Show Lecture.ImplicitInclusion as a slide show.

CS253 Implicit Inclusion

Dubious Code

Is this code correct? I mean, sure, it’s factually correct, but is it a correct C++ program?

#include <iostream>

using namespace std;

int main() {
    string s = "Han shot first!\n";
    cout << s;
}

Han shot first!

Bad Code

#include <iostream>

using namespace std;

int main() {
    string s = "Han shot first!\n";
    cout << s;
}

Han shot first!

The code is not correct. A std::string was defined, but there is no #include <string>. But, still, it compiled. ☺? ☹?

Not a Cause for Celebration

• You may think, “What’s the problem? So it compiles. That doesn’t bother me. Now, if it ever fails to compile, let me know.”
• Sure, it compiles, but it doesn’t have to.
• It might stop compiling come the next compiler release.
• It might not compile using another compiler (e.g., clang++ vs. g++).

operator<<

• Consider the statement cout << s;. This meaning of << isn’t built into C++ like integer addition. Instead, somebody must have written a operator<< function that takes an ostream and a std::string.
• Where is that operator declared?
• The obvious places are <iostream> or <string>.
• Wherever it is, it has to know about both classes: ostream and string.
• If it’s in <iostream>, then <iostream> has to #include <string>.
• If it’s in <string>, then <string> has to #include <iostream>.
• If it’s in some other header file, then somebody has to #include that.

Light Dawns

#include <iostream>

using namespace std;

int main() {
    string s = "Han shot first!\n";
    cout << s;
}

Han shot first!

• Now, we understand how this can compile.
• It must be that, on this particular compiler, <iostream> includes <string>.
• Can we count on that? No!
• On another computer, <string> might #include <iostream>.
• Or, <iostream> might conditionally define operator<< only if <string> has already been included, and vice versa.

Solution

#include <iostream>
#include <string>

using namespace std;

int main() {
    string s = "Han shot first!\n";
    cout << s;
}

Han shot first!

• To write a portable program:
  • #include what you need.
  • Don’t count on an accident of implementation.
• It’s a shame that the compiler doesn’t tell us that we got it wrong.
• This is why computer programmers get the big bucks.

Theoretical

• Is this just theoretical handwringing? No.
• Compiler vendors are currently quite casual about having one header file include another.
  • It is permitted by the standard.
• For nearly every assignment, somebody turns in code that compiles on their computer, but fails on CSU computers.

Example

Consider this program, compiled on my Macbook Air under macOS:

    % cat c.cc
    #include <cmath>
    int main() {
        return isinf(0.0);
    }
    % g++ c.cc
    % ./a.out
    % g++ --version
    Configured with: --prefix=/Library/Developer/CommandLineTools/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
    Apple LLVM version 8.0.0 (clang-800.0.42.1)
    Target: x86_64-apple-darwin15.6.0
    Thread model: posix
    InstalledDir: /Library/Developer/CommandLineTools/usr/bin

Example

The same program, compiled on a CSU CS Department computer:

    % cat c.cc
    #include <cmath>
    int main() {
        return isinf(0.0);
    }
    % g++ c.cc
    c.cc: In function ‘int main()’:
    c.cc:3:12: error: ‘isinf’ was not declared in this scope
         return isinf(0.0);
        	^~~~~
    c.cc:3:12: note: suggested alternative:
    In file included from c.cc:1:0:
    /usr/include/c++/7/cmath:612:5: note:   ‘std::isinf’
         isinf(_Tp __x)
         ^~~~~
    % g++ --version
    g++ (GCC) 7.3.1 20180130 (Red Hat 7.3.1-2)

Conclusion