Header files is that thing for me. I hate that I need to create and maintain two files for every non-trivial class.
Printable View
Header files is that thing for me. I hate that I need to create and maintain two files for every non-trivial class.
Implicit conversions. They are so evil that they're nearly useless when you want to overload operators such as operator T or operator bool.
The fact that you can't make them explicit annoys me.
This isn't going to be another one of those poll threads like in GD, is it?
QuantumPete
You can cast explicitly, but there is no other use for the explicit keyword apart from how it is applied to constructors.Quote:
Originally Posted by Elysia
There isn't, and that's a shame.Quote:
Originally Posted by citizen
But I thought you like adding conversion operators to your classes? Did you change your mind?Quote:
Originally Posted by Elysia
As for me, I'm not too fond of the whole C side of C++ and the inconsistencies.
It would also be nice if all types had a standard size instead of leaving so much of the language as implementation defined.
Yes, I do. But I've had to stop because they just keep causing too many problems.Quote:
Originally Posted by cpjust
They never work as expected, so this is a major failing in the language.
Really? Wow. Color me lucky...Quote:
They never work as expected, so this is a major failing in the language.
Ditto.Quote:
As for me, I'm not too fond of the whole C side of C++ and the inconsistencies.
Soma
OK, bad wording.Quote:
Originally Posted by phantomotap
They work in a lot of more unexpected ways than you think. They can easily introduce bugs or cause ambiguity, where this is undesired.
If there was a way to control them better - such as explicit constructors, I'd be much happier.
I dont like Herbie's teachings -- but I never read his books anyway.
You do mean Schildt I hope?! O_oQuote:
I dont like Herbie's teachings -- but I never read his books anyway.
Soma
swgh must be referring to Herb Sutter :)Quote:
You do mean Schildt I hope?! O_o
requiring goto to break/continue in nested loops
No default types for template arguments, mandatory specialisation of all arguments following a one which has been specialised. This latter point implies a profound reflection about the order of the template arguments, even before coding!, and which makes code much less maintainable....
Eeek. Blasphomy! :pQuote:
Originally Posted by laserlight
Ah, but you can have default types for template arguments (I assume you mean the template types inside the template<> part?). For classes it works, but not for functions (that's a shame).Quote:
Originally Posted by MarkZWEERS
What? O_oQuote:
mandatory specialisation of all arguments following a one which has been specialised
Example?
Soma
Yeah, that's exactly what I mean....Quote:
but not for functions
Josuttis an Vandevoorde mention in their book "C++ Templates" that very, very probably the next standard will allow default types for template arguments, as well as a 'double' as non-type template argument. They say that there are no technical difficulties in doing so. Can't wait...
I agree. C++ is missing that one :)
pseudo-codeQuote:
Example?
What if I've designed my template class, later I decide to add a function 'function' with only its first argument specialised? As far as I know, after a partial specialisation everything has to be specialised.Code:template<typename T1, typename T2>
class TestClass {
....
__ function();
}
template<typename T2> // partial specialisation for first template argument, #error!
__ class<char, T2> :: function()
Henceforth, this will work:
EDIT: This is especially horrible with traits... along the road you discover over and over again the "correct" order of the argumentsCode:template<typename T1> // partial specialisation for second template argument, OK
__ class<T1, char> :: function()