more typename fixes
[prop.git] / lib-src / automata / sparsdfa.cc
blobadeffab0d0d0dd7b4fb9ce68fa732c7b8a101130
1 //////////////////////////////////////////////////////////////////////////////
2 // NOTICE:
3 //
4 // ADLib, Prop and their related set of tools and documentation are in the
5 // public domain. The author(s) of this software reserve no copyrights on
6 // the source code and any code generated using the tools. You are encouraged
7 // to use ADLib and Prop to develop software, in both academic and commercial
8 // settings, and are free to incorporate any part of ADLib and Prop into
9 // your programs.
11 // Although you are under no obligation to do so, we strongly recommend that
12 // you give away all software developed using our tools.
14 // We also ask that credit be given to us when ADLib and/or Prop are used in
15 // your programs, and that this notice be preserved intact in all the source
16 // code.
18 // This software is still under development and we welcome any suggestions
19 // and help from the users.
21 // Allen Leung
22 // 1994
23 //////////////////////////////////////////////////////////////////////////////
25 #include <AD/automata/sparsdfa.h>
26 #include <AD/contain/varstack.h>
28 //////////////////////////////////////////////////////////////////////////////
29 // Since optimal compression of the sparse table is equivalent to the
30 // problem of prune trie space minimization and is
31 // NP-complete(\cite{NP}, page 226), we'll use a simple first-fit
32 // strategy instead.
34 // The first-fit compression algorithm treats singleton states
35 // (i.e. states with only one out transition) specially. Singleton states
36 // transitions are saved on the stack |singletons| and are processed at
37 // the end of the run to fill up the remaining gaps.
38 //////////////////////////////////////////////////////////////////////////////
40 ////////////////////////////////////////////////////////////////////////
41 // Make hidden types visible
42 ////////////////////////////////////////////////////////////////////////
43 typedef SparseDFA::Symbol Symbol;
44 typedef SparseDFA::State State;
46 //////////////////////////////////////////////////////////////////////////
47 // Internally, class SparseDFA contains a stack of states with
48 // single non-error transition. These types of states are treated
49 // as a special case to during table compression.
50 //////////////////////////////////////////////////////////////////////////
51 struct SingleTrans {
52 State state; Symbol c; State dest;
53 SingleTrans() {}
54 SingleTrans(State s, Symbol a, State d) : state(s), c(a), dest(d) {}
57 //////////////////////////////////////////////////////////////////////////
58 // Stack of states with only 1 out transition
59 //////////////////////////////////////////////////////////////////////////
60 struct SparseDFA_Impl {
61 VarStack <SingleTrans> singletons;
64 ////////////////////////////////////////////////////////////////////////
65 // Constructor and destructor
66 ////////////////////////////////////////////////////////////////////////
67 SparseDFA:: SparseDFA () : impl(0) {}
68 SparseDFA::~SparseDFA () { delete impl; }
70 ////////////////////////////////////////////////////////////////////////
71 // Start generating tables; just clear the stack that memorize
72 // singleton transitions
73 ////////////////////////////////////////////////////////////////////////
74 void SparseDFA::start()
75 { delete impl; impl = new SparseDFA_Impl; impl->singletons.clear();
76 offset_base = 0;
79 ////////////////////////////////////////////////////////////////////////
80 // Finish generating tables. We'll go thru the singletons transitions
81 // emit them at this point
82 ////////////////////////////////////////////////////////////////////////
83 void SparseDFA::finish()
84 { //
85 // Fit all singleton transitions into the gaps. O(n) complexity.
87 register int offset, limit;
88 VarStack <SingleTrans>& singletons = impl->singletons;
89 for (offset = - min_symbol; ! singletons.is_empty(); ) {
90 SingleTrans& t = singletons.pop();
91 for (;;) {
92 for (limit = table_size - max_symbol - 1; offset < limit; offset++) {
93 if (check[offset + t.c] == error_state) {
94 base[t.state] = offset;
95 check[offset + t.c] = t.state;
96 next[offset + t.c] = t.dest;
97 if (offset + max_symbol > max_check)
98 max_check = offset + max_symbol;
99 goto next_symbol;
103 // All available gaps are filled. We'll have to increase
104 // the table size to fit in the rest. We'll have a perfect fit.
106 grow_tables( singletons.size() + max_symbol - min_symbol + 1);
107 limit = table_size;
108 max_check = table_size - 1;
110 next_symbol: ;
113 /////////////////////////////////////////////////////////////////////
114 // Clean up
115 /////////////////////////////////////////////////////////////////////
116 delete impl; impl = 0;
119 ////////////////////////////////////////////////////////////////////////
120 // Add a new state to the automaton
121 ////////////////////////////////////////////////////////////////////////
122 void SparseDFA::add_state
123 ( State s, int fan_out, const Symbol symbols[], const State delta[] )
124 { register int i, offset;
127 // Increase the size of the state to base offset table if necessary.
129 if (s > max_state) max_state = s;
130 if (s - error_state + 1 > number_of_states)
131 { grow_states(states() * 3 / 2 + 8);
135 // Singletons are saved onto a stack and processed at the end of the run.
137 if (fan_out == 0) return;
138 if (fan_out == 1) {
139 impl->singletons.push(SingleTrans(s,symbols[0],delta[0]));
140 return;
144 // Now fit in the rest using a simple first fit strategy. Worst case
145 // complexity is O(n^2). However, since the transitions are sparse
146 // we expect the average complexity to be better.
148 { register int i, limit;
149 for (i = offset_base, limit = table_size; i < limit; i++)
150 { if (check[i] == error_state) break; }
151 offset_base = i;
154 for (offset = - min_symbol + offset_base;; ) {
155 register int limit = table_size - max_symbol - 1;
156 if (limit <= offset) grow_tables(table_size * 3 / 2 + offset - limit);
159 // Use linear search on the tables
162 for ( ; offset < limit; offset++) {
163 for (i = fan_out - 1; i >= 0; i--)
164 if (check[offset + symbols[i]] != error_state) goto try_again;
166 // Found the right spot
168 for (i = fan_out - 1; i >= 0; i--) {
169 check[offset + symbols[i]] = s;
170 next[offset + symbols[i]] = delta[i];
172 base[s] = offset;
173 if (offset + max_symbol > max_check)
174 max_check = offset + max_symbol;
175 return;
176 try_again: ;