Problem with overloading member functions
I have written a class like this:
Recently, I changed the data type of one member variable from
But, when I call
I tried to fix it by providing a
|
|
Recently, I changed the data type of one member variable from
int to long, and now I want the member functions to take a double or a long. I tried doing this: |
|
But, when I call
Foo::bar() with an integer (e.g. x.bar(1)), the compiler tells me that both of the overloads meet the specified type, and that it doesn't know to cast "10" into a long or double.I tried to fix it by providing a
void Foo::bar(const int&), and call Foo::bar(const long&) in it with the parameter casted to a long. It worked, but in my real code there are more than a few of functions overloaded like that. Do I have to do the same thing to every function, or is there a better alternative?
Last edited on
Without seeing the actual functions it is hard to advise as to the correct approach.
What is the difference between the two functions (the double, and the "integer" version)?
Have you considered a template class?
What is the difference between the two functions (the double, and the "integer" version)?
Have you considered a template class?
Last edited on
It's supposed to be a container class for storing numbers, floating-point and integer in one. Here's a part of it:
There is an integer and a floating-point overload because I want it to accept both of them. It should be able to convert between the two. Before changing the data types storing the values are
|
|
There is an integer and a floating-point overload because I want it to accept both of them. It should be able to convert between the two. Before changing the data types storing the values are
int and double. Now I changed it to long and double.
| Without seeing the actual functions it is hard to advise as to the correct approach. |
| Have you considered a template class? |
Well here's the .h file if you want to see it (before changing to
I've tried template classes, but finally gave up because (at least I think) they are not practical in this case, and the fact that I cannot separate the function definitions from the declarations into .h and .cpp files anymore if I use templates.
long): |
|
I've tried template classes, but finally gave up because (at least I think) they are not practical in this case, and the fact that I cannot separate the function definitions from the declarations into .h and .cpp files anymore if I use templates.
Oh and here's the .cpp file: (part 1)
|
|
part 2:
|
|
| I've tried template classes, but finally gave up because (at least I think) they are not practical in this case, |
Why do you think they are not practical in this case? At a quick glance this class looks like an excellent candidate for using templates.
| and the fact that I cannot separate the function definitions from the declarations into .h and .cpp files anymore if I use templates. |
And why do you think this is a problem?
If you were to use templates your code would probably decrease by a factor of about three, so that should make implementing your class easier in the long run.
I used templates in another way... When I tried templates, I made the arguments what data type should the number be stored in if it is an integer and what it should be if it is a floating-point. Do you mean to use templates like this:
?
I'll try that. Seems like a great idea. Thanks!
|
|
?
I'll try that. Seems like a great idea. Thanks!
Hi,
It seems you have an operator that calls a function (with the same name) which calls another function which calls the constructor. For example
I am guessing if you use templates your code will reduce by a factor of 12.
But really what you have done is to overload all the language operators in order to partially achieve something the language does anyway. What is the advantage of this
It seems you have an operator that calls a function (with the same name) which calls another function which calls the constructor. For example
operator%= calls modEquals(num), which calls this->mod(num), which calls this->mod(Number(num)) ( including the constructor call), which finally returns iVal % num.iVal;I am guessing if you use templates your code will reduce by a factor of 12.
But really what you have done is to overload all the language operators in order to partially achieve something the language does anyway. What is the advantage of this
Number class that stores either a double or an int? The language already does the right thing with num1 operator num2; using a predefined set of rules, never mind what the type or operator is.
@TheIdeasMan By creating this
Number class I hope to achieve a generic Number class that combines both integers and floating-points like those in JavaScript. This way, in some places where you are not sure if the user would enter an integer or floating-point, you can use a Number, which would completely eliminate floating-point precision errors if the user entered an integer.
| This way, in some places where you are not sure if the user would enter an integer or floating-point, ..... |
There are things called
unions which can hold one of a selected type at a time.http://en.cppreference.com/w/cpp/language/union
But as I said there is already a well defined set rules to cope with mixing of integral and floating point types, so using a union probably isn't worth it either. There is also well defined methods of dealing with FP values properly. And we have templates, which one would use for functions and classes, not for the numbers themselves.
| ..... which would completely eliminate floating-point precision errors if the user entered an integer. |
I doubt that you will achieve that with the built-in types, static-casting an int is no different than assigning an int to a double. There will always be numbers that don't convert exactly (even if say one uses a FP with 100 sf ), that is the nature of the current floating point methodology of binary fractions. There are exact decimal types, gnu has a decimal class (#include /decimal/decimal.h if you have it)
There is also boost::any http://www.boost.org/doc/libs/1_61_0/doc/html/any.html
I tried to use templates on the constructors, like this:
(The function declarations and definitions are in the same file now)
But I got two errors (actually, a whole bunch of errors, but I figured all of them were caused by the two) saying "Error C2988 unrecognizable template declaration/definition" and when I double click it it points to
What am I doing wrong here?
P.S. If you are wondering what
|
|
(The function declarations and definitions are in the same file now)
But I got two errors (actually, a whole bunch of errors, but I figured all of them were caused by the two) saying "Error C2988 unrecognizable template declaration/definition" and when I double click it it points to
Number::Number<double>(const double& num) { and Number::Number<Number>(const Number& otherNum) : minFloatingPointAmount(otherNum.minFloatingPointAmount) {, but not Number::Number<double>(const double& num, const double& minFloatAmount) : minFloatingPointAmount(minFloatAmount) {. And, at the end of the file IntelliSense tells me that "Error: PCH warning: header stop needs to be at global scope. An IntelliSense PCH file was not generated."What am I doing wrong here?
P.S. If you are wondering what
minFloatingPointAmount is, I updated the code and here's a part of the new .h file: |
|
const double minFloatingPointAmount = 1e-7;Sorry dude but that is going to render your doubles quite ineffective: they may as well be floats ! Are you aware of how many significant figures doubles have compared to float?
There is also boost::variant
http://www.boost.org/doc/libs/1_61_0/doc/html/variant.html
This is also handy as a tutorial:
http://theboostcpplibraries.com/
http://www.boost.org/doc/libs/1_61_0/doc/html/variant.html
This is also handy as a tutorial:
http://theboostcpplibraries.com/
Well, thanks. From what you've said I figured that the class I'm writing is quite useless, and now I've stopped working on it anyway. Thanks for all the advice though.
Last edited on
Topic archived. No new replies allowed.