Boost C++ Libraries

The macro can be used at namespace scope, if there is some requirement must always be true; generally this means some platform specific requirement. Suppose we require that int be at least a 32-bit integral type, and that wchar_t be an unsigned type. We can verify this at compile time as follows:

#include <climits>
#include <cwchar>
#include <limits>
#include <boost/static_assert.hpp>

namespace my_conditions {

   BOOST_STATIC_ASSERT(std::numeric_limits<int>::digits >= 32);
   BOOST_STATIC_ASSERT(WCHAR_MIN >= 0);

} // namespace my_conditions

The use of the namespace my_conditions here requires some comment. The macro BOOST_STATIC_ASSERT works by generating an typedef declaration, and since the typedef must have a name, the macro generates one automatically by mangling a stub name with the value of __LINE__. When BOOST_STATIC_ASSERT is used at either class or function scope then each use of BOOST_STATIC_ASSERT is guaranteed to produce a name unique to that scope (provided you only use the macro once on each line). However when used in a header at namespace scope, that namespace can be continued over multiple headers, each of which may have their own static assertions, and on the "same" lines, thereby generating duplicate declarations. In theory the compiler should silently ignore duplicate typedef declarations, however many do not do so (and even if they do they are entitled to emit warnings in such cases). To avoid potential problems, if you use BOOST_STATIC_ASSERT in a header and at namespace scope, then enclose them in a namespace unique to that header.

The macro is typically used at function scope inside template functions, when the template arguments need checking. Imagine that we have an iterator-based algorithm that requires random access iterators. If the algorithm is instantiated with iterators that do not meet our requirements then an error will be generated eventually, but this may be nested deep inside several templates, making it hard for the user to determine what went wrong. One option is to add a static assertion at the top level of the template, in that case if the condition is not met, then an error will be generated in a way that makes it reasonably obvious to the user that the template is being misused.

#include <iterator>
#include <boost/static_assert.hpp>
#include <boost/type_traits.hpp>

template <class RandomAccessIterator >
RandomAccessIterator foo(RandomAccessIterator from, 
                         RandomAccessIterator to)
{
   // this template can only be used with
   // random access iterators...
   typedef typename std::iterator_traits< 
         RandomAccessIterator >::iterator_category cat;
   BOOST_STATIC_ASSERT(
      (boost::is_convertible<
         cat, 
         const std::random_access_iterator_tag&>::value));
   //
   // detail goes here...
   return from;
}

A couple of footnotes are in order here: the extra set of parenthesis around the assert, is to prevent the comma inside the is_convertible template being interpreted by the preprocessor as a macro argument separator; the target type for is_convertible is a reference type, as some compilers have problems using is_convertible when the conversion is via a user defined constructor (in any case there is no guarantee that the iterator tag classes are copy-constructible).

The macro is typically used inside classes that are templates. Suppose we have a template-class that requires an unsigned integral type with at least 16-bits of precision as a template argument, we can achieve this using something like this:

#include <limits>
#include <boost/static_assert.hpp>

template <class UnsignedInt>
class myclass
{
private:
   BOOST_STATIC_ASSERT((std::numeric_limits<UnsignedInt>::digits >= 16)
                        && std::numeric_limits<UnsignedInt>::is_specialized
                        && std::numeric_limits<UnsignedInt>::is_integer
                        && !std::numeric_limits<UnsignedInt>::is_signed);
public:
   /* details here */
};

Normally static assertions when used inside a class or function template, will not be instantiated until the template in which it is used is instantiated. However, there is one potential problem to watch out for: if the static assertion is not dependent upon one or more template parameters, then the compiler is permitted to evaluate the static assertion at the point it is first seen, irrespective of whether the template is ever instantiated, for example:

template <class T>
struct must_not_be_instantiated
{   
   BOOST_STATIC_ASSERT(false);
};

Will produce a compiler error with some compilers (for example Intel 8.1 or gcc 3.4), regardless of whether the template is ever instantiated. A workaround in cases like this is to force the assertion to be dependent upon a template parameter:

template <class T>
struct must_not_be_instantiated
{   
   // this will be triggered if this type is instantiated
   BOOST_STATIC_ASSERT(sizeof(T) == 0); 
};