///////////////////////////////////////////////////////////////////////////// // // (C) Copyright Ion Gaztanaga 2006-2013 // // 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) // // See http://www.boost.org/libs/intrusive for documentation. // ///////////////////////////////////////////////////////////////////////////// #ifndef BOOST_INTRUSIVE_HASHTABLE_HPP #define BOOST_INTRUSIVE_HASHTABLE_HPP #include //std C++ #include //std::equal_to #include //std::pair #include //std::swap, std::lower_bound, std::upper_bound #include //std::size_t //boost #include #include #include #include //General intrusive utilities #include #include #include #include #include #include #include //Implementation utilities #include #include #include #include #include namespace boost { namespace intrusive { /// @cond struct hash_bool_flags { static const std::size_t unique_keys_pos = 1u; static const std::size_t constant_time_size_pos = 2u; static const std::size_t power_2_buckets_pos = 4u; static const std::size_t cache_begin_pos = 8u; static const std::size_t compare_hash_pos = 16u; static const std::size_t incremental_pos = 32u; }; namespace detail { template struct get_slist_impl_from_supposed_value_traits { typedef typename detail::get_real_value_traits ::type real_value_traits; typedef typename detail::get_node_traits ::type node_traits; typedef typename get_slist_impl ::type >::type type; }; template struct unordered_bucket_impl { typedef typename get_slist_impl_from_supposed_value_traits ::type slist_impl; typedef detail::bucket_impl implementation_defined; typedef implementation_defined type; }; template struct unordered_bucket_ptr_impl { typedef typename detail::get_node_traits ::type::node_ptr node_ptr; typedef typename unordered_bucket_impl ::type bucket_type; typedef typename pointer_traits ::template rebind_pointer < bucket_type >::type implementation_defined; typedef implementation_defined type; }; template struct store_hash_bool { template struct two_or_three {one _[2 + Add];}; template static one test(...); template static two_or_three test (int); static const std::size_t value = sizeof(test(0)); }; template struct store_hash_is_true { static const bool value = store_hash_bool::value > sizeof(one)*2; }; template struct optimize_multikey_bool { template struct two_or_three {one _[2 + Add];}; template static one test(...); template static two_or_three test (int); static const std::size_t value = sizeof(test(0)); }; template struct optimize_multikey_is_true { static const bool value = optimize_multikey_bool::value > sizeof(one)*2; }; struct insert_commit_data_impl { std::size_t hash; }; template inline typename pointer_traits::template rebind_pointer::type dcast_bucket_ptr(const SlistNodePtr &p) { typedef typename pointer_traits::template rebind_pointer::type node_ptr; return pointer_traits::pointer_to(static_cast(*p)); } template struct group_functions { typedef NodeTraits node_traits; typedef unordered_group_adapter group_traits; typedef typename node_traits::node_ptr node_ptr; typedef typename node_traits::node node; typedef typename reduced_slist_node_traits ::type reduced_node_traits; typedef typename reduced_node_traits::node_ptr slist_node_ptr; typedef typename reduced_node_traits::node slist_node; typedef circular_slist_algorithms group_algorithms; static slist_node_ptr get_bucket_before_begin (const slist_node_ptr &bucket_beg, const slist_node_ptr &bucket_end, const node_ptr &p) { //First find the last node of p's group. //This requires checking the first node of the next group or //the bucket node. node_ptr prev_node = p; node_ptr nxt(node_traits::get_next(p)); while(!(bucket_beg <= nxt && nxt <= bucket_end) && (group_traits::get_next(nxt) == prev_node)){ prev_node = nxt; nxt = node_traits::get_next(nxt); } //If we've reached the bucket node just return it. if(bucket_beg <= nxt && nxt <= bucket_end){ return nxt; } //Otherwise, iterate using group links until the bucket node node_ptr first_node_of_group = nxt; node_ptr last_node_group = group_traits::get_next(first_node_of_group); slist_node_ptr possible_end = node_traits::get_next(last_node_group); while(!(bucket_beg <= possible_end && possible_end <= bucket_end)){ first_node_of_group = detail::dcast_bucket_ptr(possible_end); last_node_group = group_traits::get_next(first_node_of_group); possible_end = node_traits::get_next(last_node_group); } return possible_end; } static node_ptr get_prev_to_first_in_group(const slist_node_ptr &bucket_node, const node_ptr &first_in_group) { //Just iterate using group links and obtain the node //before "first_in_group)" node_ptr prev_node = detail::dcast_bucket_ptr(bucket_node); node_ptr nxt(node_traits::get_next(prev_node)); while(nxt != first_in_group){ prev_node = group_traits::get_next(nxt); nxt = node_traits::get_next(prev_node); } return prev_node; } static node_ptr get_first_in_group_of_last_in_group(const node_ptr &last_in_group) { //Just iterate using group links and obtain the node //before "last_in_group" node_ptr possible_first = group_traits::get_next(last_in_group); node_ptr possible_first_prev = group_traits::get_next(possible_first); // The deleted node is at the end of the group, so the // node in the group pointing to it is at the beginning // of the group. Find that to change its pointer. while(possible_first_prev != last_in_group){ possible_first = possible_first_prev; possible_first_prev = group_traits::get_next(possible_first); } return possible_first; } static void erase_from_group(const slist_node_ptr &end_ptr, const node_ptr &to_erase_ptr, detail::true_) { node_ptr nxt_ptr(node_traits::get_next(to_erase_ptr)); node_ptr prev_in_group_ptr(group_traits::get_next(to_erase_ptr)); bool last_in_group = (end_ptr == nxt_ptr) || (group_traits::get_next(nxt_ptr) != to_erase_ptr); bool is_first_in_group = node_traits::get_next(prev_in_group_ptr) != to_erase_ptr; if(is_first_in_group && last_in_group){ group_algorithms::init(to_erase_ptr); } else if(is_first_in_group){ group_algorithms::unlink_after(nxt_ptr); } else if(last_in_group){ node_ptr first_in_group = get_first_in_group_of_last_in_group(to_erase_ptr); group_algorithms::unlink_after(first_in_group); } else{ group_algorithms::unlink_after(nxt_ptr); } } static void erase_from_group(const slist_node_ptr&, const node_ptr&, detail::false_) {} static node_ptr get_last_in_group(const node_ptr &first_in_group, detail::true_) { return group_traits::get_next(first_in_group); } static node_ptr get_last_in_group(const node_ptr &n, detail::false_) { return n; } static void init_group(const node_ptr &n, true_) { group_algorithms::init(n); } static void init_group(const node_ptr &, false_) {} static void insert_in_group(const node_ptr &first_in_group, const node_ptr &n, true_) { if(first_in_group){ if(group_algorithms::unique(first_in_group)) group_algorithms::link_after(first_in_group, n); else{ group_algorithms::link_after(group_algorithms::node_traits::get_next(first_in_group), n); } } else{ group_algorithms::init_header(n); } } static slist_node_ptr get_previous_and_next_in_group ( const slist_node_ptr &i, node_ptr &nxt_in_group //If first_end_ptr == last_end_ptr, then first_end_ptr is the bucket of i //Otherwise first_end_ptr is the first bucket and last_end_ptr the last one. , const slist_node_ptr &first_end_ptr, const slist_node_ptr &last_end_ptr) { slist_node_ptr prev; node_ptr elem(detail::dcast_bucket_ptr(i)); //It's the last in group if the next_node is a bucket slist_node_ptr nxt(node_traits::get_next(elem)); bool last_in_group = (first_end_ptr <= nxt && nxt <= last_end_ptr)/* || (group_traits::get_next(nxt) != elem)*/; //It's the first in group if group_previous's next_node is not //itself, as group list does not link bucket node_ptr prev_in_group(group_traits::get_next(elem)); bool first_in_group = node_traits::get_next(prev_in_group) != elem; if(first_in_group){ node_ptr start_pos; if(last_in_group){ start_pos = elem; nxt_in_group = node_ptr(); } else{ start_pos = prev_in_group; nxt_in_group = node_traits::get_next(elem); } slist_node_ptr bucket_node; if(first_end_ptr != last_end_ptr){ bucket_node = group_functions::get_bucket_before_begin (first_end_ptr, last_end_ptr, start_pos); } else{ bucket_node = first_end_ptr; } prev = group_functions::get_prev_to_first_in_group(bucket_node, elem); } else{ if(last_in_group){ nxt_in_group = group_functions::get_first_in_group_of_last_in_group(elem); } else{ nxt_in_group = node_traits::get_next(elem); } prev = group_traits::get_next(elem); } return prev; } static void insert_in_group(const node_ptr&, const node_ptr&, false_) {} }; template class incremental_rehash_rollback { private: typedef BucketType bucket_type; typedef SplitTraits split_traits; incremental_rehash_rollback(); incremental_rehash_rollback & operator=(const incremental_rehash_rollback &); incremental_rehash_rollback (const incremental_rehash_rollback &); public: incremental_rehash_rollback (bucket_type &source_bucket, bucket_type &destiny_bucket, split_traits &split_traits) : source_bucket_(source_bucket), destiny_bucket_(destiny_bucket) , split_traits_(split_traits), released_(false) {} void release() { released_ = true; } ~incremental_rehash_rollback() { if(!released_){ //If an exception is thrown, just put all moved nodes back in the old bucket //and move back the split mark. destiny_bucket_.splice_after(destiny_bucket_.before_begin(), source_bucket_); split_traits_.decrement(); } } private: bucket_type &source_bucket_; bucket_type &destiny_bucket_; split_traits &split_traits_; bool released_; }; template struct node_functions { static void store_hash(typename NodeTraits::node_ptr p, std::size_t h, true_) { return NodeTraits::set_hash(p, h); } static void store_hash(typename NodeTraits::node_ptr, std::size_t, false_) {} }; inline std::size_t hash_to_bucket(std::size_t hash_value, std::size_t bucket_cnt, detail::false_) { return hash_value % bucket_cnt; } inline std::size_t hash_to_bucket(std::size_t hash_value, std::size_t bucket_cnt, detail::true_) { return hash_value & (bucket_cnt - 1); } template inline std::size_t hash_to_bucket_split(std::size_t hash_value, std::size_t bucket_cnt, std::size_t split) { std::size_t bucket_number = detail::hash_to_bucket(hash_value, bucket_cnt, detail::bool_()); if(Incremental) if(bucket_number >= split) bucket_number -= bucket_cnt/2; return bucket_number; } } //namespace detail { //!This metafunction will obtain the type of a bucket //!from the value_traits or hook option to be used with //!a hash container. template struct unordered_bucket : public detail::unordered_bucket_impl ::proto_value_traits > {}; //!This metafunction will obtain the type of a bucket pointer //!from the value_traits or hook option to be used with //!a hash container. template struct unordered_bucket_ptr : public detail::unordered_bucket_ptr_impl ::proto_value_traits > {}; //!This metafunction will obtain the type of the default bucket traits //!(when the user does not specify the bucket_traits<> option) from the //!value_traits or hook option to be used with //!a hash container. template struct unordered_default_bucket_traits { typedef typename ValueTraitsOrHookOption:: template pack::proto_value_traits supposed_value_traits; typedef typename detail:: get_slist_impl_from_supposed_value_traits ::type slist_impl; typedef detail::bucket_traits_impl implementation_defined; typedef implementation_defined type; }; struct default_bucket_traits; struct hashtable_defaults { typedef detail::default_hashtable_hook proto_value_traits; typedef std::size_t size_type; typedef void equal; typedef void hash; typedef default_bucket_traits bucket_traits; static const bool constant_time_size = true; static const bool power_2_buckets = false; static const bool cache_begin = false; static const bool compare_hash = false; static const bool incremental = false; }; template struct downcast_node_to_value_t : public detail::node_to_value { typedef detail::node_to_value base_t; typedef typename base_t::result_type result_type; typedef RealValueTraits real_value_traits; typedef typename detail::get_slist_impl ::type >::type slist_impl; typedef typename detail::add_const_if_c ::type & first_argument_type; typedef typename detail::add_const_if_c < typename RealValueTraits::node_traits::node , IsConst>::type & intermediate_argument_type; typedef typename pointer_traits :: template rebind_pointer ::type const_real_value_traits_ptr; downcast_node_to_value_t(const const_real_value_traits_ptr &ptr) : base_t(ptr) {} result_type operator()(first_argument_type arg) const { return this->base_t::operator()(static_cast(arg)); } }; template struct node_cast_adaptor : private detail::ebo_functor_holder { typedef detail::ebo_functor_holder base_t; typedef typename pointer_traits::element_type slist_node; typedef typename pointer_traits::element_type node; template node_cast_adaptor(const ConvertibleToF &c2f, const RealValuTraits *traits) : base_t(base_t(c2f, traits)) {} typename base_t::node_ptr operator()(const slist_node &to_clone) { return base_t::operator()(static_cast(to_clone)); } void operator()(SlistNodePtr to_clone) { base_t::operator()(pointer_traits::pointer_to(static_cast(*to_clone))); } }; static const std::size_t hashtable_data_bool_flags_mask = ( hash_bool_flags::cache_begin_pos | hash_bool_flags::constant_time_size_pos | hash_bool_flags::incremental_pos ); template struct bucket_plus_vtraits : public ValueTraits { typedef BucketTraits bucket_traits; typedef ValueTraits value_traits; static const bool external_value_traits = detail::external_value_traits_bool_is_true::value; static const bool external_bucket_traits = detail::external_bucket_traits_bool_is_true::value; typedef typename detail::get_real_value_traits::type real_value_traits; static const bool safemode_or_autounlink = is_safe_autounlink::value; typedef typename detail::eval_if_c < external_bucket_traits , detail::eval_bucket_traits , detail::identity >::type real_bucket_traits; typedef typename detail::get_slist_impl_from_supposed_value_traits ::type slist_impl; template bucket_plus_vtraits(const ValueTraits &val_traits, BOOST_FWD_REF(BucketTraitsType) b_traits) : ValueTraits(val_traits), bucket_traits_(::boost::forward(b_traits)) {} bucket_plus_vtraits & operator =(const bucket_plus_vtraits &x) { bucket_traits_ = x.bucket_traits_; return *this; } //real_value_traits // const real_value_traits &priv_real_value_traits(detail::false_) const { return *this; } const real_value_traits &priv_real_value_traits(detail::true_) const { return this->get_value_traits(*this); } real_value_traits &priv_real_value_traits(detail::false_) { return *this; } real_value_traits &priv_real_value_traits(detail::true_) { return this->get_value_traits(*this); } const real_value_traits &priv_real_value_traits() const { return this->priv_real_value_traits(detail::bool_()); } real_value_traits &priv_real_value_traits() { return this->priv_real_value_traits(detail::bool_()); } typedef typename pointer_traits:: template rebind_pointer::type const_real_value_traits_ptr; const_real_value_traits_ptr real_value_traits_ptr() const { return pointer_traits::pointer_to(this->priv_real_value_traits()); } //real_bucket_traits // const real_bucket_traits &priv_real_bucket_traits(detail::false_) const { return this->bucket_traits_; } const real_bucket_traits &priv_real_bucket_traits(detail::true_) const { return this->bucket_traits_.get_bucket_traits(*this); } real_bucket_traits &priv_real_bucket_traits(detail::false_) { return bucket_traits_; } real_bucket_traits &priv_real_bucket_traits(detail::true_) { return this->get_bucket_traits(*this); } const real_bucket_traits &priv_real_bucket_traits() const { return this->priv_real_bucket_traits(detail::bool_()); } real_bucket_traits &priv_real_bucket_traits() { return this->priv_real_bucket_traits(detail::bool_()); } //bucket_value_traits // const bucket_plus_vtraits &get_bucket_value_traits() const { return *this; } bucket_plus_vtraits &get_bucket_value_traits() { return *this; } typedef typename pointer_traits:: template rebind_pointer::type const_bucket_value_traits_ptr; const_bucket_value_traits_ptr bucket_value_traits_ptr() const { return pointer_traits::pointer_to(this->get_bucket_value_traits()); } //value traits // const value_traits &priv_value_traits() const { return *this; } value_traits &priv_value_traits() { return *this; } //bucket_traits // const bucket_traits &priv_bucket_traits() const { return this->bucket_traits_; } bucket_traits &priv_bucket_traits() { return this->bucket_traits_; } //operations typedef typename detail::unordered_bucket_ptr_impl::type bucket_ptr; bucket_ptr priv_bucket_pointer() const { return this->priv_real_bucket_traits().bucket_begin(); } typename slist_impl::size_type priv_bucket_count() const { return this->priv_real_bucket_traits().bucket_count(); } bucket_ptr priv_invalid_bucket() const { const real_bucket_traits &rbt = this->priv_real_bucket_traits(); return rbt.bucket_begin() + rbt.bucket_count(); } typedef typename real_value_traits::node_traits node_traits; typedef unordered_group_adapter group_traits; typedef typename slist_impl::iterator siterator; typedef typename slist_impl::size_type size_type; typedef detail::bucket_impl bucket_type; typedef detail::group_functions group_functions_t; typedef typename slist_impl::node_algorithms node_algorithms; typedef typename slist_impl::node_ptr slist_node_ptr; typedef typename node_traits::node_ptr node_ptr; typedef typename node_traits::node node; typedef typename real_value_traits::value_type value_type; typedef circular_slist_algorithms group_algorithms; /* siterator priv_invalid_local_it() const { return this->priv_invalid_bucket()->end(); } */ siterator priv_invalid_local_it() const { return this->priv_real_bucket_traits().bucket_begin()->before_begin(); } /// static siterator priv_get_last(bucket_type &b, detail::true_) //optimize multikey { //First find the last node of p's group. //This requires checking the first node of the next group or //the bucket node. slist_node_ptr end_ptr(b.end().pointed_node()); node_ptr possible_end(node_traits::get_next( detail::dcast_bucket_ptr(end_ptr))); node_ptr last_node_group(possible_end); while(end_ptr != possible_end){ last_node_group = group_traits::get_next(detail::dcast_bucket_ptr(possible_end)); possible_end = node_traits::get_next(last_node_group); } return bucket_type::s_iterator_to(*last_node_group); } static siterator priv_get_last(bucket_type &b, detail::false_) //NOT optimize multikey { return b.previous(b.end()); } static siterator priv_get_previous(bucket_type &b, siterator i, detail::true_) //optimize multikey { node_ptr elem(detail::dcast_bucket_ptr(i.pointed_node())); node_ptr prev_in_group(group_traits::get_next(elem)); bool first_in_group = node_traits::get_next(prev_in_group) != elem; typename bucket_type::node &n = first_in_group ? *group_functions_t::get_prev_to_first_in_group(b.end().pointed_node(), elem) : *group_traits::get_next(elem) ; return bucket_type::s_iterator_to(n); } static siterator priv_get_previous(bucket_type &b, siterator i, detail::false_) //NOT optimize multikey { return b.previous(i); } static void priv_clear_group_nodes(bucket_type &b, detail::true_) //optimize multikey { siterator it(b.begin()), itend(b.end()); while(it != itend){ node_ptr to_erase(detail::dcast_bucket_ptr(it.pointed_node())); ++it; group_algorithms::init(to_erase); } } static void priv_clear_group_nodes(bucket_type &, detail::false_) //NOT optimize multikey {} std::size_t priv_get_bucket_num_no_hash_store(siterator it, detail::true_) //optimize multikey { const bucket_ptr f(this->priv_bucket_pointer()), l(f + this->priv_bucket_count() - 1); slist_node_ptr bb = group_functions_t::get_bucket_before_begin ( f->end().pointed_node() , l->end().pointed_node() , detail::dcast_bucket_ptr(it.pointed_node())); //Now get the bucket_impl from the iterator const bucket_type &b = static_cast (bucket_type::slist_type::container_from_end_iterator(bucket_type::s_iterator_to(*bb))); //Now just calculate the index b has in the bucket array return static_cast(&b - &*f); } std::size_t priv_get_bucket_num_no_hash_store(siterator it, detail::false_) //NO optimize multikey { bucket_ptr f(this->priv_bucket_pointer()), l(f + this->priv_bucket_count() - 1); slist_node_ptr first_ptr(f->cend().pointed_node()) , last_ptr(l->cend().pointed_node()); //The end node is embedded in the singly linked list: //iterate until we reach it. while(!(first_ptr <= it.pointed_node() && it.pointed_node() <= last_ptr)){ ++it; } //Now get the bucket_impl from the iterator const bucket_type &b = static_cast (bucket_type::container_from_end_iterator(it)); //Now just calculate the index b has in the bucket array return static_cast(&b - &*f); } static std::size_t priv_stored_hash(slist_node_ptr n, detail::true_) //store_hash { return node_traits::get_hash(detail::dcast_bucket_ptr(n)); } static std::size_t priv_stored_hash(slist_node_ptr, detail::false_) //NO store_hash { //This code should never be reached! BOOST_INTRUSIVE_INVARIANT_ASSERT(0); return 0; } node &priv_value_to_node(value_type &v) { return *this->priv_real_value_traits().to_node_ptr(v); } const node &priv_value_to_node(const value_type &v) const { return *this->priv_real_value_traits().to_node_ptr(v); } value_type &priv_value_from_slist_node(slist_node_ptr n) { return *this->priv_real_value_traits().to_value_ptr(detail::dcast_bucket_ptr(n)); } const value_type &priv_value_from_slist_node(slist_node_ptr n) const { return *this->priv_real_value_traits().to_value_ptr(detail::dcast_bucket_ptr(n)); } bucket_traits bucket_traits_; }; template struct bucket_hash_t : public detail::ebo_functor_holder ::real_value_traits::value_type >::type > , bucket_plus_vtraits { typedef typename bucket_plus_vtraits::real_value_traits real_value_traits; typedef typename real_value_traits::value_type value_type; typedef typename real_value_traits::node_traits node_traits; typedef typename get_hash< VoidOrKeyHash, value_type>::type hasher; typedef BucketTraits bucket_traits; typedef bucket_plus_vtraits bucket_plus_vtraits_t; template bucket_hash_t(const ValueTraits &val_traits, BOOST_FWD_REF(BucketTraitsType) b_traits, const hasher & h) : detail::ebo_functor_holder(h), bucket_plus_vtraits_t(val_traits, ::boost::forward(b_traits)) {} const hasher &priv_hasher() const { return this->detail::ebo_functor_holder::get(); } hasher &priv_hasher() { return this->detail::ebo_functor_holder::get(); } std::size_t priv_stored_or_compute_hash(const value_type &v, detail::true_) const //For store_hash == true { return node_traits::get_hash(this->priv_real_value_traits().to_node_ptr(v)); } std::size_t priv_stored_or_compute_hash(const value_type &v, detail::false_) const //For store_hash == false { return this->priv_hasher()(v); } }; template struct bucket_hash_equal_t : public detail::ebo_functor_holder //equal ::real_value_traits::value_type >::type > , bucket_hash_t { typedef bucket_hash_t bucket_hash_type; typedef typename bucket_plus_vtraits::real_value_traits real_value_traits; typedef typename get_equal_to< VoidOrKeyEqual , typename real_value_traits::value_type >::type value_equal; typedef typename bucket_hash_type::hasher hasher; typedef BucketTraits bucket_traits; typedef bucket_hash_t buckethash_t; typedef typename bucket_hash_type::real_bucket_traits real_bucket_traits; typedef typename bucket_hash_type::slist_impl slist_impl; typedef typename slist_impl::size_type size_type; typedef typename slist_impl::iterator siterator; typedef detail::bucket_impl bucket_type; typedef typename detail::unordered_bucket_ptr_impl::type bucket_ptr; template bucket_hash_equal_t(const ValueTraits &val_traits, BOOST_FWD_REF(BucketTraitsType) b_traits, const hasher & h, const value_equal &e) : detail::ebo_functor_holder(e) , buckethash_t(val_traits, ::boost::forward(b_traits), h) {} bucket_ptr priv_get_cache() { return this->priv_bucket_pointer(); } void priv_set_cache(const bucket_ptr &) {} size_type priv_get_cache_bucket_num() { return 0u; } void priv_initialize_cache() {} void priv_swap_cache(bucket_hash_equal_t &) {} siterator priv_begin() const { size_type n = 0; size_type bucket_cnt = this->priv_bucket_count(); for (n = 0; n < bucket_cnt; ++n){ bucket_type &b = this->priv_bucket_pointer()[n]; if(!b.empty()){ return b.begin(); } } return this->priv_invalid_local_it(); } void priv_insertion_update_cache(size_type) {} void priv_erasure_update_cache_range(size_type, size_type) {} void priv_erasure_update_cache() {} const value_equal &priv_equal() const { return this->detail::ebo_functor_holder::get(); } value_equal &priv_equal() { return this->detail::ebo_functor_holder::get(); } }; template //cache_begin == true version struct bucket_hash_equal_t : public detail::ebo_functor_holder //equal ::real_value_traits::value_type >::type > , public bucket_hash_t { typedef bucket_hash_t bucket_hash_type; typedef typename get_equal_to< VoidOrKeyEqual , typename bucket_plus_vtraits::real_value_traits::value_type >::type value_equal; typedef typename bucket_hash_type::hasher hasher; typedef BucketTraits bucket_traits; typedef typename bucket_hash_type::slist_impl::size_type size_type; typedef typename bucket_hash_type::slist_impl::iterator siterator; template bucket_hash_equal_t(const ValueTraits &val_traits, BOOST_FWD_REF(BucketTraitsType) b_traits, const hasher & h, const value_equal &e) : detail::ebo_functor_holder(e) , bucket_hash_type(val_traits, ::boost::forward(b_traits), h) {} typedef typename detail::unordered_bucket_ptr_impl ::type bucket_ptr; bucket_ptr &priv_get_cache() { return cached_begin_; } const bucket_ptr &priv_get_cache() const { return cached_begin_; } void priv_set_cache(const bucket_ptr &p) { cached_begin_ = p; } std::size_t priv_get_cache_bucket_num() { return this->cached_begin_ - this->priv_bucket_pointer(); } void priv_initialize_cache() { this->cached_begin_ = this->priv_invalid_bucket(); } void priv_swap_cache(bucket_hash_equal_t &other) { std::swap(this->cached_begin_, other.cached_begin_); } siterator priv_begin() const { if(this->cached_begin_ == this->priv_invalid_bucket()){ return this->priv_invalid_local_it(); } else{ return this->cached_begin_->begin(); } } void priv_insertion_update_cache(size_type insertion_bucket) { bucket_ptr p = this->priv_bucket_pointer() + insertion_bucket; if(p < this->cached_begin_){ this->cached_begin_ = p; } } const value_equal &priv_equal() const { return this->detail::ebo_functor_holder::get(); } value_equal &priv_equal() { return this->detail::ebo_functor_holder::get(); } void priv_erasure_update_cache_range(size_type first_bucket_num, size_type last_bucket_num) { //If the last bucket is the end, the cache must be updated //to the last position if all if(this->priv_get_cache_bucket_num() == first_bucket_num && this->priv_bucket_pointer()[first_bucket_num].empty() ){ this->priv_set_cache(this->priv_bucket_pointer() + last_bucket_num); this->priv_erasure_update_cache(); } } void priv_erasure_update_cache() { if(this->cached_begin_ != this->priv_invalid_bucket()){ size_type current_n = this->priv_get_cache() - this->priv_bucket_pointer(); for( const size_type num_buckets = this->priv_bucket_count() ; current_n < num_buckets ; ++current_n, ++this->priv_get_cache()){ if(!this->priv_get_cache()->empty()){ return; } } this->priv_initialize_cache(); } } private: bucket_ptr cached_begin_; }; template struct hashdata_internal : public detail::size_holder< 0 != (BoolFlags & hash_bool_flags::incremental_pos), SizeType, int> //split_traits , public bucket_hash_equal_t < VoidOrKeyHash, VoidOrKeyEqual, ValueTraits, BucketTraits , 0 != (BoolFlags & hash_bool_flags::cache_begin_pos) > { typedef bucket_hash_equal_t < VoidOrKeyHash, VoidOrKeyEqual, ValueTraits, BucketTraits , 0 != (BoolFlags & hash_bool_flags::cache_begin_pos) > bucket_hash_equal_type; typedef typename bucket_hash_equal_type::value_equal value_equal; typedef typename bucket_hash_equal_type::hasher hasher; typedef bucket_plus_vtraits bucket_plus_vtraits_t; typedef typename bucket_plus_vtraits_t::size_type size_type; typedef typename bucket_plus_vtraits_t::bucket_ptr bucket_ptr; static const bool optimize_multikey = detail::optimize_multikey_is_true::value; typedef detail::bool_ optimize_multikey_t; template hashdata_internal(const ValueTraits &val_traits, BOOST_FWD_REF(BucketTraitsType) b_traits, const hasher & h, const value_equal &e) : bucket_hash_equal_type(val_traits, ::boost::forward(b_traits), h, e) {} typedef detail::size_holder <0 != (BoolFlags & hash_bool_flags::incremental_pos), SizeType, int> split_traits; split_traits &priv_split_traits() { return *this; } const split_traits &priv_split_traits() const { return *this; } }; template struct hashtable_data_t : public detail::size_holder< 0 != (BoolFlags & hash_bool_flags::constant_time_size_pos), SizeType> //size_traits , public hashdata_internal < SizeType, BoolFlags & (hash_bool_flags::incremental_pos | hash_bool_flags::cache_begin_pos) , VoidOrKeyHash, VoidOrKeyEqual, ValueTraits, BucketTraits> { static const std::size_t bool_flags = BoolFlags; typedef detail::size_holder < 0 != (BoolFlags & hash_bool_flags::constant_time_size_pos) , SizeType> size_traits; typedef hashdata_internal < SizeType, BoolFlags & (hash_bool_flags::incremental_pos | hash_bool_flags::cache_begin_pos) , VoidOrKeyHash, VoidOrKeyEqual, ValueTraits, BucketTraits> internal_type; typedef ValueTraits value_traits; typedef typename internal_type::value_equal value_equal; typedef typename internal_type::hasher hasher; typedef BucketTraits bucket_traits; typedef bucket_plus_vtraits bucket_plus_vtraits_t; static const bool external_value_traits = detail::external_value_traits_bool_is_true::value; static const bool external_bucket_traits = bucket_plus_vtraits_t::external_bucket_traits; typedef typename bucket_plus_vtraits_t::real_value_traits real_value_traits; typedef typename bucket_plus_vtraits_t::real_bucket_traits real_bucket_traits; size_traits &priv_size_traits() { return *this; } const size_traits &priv_size_traits() const { return *this; } template hashtable_data_t( BOOST_FWD_REF(BucketTraitsType) b_traits, const hasher & h , const value_equal &e, const value_traits &val_traits) : size_traits() , internal_type(val_traits, ::boost::forward(b_traits), h, e) {} }; /// @endcond //! The class template hashtable is an intrusive hash table container, that //! is used to construct intrusive unordered_set and unordered_multiset containers. The //! no-throw guarantee holds only, if the VoidOrKeyEqual object and Hasher don't throw. //! //! hashtable is a semi-intrusive container: each object to be stored in the //! container must contain a proper hook, but the container also needs //! additional auxiliary memory to work: hashtable needs a pointer to an array //! of type `bucket_type` to be passed in the constructor. This bucket array must //! have at least the same lifetime as the container. This makes the use of //! hashtable more complicated than purely intrusive containers. //! `bucket_type` is default-constructible, copyable and assignable //! //! The template parameter \c T is the type to be managed by the container. //! The user can specify additional options and if no options are provided //! default options are used. //! //! The container supports the following options: //! \c base_hook<>/member_hook<>/value_traits<>, //! \c constant_time_size<>, \c size_type<>, \c hash<> and \c equal<> //! \c bucket_traits<>, power_2_buckets<>, cache_begin<> and incremental<>. //! //! hashtable only provides forward iterators but it provides 4 iterator types: //! iterator and const_iterator to navigate through the whole container and //! local_iterator and const_local_iterator to navigate through the values //! stored in a single bucket. Local iterators are faster and smaller. //! //! It's not recommended to use non constant-time size hashtables because several //! key functions, like "empty()", become non-constant time functions. Non //! constant_time size hashtables are mainly provided to support auto-unlink hooks. //! //! hashtables, does not make automatic rehashings nor //! offers functions related to a load factor. Rehashing can be explicitly requested //! and the user must provide a new bucket array that will be used from that moment. //! //! Since no automatic rehashing is done, iterators are never invalidated when //! inserting or erasing elements. Iterators are only invalidated when rehashing. #if defined(BOOST_INTRUSIVE_DOXYGEN_INVOKED) template #else template #endif class hashtable_impl : public hashtable_data_t < SizeType , BoolFlags & hashtable_data_bool_flags_mask , VoidOrKeyHash, VoidOrKeyEqual, ValueTraits, BucketTraits> , private detail::clear_on_destructor_base < hashtable_impl , true //To always clear the bucket array //is_safe_autounlink::type::link_mode>::value > { template friend class detail::clear_on_destructor_base; public: typedef ValueTraits value_traits; typedef hashtable_data_t < SizeType , BoolFlags & hashtable_data_bool_flags_mask , VoidOrKeyHash, VoidOrKeyEqual, ValueTraits, BucketTraits> data_type; /// @cond static const bool external_value_traits = data_type::external_value_traits; static const bool external_bucket_traits = data_type::external_bucket_traits; typedef BucketTraits bucket_traits; typedef typename data_type::real_bucket_traits real_bucket_traits; typedef typename data_type::real_value_traits real_value_traits; typedef typename detail::get_slist_impl ::type >::type slist_impl; typedef bucket_plus_vtraits bucket_plus_vtraits_t; typedef typename bucket_plus_vtraits_t::const_real_value_traits_ptr const_real_value_traits_ptr; /// @endcond typedef typename real_value_traits::pointer pointer; typedef typename real_value_traits::const_pointer const_pointer; typedef typename real_value_traits::value_type value_type; typedef typename pointer_traits::reference reference; typedef typename pointer_traits::reference const_reference; typedef typename pointer_traits::difference_type difference_type; typedef SizeType size_type; typedef value_type key_type; typedef typename data_type::value_equal key_equal; typedef typename data_type::hasher hasher; typedef detail::bucket_impl bucket_type; typedef typename pointer_traits ::template rebind_pointer < bucket_type >::type bucket_ptr; typedef typename slist_impl::iterator siterator; typedef typename slist_impl::const_iterator const_siterator; typedef hashtable_iterator iterator; typedef hashtable_iterator const_iterator; typedef typename real_value_traits::node_traits node_traits; typedef typename node_traits::node node; typedef typename pointer_traits ::template rebind_pointer < node >::type node_ptr; typedef typename pointer_traits ::template rebind_pointer < const node >::type const_node_ptr; typedef typename slist_impl::node_algorithms node_algorithms; static const bool stateful_value_traits = detail::is_stateful_value_traits::value; static const bool store_hash = detail::store_hash_is_true::value; static const bool unique_keys = 0 != (BoolFlags & hash_bool_flags::unique_keys_pos); static const bool constant_time_size = 0 != (BoolFlags & hash_bool_flags::constant_time_size_pos); static const bool cache_begin = 0 != (BoolFlags & hash_bool_flags::cache_begin_pos); static const bool compare_hash = 0 != (BoolFlags & hash_bool_flags::compare_hash_pos); static const bool incremental = 0 != (BoolFlags & hash_bool_flags::incremental_pos); static const bool power_2_buckets = incremental || (0 != (BoolFlags & hash_bool_flags::power_2_buckets_pos)); static const bool optimize_multikey = detail::optimize_multikey_is_true::value && !unique_keys; /// @cond private: //Configuration error: compare_hash<> can't be specified without store_hash<> //See documentation for more explanations BOOST_STATIC_ASSERT((!compare_hash || store_hash)); typedef typename slist_impl::node_ptr slist_node_ptr; typedef typename pointer_traits ::template rebind_pointer < void >::type void_pointer; //We'll define group traits, but these won't be instantiated if //optimize_multikey is not true typedef unordered_group_adapter group_traits; typedef circular_slist_algorithms group_algorithms; typedef detail::bool_ store_hash_t; typedef detail::bool_ optimize_multikey_t; typedef detail::bool_ cache_begin_t; typedef detail::bool_ power_2_buckets_t; typedef detail::size_holder size_traits; typedef detail::size_holder split_traits; typedef detail::group_functions group_functions_t; typedef detail::node_functions node_functions_t; private: //noncopyable, movable BOOST_MOVABLE_BUT_NOT_COPYABLE(hashtable_impl) static const bool safemode_or_autounlink = is_safe_autounlink::value; //Constant-time size is incompatible with auto-unlink hooks! BOOST_STATIC_ASSERT(!(constant_time_size && ((int)real_value_traits::link_mode == (int)auto_unlink))); //Cache begin is incompatible with auto-unlink hooks! BOOST_STATIC_ASSERT(!(cache_begin && ((int)real_value_traits::link_mode == (int)auto_unlink))); template node_cast_adaptor< detail::node_disposer , slist_node_ptr, node_ptr > make_node_disposer(const Disposer &disposer) const { return node_cast_adaptor < detail::node_disposer , slist_node_ptr, node_ptr > (disposer, &this->priv_real_value_traits()); } /// @endcond public: typedef detail::insert_commit_data_impl insert_commit_data; typedef detail::transform_iterator < typename slist_impl::iterator , downcast_node_to_value_t < real_value_traits , false> > local_iterator; typedef detail::transform_iterator < typename slist_impl::iterator , downcast_node_to_value_t < real_value_traits , true> > const_local_iterator; public: //! Requires: buckets must not be being used by any other resource. //! //! Effects: Constructs an empty unordered_set, storing a reference //! to the bucket array and copies of the key_hasher and equal_func functors. //! //! Complexity: Constant. //! //! Throws: If value_traits::node_traits::node //! constructor throws (this does not happen with predefined Boost.Intrusive hooks) //! or the copy constructor or invocation of hash_func or equal_func throws. //! //! Notes: buckets array must be disposed only after //! *this is disposed. explicit hashtable_impl ( const bucket_traits &b_traits , const hasher & hash_func = hasher() , const key_equal &equal_func = key_equal() , const value_traits &v_traits = value_traits()) : data_type(b_traits, hash_func, equal_func, v_traits) { this->priv_initialize_buckets(); this->priv_size_traits().set_size(size_type(0)); size_type bucket_sz = this->priv_bucket_count(); BOOST_INTRUSIVE_INVARIANT_ASSERT(bucket_sz != 0); //Check power of two bucket array if the option is activated BOOST_INTRUSIVE_INVARIANT_ASSERT (!power_2_buckets || (0 == (bucket_sz & (bucket_sz-1)))); this->priv_split_traits().set_size(bucket_sz>>1); } //! Effects: to-do //! hashtable_impl(BOOST_RV_REF(hashtable_impl) x) : data_type( ::boost::move(x.priv_bucket_traits()) , ::boost::move(x.priv_hasher()) , ::boost::move(x.priv_equal()) , ::boost::move(x.priv_value_traits()) ) { this->priv_swap_cache(x); x.priv_initialize_cache(); if(constant_time_size){ this->priv_size_traits().set_size(size_type(0)); this->priv_size_traits().set_size(x.priv_size_traits().get_size()); x.priv_size_traits().set_size(size_type(0)); } if(incremental){ this->priv_split_traits().set_size(x.priv_split_traits().get_size()); x.priv_split_traits().set_size(size_type(0)); } } //! Effects: to-do //! hashtable_impl& operator=(BOOST_RV_REF(hashtable_impl) x) { this->swap(x); return *this; } #if defined(BOOST_INTRUSIVE_DOXYGEN_INVOKED) //! Effects: Detaches all elements from this. The objects in the unordered_set //! are not deleted (i.e. no destructors are called). //! //! Complexity: Linear to the number of elements in the unordered_set, if //! it's a safe-mode or auto-unlink value. Otherwise constant. //! //! Throws: Nothing. ~hashtable_impl() {} #endif //! Effects: Returns an iterator pointing to the beginning of the unordered_set. //! //! Complexity: Amortized constant time. //! Worst case (empty unordered_set): O(this->bucket_count()) //! //! Throws: Nothing. iterator begin() { return iterator(this->priv_begin(), &this->get_bucket_value_traits()); } //! Effects: Returns a const_iterator pointing to the beginning //! of the unordered_set. //! //! Complexity: Amortized constant time. //! Worst case (empty unordered_set): O(this->bucket_count()) //! //! Throws: Nothing. const_iterator begin() const { return this->cbegin(); } //! Effects: Returns a const_iterator pointing to the beginning //! of the unordered_set. //! //! Complexity: Amortized constant time. //! Worst case (empty unordered_set): O(this->bucket_count()) //! //! Throws: Nothing. const_iterator cbegin() const { return const_iterator(this->priv_begin(), &this->get_bucket_value_traits()); } //! Effects: Returns an iterator pointing to the end of the unordered_set. //! //! Complexity: Constant. //! //! Throws: Nothing. iterator end() { return iterator(this->priv_invalid_local_it(), 0); } //! Effects: Returns a const_iterator pointing to the end of the unordered_set. //! //! Complexity: Constant. //! //! Throws: Nothing. const_iterator end() const { return this->cend(); } //! Effects: Returns a const_iterator pointing to the end of the unordered_set. //! //! Complexity: Constant. //! //! Throws: Nothing. const_iterator cend() const { return const_iterator(this->priv_invalid_local_it(), 0); } //! Effects: Returns the hasher object used by the unordered_set. //! //! Complexity: Constant. //! //! Throws: If hasher copy-constructor throws. hasher hash_function() const { return this->priv_hasher(); } //! Effects: Returns the key_equal object used by the unordered_set. //! //! Complexity: Constant. //! //! Throws: If key_equal copy-constructor throws. key_equal key_eq() const { return this->priv_equal(); } //! Effects: Returns true if the container is empty. //! //! Complexity: if constant-time size and cache_begin options are disabled, //! average constant time (worst case, with empty() == true: O(this->bucket_count()). //! Otherwise constant. //! //! Throws: Nothing. bool empty() const { if(constant_time_size){ return !this->size(); } else if(cache_begin){ return this->begin() == this->end(); } else{ size_type bucket_cnt = this->priv_bucket_count(); const bucket_type *b = boost::intrusive::detail::to_raw_pointer(this->priv_bucket_pointer()); for (size_type n = 0; n < bucket_cnt; ++n, ++b){ if(!b->empty()){ return false; } } return true; } } //! Effects: Returns the number of elements stored in the unordered_set. //! //! Complexity: Linear to elements contained in *this if //! constant_time_size is false. Constant-time otherwise. //! //! Throws: Nothing. size_type size() const { if(constant_time_size) return this->priv_size_traits().get_size(); else{ size_type len = 0; size_type bucket_cnt = this->priv_bucket_count(); const bucket_type *b = boost::intrusive::detail::to_raw_pointer(this->priv_bucket_pointer()); for (size_type n = 0; n < bucket_cnt; ++n, ++b){ len += b->size(); } return len; } } //! Requires: the hasher and the equality function unqualified swap //! call should not throw. //! //! Effects: Swaps the contents of two unordered_sets. //! Swaps also the contained bucket array and equality and hasher functors. //! //! Complexity: Constant. //! //! Throws: If the swap() call for the comparison or hash functors //! found using ADL throw. Basic guarantee. void swap(hashtable_impl& other) { using std::swap; //These can throw swap(this->priv_equal(), other.priv_equal()); swap(this->priv_hasher(), other.priv_hasher()); //These can't throw swap(this->priv_bucket_traits(), other.priv_bucket_traits()); swap(this->priv_value_traits(), other.priv_value_traits()); this->priv_swap_cache(other); if(constant_time_size){ size_type backup = this->priv_size_traits().get_size(); this->priv_size_traits().set_size(other.priv_size_traits().get_size()); other.priv_size_traits().set_size(backup); } if(incremental){ size_type backup = this->priv_split_traits().get_size(); this->priv_split_traits().set_size(other.priv_split_traits().get_size()); other.priv_split_traits().set_size(backup); } } //! Requires: Disposer::operator()(pointer) shouldn't throw //! Cloner should yield to nodes that compare equal and produce the same //! hash than the original node. //! //! Effects: Erases all the elements from *this //! calling Disposer::operator()(pointer), clones all the //! elements from src calling Cloner::operator()(const_reference ) //! and inserts them on *this. The hash function and the equality //! predicate are copied from the source. //! //! If store_hash option is true, this method does not use the hash function. //! //! If any operation throws, all cloned elements are unlinked and disposed //! calling Disposer::operator()(pointer). //! //! Complexity: Linear to erased plus inserted elements. //! //! Throws: If cloner or hasher throw or hash or equality predicate copying //! throws. Basic guarantee. template void clone_from(const hashtable_impl &src, Cloner cloner, Disposer disposer) { this->clear_and_dispose(disposer); if(!constant_time_size || !src.empty()){ const size_type src_bucket_count = src.bucket_count(); const size_type dst_bucket_count = this->bucket_count(); //Check power of two bucket array if the option is activated BOOST_INTRUSIVE_INVARIANT_ASSERT (!power_2_buckets || (0 == (src_bucket_count & (src_bucket_count-1)))); BOOST_INTRUSIVE_INVARIANT_ASSERT (!power_2_buckets || (0 == (dst_bucket_count & (dst_bucket_count-1)))); //If src bucket count is bigger or equal, structural copy is possible if(!incremental && (src_bucket_count >= dst_bucket_count)){ //First clone the first ones const bucket_ptr src_buckets = src.priv_bucket_pointer(); const bucket_ptr dst_buckets = this->priv_bucket_pointer(); size_type constructed; typedef node_cast_adaptor< detail::node_disposer , slist_node_ptr, node_ptr > NodeDisposer; typedef node_cast_adaptor< detail::node_cloner , slist_node_ptr, node_ptr > NodeCloner; NodeDisposer node_disp(disposer, &this->priv_real_value_traits()); detail::exception_array_disposer rollback(dst_buckets[0], node_disp, constructed); for( constructed = 0 ; constructed < dst_bucket_count ; ++constructed){ dst_buckets[constructed].clone_from ( src_buckets[constructed] , NodeCloner(cloner, &this->priv_real_value_traits()), node_disp); } if(src_bucket_count != dst_bucket_count){ //Now insert the remaining ones using the modulo trick for(//"constructed" comes from the previous loop ; constructed < src_bucket_count ; ++constructed){ bucket_type &dst_b = dst_buckets[detail::hash_to_bucket_split(constructed, dst_bucket_count, dst_bucket_count)]; bucket_type &src_b = src_buckets[constructed]; for( siterator b(src_b.begin()), e(src_b.end()) ; b != e ; ++b){ dst_b.push_front(*(NodeCloner(cloner, &this->priv_real_value_traits())(*b.pointed_node()))); } } } this->priv_hasher() = src.priv_hasher(); this->priv_equal() = src.priv_equal(); rollback.release(); this->priv_size_traits().set_size(src.priv_size_traits().get_size()); this->priv_split_traits().set_size(dst_bucket_count); this->priv_insertion_update_cache(0u); this->priv_erasure_update_cache(); } else if(store_hash){ //Unlike previous cloning algorithm, this can throw //if cloner, hasher or comparison functor throw const_iterator b(src.begin()), e(src.end()); detail::exception_disposer rollback(*this, disposer); for(; b != e; ++b){ std::size_t hash_value = this->priv_stored_or_compute_hash(*b, store_hash_t());; this->priv_insert_equal_with_hash(*cloner(*b), hash_value); } rollback.release(); } else{ //Unlike previous cloning algorithm, this can throw //if cloner, hasher or comparison functor throw const_iterator b(src.begin()), e(src.end()); detail::exception_disposer rollback(*this, disposer); for(; b != e; ++b){ this->insert_equal(*cloner(*b)); } rollback.release(); } } } //! Requires: value must be an lvalue //! //! Effects: Inserts the value into the unordered_set. //! //! Returns: An iterator to the inserted value. //! //! Complexity: Average case O(1), worst case O(this->size()). //! //! Throws: If the internal hasher or the equality functor throws. Strong guarantee. //! //! Note: Does not affect the validity of iterators and references. //! No copy-constructors are called. iterator insert_equal(reference value) { size_type bucket_num; std::size_t hash_value; siterator prev; siterator it = this->priv_find (value, this->priv_hasher(), this->priv_equal(), bucket_num, hash_value, prev); return this->priv_insert_equal_find(value, bucket_num, hash_value, it); } //! Requires: Dereferencing iterator must yield an lvalue //! of type value_type. //! //! Effects: Equivalent to this->insert_equal(t) for each element in [b, e). //! //! Complexity: Average case O(N), where N is std::distance(b, e). //! Worst case O(N*this->size()). //! //! Throws: If the internal hasher or the equality functor throws. Basic guarantee. //! //! Note: Does not affect the validity of iterators and references. //! No copy-constructors are called. template void insert_equal(Iterator b, Iterator e) { for (; b != e; ++b) this->insert_equal(*b); } //! Requires: value must be an lvalue //! //! Effects: Tries to inserts value into the unordered_set. //! //! Returns: If the value //! is not already present inserts it and returns a pair containing the //! iterator to the new value and true. If there is an equivalent value //! returns a pair containing an iterator to the already present value //! and false. //! //! Complexity: Average case O(1), worst case O(this->size()). //! //! Throws: If the internal hasher or the equality functor throws. Strong guarantee. //! //! Note: Does not affect the validity of iterators and references. //! No copy-constructors are called. std::pair insert_unique(reference value) { insert_commit_data commit_data; std::pair ret = this->insert_unique_check (value, this->priv_hasher(), this->priv_equal(), commit_data); if(!ret.second) return ret; return std::pair (this->insert_unique_commit(value, commit_data), true); } //! Requires: Dereferencing iterator must yield an lvalue //! of type value_type. //! //! Effects: Equivalent to this->insert_unique(t) for each element in [b, e). //! //! Complexity: Average case O(N), where N is std::distance(b, e). //! Worst case O(N*this->size()). //! //! Throws: If the internal hasher or the equality functor throws. Basic guarantee. //! //! Note: Does not affect the validity of iterators and references. //! No copy-constructors are called. template void insert_unique(Iterator b, Iterator e) { for (; b != e; ++b) this->insert_unique(*b); } //! Requires: "hash_func" must be a hash function that induces //! the same hash values as the stored hasher. The difference is that //! "hash_func" hashes the given key instead of the value_type. //! //! "equal_func" must be a equality function that induces //! the same equality as key_equal. The difference is that //! "equal_func" compares an arbitrary key with the contained values. //! //! Effects: Checks if a value can be inserted in the unordered_set, using //! a user provided key instead of the value itself. //! //! Returns: If there is an equivalent value //! returns a pair containing an iterator to the already present value //! and false. If the value can be inserted returns true in the returned //! pair boolean and fills "commit_data" that is meant to be used with //! the "insert_commit" function. //! //! Complexity: Average case O(1), worst case O(this->size()). //! //! Throws: If hash_func or equal_func throw. Strong guarantee. //! //! Notes: This function is used to improve performance when constructing //! a value_type is expensive: if there is an equivalent value //! the constructed object must be discarded. Many times, the part of the //! node that is used to impose the hash or the equality is much cheaper to //! construct than the value_type and this function offers the possibility to //! use that the part to check if the insertion will be successful. //! //! If the check is successful, the user can construct the value_type and use //! "insert_commit" to insert the object in constant-time. //! //! "commit_data" remains valid for a subsequent "insert_commit" only if no more //! objects are inserted or erased from the unordered_set. //! //! After a successful rehashing insert_commit_data remains valid. template std::pair insert_unique_check ( const KeyType &key , KeyHasher hash_func , KeyValueEqual equal_func , insert_commit_data &commit_data) { size_type bucket_num; siterator prev; siterator prev_pos = this->priv_find(key, hash_func, equal_func, bucket_num, commit_data.hash, prev); bool success = prev_pos == this->priv_invalid_local_it(); if(success){ prev_pos = prev; } return std::pair(iterator(prev_pos, &this->get_bucket_value_traits()),success); } //! Requires: value must be an lvalue of type value_type. commit_data //! must have been obtained from a previous call to "insert_check". //! No objects should have been inserted or erased from the unordered_set between //! the "insert_check" that filled "commit_data" and the call to "insert_commit". //! //! Effects: Inserts the value in the unordered_set using the information obtained //! from the "commit_data" that a previous "insert_check" filled. //! //! Returns: An iterator to the newly inserted object. //! //! Complexity: Constant time. //! //! Throws: Nothing. //! //! Notes: This function has only sense if a "insert_check" has been //! previously executed to fill "commit_data". No value should be inserted or //! erased between the "insert_check" and "insert_commit" calls. //! //! After a successful rehashing insert_commit_data remains valid. iterator insert_unique_commit(reference value, const insert_commit_data &commit_data) { size_type bucket_num = this->priv_hash_to_bucket(commit_data.hash); bucket_type &b = this->priv_bucket_pointer()[bucket_num]; this->priv_size_traits().increment(); node_ptr n = pointer_traits::pointer_to(this->priv_value_to_node(value)); node_functions_t::store_hash(n, commit_data.hash, store_hash_t()); if(safemode_or_autounlink) BOOST_INTRUSIVE_SAFE_HOOK_DEFAULT_ASSERT(node_algorithms::unique(n)); this->priv_insertion_update_cache(bucket_num); group_functions_t::insert_in_group(node_ptr(), n, optimize_multikey_t()); return iterator(b.insert_after(b.before_begin(), *n), &this->get_bucket_value_traits()); } //! Effects: Erases the element pointed to by i. //! //! Complexity: Average case O(1), worst case O(this->size()). //! //! Throws: Nothing. //! //! Note: Invalidates the iterators (but not the references) //! to the erased element. No destructors are called. void erase(const_iterator i) { this->erase_and_dispose(i, detail::null_disposer()); } //! Effects: Erases the range pointed to by b end e. //! //! Complexity: Average case O(std::distance(b, e)), //! worst case O(this->size()). //! //! Throws: Nothing. //! //! Note: Invalidates the iterators (but not the references) //! to the erased elements. No destructors are called. void erase(const_iterator b, const_iterator e) { this->erase_and_dispose(b, e, detail::null_disposer()); } //! Effects: Erases all the elements with the given value. //! //! Returns: The number of erased elements. //! //! Complexity: Average case O(this->count(value)). //! Worst case O(this->size()). //! //! Throws: If the internal hasher or the equality functor throws. //! Basic guarantee. //! //! Note: Invalidates the iterators (but not the references) //! to the erased elements. No destructors are called. size_type erase(const_reference value) { return this->erase(value, this->priv_hasher(), this->priv_equal()); } //! Requires: "hash_func" must be a hash function that induces //! the same hash values as the stored hasher. The difference is that //! "hash_func" hashes the given key instead of the value_type. //! //! "equal_func" must be a equality function that induces //! the same equality as key_equal. The difference is that //! "equal_func" compares an arbitrary key with the contained values. //! //! Effects: Erases all the elements that have the same hash and //! compare equal with the given key. //! //! Returns: The number of erased elements. //! //! Complexity: Average case O(this->count(value)). //! Worst case O(this->size()). //! //! Throws: If hash_func or equal_func throw. Basic guarantee. //! //! Note: Invalidates the iterators (but not the references) //! to the erased elements. No destructors are called. template size_type erase(const KeyType& key, KeyHasher hash_func, KeyValueEqual equal_func) { return this->erase_and_dispose(key, hash_func, equal_func, detail::null_disposer()); } //! Requires: Disposer::operator()(pointer) shouldn't throw. //! //! Effects: Erases the element pointed to by i. //! Disposer::operator()(pointer) is called for the removed element. //! //! Complexity: Average case O(1), worst case O(this->size()). //! //! Throws: Nothing. //! //! Note: Invalidates the iterators //! to the erased elements. template void erase_and_dispose(const_iterator i, Disposer disposer /// @cond , typename detail::enable_if_c::value >::type * = 0 /// @endcond ) { this->priv_erase(i, disposer, optimize_multikey_t()); this->priv_size_traits().decrement(); this->priv_erasure_update_cache(); } //! Requires: Disposer::operator()(pointer) shouldn't throw. //! //! Effects: Erases the range pointed to by b end e. //! Disposer::operator()(pointer) is called for the removed elements. //! //! Complexity: Average case O(std::distance(b, e)), //! worst case O(this->size()). //! //! Throws: Nothing. //! //! Note: Invalidates the iterators //! to the erased elements. template void erase_and_dispose(const_iterator b, const_iterator e, Disposer disposer) { if(b != e){ //Get the bucket number and local iterator for both iterators siterator first_local_it(b.slist_it()); size_type first_bucket_num = this->priv_get_bucket_num(first_local_it); const bucket_ptr buck_ptr = this->priv_bucket_pointer(); siterator before_first_local_it = this->priv_get_previous(buck_ptr[first_bucket_num], first_local_it); size_type last_bucket_num; siterator last_local_it; //For the end iterator, we will assign the end iterator //of the last bucket if(e == this->end()){ last_bucket_num = this->bucket_count() - 1; last_local_it = buck_ptr[last_bucket_num].end(); } else{ last_local_it = e.slist_it(); last_bucket_num = this->priv_get_bucket_num(last_local_it); } this->priv_erase_range(before_first_local_it, first_bucket_num, last_local_it, last_bucket_num, disposer); this->priv_erasure_update_cache_range(first_bucket_num, last_bucket_num); } } //! Requires: Disposer::operator()(pointer) shouldn't throw. //! //! Effects: Erases all the elements with the given value. //! Disposer::operator()(pointer) is called for the removed elements. //! //! Returns: The number of erased elements. //! //! Complexity: Average case O(this->count(value)). //! Worst case O(this->size()). //! //! Throws: If the internal hasher or the equality functor throws. //! Basic guarantee. //! //! Note: Invalidates the iterators (but not the references) //! to the erased elements. No destructors are called. template size_type erase_and_dispose(const_reference value, Disposer disposer) { return this->erase_and_dispose(value, this->priv_hasher(), this->priv_equal(), disposer); } //! Requires: Disposer::operator()(pointer) shouldn't throw. //! //! Effects: Erases all the elements with the given key. //! according to the comparison functor "equal_func". //! Disposer::operator()(pointer) is called for the removed elements. //! //! Returns: The number of erased elements. //! //! Complexity: Average case O(this->count(value)). //! Worst case O(this->size()). //! //! Throws: If hash_func or equal_func throw. Basic guarantee. //! //! Note: Invalidates the iterators //! to the erased elements. template size_type erase_and_dispose(const KeyType& key, KeyHasher hash_func ,KeyValueEqual equal_func, Disposer disposer) { size_type bucket_num; std::size_t h; siterator prev; siterator it = this->priv_find(key, hash_func, equal_func, bucket_num, h, prev); bool success = it != this->priv_invalid_local_it(); size_type cnt(0); if(!success){ return 0; } else if(optimize_multikey){ siterator last = bucket_type::s_iterator_to (*node_traits::get_next(group_functions_t::get_last_in_group (detail::dcast_bucket_ptr(it.pointed_node()), optimize_multikey_t()))); this->priv_erase_range_impl(bucket_num, prev, last, disposer, cnt); } else{ //If found erase all equal values bucket_type &b = this->priv_bucket_pointer()[bucket_num]; for(siterator end_sit = b.end(); it != end_sit; ++cnt, ++it){ slist_node_ptr n(it.pointed_node()); const value_type &v = this->priv_value_from_slist_node(n); if(compare_hash){ std::size_t vh = this->priv_stored_or_compute_hash(v, store_hash_t()); if(h != vh || !equal_func(key, v)){ break; } } else if(!equal_func(key, v)){ break; } this->priv_size_traits().decrement(); } b.erase_after_and_dispose(prev, it, make_node_disposer(disposer)); } this->priv_erasure_update_cache(); return cnt; } //! Effects: Erases all of the elements. //! //! Complexity: Linear to the number of elements on the container. //! if it's a safe-mode or auto-unlink value_type. Constant time otherwise. //! //! Throws: Nothing. //! //! Note: Invalidates the iterators (but not the references) //! to the erased elements. No destructors are called. void clear() { this->priv_clear_buckets(); this->priv_size_traits().set_size(size_type(0)); } //! Requires: Disposer::operator()(pointer) shouldn't throw. //! //! Effects: Erases all of the elements. //! //! Complexity: Linear to the number of elements on the container. //! Disposer::operator()(pointer) is called for the removed elements. //! //! Throws: Nothing. //! //! Note: Invalidates the iterators (but not the references) //! to the erased elements. No destructors are called. template void clear_and_dispose(Disposer disposer) { if(!constant_time_size || !this->empty()){ size_type num_buckets = this->bucket_count(); bucket_ptr b = this->priv_bucket_pointer(); for(; num_buckets--; ++b){ b->clear_and_dispose(make_node_disposer(disposer)); } this->priv_size_traits().set_size(size_type(0)); } this->priv_initialize_cache(); } //! Effects: Returns the number of contained elements with the given value //! //! Complexity: Average case O(1), worst case O(this->size()). //! //! Throws: If the internal hasher or the equality functor throws. size_type count(const_reference value) const { return this->count(value, this->priv_hasher(), this->priv_equal()); } //! Requires: "hash_func" must be a hash function that induces //! the same hash values as the stored hasher. The difference is that //! "hash_func" hashes the given key instead of the value_type. //! //! "equal_func" must be a equality function that induces //! the same equality as key_equal. The difference is that //! "equal_func" compares an arbitrary key with the contained values. //! //! Effects: Returns the number of contained elements with the given key //! //! Complexity: Average case O(1), worst case O(this->size()). //! //! Throws: If hash_func or equal throw. template size_type count(const KeyType &key, const KeyHasher &hash_func, const KeyValueEqual &equal_func) const { size_type bucket_n1, bucket_n2, cnt; this->priv_equal_range(key, hash_func, equal_func, bucket_n1, bucket_n2, cnt); return cnt; } //! Effects: Finds an iterator to the first element is equal to //! "value" or end() if that element does not exist. //! //! Complexity: Average case O(1), worst case O(this->size()). //! //! Throws: If the internal hasher or the equality functor throws. iterator find(const_reference value) { return this->find(value, this->priv_hasher(), this->priv_equal()); } //! Requires: "hash_func" must be a hash function that induces //! the same hash values as the stored hasher. The difference is that //! "hash_func" hashes the given key instead of the value_type. //! //! "equal_func" must be a equality function that induces //! the same equality as key_equal. The difference is that //! "equal_func" compares an arbitrary key with the contained values. //! //! Effects: Finds an iterator to the first element whose key is //! "key" according to the given hash and equality functor or end() if //! that element does not exist. //! //! Complexity: Average case O(1), worst case O(this->size()). //! //! Throws: If hash_func or equal_func throw. //! //! Note: This function is used when constructing a value_type //! is expensive and the value_type can be compared with a cheaper //! key type. Usually this key is part of the value_type. template iterator find(const KeyType &key, KeyHasher hash_func, KeyValueEqual equal_func) { size_type bucket_n; std::size_t hash; siterator prev; siterator local_it = this->priv_find(key, hash_func, equal_func, bucket_n, hash, prev); return iterator(local_it, &this->get_bucket_value_traits()); } //! Effects: Finds a const_iterator to the first element whose key is //! "key" or end() if that element does not exist. //! //! Complexity: Average case O(1), worst case O(this->size()). //! //! Throws: If the internal hasher or the equality functor throws. const_iterator find(const_reference value) const { return this->find(value, this->priv_hasher(), this->priv_equal()); } //! Requires: "hash_func" must be a hash function that induces //! the same hash values as the stored hasher. The difference is that //! "hash_func" hashes the given key instead of the value_type. //! //! "equal_func" must be a equality function that induces //! the same equality as key_equal. The difference is that //! "equal_func" compares an arbitrary key with the contained values. //! //! Effects: Finds an iterator to the first element whose key is //! "key" according to the given hasher and equality functor or end() if //! that element does not exist. //! //! Complexity: Average case O(1), worst case O(this->size()). //! //! Throws: If hash_func or equal_func throw. //! //! Note: This function is used when constructing a value_type //! is expensive and the value_type can be compared with a cheaper //! key type. Usually this key is part of the value_type. template const_iterator find (const KeyType &key, KeyHasher hash_func, KeyValueEqual equal_func) const { size_type bucket_n; std::size_t hash_value; siterator prev; siterator sit = this->priv_find(key, hash_func, equal_func, bucket_n, hash_value, prev); return const_iterator(sit, &this->get_bucket_value_traits()); } //! Effects: Returns a range containing all elements with values equivalent //! to value. Returns std::make_pair(this->end(), this->end()) if no such //! elements exist. //! //! Complexity: Average case O(this->count(value)). Worst case O(this->size()). //! //! Throws: If the internal hasher or the equality functor throws. std::pair equal_range(const_reference value) { return this->equal_range(value, this->priv_hasher(), this->priv_equal()); } //! Requires: "hash_func" must be a hash function that induces //! the same hash values as the stored hasher. The difference is that //! "hash_func" hashes the given key instead of the value_type. //! //! "equal_func" must be a equality function that induces //! the same equality as key_equal. The difference is that //! "equal_func" compares an arbitrary key with the contained values. //! //! Effects: Returns a range containing all elements with equivalent //! keys. Returns std::make_pair(this->end(), this->end()) if no such //! elements exist. //! //! Complexity: Average case O(this->count(key, hash_func, equal_func)). //! Worst case O(this->size()). //! //! Throws: If hash_func or the equal_func throw. //! //! Note: This function is used when constructing a value_type //! is expensive and the value_type can be compared with a cheaper //! key type. Usually this key is part of the value_type. template std::pair equal_range (const KeyType &key, KeyHasher hash_func, KeyValueEqual equal_func) { size_type bucket_n1, bucket_n2, cnt; std::pair ret = this->priv_equal_range (key, hash_func, equal_func, bucket_n1, bucket_n2, cnt); return std::pair (iterator(ret.first, &this->get_bucket_value_traits()), iterator(ret.second, &this->get_bucket_value_traits())); } //! Effects: Returns a range containing all elements with values equivalent //! to value. Returns std::make_pair(this->end(), this->end()) if no such //! elements exist. //! //! Complexity: Average case O(this->count(value)). Worst case O(this->size()). //! //! Throws: If the internal hasher or the equality functor throws. std::pair equal_range(const_reference value) const { return this->equal_range(value, this->priv_hasher(), this->priv_equal()); } //! Requires: "hash_func" must be a hash function that induces //! the same hash values as the stored hasher. The difference is that //! "hash_func" hashes the given key instead of the value_type. //! //! "equal_func" must be a equality function that induces //! the same equality as key_equal. The difference is that //! "equal_func" compares an arbitrary key with the contained values. //! //! Effects: Returns a range containing all elements with equivalent //! keys. Returns std::make_pair(this->end(), this->end()) if no such //! elements exist. //! //! Complexity: Average case O(this->count(key, hash_func, equal_func)). //! Worst case O(this->size()). //! //! Throws: If the hasher or equal_func throw. //! //! Note: This function is used when constructing a value_type //! is expensive and the value_type can be compared with a cheaper //! key type. Usually this key is part of the value_type. template std::pair equal_range (const KeyType &key, KeyHasher hash_func, KeyValueEqual equal_func) const { size_type bucket_n1, bucket_n2, cnt; std::pair ret = this->priv_equal_range(key, hash_func, equal_func, bucket_n1, bucket_n2, cnt); return std::pair (const_iterator(ret.first, &this->get_bucket_value_traits()), const_iterator(ret.second, &this->get_bucket_value_traits())); } //! Requires: value must be an lvalue and shall be in a unordered_set of //! appropriate type. Otherwise the behavior is undefined. //! //! Effects: Returns: a valid iterator belonging to the unordered_set //! that points to the value //! //! Complexity: Constant. //! //! Throws: If the internal hash function throws. iterator iterator_to(reference value) { return iterator(bucket_type::s_iterator_to(this->priv_value_to_node(value)), &this->get_bucket_value_traits()); } //! Requires: value must be an lvalue and shall be in a unordered_set of //! appropriate type. Otherwise the behavior is undefined. //! //! Effects: Returns: a valid const_iterator belonging to the //! unordered_set that points to the value //! //! Complexity: Constant. //! //! Throws: If the internal hash function throws. const_iterator iterator_to(const_reference value) const { siterator sit = bucket_type::s_iterator_to(const_cast(this->priv_value_to_node(value))); return const_iterator(sit, &this->get_bucket_value_traits()); } //! Requires: value must be an lvalue and shall be in a unordered_set of //! appropriate type. Otherwise the behavior is undefined. //! //! Effects: Returns: a valid local_iterator belonging to the unordered_set //! that points to the value //! //! Complexity: Constant. //! //! Throws: Nothing. //! //! Note: This static function is available only if the value traits //! is stateless. static local_iterator s_local_iterator_to(reference value) { BOOST_STATIC_ASSERT((!stateful_value_traits)); siterator sit = bucket_type::s_iterator_to(((hashtable_impl*)0)->priv_value_to_node(value)); return local_iterator(sit, const_real_value_traits_ptr()); } //! Requires: value must be an lvalue and shall be in a unordered_set of //! appropriate type. Otherwise the behavior is undefined. //! //! Effects: Returns: a valid const_local_iterator belonging to //! the unordered_set that points to the value //! //! Complexity: Constant. //! //! Throws: Nothing. //! //! Note: This static function is available only if the value traits //! is stateless. static const_local_iterator s_local_iterator_to(const_reference value) { BOOST_STATIC_ASSERT((!stateful_value_traits)); siterator sit = bucket_type::s_iterator_to(((hashtable_impl*)0)->priv_value_to_node(const_cast(value))); return const_local_iterator(sit, const_real_value_traits_ptr()); } //! Requires: value must be an lvalue and shall be in a unordered_set of //! appropriate type. Otherwise the behavior is undefined. //! //! Effects: Returns: a valid local_iterator belonging to the unordered_set //! that points to the value //! //! Complexity: Constant. //! //! Throws: Nothing. local_iterator local_iterator_to(reference value) { siterator sit = bucket_type::s_iterator_to(this->priv_value_to_node(value)); return local_iterator(sit, this->real_value_traits_ptr()); } //! Requires: value must be an lvalue and shall be in a unordered_set of //! appropriate type. Otherwise the behavior is undefined. //! //! Effects: Returns: a valid const_local_iterator belonging to //! the unordered_set that points to the value //! //! Complexity: Constant. //! //! Throws: Nothing. const_local_iterator local_iterator_to(const_reference value) const { siterator sit = bucket_type::s_iterator_to (const_cast(this->priv_value_to_node(value))); return const_local_iterator(sit, this->real_value_traits_ptr()); } //! Effects: Returns the number of buckets passed in the constructor //! or the last rehash function. //! //! Complexity: Constant. //! //! Throws: Nothing. size_type bucket_count() const { return this->priv_bucket_count(); } //! Requires: n is in the range [0, this->bucket_count()). //! //! Effects: Returns the number of elements in the nth bucket. //! //! Complexity: Constant. //! //! Throws: Nothing. size_type bucket_size(size_type n) const { return this->priv_bucket_pointer()[n].size(); } //! Effects: Returns the index of the bucket in which elements //! with keys equivalent to k would be found, if any such element existed. //! //! Complexity: Constant. //! //! Throws: If the hash functor throws. //! //! Note: the return value is in the range [0, this->bucket_count()). size_type bucket(const key_type& k) const { return this->bucket(k, this->priv_hasher()); } //! Requires: "hash_func" must be a hash function that induces //! the same hash values as the stored hasher. The difference is that //! "hash_func" hashes the given key instead of the value_type. //! //! Effects: Returns the index of the bucket in which elements //! with keys equivalent to k would be found, if any such element existed. //! //! Complexity: Constant. //! //! Throws: If hash_func throws. //! //! Note: the return value is in the range [0, this->bucket_count()). template size_type bucket(const KeyType& k, const KeyHasher &hash_func) const { return this->priv_hash_to_bucket(hash_func(k)); } //! Effects: Returns the bucket array pointer passed in the constructor //! or the last rehash function. //! //! Complexity: Constant. //! //! Throws: Nothing. bucket_ptr bucket_pointer() const { return this->priv_bucket_pointer(); } //! Requires: n is in the range [0, this->bucket_count()). //! //! Effects: Returns a local_iterator pointing to the beginning //! of the sequence stored in the bucket n. //! //! Complexity: Constant. //! //! Throws: Nothing. //! //! Note: [this->begin(n), this->end(n)) is a valid range //! containing all of the elements in the nth bucket. local_iterator begin(size_type n) { return local_iterator(this->priv_bucket_pointer()[n].begin(), this->real_value_traits_ptr()); } //! Requires: n is in the range [0, this->bucket_count()). //! //! Effects: Returns a const_local_iterator pointing to the beginning //! of the sequence stored in the bucket n. //! //! Complexity: Constant. //! //! Throws: Nothing. //! //! Note: [this->begin(n), this->end(n)) is a valid range //! containing all of the elements in the nth bucket. const_local_iterator begin(size_type n) const { return this->cbegin(n); } //! Requires: n is in the range [0, this->bucket_count()). //! //! Effects: Returns a const_local_iterator pointing to the beginning //! of the sequence stored in the bucket n. //! //! Complexity: Constant. //! //! Throws: Nothing. //! //! Note: [this->begin(n), this->end(n)) is a valid range //! containing all of the elements in the nth bucket. const_local_iterator cbegin(size_type n) const { siterator sit = const_cast(this->priv_bucket_pointer()[n]).begin(); return const_local_iterator(sit, this->real_value_traits_ptr()); } //! Requires: n is in the range [0, this->bucket_count()). //! //! Effects: Returns a local_iterator pointing to the end //! of the sequence stored in the bucket n. //! //! Complexity: Constant. //! //! Throws: Nothing. //! //! Note: [this->begin(n), this->end(n)) is a valid range //! containing all of the elements in the nth bucket. local_iterator end(size_type n) { return local_iterator(this->priv_bucket_pointer()[n].end(), this->real_value_traits_ptr()); } //! Requires: n is in the range [0, this->bucket_count()). //! //! Effects: Returns a const_local_iterator pointing to the end //! of the sequence stored in the bucket n. //! //! Complexity: Constant. //! //! Throws: Nothing. //! //! Note: [this->begin(n), this->end(n)) is a valid range //! containing all of the elements in the nth bucket. const_local_iterator end(size_type n) const { return this->cend(n); } //! Requires: n is in the range [0, this->bucket_count()). //! //! Effects: Returns a const_local_iterator pointing to the end //! of the sequence stored in the bucket n. //! //! Complexity: Constant. //! //! Throws: Nothing. //! //! Note: [this->begin(n), this->end(n)) is a valid range //! containing all of the elements in the nth bucket. const_local_iterator cend(size_type n) const { return const_local_iterator ( const_cast(this->priv_bucket_pointer()[n]).end() , this->real_value_traits_ptr()); } //! Requires: new_bucket_traits can hold a pointer to a new bucket array //! or the same as the old bucket array with a different length. new_size is the length of the //! the array pointed by new_buckets. If new_bucket_traits.bucket_begin() == this->bucket_pointer() //! new_bucket_traits.bucket_count() can be bigger or smaller than this->bucket_count(). //! 'new_bucket_traits' copy constructor should not throw. //! //! Effects: Updates the internal reference with the new bucket, erases //! the values from the old bucket and inserts then in the new one. //! Bucket traits hold by *this is assigned from new_bucket_traits. //! If the container is configured as incremental<>, the split bucket is set //! to the new bucket_count(). //! //! If store_hash option is true, this method does not use the hash function. //! //! Complexity: Average case linear in this->size(), worst case quadratic. //! //! Throws: If the hasher functor throws. Basic guarantee. void rehash(const bucket_traits &new_bucket_traits) { const bucket_ptr new_buckets = new_bucket_traits.bucket_begin(); size_type new_bucket_count = new_bucket_traits.bucket_count(); const bucket_ptr old_buckets = this->priv_bucket_pointer(); size_type old_bucket_count = this->priv_bucket_count(); //Check power of two bucket array if the option is activated BOOST_INTRUSIVE_INVARIANT_ASSERT (!power_2_buckets || (0 == (new_bucket_count & (new_bucket_count-1u)))); size_type n = this->priv_get_cache_bucket_num(); const bool same_buffer = old_buckets == new_buckets; //If the new bucket length is a common factor //of the old one we can avoid hash calculations. const bool fast_shrink = (!incremental) && (old_bucket_count > new_bucket_count) && (power_2_buckets ||(old_bucket_count % new_bucket_count) == 0); //If we are shrinking the same bucket array and it's //is a fast shrink, just rehash the last nodes size_type new_first_bucket_num = new_bucket_count; if(same_buffer && fast_shrink && (n < new_bucket_count)){ n = new_bucket_count; new_first_bucket_num = this->priv_get_cache_bucket_num(); } //Anti-exception stuff: they destroy the elements if something goes wrong. //If the source and destination buckets are the same, the second rollback function //is harmless, because all elements have been already unlinked and destroyed typedef detail::init_disposer NodeDisposer; NodeDisposer node_disp; bucket_type & newbuck = new_buckets[0]; bucket_type & oldbuck = old_buckets[0]; detail::exception_array_disposer rollback1(newbuck, node_disp, new_bucket_count); detail::exception_array_disposer rollback2(oldbuck, node_disp, old_bucket_count); //Put size in a safe value for rollback exception size_type size_backup = this->priv_size_traits().get_size(); this->priv_size_traits().set_size(0); //Put cache to safe position this->priv_initialize_cache(); this->priv_insertion_update_cache(size_type(0u)); //Iterate through nodes for(; n < old_bucket_count; ++n){ bucket_type &old_bucket = old_buckets[n]; if(!fast_shrink){ siterator before_i(old_bucket.before_begin()); siterator end_sit(old_bucket.end()); siterator i(old_bucket.begin()); for(;i != end_sit; ++i){ const value_type &v = this->priv_value_from_slist_node(i.pointed_node()); const std::size_t hash_value = this->priv_stored_or_compute_hash(v, store_hash_t()); const size_type new_n = detail::hash_to_bucket_split(hash_value, new_bucket_count, new_bucket_count); if(cache_begin && new_n < new_first_bucket_num) new_first_bucket_num = new_n; siterator last = bucket_type::s_iterator_to (*group_functions_t::get_last_in_group (detail::dcast_bucket_ptr(i.pointed_node()), optimize_multikey_t())); if(same_buffer && new_n == n){ before_i = last; } else{ bucket_type &new_b = new_buckets[new_n]; new_b.splice_after(new_b.before_begin(), old_bucket, before_i, last); } i = before_i; } } else{ const size_type new_n = detail::hash_to_bucket_split(n, new_bucket_count, new_bucket_count); if(cache_begin && new_n < new_first_bucket_num) new_first_bucket_num = new_n; bucket_type &new_b = new_buckets[new_n]; if(!old_bucket.empty()){ new_b.splice_after( new_b.before_begin() , old_bucket , old_bucket.before_begin() , hashtable_impl::priv_get_last(old_bucket)); } } } this->priv_size_traits().set_size(size_backup); this->priv_split_traits().set_size(new_bucket_count); this->priv_real_bucket_traits() = new_bucket_traits; this->priv_initialize_cache(); this->priv_insertion_update_cache(new_first_bucket_num); rollback1.release(); rollback2.release(); } //! Requires: //! //! Effects: //! //! Complexity: //! //! Throws: //! //! Note: this method is only available if incremental option is activated. bool incremental_rehash(bool grow = true) { //This function is only available for containers with incremental hashing BOOST_STATIC_ASSERT(( incremental && power_2_buckets )); const size_type split_idx = this->priv_split_traits().get_size(); const size_type bucket_cnt = this->priv_bucket_count(); const bucket_ptr buck_ptr = this->priv_bucket_pointer(); if(grow){ //Test if the split variable can be changed if(split_idx >= bucket_cnt) return false; const size_type bucket_to_rehash = split_idx - bucket_cnt/2; bucket_type &old_bucket = buck_ptr[bucket_to_rehash]; siterator before_i(old_bucket.before_begin()); const siterator end_sit(old_bucket.end()); siterator i(old_bucket.begin()); this->priv_split_traits().increment(); //Anti-exception stuff: if an exception is thrown while //moving elements from old_bucket to the target bucket, all moved //elements are moved back to the original one. detail::incremental_rehash_rollback rollback ( buck_ptr[split_idx], old_bucket, this->priv_split_traits()); for(;i != end_sit; ++i){ const value_type &v = this->priv_value_from_slist_node(i.pointed_node()); const std::size_t hash_value = this->priv_stored_or_compute_hash(v, store_hash_t()); const size_type new_n = this->priv_hash_to_bucket(hash_value); siterator last = bucket_type::s_iterator_to (*group_functions_t::get_last_in_group (detail::dcast_bucket_ptr(i.pointed_node()), optimize_multikey_t())); if(new_n == bucket_to_rehash){ before_i = last; } else{ bucket_type &new_b = buck_ptr[new_n]; new_b.splice_after(new_b.before_begin(), old_bucket, before_i, last); } i = before_i; } rollback.release(); this->priv_erasure_update_cache(); return true; } else{ //Test if the split variable can be changed if(split_idx <= bucket_cnt/2) return false; const size_type target_bucket_num = split_idx - 1 - bucket_cnt/2; bucket_type &target_bucket = buck_ptr[target_bucket_num]; bucket_type &source_bucket = buck_ptr[split_idx-1]; target_bucket.splice_after(target_bucket.cbefore_begin(), source_bucket); this->priv_split_traits().decrement(); this->priv_insertion_update_cache(target_bucket_num); return true; } } //! Effects: If new_bucket_traits.bucket_count() is not //! this->bucket_count()/2 or this->bucket_count()*2, or //! this->split_bucket() != new_bucket_traits.bucket_count() returns false //! and does nothing. //! //! Otherwise, copy assigns new_bucket_traits to the internal bucket_traits //! and transfers all the objects from old buckets to the new ones. //! //! Complexity: Linear to size(). //! //! Throws: Nothing //! //! Note: this method is only available if incremental option is activated. bool incremental_rehash(const bucket_traits &new_bucket_traits) { //This function is only available for containers with incremental hashing BOOST_STATIC_ASSERT(( incremental && power_2_buckets )); size_type new_bucket_traits_size = new_bucket_traits.bucket_count(); size_type cur_bucket_traits = this->priv_bucket_count(); if(new_bucket_traits_size/2 != cur_bucket_traits && new_bucket_traits_size != cur_bucket_traits/2){ return false; } const size_type split_idx = this->split_count(); if(new_bucket_traits_size/2 == cur_bucket_traits){ //Test if the split variable can be changed if(!(split_idx >= cur_bucket_traits)) return false; } else{ //Test if the split variable can be changed if(!(split_idx <= cur_bucket_traits/2)) return false; } const size_type ini_n = this->priv_get_cache_bucket_num(); const bucket_ptr old_buckets = this->priv_bucket_pointer(); this->priv_real_bucket_traits() = new_bucket_traits; if(new_bucket_traits.bucket_begin() != old_buckets){ for(size_type n = ini_n; n < split_idx; ++n){ bucket_type &new_bucket = new_bucket_traits.bucket_begin()[n]; bucket_type &old_bucket = old_buckets[n]; new_bucket.splice_after(new_bucket.cbefore_begin(), old_bucket); } //Put cache to safe position this->priv_initialize_cache(); this->priv_insertion_update_cache(ini_n); } return true; } //! Requires: //! //! Effects: //! //! Complexity: //! //! Throws: size_type split_count() const { //This function is only available if incremental hashing is activated BOOST_STATIC_ASSERT(( incremental && power_2_buckets )); return this->priv_split_traits().get_size(); } //! Effects: Returns the nearest new bucket count optimized for //! the container that is bigger or equal than n. This suggestion can be //! used to create bucket arrays with a size that will usually improve //! container's performance. If such value does not exist, the //! higher possible value is returned. //! //! Complexity: Amortized constant time. //! //! Throws: Nothing. static size_type suggested_upper_bucket_count(size_type n) { const std::size_t *primes = &detail::prime_list_holder<0>::prime_list[0]; const std::size_t *primes_end = primes + detail::prime_list_holder<0>::prime_list_size; std::size_t const* bound = std::lower_bound(primes, primes_end, n); if(bound == primes_end) --bound; return size_type(*bound); } //! Effects: Returns the nearest new bucket count optimized for //! the container that is smaller or equal than n. This suggestion can be //! used to create bucket arrays with a size that will usually improve //! container's performance. If such value does not exist, the //! lowest possible value is returned. //! //! Complexity: Amortized constant time. //! //! Throws: Nothing. static size_type suggested_lower_bucket_count(size_type n) { const std::size_t *primes = &detail::prime_list_holder<0>::prime_list[0]; const std::size_t *primes_end = primes + detail::prime_list_holder<0>::prime_list_size; size_type const* bound = std::upper_bound(primes, primes_end, n); if(bound != primes) --bound; return size_type(*bound); } /// @cond private: void priv_clear_buckets(bucket_ptr buckets_ptr, size_type bucket_cnt) { for(; bucket_cnt--; ++buckets_ptr){ if(safemode_or_autounlink){ bucket_plus_vtraits_t::priv_clear_group_nodes(*buckets_ptr, optimize_multikey_t()); buckets_ptr->clear_and_dispose(detail::init_disposer()); } else{ buckets_ptr->clear(); } } this->priv_initialize_cache(); } void priv_initialize_buckets() { this->priv_clear_buckets(this->priv_bucket_pointer(), this->priv_bucket_count()); } void priv_clear_buckets() { this->priv_clear_buckets ( this->priv_get_cache() , this->priv_bucket_count() - (this->priv_get_cache() - this->priv_bucket_pointer())); } std::size_t priv_hash_to_bucket(std::size_t hash_value) const { return detail::hash_to_bucket_split (hash_value, this->priv_real_bucket_traits().bucket_count(), this->priv_split_traits().get_size()); } template void priv_erase_range_impl (size_type bucket_num, siterator before_first_it, siterator end_sit, Disposer disposer, size_type &num_erased) { const bucket_ptr buckets = this->priv_bucket_pointer(); bucket_type &b = buckets[bucket_num]; if(before_first_it == b.before_begin() && end_sit == b.end()){ this->priv_erase_range_impl(bucket_num, 1, disposer, num_erased); } else{ num_erased = 0; siterator to_erase(before_first_it); ++to_erase; slist_node_ptr end_ptr = end_sit.pointed_node(); while(to_erase != end_sit){ group_functions_t::erase_from_group(end_ptr, detail::dcast_bucket_ptr(to_erase.pointed_node()), optimize_multikey_t()); to_erase = b.erase_after_and_dispose(before_first_it, make_node_disposer(disposer)); ++num_erased; } this->priv_size_traits().set_size(this->priv_size_traits().get_size()-num_erased); } } template void priv_erase_range_impl (size_type first_bucket_num, size_type num_buckets, Disposer disposer, size_type &num_erased) { //Now fully clear the intermediate buckets const bucket_ptr buckets = this->priv_bucket_pointer(); num_erased = 0; for(size_type i = first_bucket_num; i < (num_buckets + first_bucket_num); ++i){ bucket_type &b = buckets[i]; siterator b_begin(b.before_begin()); siterator nxt(b_begin); ++nxt; siterator end_sit(b.end()); while(nxt != end_sit){ group_functions_t::init_group(detail::dcast_bucket_ptr(nxt.pointed_node()), optimize_multikey_t()); nxt = b.erase_after_and_dispose (b_begin, make_node_disposer(disposer)); this->priv_size_traits().decrement(); ++num_erased; } } } template void priv_erase_range( siterator before_first_it, size_type first_bucket , siterator last_it, size_type last_bucket , Disposer disposer) { size_type num_erased; if (first_bucket == last_bucket){ this->priv_erase_range_impl(first_bucket, before_first_it, last_it, disposer, num_erased); } else { bucket_type *b = (&this->priv_bucket_pointer()[0]); this->priv_erase_range_impl(first_bucket, before_first_it, b[first_bucket].end(), disposer, num_erased); if(size_type n = (last_bucket - first_bucket - 1)) this->priv_erase_range_impl(first_bucket + 1, n, disposer, num_erased); this->priv_erase_range_impl(last_bucket, b[last_bucket].before_begin(), last_it, disposer, num_erased); } } std::size_t priv_get_bucket_num(siterator it) { return this->priv_get_bucket_num_hash_dispatch(it, store_hash_t()); } std::size_t priv_get_bucket_num_hash_dispatch(siterator it, detail::true_) //store_hash { return this->priv_hash_to_bucket (this->priv_stored_hash(it.pointed_node(), store_hash_t())); } std::size_t priv_get_bucket_num_hash_dispatch(siterator it, detail::false_) //NO store_hash { return this->priv_get_bucket_num_no_hash_store(it, optimize_multikey_t()); } static siterator priv_get_previous(bucket_type &b, siterator i) { return bucket_plus_vtraits_t::priv_get_previous(b, i, optimize_multikey_t()); } static siterator priv_get_last(bucket_type &b) { return bucket_plus_vtraits_t::priv_get_last(b, optimize_multikey_t()); } template void priv_erase(const_iterator i, Disposer disposer, detail::true_) { slist_node_ptr elem(i.slist_it().pointed_node()); slist_node_ptr f_bucket_end, l_bucket_end; if(store_hash){ f_bucket_end = l_bucket_end = (this->priv_bucket_pointer() [this->priv_hash_to_bucket (this->priv_stored_hash(elem, store_hash_t())) ]).before_begin().pointed_node(); } else{ f_bucket_end = this->priv_bucket_pointer()->cend().pointed_node(); l_bucket_end = f_bucket_end + this->priv_bucket_count() - 1; } node_ptr nxt_in_group; siterator prev = bucket_type::s_iterator_to (*group_functions_t::get_previous_and_next_in_group ( elem, nxt_in_group, f_bucket_end, l_bucket_end) ); bucket_type::s_erase_after_and_dispose(prev, make_node_disposer(disposer)); if(nxt_in_group) group_algorithms::unlink_after(nxt_in_group); if(safemode_or_autounlink) group_algorithms::init(detail::dcast_bucket_ptr(elem)); } template void priv_erase(const_iterator i, Disposer disposer, detail::false_) { siterator to_erase(i.slist_it()); bucket_type &b = this->priv_bucket_pointer()[this->priv_get_bucket_num(to_erase)]; siterator prev(this->priv_get_previous(b, to_erase)); b.erase_after_and_dispose(prev, make_node_disposer(disposer)); } template siterator priv_find ( const KeyType &key, KeyHasher hash_func , KeyValueEqual equal_func, size_type &bucket_number, std::size_t &h, siterator &previt) const { h = hash_func(key); return this->priv_find_with_hash(key, equal_func, bucket_number, h, previt); } template siterator priv_find_with_hash ( const KeyType &key, KeyValueEqual equal_func, size_type &bucket_number, const std::size_t h, siterator &previt) const { bucket_number = this->priv_hash_to_bucket(h); bucket_type &b = this->priv_bucket_pointer()[bucket_number]; previt = b.before_begin(); if(constant_time_size && this->empty()){ return this->priv_invalid_local_it(); } siterator it = previt; ++it; while(it != b.end()){ const value_type &v = this->priv_value_from_slist_node(it.pointed_node()); if(compare_hash){ std::size_t vh = this->priv_stored_or_compute_hash(v, store_hash_t()); if(h == vh && equal_func(key, v)){ return it; } } else if(equal_func(key, v)){ return it; } if(optimize_multikey){ previt = bucket_type::s_iterator_to (*group_functions_t::get_last_in_group (detail::dcast_bucket_ptr(it.pointed_node()), optimize_multikey_t())); it = previt; } else{ previt = it; } ++it; } previt = b.before_begin(); return this->priv_invalid_local_it(); } iterator priv_insert_equal_with_hash(reference value, std::size_t hash_value) { size_type bucket_num; siterator prev; siterator it = this->priv_find_with_hash (value, this->priv_equal(), bucket_num, hash_value, prev); return this->priv_insert_equal_find(value, bucket_num, hash_value, it); } iterator priv_insert_equal_find(reference value, size_type bucket_num, std::size_t hash_value, siterator it) { bucket_type &b = this->priv_bucket_pointer()[bucket_num]; bool found_equal = it != this->priv_invalid_local_it(); if(!found_equal){ it = b.before_begin(); } //Now store hash if needed node_ptr n = pointer_traits::pointer_to(this->priv_value_to_node(value)); node_functions_t::store_hash(n, hash_value, store_hash_t()); //Checks for some modes if(safemode_or_autounlink) BOOST_INTRUSIVE_SAFE_HOOK_DEFAULT_ASSERT(node_algorithms::unique(n)); //Shortcut for optimize_multikey cases if(optimize_multikey){ node_ptr first_in_group = found_equal ? detail::dcast_bucket_ptr(it.pointed_node()) : node_ptr(); group_functions_t::insert_in_group(first_in_group, n, optimize_multikey_t()); } //Update cache and increment size if needed this->priv_insertion_update_cache(bucket_num); this->priv_size_traits().increment(); //Insert the element in the bucket after it return iterator(b.insert_after(it, *n), &this->get_bucket_value_traits()); } template std::pair priv_equal_range ( const KeyType &key , KeyHasher hash_func , KeyValueEqual equal_func , size_type &bucket_number_first , size_type &bucket_number_second , size_type &cnt) const { std::size_t h; cnt = 0; siterator prev; //Let's see if the element is present std::pair to_return ( this->priv_find(key, hash_func, equal_func, bucket_number_first, h, prev) , this->priv_invalid_local_it()); if(to_return.first == to_return.second){ bucket_number_second = bucket_number_first; return to_return; } { //If it's present, find the first that it's not equal in //the same bucket bucket_type &b = this->priv_bucket_pointer()[bucket_number_first]; siterator it = to_return.first; if(optimize_multikey){ to_return.second = bucket_type::s_iterator_to (*node_traits::get_next(group_functions_t::get_last_in_group (detail::dcast_bucket_ptr(it.pointed_node()), optimize_multikey_t()))); cnt = std::distance(it, to_return.second); if(to_return.second != b.end()){ bucket_number_second = bucket_number_first; return to_return; } } else{ ++cnt; ++it; while(it != b.end()){ const value_type &v = this->priv_value_from_slist_node(it.pointed_node()); if(compare_hash){ std::size_t hv = this->priv_stored_or_compute_hash(v, store_hash_t()); if(hv != h || !equal_func(key, v)){ to_return.second = it; bucket_number_second = bucket_number_first; return to_return; } } else if(!equal_func(key, v)){ to_return.second = it; bucket_number_second = bucket_number_first; return to_return; } ++it; ++cnt; } } } //If we reached the end, find the first, non-empty bucket for(bucket_number_second = bucket_number_first+1 ; bucket_number_second != this->priv_bucket_count() ; ++bucket_number_second){ bucket_type &b = this->priv_bucket_pointer()[bucket_number_second]; if(!b.empty()){ to_return.second = b.begin(); return to_return; } } //Otherwise, return the end node to_return.second = this->priv_invalid_local_it(); return to_return; } /// @endcond }; /// @cond #if !defined(BOOST_INTRUSIVE_VARIADIC_TEMPLATES) template < class T , bool UniqueKeys , class PackedOptions > #else template #endif struct make_real_bucket_traits { //Real value traits must be calculated from options typedef typename detail::get_value_traits ::type value_traits; /* static const bool resizable_bucket_traits = detail::resizable_bool_is_true::value;*/ typedef typename detail::get_real_value_traits::type real_value_traits; typedef typename PackedOptions::bucket_traits specified_bucket_traits; //Real bucket traits must be calculated from options and calculated value_traits typedef typename detail::get_slist_impl ::type >::type slist_impl; typedef typename detail::if_c< detail::is_same < specified_bucket_traits , default_bucket_traits >::value , detail::bucket_traits_impl , specified_bucket_traits >::type type; }; /// @endcond //! Helper metafunction to define a \c hashtable that yields to the same type when the //! same options (either explicitly or implicitly) are used. #if defined(BOOST_INTRUSIVE_DOXYGEN_INVOKED) || defined(BOOST_INTRUSIVE_VARIADIC_TEMPLATES) template #else template #endif struct make_hashtable { /// @cond typedef typename pack_options < hashtable_defaults, #if !defined(BOOST_INTRUSIVE_VARIADIC_TEMPLATES) O1, O2, O3, O4, O5, O6, O7, O8, O9, O10 #else Options... #endif >::type packed_options; typedef typename detail::get_value_traits ::type value_traits; typedef typename make_real_bucket_traits ::type real_bucket_traits; typedef hashtable_impl < value_traits , typename packed_options::hash , typename packed_options::equal , typename packed_options::size_type , real_bucket_traits , (std::size_t(false)*hash_bool_flags::unique_keys_pos) | (std::size_t(packed_options::constant_time_size)*hash_bool_flags::constant_time_size_pos) | (std::size_t(packed_options::power_2_buckets)*hash_bool_flags::power_2_buckets_pos) | (std::size_t(packed_options::cache_begin)*hash_bool_flags::cache_begin_pos) | (std::size_t(packed_options::compare_hash)*hash_bool_flags::compare_hash_pos) | (std::size_t(packed_options::incremental)*hash_bool_flags::incremental_pos) > implementation_defined; /// @endcond typedef implementation_defined type; }; #if !defined(BOOST_INTRUSIVE_DOXYGEN_INVOKED) #if defined(BOOST_INTRUSIVE_VARIADIC_TEMPLATES) template #else template #endif class hashtable : public make_hashtable::type { typedef typename make_hashtable::type Base; BOOST_MOVABLE_BUT_NOT_COPYABLE(hashtable) public: typedef typename Base::value_traits value_traits; typedef typename Base::real_value_traits real_value_traits; typedef typename Base::iterator iterator; typedef typename Base::const_iterator const_iterator; typedef typename Base::bucket_ptr bucket_ptr; typedef typename Base::size_type size_type; typedef typename Base::hasher hasher; typedef typename Base::bucket_traits bucket_traits; typedef typename Base::key_equal key_equal; //Assert if passed value traits are compatible with the type BOOST_STATIC_ASSERT((detail::is_same::value)); explicit hashtable ( const bucket_traits &b_traits , const hasher & hash_func = hasher() , const key_equal &equal_func = key_equal() , const value_traits &v_traits = value_traits()) : Base(b_traits, hash_func, equal_func, v_traits) {} hashtable(BOOST_RV_REF(hashtable) x) : Base(::boost::move(static_cast(x))) {} hashtable& operator=(BOOST_RV_REF(hashtable) x) { return static_cast(this->Base::operator=(::boost::move(static_cast(x)))); } }; #endif } //namespace intrusive } //namespace boost #include #endif //BOOST_INTRUSIVE_HASHTABLE_HPP