...one of the most highly
regarded and expertly designed C++ library projects in the
world.
— Herb Sutter and Andrei
Alexandrescu, C++
Coding Standards
The C++ Standard library lacks an important feature, namely the support for any formatted output of containers. Sure, it's fairly easy to write a custom routine to output a specific container, but doing so over and over again is tedious at best. In this section we will demonstrate some more of the capabilities of Spirit.Karma for generating output from arbitrary STL containers. We will build on the example presented in an earlier section (see Warming Up).
The full source code of the example shown in this section can be found here: num_list2.cpp.
This time we take advantage of Karma's List
(%
) operator. The semantics
of the list operator are fully equivalent to the semantics of the sequence
we used before. The generator expression
double_ << *(',' << double_)
is semantically equivalent to the generator expression
double_ % ','
simplifying the overall code. The list operator's attribute is compatible
with any STL container as well. For a change we use a std::vector<double>
instead of the std::list<double>
we used before. Additionally, the routine
generate_numbers
takes
the container as a template parameter, so it will now work with any STL
container holding double
numbers.
template <typename OutputIterator, typename Container> bool generate_numbers(OutputIterator& sink, Container const& v) { using boost::spirit::karma::double_; using boost::spirit::karma::generate_delimited; using boost::spirit::ascii::space; bool r = generate_delimited( sink, // destination: output iterator double_ % ',', // the generator space, // the delimiter-generator v // the data to output ); return r; }
Note | |
---|---|
Despite the container being a template parameter, the Spirit.Karma
formatting expression ( |
The output routine developed above is still not generically usable for
all types of STL containers and for arbitrary elements stored in them.
In order to be usable the items stored in the container still need to be
convertible to a double
. Fortunately
Spirit.Karma is capable to output arbitrary data types
while using the same format description expression. It implements the
stream
generators which are able to consume any attribute type as long as a matching
standard streaming operator is defined. I.e. for any attribute type Attrib
a function:
std::ostream& operator<< (std::ostream&, Attrib const&);
needs to be available. The stream
generator will use the
standard streaming operator to generate the output.
The following example modifies the code shown above to utilize the stream
operator, which makes it compatible with almost any data type. We implement
a custom data type complex
to demonstrate this. The example shows how it is possible to integrate
this (or any other) custom data type into the Spirit.Karma
generator framework.
This is the custom data structure together with the required standard streaming operator:
// a simple complex number representation z = a + bi struct complex { complex (double a, double b = 0.0) : a(a), b(b) {} double a; double b; }; // the streaming operator for the type complex std::ostream& operator<< (std::ostream& os, complex const& z) { os << "{" << z.a << "," << z.b << "}"; return os; }
And this is the actual call to generate the output from a vector of those.
This time we interleave the generated output with newline breaks (see
eol
),
putting each complex number onto a separate line:
template <typename OutputIterator, typename Container> bool generate_numbers(OutputIterator& sink, Container const& v) { using boost::spirit::karma::stream; using boost::spirit::karma::generate; using boost::spirit::karma::eol; bool r = generate( sink, // destination: output iterator stream % eol, // the generator v // the data to output ); return r; }
The code shown is fully generic and can be used with any STL container as long as the data items stored in that container implement the standard streaming operator.
The full source code of the example presented in this section can be found here: num_list3.cpp.