Or are you making a[/color]
> > stylistic[color=darkred]
> > > > comment?
> > > >
> > > > john
> > >
> > > No, stylistic comment, but:
> > >
> > > - strlen returns a character count, memmove takes a byte count,[/color]
> > multiplying[color=darkred]
> > > simply converts from char count to bytes.[/color]
> >
> > And - as John said - they are exactly the same.[color=darkred]
> > >
> > > - More importantly, if the user changes from ASCII to Unicode (or[/color][/color][/color]
other[color=blue][color=green]
> > mbcs),[color=darkred]
> > > then their code is much more portable/convertible.[/color]
> >
> > Well... that could be argued. The code probably deserves a major review[/color][/color]
if[color=blue][color=green]
> > not using plain chars.[/color]
>
> No disagreement for changes to existing code -- it should be reviewed in
> detail.
>
> However, any new code should be built to be as portable to non _char_[/color]
character[color=blue]
> types. Duty now for the future.[/color]

Of course. But the code shown was not in any way in a condition that you
might consider porting it anyway.
[color=blue]
>
>[color=green][color=darkred]
> > >
> > > In this day and age, it is no longer safe to make assumptions that
> > > sizeof('charact er type') == 1[/color]
> >
> > /Peter[/color][/color]