Home | Libraries | People | FAQ | More |
Copyright © 2000, 2005 Steve Cleary and John Maddock
Distributed under the Boost Software License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt )
Table of Contents
This manual is also available in printer friendly PDF format.
The header <boost/static_assert.hpp>
supplies two macros:
BOOST_STATIC_ASSERT(x) BOOST_STATIC_ASSERT_MSG(x, msg)
Both generate a compile time error message if the integral-constant-expression
x
is not true. In other words,
they are the compile time equivalent of the assert macro; this is sometimes
known as a "compile-time-assertion", but will be called a "static
assertion" throughout these docs. Note that if the condition is true
, then the macros will generate neither
code nor data - and the macros can also be used at either namespace, class
or function scope. When used in a template, the static assertion will be evaluated
at the time the template is instantiated; this is particularly useful for validating
template parameters.
If the C++0x static_assert
feature
is available, both macros will use it. For BOOST_STATIC_ASSERT(x)
,
the error message will be a stringized version of x
.
For BOOST_STATIC_ASSERT_MSG(x,
msg)
,
the error message will be the msg
string.
If the C++0x static_assert
feature
is not available, BOOST_STATIC_ASSERT_MSG(x,
msg)
will be treated as BOOST_STATIC_ASSERT(x)
.
The material that follows assumes the C++0x static_assert
feature is not available.
One of the aims of BOOST_STATIC_ASSERT
is to generate readable error messages. These immediately tell the user that
a library is being used in a manner that is not supported. While error messages
obviously differ from compiler to compiler, but you should see something like:
Illegal use of STATIC_ASSERTION_FAILURE<false>
Which is intended to at least catch the eye!
You can use BOOST_STATIC_ASSERT
at any place where you can place a declaration, that is at class, function
or namespace scope, this is illustrated by the following examples:
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_MSG(std::numeric_limits<UnsignedInt>::is_specialized, "myclass can only be specialized for types with numeric_limits support."); BOOST_STATIC_ASSERT_MSG(std::numeric_limits<UnsignedInt>::digits >= 16, "Template argument UnsignedInt must have at least 16 bits precision.") BOOST_STATIC_ASSERT_MSG(std::numeric_limits<UnsignedInt>::is_integer, "Template argument UnsignedInt must be an integer."); BOOST_STATIC_ASSERT_MSG(!std::numeric_limits<UnsignedInt>::is_signed, "Template argument UnsignedInt must not be 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); };