bec0f07fee18448987bbc8feef08cc8bd2b07849
[nit.git] / contrib / nitcc / src / autom.nit
1 # This file is part of NIT ( http://www.nitlanguage.org ).
2 #
3 # Licensed under the Apache License, Version 2.0 (the "License");
4 # you may not use this file except in compliance with the License.
5 # You may obtain a copy of the License at
6 #
7 # http://www.apache.org/licenses/LICENSE-2.0
8 #
9 # Unless required by applicable law or agreed to in writing, software
10 # distributed under the License is distributed on an "AS IS" BASIS,
11 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 # See the License for the specific language governing permissions and
13 # limitations under the License.
14
15 # Finite automaton (NFA & DFA)
16 module autom
17
18 # For the class Token
19 import grammar
20
21 # A finite automaton
22 class Automaton
23 # The start state
24 var start: State is noinit
25
26 # State that are accept states
27 var accept = new Array[State]
28
29 # All states
30 var states = new Array[State]
31
32 # Tokens associated on accept states.
33 # Use `add_tag` to update
34 var tags = new HashMap[State, Set[Token]]
35
36 # Accept states associated on tokens.
37 # Use `add_tag` to update
38 var retrotags = new HashMap[Token, Set[State]]
39
40 # Tag all accept states
41 fun tag_accept(t: Token)
42 do
43 for s in accept do add_tag(s, t)
44 end
45
46 # Add a token to a state
47 fun add_tag(s: State, t: Token)
48 do
49 if not tags.has_key(s) then
50 var set = new ArraySet[Token]
51 tags[s] = set
52 set.add t
53 else
54 tags[s].add t
55 end
56
57 if not retrotags.has_key(t) then
58 var set = new ArraySet[State]
59 retrotags[t] = set
60 set.add s
61 else
62 retrotags[t].add s
63 end
64
65 assert tags[s].has(t)
66 assert retrotags[t].has(s)
67 end
68
69 # Remove all occurrences of a tag in an automaton
70 fun clear_tag(t: Token)
71 do
72 if not retrotags.has_key(t) then return
73 for s in retrotags[t] do
74 if not tags.has_key(s) then continue
75 tags[s].remove(t)
76 if tags[s].is_empty then tags.keys.remove(s)
77 end
78 retrotags.keys.remove(t)
79 end
80
81 # Remove tokens from conflicting state according the inclusion of language.
82 # REQUIRE: self isa DFA automaton
83 fun solve_token_inclusion
84 do
85 for s, ts in tags do
86 if ts.length <= 1 then continue
87 var losers = new Array[Token]
88 for t1 in ts do
89 for t2 in ts do
90 if t1 == t2 then continue
91 if retrotags[t1].length > retrotags[t2].length and retrotags[t1].has_all(retrotags[t2]) then
92 losers.add(t1)
93 break
94 end
95 end
96 end
97 for t in losers do
98 ts.remove(t)
99 retrotags[t].remove s
100 end
101 end
102 end
103
104 # Initialize a new automaton for the empty language.
105 # One state, no accept, no transition.
106 init empty
107 do
108 var state = new State
109 start = state
110 states.add state
111 end
112
113 # Initialize a new automaton for the empty-string language.
114 # One state, is accept, no transition.
115 init epsilon
116 do
117 var state = new State
118 start = state
119 accept.add state
120 states.add state
121 end
122
123 # Initialize a new automation for the language that accepts only a single symbol.
124 # Two state, the second is accept, one transition on `symbol`.
125 init atom(symbol: Int)
126 do
127 var s = new State
128 var a = new State
129 var sym = new TSymbol(symbol, symbol)
130 s.add_trans(a, sym)
131 start = s
132 accept.add a
133 states.add s
134 states.add a
135 end
136
137 # Initialize a new automation for the language that accepts only a range of symbols
138 # Two state, the second is accept, one transition for `from` to `to`
139 init cla(first: Int, last: nullable Int)
140 do
141 var s = new State
142 var a = new State
143 var sym = new TSymbol(first, last)
144 s.add_trans(a, sym)
145 start = s
146 accept.add a
147 states.add s
148 states.add a
149 end
150
151 # Concatenate `other` to `self`.
152 # Other is modified and invalidated.
153 fun concat(other: Automaton)
154 do
155 var s2 = other.start
156 for a1 in accept do
157 a1.add_trans(s2, null)
158 end
159 accept = other.accept
160 states.add_all other.states
161 end
162
163 # `self` become the alternation of `self` and `other`.
164 # `other` is modified and invalidated.
165 fun alternate(other: Automaton)
166 do
167 var s = new State
168 var a = new State
169 s.add_trans(start, null)
170 for a1 in accept do
171 a1.add_trans(a, null)
172 end
173 s.add_trans(other.start, null)
174 for a2 in other.accept do
175 a2.add_trans(a, null)
176 accept.add(a2)
177 end
178
179 start = s
180 accept = [a]
181
182 states.add s
183 states.add a
184 states.add_all other.states
185 end
186
187 # Return a new automaton that recognize `self` but not `other`.
188 # For a theoretical POV, this is the subtraction of languages.
189 # Note: the implementation use `to_dfa` internally, so the theoretical complexity is not cheap.
190 fun except(other: Automaton): Automaton
191 do
192 var ta = new Token("1")
193 self.tag_accept(ta)
194 var tb = new Token("2")
195 other.tag_accept(tb)
196
197 var c = new Automaton.empty
198 c.absorb(self)
199 c.absorb(other)
200 c = c.to_dfa
201 c.accept.clear
202 for s in c.retrotags[ta] do
203 if not c.tags[s].has(tb) then
204 c.accept.add(s)
205 end
206 end
207 c.clear_tag(ta)
208 c.clear_tag(tb)
209 return c
210 end
211
212 # `self` absorbs all states, transitions, tags, and acceptations of `other`.
213 # An epsilon transition is added between `self.start` and `other.start`.
214 fun absorb(other: Automaton)
215 do
216 states.add_all other.states
217 start.add_trans(other.start, null)
218 for s, ts in other.tags do for t in ts do add_tag(s, t)
219 accept.add_all other.accept
220 end
221
222 # Do the Kleene closure (*) on self
223 fun close
224 do
225 for a1 in accept do
226 a1.add_trans(start, null)
227 start.add_trans(a1, null)
228 end
229 end
230
231 # Do the + on self
232 fun plus
233 do
234 for a1 in accept do
235 a1.add_trans(start, null)
236 end
237 end
238
239 # Do the ? on self
240 fun optionnal
241 do
242 alternate(new Automaton.epsilon)
243 end
244
245 # Remove all transitions on a given symbol
246 fun minus_sym(symbol: TSymbol)
247 do
248 var f = symbol.first
249 var l = symbol.last
250 for s in states do
251 for t in s.outs.to_a do
252 if t.symbol == null then continue
253
254 # Check overlaps
255 var tf = t.symbol.as(not null).first
256 var tl = t.symbol.as(not null).last
257 if l != null and tf > l then continue
258 if tl != null and f > tl then continue
259
260 t.delete
261
262 # Add left and right part if non empty
263 if tf < f then
264 var sym = new TSymbol(tf,f-1)
265 s.add_trans(t.to, sym)
266 end
267 if l != null then
268 if tl == null then
269 var sym = new TSymbol(l+1, null)
270 s.add_trans(t.to, sym)
271 else if tl > l then
272 var sym = new TSymbol(l+1, tl)
273 s.add_trans(t.to, sym)
274 end
275 end
276 end
277 end
278 end
279
280 # Fully duplicate an automaton
281 fun dup: Automaton
282 do
283 var res = new Automaton.empty
284 var map = new HashMap[State, State]
285 map[start] = res.start
286 for s in states do
287 if s == start then continue
288 var s2 = new State
289 map[s] = s2
290 res.states.add(s2)
291 end
292 for s in accept do
293 res.accept.add map[s]
294 end
295 for s, ts in tags do for t in ts do
296 res.add_tag(map[s], t)
297 end
298 for s in states do
299 for t in s.outs do
300 map[s].add_trans(map[t.to], t.symbol)
301 end
302 end
303 return res
304 end
305
306 # Reverse an automaton in place
307 fun reverse
308 do
309 for s in states do
310 var tmp = s.ins
311 s.ins = s.outs
312 s.outs = tmp
313 for t in s.outs do
314 var tmp2 = t.from
315 t.from = t.to
316 t.to = tmp2
317 end
318 end
319 var st = start
320 if accept.length == 1 then
321 start = accept.first
322 else
323 var st2 = new State
324 start = st2
325 states.add(st2)
326
327 for s in accept do
328 st2.add_trans(s, null)
329 end
330 end
331 accept.clear
332 accept.add(st)
333 end
334
335 # Remove states (and transitions) that does not reach an accept state
336 fun trim
337 do
338 # Good states are those we want to keep
339 var goods = new HashSet[State]
340 goods.add_all(accept)
341
342 var todo = accept.to_a
343
344 # Propagate goodness
345 while not todo.is_empty do
346 var s = todo.pop
347 for t in s.ins do
348 var s2 = t.from
349 if goods.has(s2) then continue
350 goods.add(s2)
351 todo.add(s2)
352 end
353 end
354
355 # What are the bad state then?
356 var bads = new Array[State]
357 for s in states do
358 if not goods.has(s) then bads.add(s)
359 end
360
361 # Remove their transitions
362 for s in bads do
363 for t in s.ins.to_a do t.delete
364 for t in s.outs.to_a do t.delete
365 end
366
367 # Keep only the good stuff
368 states.clear
369 states.add_all(goods)
370 states.add(start)
371 end
372
373 # Generate a minimal DFA
374 # REQUIRE: self is a DFA
375 fun to_minimal_dfa: Automaton
376 do
377 assert_valid
378
379 trim
380
381 # Graph of known distinct states.
382 var distincts = new HashMap[State, Set[State]]
383 for s in states do
384 distincts[s] = new HashSet[State]
385 end
386
387 # split accept states.
388 # An accept state is distinct with a non accept state.
389 for s1 in accept do
390 for s2 in states do
391 if distincts[s1].has(s2) then continue
392 if not accept.has(s2) then
393 distincts[s1].add(s2)
394 distincts[s2].add(s1)
395 continue
396 end
397 if tags.get_or_null(s1) != tags.get_or_null(s2) then
398 distincts[s1].add(s2)
399 distincts[s2].add(s1)
400 continue
401 end
402 end
403 end
404
405 # Fixed point algorithm.
406 # * Get 2 states s1 and s2 not yet distinguished.
407 # * Get a symbol w.
408 # * If s1.trans(w) and s2.trans(w) are distinguished, then
409 # distinguish s1 and s2.
410 var changed = true
411 var ints = new Array[Int] # List of symbols to check
412 while changed do
413 changed = false
414 for s1 in states do for s2 in states do
415 if distincts[s1].has(s2) then continue
416
417 # The transitions use intervals. Therefore, for the states s1 and s2,
418 # we need to check only the meaningful symbols. They are the `first`
419 # symbol of each interval and the first one after the interval (`last+1`).
420 ints.clear
421 # Check only `s1`; `s2` will be checked later when s1 and s2 are switched.
422 for t in s1.outs do
423 var sym = t.symbol
424 assert sym != null
425 ints.add sym.first
426 var l = sym.last
427 if l != null then ints.add l + 1
428 end
429
430 # Check each symbol
431 for i in ints do
432 var ds1 = s1.trans(i)
433 var ds2 = s2.trans(i)
434 if ds1 == ds2 then continue
435 if ds1 != null and ds2 != null and not distincts[ds1].has(ds2) then continue
436 distincts[s1].add(s2)
437 distincts[s2].add(s1)
438 changed = true
439 break
440 end
441 end
442 end
443
444 # We need to unify not-distinguished states.
445 # Just add an epsilon-transition and DFAize the automaton.
446 for s1 in states do for s2 in states do
447 if distincts[s1].has(s2) then continue
448 s1.add_trans(s2, null)
449 end
450
451 return to_dfa
452 end
453
454 # Assert that `self` is a valid automaton or abort
455 fun assert_valid
456 do
457 assert states.has(start)
458 assert states.has_all(accept)
459 for s in states do
460 for t in s.outs do assert states.has(t.to)
461 for t in s.ins do assert states.has(t.from)
462 end
463 assert states.has_all(tags.keys)
464 for t, ss in retrotags do
465 assert states.has_all(ss)
466 end
467 end
468
469 # Produce a graphviz string from the automatom
470 #
471 # Set `merge_transitions = false` to generate one edge by transition (default true).
472 fun to_dot(merge_transitions: nullable Bool): Writable
473 do
474 var names = new HashMap[State, String]
475 var ni = 0
476 for s in states do
477 names[s] = ni.to_s
478 ni += 1
479 end
480
481 var f = new Buffer
482 f.append("digraph g \{\n")
483 f.append("rankdir=LR;")
484
485 var state_nb = 0
486 for s in states do
487 f.append("s{names[s]}[shape=circle")
488 #f.write("label=\"\",")
489 if accept.has(s) then
490 f.append(",shape=doublecircle")
491 end
492 if tags.has_key(s) then
493 f.append(",label=\"")
494 for token in tags[s] do
495 f.append("{token.name.escape_to_dot}\\n")
496 end
497 f.append("\"")
498 else
499 f.append(",label=\"{state_nb}\"")
500 end
501 f.append("];\n")
502 var outs = new HashMap[State, Array[nullable TSymbol]]
503 for t in s.outs do
504 var a
505 var s2 = t.to
506 var c = t.symbol
507 if outs.has_key(s2) then
508 a = outs[s2]
509 else
510 a = new Array[nullable TSymbol]
511 outs[s2] = a
512 end
513 a.add(c)
514 end
515 for s2, a in outs do
516 var labe = ""
517 for c in a do
518 if merge_transitions == false then labe = ""
519 if not labe.is_empty then labe += "\n"
520 if c == null then
521 labe += "ε"
522 else
523 labe += c.to_s
524 end
525 if merge_transitions == false then
526 f.append("s{names[s]}->s{names[s2]} [label=\"{labe.escape_to_dot}\"];\n")
527 end
528 end
529 if merge_transitions or else true then
530 f.append("s{names[s]}->s{names[s2]} [label=\"{labe.escape_to_c}\"];\n")
531 end
532 end
533 state_nb += 1
534 end
535 f.append("empty->s{names[start]}; empty[label=\"\",shape=none];\n")
536 f.append("\}\n")
537 return f
538 end
539
540 # Transform a NFA to a DFA.
541 # note: the DFA is not minimized.
542 fun to_dfa: Automaton
543 do
544 assert_valid
545
546 trim
547
548 var dfa = new Automaton.empty
549 var n2d = new ArrayMap[Set[State], State]
550 var seen = new ArraySet[Set[State]]
551 var alphabet = new HashSet[Int]
552 var st = eclosure([start])
553 var todo = [st]
554 n2d[st] = dfa.start
555 seen.add(st)
556 while not todo.is_empty do
557 var nfa_states = todo.pop
558 #print "* work on {nfa_states.inspect}={nfa_states} (remains {todo.length}/{seen.length})"
559 var dfa_state = n2d[nfa_states]
560 alphabet.clear
561 for s in nfa_states do
562 # Collect important values to build the alphabet
563 for t in s.outs do
564 var sym = t.symbol
565 if sym == null then continue
566 alphabet.add(sym.first)
567 var l = sym.last
568 if l != null then alphabet.add(l)
569 end
570
571 # Mark accept and tags
572 if accept.has(s) then
573 if tags.has_key(s) then
574 for t in tags[s] do
575 dfa.add_tag(dfa_state, t)
576 end
577 end
578 dfa.accept.add(dfa_state)
579 end
580 end
581
582 # From the important values, build a sequence of TSymbols
583 var a = alphabet.to_a
584 default_comparator.sort(a)
585 var tsyms = new Array[TSymbol]
586 var last = 0
587 for i in a do
588 if last > 0 and last <= i-1 then
589 tsyms.add(new TSymbol(last,i-1))
590 end
591 tsyms.add(new TSymbol(i,i))
592 last = i+1
593 end
594 if last > 0 then
595 tsyms.add(new TSymbol(last,null))
596 end
597 #print "Alphabet: {tsyms.join(", ")}"
598
599 var lastst: nullable Transition = null
600 for sym in tsyms do
601 var nfa_dest = eclosure(trans(nfa_states, sym.first))
602 if nfa_dest.is_empty then
603 lastst = null
604 continue
605 end
606 #print "{nfa_states} -> {sym} -> {nfa_dest}"
607 var dfa_dest
608 if seen.has(nfa_dest) then
609 #print "* reuse {nfa_dest.inspect}={nfa_dest}"
610 dfa_dest = n2d[nfa_dest]
611 else
612 #print "* new {nfa_dest.inspect}={nfa_dest}"
613 dfa_dest = new State
614 dfa.states.add(dfa_dest)
615 n2d[nfa_dest] = dfa_dest
616 todo.add(nfa_dest)
617 seen.add(nfa_dest)
618 end
619 if lastst != null and lastst.to == dfa_dest then
620 lastst.symbol.as(not null).last = sym.last
621 else
622 lastst = dfa_state.add_trans(dfa_dest, sym)
623 end
624 end
625 end
626 return dfa
627 end
628
629 # Transform a NFA to a epsilonless NFA.
630 fun to_nfa_noe: Automaton
631 do
632 assert_valid
633
634 trim
635
636 var dfa = new Automaton.empty
637 var n2d = new ArrayMap[Set[State], State]
638 var seen = new ArraySet[Set[State]]
639 var st = eclosure([start])
640 var todo = [st]
641 n2d[st] = dfa.start
642 seen.add(st)
643 while not todo.is_empty do
644 var nfa_states = todo.pop
645 #print "* work on {nfa_states.inspect}={nfa_states} (remains {todo.length}/{seen.length})"
646 var dfa_state = n2d[nfa_states]
647 for s in nfa_states do
648 for t in s.outs do
649 if t.symbol == null then continue
650 var nfa_dest = eclosure([t.to])
651 #print "{nfa_states} -> {sym} -> {nfa_dest}"
652 var dfa_dest
653 if seen.has(nfa_dest) then
654 #print "* reuse {nfa_dest.inspect}={nfa_dest}"
655 dfa_dest = n2d[nfa_dest]
656 else
657 #print "* new {nfa_dest.inspect}={nfa_dest}"
658 dfa_dest = new State
659 dfa.states.add(dfa_dest)
660 n2d[nfa_dest] = dfa_dest
661 todo.add(nfa_dest)
662 seen.add(nfa_dest)
663 end
664 dfa_state.add_trans(dfa_dest, t.symbol)
665 end
666
667 # Mark accept and tags
668 if accept.has(s) then
669 if tags.has_key(s) then
670 for t in tags[s] do
671 dfa.add_tag(dfa_state, t)
672 end
673 end
674 dfa.accept.add(dfa_state)
675 end
676 end
677 end
678 return dfa
679 end
680
681 # Epsilon-closure on a state of states.
682 # Used by `to_dfa`.
683 private fun eclosure(states: Collection[State]): Set[State]
684 do
685 var res = new ArraySet[State]
686 res.add_all(states)
687 var todo = states.to_a
688 while not todo.is_empty do
689 var s = todo.pop
690 for t in s.outs do
691 if t.symbol != null then continue
692 var to = t.to
693 if res.has(to) then continue
694 res.add(to)
695 todo.add(to)
696 end
697 end
698 return res
699 end
700
701 # Trans on a set of states.
702 # Used by `to_dfa`.
703 fun trans(states: Collection[State], symbol: Int): Set[State]
704 do
705 var res = new ArraySet[State]
706 for s in states do
707 for t in s.outs do
708 var sym = t.symbol
709 if sym == null then continue
710 if sym.first > symbol then continue
711 var l = sym.last
712 if l != null and l < symbol then continue
713 var to = t.to
714 if res.has(to) then continue
715 res.add(to)
716 end
717 end
718 return res
719 end
720
721 # Generate the Nit source code of the lexer.
722 # `filepath` is the name of the output file.
723 # `parser` is the name of the parser module (used to import the token classes).
724 fun gen_to_nit(filepath: String, name: String, parser: nullable String)
725 do
726 var gen = new DFAGenerator(filepath, name, self, parser)
727 gen.gen_to_nit
728 end
729 end
730
731 # Generate the Nit source code of the lexer
732 private class DFAGenerator
733 var filepath: String
734 var name: String
735 var automaton: Automaton
736 var parser: nullable String
737
738 var out: Writer is noinit
739
740 init do
741 self.out = new FileWriter.open(filepath)
742 end
743
744 fun add(s: String) do out.write(s)
745
746 fun gen_to_nit
747 do
748 var names = new HashMap[State, String]
749 var i = 0
750 for s in automaton.states do
751 names[s] = i.to_s
752 i += 1
753 end
754
755 add "# Lexer generated by nitcc for the grammar {name}\n"
756 add "module {name}_lexer is generated, no_warning \"missing-doc\"\n"
757 add("import nitcc_runtime\n")
758
759 var p = parser
760 if p != null then add("import {p}\n")
761
762 add("class Lexer_{name}\n")
763 add("\tsuper Lexer\n")
764 add("\tredef fun start_state do return dfastate_{names[automaton.start]}\n")
765 add("end\n")
766
767 for s in automaton.states do
768 var n = names[s]
769 add("private fun dfastate_{n}: DFAState{n} do return once new DFAState{n}\n")
770 end
771
772 add("class MyNToken\n")
773 add("\tsuper NToken\n")
774 add("end\n")
775
776 for s in automaton.states do
777 var n = names[s]
778 add("private class DFAState{n}\n")
779 add("\tsuper DFAState\n")
780 if automaton.accept.has(s) then
781 var token
782 if automaton.tags.has_key(s) then
783 token = automaton.tags[s].first
784 else
785 token = null
786 end
787 add("\tredef fun is_accept do return true\n")
788 var is_ignored = false
789 if token != null and token.name == "Ignored" then
790 is_ignored = true
791 add("\tredef fun is_ignored do return true\n")
792 end
793 add("\tredef fun make_token(position, source) do\n")
794 if is_ignored then
795 add("\t\treturn null\n")
796 else
797 if token == null then
798 add("\t\tvar t = new MyNToken\n")
799 add("\t\tt.text = position.extract(source)\n")
800 else
801 add("\t\tvar t = new {token.cname}\n")
802 var ttext = token.text
803 if ttext == null then
804 add("\t\tt.text = position.extract(source)\n")
805 else
806 add("\t\tt.text = \"{ttext.escape_to_nit}\"\n")
807 end
808 end
809 add("\t\tt.position = position\n")
810 add("\t\treturn t\n")
811 end
812 add("\tend\n")
813 end
814 var trans = new ArrayMap[TSymbol, State]
815 for t in s.outs do
816 var sym = t.symbol
817 assert sym != null
818 trans[sym] = t.to
819 end
820 if trans.is_empty then
821 # Do nothing, inherit the trans
822 else
823 add("\tredef fun trans(char) do\n")
824
825 # Collect the sequence of tests in the dispatch sequence
826 # The point here is that for each transition, there is a first and a last
827 # So holes have to be identified
828 var dispatch = new HashMap[Int, nullable State]
829 var haslast: nullable State = null
830
831 var last = -1
832 for sym, next in trans do
833 assert haslast == null
834 assert sym.first > last
835 if sym.first > last + 1 then
836 dispatch[sym.first-1] = null
837 end
838 var l = sym.last
839 if l == null then
840 haslast = next
841 else
842 dispatch[l] = next
843 last = l
844 end
845 end
846
847 if dispatch.is_empty and haslast != null then
848 # Only one transition that accepts everything (quite rare)
849 else
850 # We need to check
851 add("\t\tvar c = char.code_point\n")
852 end
853
854 # Generate a sequence of `if` for the dispatch
855 if haslast != null and last >= 0 then
856 # Special case: handle up-bound first if not an error
857 add("\t\tif c > {last} then return dfastate_{names[haslast]}\n")
858 # previous become the new last case
859 haslast = dispatch[last]
860 dispatch.keys.remove(last)
861 end
862 for c, next in dispatch do
863 if next == null then
864 add("\t\tif c <= {c} then return null\n")
865 else
866 add("\t\tif c <= {c} then return dfastate_{names[next]}\n")
867 end
868 end
869 if haslast == null then
870 add("\t\treturn null\n")
871 else
872 add("\t\treturn dfastate_{names[haslast]}\n")
873 end
874
875 add("\tend\n")
876 end
877 add("end\n")
878 end
879
880 self.out.close
881 end
882 end
883
884 redef class Token
885 # The associated text (if any, ie defined in the parser part)
886 var text: nullable String is noautoinit, writable
887 end
888
889 # A state in a finite automaton
890 class State
891 # Outgoing transitions
892 var outs = new Array[Transition]
893
894 # Ingoing transitions
895 var ins = new Array[Transition]
896
897 # Add a transitions to `to` on `symbol` (null means epsilon)
898 fun add_trans(to: State, symbol: nullable TSymbol): Transition
899 do
900 var t = new Transition(self, to, symbol)
901 outs.add(t)
902 to.ins.add(t)
903 return t
904 end
905
906 # Get the first state following the transition `i`.
907 # Null if no transition for `i`.
908 fun trans(i: Int): nullable State
909 do
910 for t in outs do
911 var sym = t.symbol
912 assert sym != null
913 var f = sym.first
914 var l = sym.last
915 if i < f then continue
916 if l != null and i > l then continue
917 return t.to
918 end
919 return null
920 end
921 end
922
923 # A range of symbols on a transition
924 class TSymbol
925 # The first symbol in the range
926 var first: Int
927
928 # The last symbol if any.
929 #
930 # `null` means infinity.
931 var last: nullable Int
932
933 redef fun to_s
934 do
935 var res
936 var f = first
937 if f <= 32 then
938 res = "#{f}"
939 else
940 res = f.code_point.to_s
941 end
942 var l = last
943 if f == l then return res
944 res += " .. "
945 if l == null then return res
946 if l <= 32 or l >= 127 then return res + "#{l}"
947 return res + l.code_point.to_s
948 end
949 end
950
951 # A transition in a finite automaton
952 class Transition
953 # The source state
954 var from: State
955 # The destination state
956 var to: State
957 # The symbol on the transition (null means epsilon)
958 var symbol: nullable TSymbol
959
960 # Remove the transition from the automaton.
961 # Detach from `from` and `to`.
962 fun delete
963 do
964 from.outs.remove(self)
965 to.ins.remove(self)
966 end
967 end