www.digitalmars.com         C & C++   DMDScript  

D - Operators shl_r, shr_r, ushr_r

reply "Walter" <walter digitalmars.com> writes:
These are there so you can overload:

    5 << a

but really, should this be allowable? The right operand of a shift is always
an integral value. Yes, I know that disallowing this will preclude an
iostream-style use of << and >>, but I'll go out on a limb and say that <<
and >> should not be used for that.
Aug 21 2002
next sibling parent reply Pavel Minayev <evilone omen.ru> writes:
On Wed, 21 Aug 2002 13:34:30 -0700 "Walter" <walter digitalmars.com> wrote:


 These are there so you can overload:
 
     5 << a
 
 but really, should this be allowable? The right operand of a shift is always



Hm, why not? It might turn out to have some better uses than C++-like 
streaming...
Aug 21 2002
parent "Walter" <walter digitalmars.com> writes:
"Pavel Minayev" <evilone omen.ru> wrote in message
news:CFN374900584580903 news.digitalmars.com...

 On Wed, 21 Aug 2002 13:34:30 -0700 "Walter" <walter digitalmars.com>




 These are there so you can overload:
     5 << a
 but really, should this be allowable? The right operand of a shift is







 Hm, why not? It might turn out to have some better uses than C++-like
 streaming...



The C++ iostreams is, to me, a good illustration of why overloading of
arithmetic operators should only be used to implement user-defined
arithmetic types. In the spirit of add() being commutative and cmp() being
usable for <, <=, > and >=, it makes sense to disallow overloading of the
right operand of a shift.

BTW, it does make some sort of sense to overload ~ for stream operations,
after all, one could think of it as concatenating a series of strings into a
stream.
Aug 21 2002
prev sibling parent "Martin M. Pedersen" <mmp www.moeller-pedersen.dk> writes:
Hi,

"Walter" <walter digitalmars.com> wrote in message
news:ak0tnt$ntf$1 digitaldaemon.com...

 These are there so you can overload:
     5 << a
 but really, should this be allowable? The right operand of a shift is




 an integral value.



I think it should. It would allow a BigNum class to implement the same
sematics as for the primitive integrals.

Regards,
Martin M. Pedersen
Aug 21 2002