xref: /third_party/boost/libs/spirit/doc/support/multi_pass.qbk

[/==============================================================================
    Copyright (C) 2001-2011 Hartmut Kaiser
    Copyright (C) 2001-2011 Joel de Guzman
    Copyright (C) 2001-2002 Daniel C. Nuffer

    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)
===============================================================================/]

[section:multi_pass The multi pass iterator]

Backtracking in __qi__ requires the use of the following types of iterator:
forward, bidirectional, or random access. Because of backtracking, input
iterators cannot be used. Therefore, the standard library classes
`std::istreambuf_iterator` and `std::istream_iterator`, that fall under the
category of input iterators, cannot be used. Another input iterator that is of
interest is one that wraps a lexer, such as LEX.

[note In general, __qi__ generates recursive descent parser which require
      backtracking parsers by design. For this reason we need to provide at
      least forward iterators to any of __qi__'s API functions. This is not an
      absolute requirement though. In the future, we shall see more
      deterministic parsers that require no more than 1 character (token) of
      lookahead. Such parsers allow us to use input iterators such as the
      `std::istream_iterator` as is. ]

Backtracking can be implemented only if we are allowed to save an iterator
position, i.e. making a copy of the current iterator. Unfortunately, with an
input iterator, there is no way to do so, and thus input iterators will not
work with backtracking in __qi__. One solution to this problem is to simply
load all the data to be parsed into a container, such as a vector or deque,
and then pass the begin and end of the container to __qi__. This method can be
too memory intensive for certain applications, which is why the `multi_pass`
iterator was created.

[heading Using the multi_pass]

The `multi_pass` iterator will convert any input iterator into a forward
iterator suitable for use with __qi__. `multi_pass` will buffer data when
needed and will discard the buffer when its contents is not needed anymore.
This happens either if only one copy of the iterator exists or if no
backtracking can occur.

A grammar must be designed with care if the `multi_pass` iterator is used.
Any rule that may need to backtrack, such as one that contains an alternative,
will cause data to be buffered. The rules that are optimal to use are
repetition constructs (as kleene and plus).

Sequences of the form `a >> b` will buffer data as well. This is different from
the behavior of __classic__ but for a good reason. Sequences need to reset the
current iterator to its initial state if one of the components of a sequence
fails to match. To compensate for this behavior we added functionality to
the `expect` parsers (i.e. constructs like `a > b`). Expectation points introduce
deterministic points into the grammar ensuring no backtracking can occur if
they match. For this reason we clear the buffers of any multi_pass iterator
on each expectation point, ensuring minimal buffer content even for large
grammars.

[important  If you use an error handler in conjunction with the `expect` parser
            while utilizing a `multi_pass` iterator and you intend to use
            the error handler to force a `retry` or a `fail` (see the
            description of error handlers - __fixme__: insert link), then
            you need to instantiate the error handler using `retry` or `fail`,
            for instance:
``
        rule r<iterator_type> r;
        on_error<retry>(r, std::cout << phoenix::val("Error!"));
``
            If you fail to do so the resulting code will trigger an assert
            statement at runtime.]

Any rule that repeats, such as kleene_star (`*a`) or positive such as (`+a`),
will only buffer the data for the current repetition.

In typical grammars, ambiguity and therefore lookahead is often localized. In
fact, many well designed languages are fully deterministic and require no
lookahead at all. Peeking at the first character from the input will
immediately determine the alternative branch to take. Yet, even with highly
ambiguous grammars, alternatives are often of the form `*(a | b | c | d)`.
The input iterator moves on and is never stuck at the beginning. Let's look at
a Pascal snippet for example:

    program =
            programHeading >> block >> '.'
        ;

    block =
           *(   labelDeclarationPart
            |   constantDefinitionPart
            |   typeDefinitionPart
            |   variableDeclarationPart
            |   procedureAndFunctionDeclarationPart
            )
        >>  statementPart
        ;

Notice the alternatives inside the Kleene star in the rule block . The rule
gobbles the input in a linear manner and throws away the past history with each
iteration. As this is fully deterministic LL(1) grammar, each failed
alternative only has to peek 1 character (token). The alternative that consumes
more than 1 character (token) is definitely a winner. After which, the Kleene
star moves on to the next.

Now, after the lecture on the features to be careful with when using
`multi_pass`, you may think that `multi_pass` is way too restrictive to use.
That's not the case. If your grammar is deterministic, you can make use of
the `flush_multi_pass` pseudo parser in your grammar to ensure that data is not
buffered when unnecessary (`flush_multi_pass` is available from the __qi__
parser __repo__).

Here we present a minimal example showing a minimal use case. The `multi_pass`
iterator is highly configurable, but the default policies have been chosen so
that its easily usable with input iterators such as `std::istreambuf_iterator`.
For the complete source code of this example please refer to
[@../../example/support/multi_pass.cpp multi_pass.cpp].

[import ../../example/support/multi_pass.cpp]
[tutorial_multi_pass]

[heading Using the flush_multi_pass parser]

The __spirit__ __repo__ contains the `flush_multi_pass` parser component.
This is usable in conjunction with the `multi_pass` iterator to minimize the
buffering. It allows to insert explicit synchronization points into your
grammar where it is safe to clear any stored input as it is ensured that no
backtracking can occur at this point anymore.

When the `flush_multi_pass` parser is used with `multi_pass`, it will call
`multi_pass::clear_queue()`. This will cause any buffered data to be erased.
This also will invalidate all other copies of multi_pass and they should not
be used. If they are, an `boost::illegal_backtracking` exception will be
thrown.

[heading The multi_pass Policies]

The `multi_pass` iterator is a templated class configurable using policies.
The description of `multi_pass` above is how it was originally implemented
(before it used policies), and is the default configuration now. But,
`multi_pass` is capable of much more. Because of the open-ended nature of
policies, you can write your own policy to make `multi_pass` behave in a way
that we never before imagined.

The multi_pass class has two template parameters:

[variablelist The multi_pass template parameters
    [[Input]    [The type multi_pass uses to acquire it's input. This is
                 typically an input iterator, or functor.]]
    [[Policies] [The combined policies to use to create an instance of a
                 multi_pass iterator. This combined policy type is described
                 below]]
]

It is possible to implement all of the required functionality of the combined
policy in a single class. But it has shown to be more convenient to split this
into four different groups of functions, i.e. four separate, but well
coordinated policies. For this reason the `multi_pass` library
implements a template `iterator_policies::default_policy` allowing to combine
several different policies, each implementing one of the functionality groups:

[table Policies needed for default_policy template
    [[Template Parameter]   [Description]]
    [[`OwnershipPolicy`]    [This policy determines how `multi_pass` deals with
                             it's shared components.]]
    [[`CheckingPolicy`]     [This policy determines how checking for invalid
                             iterators is done.]]
    [[`InputPolicy`]        [A class that defines how `multi_pass` acquires its
                             input. The `InputPolicy` is parameterized by the
                             `Input` template parameter to the `multi_pass`.]]
    [[`StoragePolicy`]      [The buffering scheme used by `multi_pass` is
                             determined and managed by the StoragePolicy.]]
]

The `multi_pass` library contains several predefined policy implementations
for each of the policy types as described above. First we will describe those
predefined types. Afterwards we will give some guidelines how you can write
your own policy implementations.

[heading Predefined policies]

All predefined `multi_pass` policies are defined in the namespace
`boost::spirit::iterator_policies`.

[table Predefined policy classes
    [[Class name]         [Description]]
    [[*InputPolicy* classes]]
    [[`input_iterator`]   [This policy directs `multi_pass` to read from an
                           input iterator of type `Input`.]]
    [[`buffering_input_iterator`]   [This policy directs `multi_pass` to read from an
                           input iterator of type `Input`. Additionally it buffers
                           the last character received from the underlying iterator.
                           This allows to wrap iterators not buffering the last
                           character on their own (as `std::istreambuf_iterator`).]]
    [[`istream`]          [This policy directs `multi_pass` to read from an
                           input stream of type `Input` (usually a
                           `std::basic_istream`).]]
    [[`lex_input`]        [This policy obtains it's input by calling yylex(),
                           which would typically be provided by a scanner
                           generated by __flex__. If you use this policy your code
                           must link against a __flex__ generated scanner.]]
    [[`functor_input`]    [This input policy obtains it's data by calling a
                           functor of type `Input`. The functor must meet
                           certain requirements. It must have a typedef called
                           `result_type` which should be the type returned
                           from `operator()`. Also, since an input policy needs
                           a way to determine when the end of input has been
                           reached, the functor must contain a static variable
                           named `eof` which is comparable to a variable of
                           `result_type`.]]
    [[`split_functor_input`][This is essentially the same as the `functor_input`
                           policy except that the (user supplied) function
                           object exposes separate `unique` and `shared` sub
                           classes, allowing to integrate the functors /unique/
                           data members with the `multi_pass` data items held
                           by each instance and its /shared/ data members will
                           be integrated with the `multi_pass` members shared
                           by all copies.]]

    [[*OwnershipPolicy* classes]]
    [[`ref_counted`]      [This class uses a reference counting scheme.
                           The `multi_pass` will delete it's shared components
                           when the count reaches zero.]]
    [[`first_owner`]      [When this policy is used, the first `multi_pass`
                           created will be the one that deletes the shared data.
                           Each copy will not take ownership of the shared data.
                           This works well for __spirit__, since no dynamic
                           allocation of iterators is done. All copies are made
                           on the stack, so the original iterator has the
                           longest lifespan.]]

    [[*CheckingPolicy* classes]]
    [[`no_check`]         [This policy does no checking at all.]]
    [[`buf_id_check`]     [This policy keeps around a buffer id, or a buffer
                           age. Every time `clear_queue()` is called on a
                           `multi_pass` iterator, it is possible that all other
                           iterators become invalid. When `clear_queue()` is
                           called, `buf_id_check` increments the buffer id.
                           When an iterator is dereferenced, this policy checks
                           that the buffer id of the iterator matches the shared
                           buffer id. This policy is most effective when used
                           together with the `split_std_deque` StoragePolicy.
                           It should not be used with the `fixed_size_queue`
                           StoragePolicy, because it will not detect iterator
                           dereferences that are out of range.]]
    [[full_check]         [This policy has not been implemented yet. When it
                           is, it will keep track of all iterators and make
                           sure that they are all valid. This will be mostly
                           useful for debugging purposes as it will incur
                           significant overhead.]]

    [[*StoragePolicy* classes]]
    [[`split_std_deque`]  [Despite its name this policy keeps all buffered data
                           in a `std::vector`. All data is stored as long as
                           there is more than one iterator. Once the iterator
                           count goes down to one, and the queue is no longer
                           needed, it is cleared, freeing up memory. The queue
                           can also be forcibly cleared by calling
                           `multi_pass::clear_queue()`.]]
    [[`fixed_size_queue<N>`][This policy keeps a circular buffer that is size
                            `N+1` and stores `N` elements. `fixed_size_queue`
                            is a template with a `std::size_t` parameter that
                            specified the queue size. It is your responsibility
                            to ensure that `N` is big enough for your parser.
                            Whenever the foremost iterator is incremented, the
                            last character of the buffer is automatically
                            erased. Currently there is no way to tell if an
                            iterator is trailing too far behind and has become
                            invalid. No dynamic allocation is done by this
                            policy during normal iterator operation, only on
                            initial construction. The memory usage of this
                            `StoragePolicy` is set at `N+1` bytes, unlike
                            `split_std_deque`, which is unbounded.]]
]

[heading Combinations: How to specify your own custom multi_pass]

The beauty of policy based designs is that you can mix and match policies to
create your own custom iterator by selecting the policies you want. Here's an
example of how to specify a custom `multi_pass` that wraps an
`std::istream_iterator<char>`, and is slightly more efficient than the default
`multi_pass` (as generated by the `make_default_multi_pass()` API function)
because it uses the `iterator_policies::first_owner` OwnershipPolicy and the
`iterator_policies::no_check` CheckingPolicy:

    typedef multi_pass<
        std::istream_iterator<char>
      , iterator_policies::default_policy<
            iterator_policies::first_owner
          , iterator_policies::no_check
          , iterator_policies::buffering_input_iterator
          , iterator_policies::split_std_deque
        >
    > first_owner_multi_pass_type;

The default template parameters for `iterator_policies::default_policy` are:

* `iterator_policies::ref_counted` OwnershipPolicy
* `iterator_policies::no_check` CheckingPolicy, if `BOOST_SPIRIT_DEBUG` is
  defined: `iterator_policies::buf_id_check` CheckingPolicy
* `iterator_policies::buffering_input_iterator` InputPolicy, and
* `iterator_policies::split_std_deque` StoragePolicy.

So if you use `multi_pass<std::istream_iterator<char> >` you will get those
pre-defined behaviors while wrapping an `std::istream_iterator<char>`.

[heading Dealing with constant look ahead]

There is one other pre-defined class called `look_ahead`. The class
`look_ahead` is another predefine `multi_pass` iterator type. It has two
template parameters: `Input`, the type of the input iterator to wrap, and a
`std::size_t N`, which specifies the size of the buffer to the
`fixed_size_queue` policy. While the default multi_pass configuration is
designed for safety, `look_ahead` is designed for speed. `look_ahead` is derived
from a multi_pass with the following policies: `input_iterator` InputPolicy,
`first_owner` OwnershipPolicy, `no_check` CheckingPolicy, and
`fixed_size_queue<N>` StoragePolicy.

This iterator is defined by including the files:

    // forwards to <boost/spirit/home/support/look_ahead.hpp>
    #include <boost/spirit/include/support_look_ahead.hpp>

Also, see __include_structure__.

[heading Reading from standard input streams]

Yet another predefined iterator for wrapping standard input streams (usually a
`std::basic_istream<>`) is called `basic_istream_iterator<Char, Traits>`. This
class is usable as a drop in replacement for `std::istream_iterator<Char, Traits>`.
Its only difference is that it is a forward iterator (instead of the
`std::istream_iterator`, which is an input iterator). `basic_istream_iterator`
is derived from a multi_pass with the following policies: `istream` InputPolicy,
`ref_counted` OwnershipPolicy, `no_check` CheckingPolicy, and
`split_std_deque` StoragePolicy.

There exists an additional predefined typedef:

    typedef basic_istream_iterator<char, std::char_traits<char> > istream_iterator;

This iterator is defined by including the files:

    // forwards to <boost/spirit/home/support/istream_iterator.hpp>
    #include <boost/spirit/include/support_istream_iterator.hpp>

Also, see __include_structure__.

[heading How to write a functor for use with the `functor_input` InputPolicy]

If you want to use the `functor_input` InputPolicy, you can write your own
function object that will supply the input to `multi_pass`. The function object
must satisfy several requirements. It must have a typedef `result_type` which
specifies the return type of its `operator()`. This is standard practice in the
STL. Also, it must supply a static variable called eof which is compared against
to know whether the input has reached the end. Last but not least the function
object must be default constructible. Here is an example:

    #include <iostream>
    #include <boost/spirit/home/qi.hpp>
    #include <boost/spirit/home/support.hpp>
    #include <boost/spirit/home/support/multi_pass.hpp>
    #include <boost/spirit/home/support/iterators/detail/functor_input_policy.hpp>

    // define the function object
    class iterate_a2m
    {
    public:
        typedef char result_type;

        iterate_a2m() : c_('A') {}
        iterate_a2m(char c) : c_(c) {}

        result_type operator()()
        {
            if (c_ == 'M')
                return eof;
            return c_++;
        }

        static result_type eof;

    private:
        char c_;
    };

    iterate_a2m::result_type iterate_a2m::eof = iterate_a2m::result_type('M');

    using namespace boost::spirit;

    // create two iterators using the define function object, one of which is
    // an end iterator
    typedef multi_pass<iterate_a2m
      , iterator_policies::first_owner
      , iterator_policies::no_check
      , iterator_policies::functor_input
      , iterator_policies::split_std_deque>
    functor_multi_pass_type;

    int main()
    {
        functor_multi_pass_type first = functor_multi_pass_type(iterate_a2m());
        functor_multi_pass_type last;

        // use the iterators: this will print "ABCDEFGHIJKL"
        while (first != last) {
            std::cout << *first;
            ++first;
        }
        std::cout << std::endl;
        return 0;
    }

[heading How to write policies for use with multi_pass]

All policies to be used with the `default_policy` template need to have two
embedded classes: `unique` and `shared`. The `unique` class needs to implement
all required functions for a particular policy type. In addition it may hold
all member data items being /unique/ for a particular instance of a `multi_pass`
(hence the name). The `shared` class does not expose any member functions
(except sometimes a constructor), but it may hold all member data items to be
/shared/ between all copies of a particular `multi_pass`.

[heading InputPolicy]

An `InputPolicy` must have the following interface:

    struct input_policy
    {
        // Input is the same type used as the first template parameter
        // while instantiating the multi_pass
        template <typename Input>
        struct unique
        {
            // these typedef's will be exposed as the multi_pass iterator
            // properties
            typedef __unspecified_type__ value_type;
            typedef __unspecified_type__ difference_type;
            typedef __unspecified_type__ distance_type;
            typedef __unspecified_type__ pointer;
            typedef __unspecified_type__ reference;

            unique() {}
            explicit unique(Input) {}

            // destroy is called whenever the last copy of a multi_pass is
            // destructed (ownership_policy::release() returned true)
            //
            //   mp:    is a reference to the whole multi_pass instance
            template <typename MultiPass>
            static void destroy(MultiPass& mp);

            // swap is called by multi_pass::swap()
            void swap(unique&);

            // get_input is called whenever the next input character/token
            // should be fetched.
            //
            //   mp:    is a reference to the whole multi_pass instance
            //
            // This method is expected to return a reference to the next
            // character/token
            template <typename MultiPass>
            static typename MultiPass::reference get_input(MultiPass& mp);

            // advance_input is called whenever the underlying input stream
            // should be advanced so that the next call to get_input will be
            // able to return the next input character/token
            //
            //   mp:    is a reference to the whole multi_pass instance
            template <typename MultiPass>
            static void advance_input(MultiPass& mp);

            // input_at_eof is called to test whether this instance is a
            // end of input iterator.
            //
            //   mp:    is a reference to the whole multi_pass instance
            //
            // This method is expected to return true if the end of input is
            // reached. It is often used in the implementation of the function
            // storage_policy::is_eof.
            template <typename MultiPass>
            static bool input_at_eof(MultiPass const& mp);

            // input_is_valid is called to verify if the parameter t represents
            // a valid input character/token
            //
            //   mp:    is a reference to the whole multi_pass instance
            //   t:     is the character/token to test for validity
            //
            // This method is expected to return true if the parameter t
            // represents a valid character/token.
            template <typename MultiPass>
            static bool input_is_valid(MultiPass const& mp, value_type const& t);
        };

        // Input is the same type used as the first template parameter passed
        // while instantiating the multi_pass
        template <typename Input>
        struct shared
        {
            explicit shared(Input) {}
        };
    };

It is possible to derive the struct `unique` from the type
`boost::spirit::detail::default_input_policy`. This type implements a minimal
sufficient interface for some of the required functions, simplifying the task
of writing a new input policy.

This class may implement a function `destroy()` being called during destruction
of the last copy of a `multi_pass`. This function should be used to free any of
the shared data items the policy might have allocated during construction of
its `shared` part. Because of the way `multi_pass` is implemented any allocated
data members in `shared` should _not_ be deep copied in a copy constructor of
`shared`.

[heading OwnershipPolicy]

The `OwnershipPolicy` must have the following interface:

    struct ownership_policy
    {
        struct unique
        {
            // destroy is called whenever the last copy of a multi_pass is
            // destructed (ownership_policy::release() returned true)
            //
            //   mp:    is a reference to the whole multi_pass instance
            template <typename MultiPass>
            static void destroy(MultiPass& mp);

            // swap is called by multi_pass::swap()
            void swap(unique&);

            // clone is called whenever a multi_pass is copied
            //
            //   mp:    is a reference to the whole multi_pass instance
            template <typename MultiPass>
            static void clone(MultiPass& mp);

            // release is called whenever a multi_pass is destroyed
            //
            //   mp:    is a reference to the whole multi_pass instance
            //
            // The method is expected to return true if the destructed
            // instance is the last copy of a particular multi_pass.
            template <typename MultiPass>
            static bool release(MultiPass& mp);

            // is_unique is called to test whether this instance is the only
            // existing copy of a particular multi_pass
            //
            //   mp:    is a reference to the whole multi_pass instance
            //
            // The method is expected to return true if this instance is unique
            // (no other copies of this multi_pass exist).
            template <typename MultiPass>
            static bool is_unique(MultiPass const& mp);
        };

        struct shared {};
    };

It is possible to derive the struct `unique` from the type
`boost::spirit::detail::default_ownership_policy`. This type implements a
minimal sufficient interface for some of the required functions, simplifying
the task of writing a new ownership policy.

This class may implement a function `destroy()` being called during destruction
of the last copy of a `multi_pass`. This function should be used to free any of
the shared data items the policy might have allocated during construction of
its `shared` part. Because of the way `multi_pass` is implemented any allocated
data members in `shared` should _not_ be deep copied in a copy constructor of
`shared`.

[heading CheckingPolicy]

The `CheckingPolicy` must have the following interface:

    struct checking_policy
    {
        struct unique
        {
            // swap is called by multi_pass::swap()
            void swap(unique&);

            // destroy is called whenever the last copy of a multi_pass is
            // destructed (ownership_policy::release() returned true)
            //
            //   mp:    is a reference to the whole multi_pass instance
            template <typename MultiPass>
            static void destroy(MultiPass& mp);

            // docheck is called before the multi_pass is dereferenced or
            // incremented.
            //
            //   mp:    is a reference to the whole multi_pass instance
            //
            // This method is expected to make sure the multi_pass instance is
            // still valid. If it is invalid an exception should be thrown.
            template <typename MultiPass>
            static void docheck(MultiPass const& mp);

            // clear_queue is called whenever the function
            // multi_pass::clear_queue is called on this instance
            //
            //   mp:    is a reference to the whole multi_pass instance
            template <typename MultiPass>
            static void clear_queue(MultiPass& mp);
        };

        struct shared {};
    };

It is possible to derive the struct `unique` from the type
`boost::spirit::detail::default_checking_policy`. This type implements a
minimal sufficient interface for some of the required functions, simplifying
the task of writing a new checking policy.

This class may implement a function `destroy()` being called during destruction
of the last copy of a `multi_pass`. This function should be used to free any of
the shared data items the policy might have allocated during construction of
its `shared` part. Because of the way `multi_pass` is implemented any allocated
data members in `shared` should _not_ be deep copied in a copy constructor of
`shared`.

[heading StoragePolicy]

A `StoragePolicy` must have the following interface:

    struct storage_policy
    {
        // Value is the same type as typename MultiPass::value_type
        template <typename Value>
        struct unique
        {
            // destroy is called whenever the last copy of a multi_pass is
            // destructed (ownership_policy::release() returned true)
            //
            //   mp:    is a reference to the whole multi_pass instance
            template <typename MultiPass>
            static void destroy(MultiPass& mp);

            // swap is called by multi_pass::swap()
            void swap(unique&);

            // dereference is called whenever multi_pass::operator*() is invoked
            //
            //   mp:    is a reference to the whole multi_pass instance
            //
            // This function is expected to return a reference to the current
            // character/token.
            template <typename MultiPass>
            static typename MultiPass::reference dereference(MultiPass const& mp);

            // increment is called whenever multi_pass::operator++ is invoked
            //
            //   mp:    is a reference to the whole multi_pass instance
            template <typename MultiPass>
            static void increment(MultiPass& mp);

            //
            //   mp:    is a reference to the whole multi_pass instance
            template <typename MultiPass>
            static void clear_queue(MultiPass& mp);

            // is_eof is called to test whether this instance is a end of input
            // iterator.
            //
            //   mp:    is a reference to the whole multi_pass instance
            //
            // This method is expected to return true if the end of input is
            // reached.
            template <typename MultiPass>
            static bool is_eof(MultiPass const& mp);

            // less_than is called whenever multi_pass::operator==() is invoked
            //
            //   mp:    is a reference to the whole multi_pass instance
            //   rhs:   is the multi_pass reference this instance is compared
            //          to
            //
            // This function is expected to return true if the current instance
            // is equal to the right hand side multi_pass instance
            template <typename MultiPass>
            static bool equal_to(MultiPass const& mp, MultiPass const& rhs);

            // less_than is called whenever multi_pass::operator<() is invoked
            //
            //   mp:    is a reference to the whole multi_pass instance
            //   rhs:   is the multi_pass reference this instance is compared
            //          to
            //
            // This function is expected to return true if the current instance
            // is less than the right hand side multi_pass instance
            template <typename MultiPass>
            static bool less_than(MultiPass const& mp, MultiPass const& rhs);
        };

        // Value is the same type as typename MultiPass::value_type
        template <typename Value>
        struct shared {};
    };

It is possible to derive the struct `unique` from the type
`boost::spirit::detail::default_storage_policy`. This type implements a
minimal sufficient interface for some of the required functions, simplifying
the task of writing a new storage policy.

This class may implement a function `destroy()` being called during destruction
of the last copy of a `multi_pass`. This function should be used to free any of
the shared data items the policy might have allocated during construction of
its `shared` part. Because of the way `multi_pass` is implemented any allocated
data members in `shared` should _not_ be deep copied in a copy constructor of
`shared`.

Generally, a `StoragePolicy` is the trickiest policy to implement. You should
study and understand the existing `StoragePolicy` classes before you try and
write your own.


[endsect]