I was told that the only event in which it's really a good idea to define functions in header files is if the function is marked inline or is a function template. Apparently,or so I was told, one of the key reasons for this is the one definition rule- a rule allowing only only one function definition per translation unit.I was told that you would get one definition of that function for every cpp file that includes that header. I am however having issues understanding how you would get one definition of that function for every cpp file in C++. I'm having a bit of trouble visualizing why that might be the case. So, would someone mind offering some insight into why that's the case? Cheers!
A bit confused in-regards to pitfalls that come along with defining functions in header files in C++
Asked
Active
Viewed 57 times
1
-
Related: [What exactly is One Definition Rule in C++?](http://stackoverflow.com/questions/4192170/what-exactly-is-one-definition-rule-in-c) – R Sahu Mar 18 '16 at 04:13
-
Do you know how `#include` works? – user253751 Mar 18 '16 at 04:14
2 Answers
3
When you #include a file, the compiler acts as if you'd just copy-pasted the file into the current file.
That is, if you have three files:
// header.h
void foo() {/* do stuff */}
// a.c
#include "header.h"
void a_func() {/* do stuff */}
// b.c
#include "header.h"
void b_func() {/* do stuff */}
then the compiler treats it as if you'd copied the contents of header.h into both a.c and b.c - that is, it behaves exactly as if you had:
// a.c
void foo() {/* do stuff */}
void a_func() {/* do stuff */}
// b.c
void foo() {/* do stuff */}
void b_func() {/* do stuff */}
Clearly, there are multiple definitions of foo here - as there's one in each file.
user253751
- 57,427
- 7
- 48
- 90
-
The ODR is definitely more advanced, but is outside the scope of the OP's question. It is closely related and is good information for suggesting further learning, but I think this answer is fine. – Jed Schaaf Mar 18 '16 at 04:35
-
@immibis Right, I am familiar with include guards. However, I'm still not really clear on how implementing functions in the header file will create a conflict with the one definition rule, but implementing them in a separate cpp file doesn't. What are the subtleties that make that the case? Thanks for taking the time to help me out! – Mrcitrusboots Mar 18 '16 at 05:20
-
@Mrcitrusboots There's another part of the one definition rule that says a non-inline function may only be defined once in the entire program. If you put the function implementation in a header file, and include that file in several translation units, then the function is defined in several translation units. – user253751 Mar 18 '16 at 09:38
-
1
For simple function such as sum, definition would be:
double sum(double a, double b) { return a + b; }
and declaration:
double sum(double a, double b);
You place declaration in header, e.g. sum.h, so later you can use sum's definition, e.g. placed in sum.cpp. Later, linker finds the definition of sum in sum.o object file.
Since #include pragma is replaced with headers' contents, if you place your function's definition in header, you'll have multiple identical functions and also linker error.
