generalised a bit and uses trie now

lib/adaptive_distinguishing_sequence.cpp

@@ -8,11 +8,8 @@
 using namespace std;
 
 adaptive_distinguishing_sequence::adaptive_distinguishing_sequence(size_t N, size_t d)
-: CI(N)
-, depth(d)
-{
-	for(size_t i = 0; i < N; ++i)
-		CI[i] = {i, i};
+: CI(N), depth(d) {
+	for (size_t i = 0; i < N; ++i) CI[i] = {i, i};
 }
 
 adaptive_distinguishing_sequence create_adaptive_distinguishing_sequence(const result & splitting_tree){

lib/seperating_family.cpp

@@ -1,4 +1,5 @@
 #include "seperating_family.hpp"
+#include "trie.hpp"
 
 #include <boost/range/algorithm.hpp>
 #include <boost/range/algorithm_ext/erase.hpp>
@@ -9,48 +10,73 @@
 
 using namespace std;
 
-seperating_family create_seperating_family(const adaptive_distinguishing_sequence & sequence, const seperating_matrix & all_pair_seperating_sequences){
-	seperating_family seperating_family(all_pair_seperating_sequences.size());
+characterization_family create_seperating_family(const adaptive_distinguishing_sequence & sequence,
+                                                 const seperating_matrix & sep_matrix) {
+	const auto N = sequence.CI.size();
+	// all words (global/local) for all states
+	vector<trie> suffixes(N);
 
+	// all global separating sequences, which we will add to a state in the end ...
+	trie all_global_separating_words;
+
+	// ... to these particualr states
+	vector<bool> state_needs_global_suffixes(N, false);
+
+
+	// First we accumulate the kind-of-UIOs and the separating words we need. We will do this with a
+	// breath first search.
 	stack<pair<word, reference_wrapper<const adaptive_distinguishing_sequence>>> work;
 	work.push({{}, sequence});
-
 	while (!work.empty()) {
 		auto word = work.top().first;
 		const adaptive_distinguishing_sequence & node = work.top().second;
 		work.pop();
 
+		// On a leaf, we need to add the accumulated word as suffix (this is more or less a UIO).
+		// And, if needed, we also need to augment the set of suffixes (for all pairs).
 		if (node.children.empty()) {
-			// add sequence to this leave
 			for (auto && p : node.CI) {
 				const auto state = p.second;
-				seperating_family[state].push_back(word);
+				suffixes[state].insert(word);
 			}
 
-			// if the leaf is not a singleton, we need the all_pair seperating seqs
 			for (auto && p : node.CI) {
 				for (auto && q : node.CI) {
 					const auto s = p.second;
 					const auto t = q.second;
 					if (s == t) continue;
-					seperating_family[s].push_back(all_pair_seperating_sequences[s][t]);
+
+					const auto & sep_word = sep_matrix[s][t];
+
+					suffixes[s].insert(sep_word);
+					all_global_separating_words.insert(sep_word);
+					state_needs_global_suffixes[s] = true;
 				}
 			}
 
 			continue;
 		}
 
-		for(auto && i : node.word)
-			word.push_back(i);
-
-		for(auto && c : node.children)
-			work.push({word, c});
+		// add some work
+		for (auto && i : node.word) word.push_back(i);        // extend the word
+		for (auto && c : node.children) work.push({word, c}); // and visit the children with word
 	}
 
-	// Remove duplicates
-	for(auto & vec : seperating_family){
-		boost::erase(vec, boost::unique<boost::return_found_end>(boost::sort(vec)));
+
+	// Then we flatten them into a characterization family.
+	characterization_family ret(N);
+
+	for (state s = 0; s < N; ++s) {
+		auto & current_suffixes = suffixes[s];
+		ret[s].local_suffixes = flatten(current_suffixes);
+
+		if (state_needs_global_suffixes[s]) {
+			all_global_separating_words.for_each(
+			    [&current_suffixes](auto w) { current_suffixes.insert(w); });
 		}
 
-	return seperating_family;
+		ret[s].global_suffixes = flatten(current_suffixes);
+	}
+
+	return ret;
 }

lib/seperating_family.hpp

@@ -4,14 +4,26 @@
 #include "seperating_matrix.hpp"
 #include "types.hpp"
 
-/*
- * Given an (incomplete) adaptive distinguishing sequence and all pair
- * seperating sequences, we can construct a seperating family (as defined
- * in Lee & Yannakakis). If the adaptive distinguishing sequence is complete,
- * then the all pair seperating sequences are not needed.
- */
+/// \brief From the LY algorithm we generate characterizations sets (as in the Chow framework)
+/// If the adaptive distinguihsing sequence is complete, then we do not need to augment the LY
+/// result. This results in a separating family, which is stronger than a characterization set.
+/// However, if it is not complete, we augment it with sequences from the Wp-method.
 
-using seperating_set = std::vector<word>;
-using seperating_family = std::vector<seperating_set>;
+/// \brief A set (belonging to some state) of characterizing sequences
+/// It contains global_suffixes which should be used for testing whether the state is correct. Once
+/// we know the states make sense, we can test the transitions with the smaller set local_suffixes.
+/// There is some redundancy in this struct, but we have plenty of memory at our disposal.
+/// Note that even the global_suffixes may really on the state (because of the adaptiveness of the
+/// LY distinguishing sequence).
+struct characterization_set {
+	std::vector<word> global_suffixes;
+	std::vector<word> local_suffixes;
+};
 
-seperating_family create_seperating_family(adaptive_distinguishing_sequence const & sequence, seperating_matrix const & all_pair_seperating_sequences);
+/// \brief A family (indexed by states) of characterizations
+using characterization_family = std::vector<characterization_set>;
+
+/// \brief Creates the characterization family from the results of the LY algorithm
+/// If the sequence is complete, we do not need the separating_matrix
+characterization_family create_seperating_family(const adaptive_distinguishing_sequence & sequence,
+                                                 const seperating_matrix & sep_matrix);

lib/trie.hpp

@@ -43,19 +43,22 @@ struct trie {
 
 	/// \brief Inserts a word given as range \p r
 	/// \returns true if the element was inserted, false if already there
-	template <typename Range> bool insert(Range const& r) {
-		return insert(begin(r), end(r));
-	}
+	template <typename Range> bool insert(Range const & r) { return insert(begin(r), end(r)); }
 
-	/// \p function is applied to all word (not to the prefixes)
+	/// \brief Applies \p function to all word (not to the prefixes)
 	template <typename Fun> void for_each(Fun && function) const {
 		std::vector<size_t> word;
 		return for_each_impl(std::forward<Fun>(function), word);
 	}
 
+	/// \brief Empties the complete set
+	void clear() {
+		count = 0;
+		branches.clear();
+	}
+
   private:
-	template <typename Fun>
-	void for_each_impl(Fun&& function, std::vector<size_t>& word) const {
+	template <typename Fun> void for_each_impl(Fun && function, std::vector<size_t> & word) const {
 		if (count == 0) {
 			const auto & cword = word;
 			function(cword); // we don't want function to modify word

src/main.cpp

@@ -183,7 +183,7 @@ int main(int argc, char *argv[]) try {
 			for(state s = 0; s < machine.graph_size; ++s){
 				const auto prefix = transfer_sequences[s];
 
-				for(auto && suffix : seperating_family[s]){
+				for(auto && suffix : seperating_family[s].local_suffixes){
 					for(auto && r : all_sequences){
 						print_word(prefix);
 						print_word(r);
@@ -226,7 +226,7 @@ int main(int argc, char *argv[]) try {
 			}
 
 			using params = uniform_int_distribution<size_t>::param_type;
-			const auto & suffixes = seperating_family[current_state];
+			const auto & suffixes = seperating_family[current_state].local_suffixes;
 			const auto & s = suffixes[suffix_selection(generator, params{0, suffixes.size()-1})];
 
 			print_word(p);