Regular-Expresso Changeset

Changeset - c827cb147cf5

Parent rev.

Child rev.

[Not reviewed]

default

0 2 0

Laman - 10 months ago 2024-06-16 16:04:44

added the finite automaton determinization

2 files changed with 130 insertions and 4 deletions:

regexp.py

src/main.rs

0 comments (0 inline, 0 general)

regexp.py

➞

Show inline comments

@@ @@ -120,73 +120,117 @@ def parse(pattern, offset=0): @@
 	i = 0
 	while i < len(pattern):
 		c = pattern[i]
 		if c == "(":
 			j = find_closing_parenthesis(pattern, i)
 			inner_content = parse(pattern[i+1:j], offset+i+1)
 			res.append(inner_content)
 			i = j+1
 		elif c == "*":
 			token = res.pop()
 			res.append(Asterisk(token))
 			i += 1
 		elif c == "+":
 			token = res.pop()
 			res.append(Plus(token))
 			i += 1
 		else:
 			res.append(Symbol(i+offset, c))
 			i += 1
 	return Chain(res)
 class Regexp:
 	def __init__(self, s):
 		(self.rules, self.end_states) = self._parse(s)
 	def __init__(self, pattern):
 		(self.rules, self.end_states) = self._parse(pattern)
 	def _parse(self, s):
 		r = parse(s)
 		rules = dict()
 		for i in r.list_first():
 			c = s[i]
 			key = (-1, c)
 			if key not in rules:
 				rules[key] = set()
 			rules[key].add(i)
 		for (i, j) in r.list_neighbours():
 			c = s[j]
 			key = (i, c)
 			if key not in rules:
 				rules[key] = set()
 			rules[key].add(j)
 		end_states = set(r.list_last())
 		if r.is_skippable:
 			end_states.add(-1)
 		return rules, end_states
 	def match(self, s):
 		current = {-1}
 		for c in s:
 			new_state = set()
 			for st in current:
 				key = (st, c)
 				if key in self.rules:
 					new_state.update(self.rules[key])
 			current = new_state
 		return any(st in self.end_states for st in current)
 	def determinize(self):
 		rules = dict()
 		end_states = {(-1,)} if -1 in self.end_states else set()
 		stack = [(-1,)]
 		processed_states = set()
 		while stack:
 			multistate = stack.pop()
 			new_rules = dict()
 			for ((st, c), target) in filter(lambda item: item[0][0] in multistate, self.rules.items()):
 				if c not in new_rules:
 					new_rules[c] = set()
 				new_rules[c].update(target)
 			for (c, target_set) in new_rules.items():
 				new_target = tuple(sorted(target_set))
 				rules[(multistate, c)] = new_target
 				if any(st in self.end_states for st in new_target):
 					end_states.add(new_target)
 				if new_target not in processed_states:
 					stack.append(new_target)
 					processed_states.add(new_target)
 		return (rules, end_states)
 class RegexpDFA:
 	def __init__(self, pattern):
 		r = Regexp(pattern)
 		(self.rules, self.end_states) = r.determinize()
 	def match(self, s):
 		st = (-1,)
 		for c in s:
 			key = (st, c)
 			if key in self.rules:
 				st = self.rules[key]
 			else:
 				return False
 		return st in self.end_states
 if __name__ == "__main__":
 	tests = ["", "a", "ab", "aabb", "abab", "abcd", "abcbcdbcd"]
 	for pattern in ["a*b*", "a+b+", "(ab)*", "(ab)+", "a((bc)*d)*"]:
 		print("#", pattern)
-		r = Regexp(pattern)
+		r = RegexpDFA(pattern)
 		for t in tests:
 			print(t, r.match(t))
 		print()

src/main.rs

➞

Show inline comments

@@ @@ -160,48 +160,71 @@ fn parse(pattern: &String, offset: usize @@
 				let inner_content = parse(&pattern[i+1..j].to_string(), offset+i+1);
 				res.push(Box::new(inner_content));
 				i = j+1;
+			}
 			'*' => {
 				let token = res.pop().unwrap();
 				res.push(Box::new(Asterisk{content: token}));
 				i += 1;
+			}
 			'+' => {
 				let token = res.pop().unwrap();
 				res.push(Box::new(Plus{content: token}));
 				i += 1;
+			}
 			c => {
 				res.push(Box::new(Symbol{position: i+offset, value: c}));
 				i += 1;
+			}
+		}
+	}
 	return Chain{content: res};
+}
 fn encode_set(set: &HashSet<usize>) -> u64 {
 	let mut res = 0;
 	for x in set.iter() {
 		res ^= 1<<x;
+	}
 	return res;
+}
 fn decode_set(x: u64) ->HashSet<usize> {
 	if x == START as u64 {return HashSet::from([START]);}
 	let mut x = x;
 	let mut res: HashSet<usize> = HashSet::new();
 	while x > 0 {
 		let y = x.trailing_zeros();
 		res.insert(y as usize);
 		x ^= 1 << y;
+	}
 	return res;
+}
 struct Regexp {
 	rules: HashMap<(usize, char), HashSet<usize>>,
 	end_states: HashSet<usize>
+}
 impl Regexp {
 	fn new(pattern: &String) -> Regexp {
 		let r = parse(pattern, 0);
 		let pattern_chars = Vec::from_iter(pattern.chars());
 		let mut rules: HashMap<(usize, char), HashSet<usize>> = HashMap::new();
 		for i in r.list_first() {
 			let c = pattern_chars[i];
 			let key = (START, c);
 			match rules.get_mut(&key) {
 				Some(set) => {set.insert(i);},
 				None => {rules.insert(key, HashSet::from([i]));}
 			};
+		}
 		for (i, j) in r.list_neighbours() {
 			let c = pattern_chars[j];
 			let key = (i, c);
 			match rules.get_mut(&key) {
@@ @@ -215,37 +238,96 @@ impl Regexp { @@
 			end_states.insert(START);
+		}
 		return Regexp{rules, end_states};
+	}
 	pub fn eval(&self, s: String) -> bool {
 		let mut multistate = HashSet::from([START]);
 		for c in s.chars() {
 			let mut new_multistate = HashSet::new();
 			for state in multistate {
 				if let Some(x) = self.rules.get(&(state, c)) {
 					new_multistate = new_multistate.union(&x).map(|&y| y).collect();
 				} else if let Some(x) = self.rules.get(&(state, '.')) {
 					new_multistate = new_multistate.union(&x).map(|&y| y).collect();
+				}
+			}
 			multistate = new_multistate;
+		}
 		return multistate.iter().any(|x| self.end_states.contains(x));
+	}
 	fn determinize(&self) -> RegexpDFA {
 		let mut rules: HashMap<(u64, char), u64> = HashMap::new();
 		let mut end_states: HashSet<u64> = HashSet::new();
 		if self.end_states.contains(&START) {end_states.insert(START as u64);}
 		let mut stack = Vec::from([START as u64]);
 		let mut processed_states = HashSet::new();
 		while !stack.is_empty() {
 			let state = stack.pop().unwrap();
 			let multistate = decode_set(state);
 			let mut new_rules: HashMap<char, HashSet<usize>> = HashMap::new();
 			for key in self.rules.keys().filter(|key| multistate.contains(&key.0)) {
 				let (_st, c) = key;
 				if !new_rules.contains_key(c) {
 					new_rules.insert(*c, HashSet::new());
+				}
 				for target in &self.rules[key] {
 					new_rules.get_mut(c).unwrap().insert(*target);
+				}
+			}
 			for (c, target_set) in new_rules.into_iter() {
 				let encoded_target = encode_set(&target_set);
 				rules.insert((state, c), encoded_target);
 				if target_set.iter().any(|st| self.end_states.contains(st)) {
 					end_states.insert(encoded_target);
+				}
 				if !processed_states.contains(&encoded_target) {
 					stack.push(encoded_target);
 					processed_states.insert(encoded_target);
+				}
+			}
+		}
 		return RegexpDFA{rules, end_states};
+	}
+}
 struct RegexpDFA {
 	rules: HashMap<(u64, char), u64>,
 	end_states: HashSet<u64>
+}
 impl RegexpDFA {
 	fn eval(&self, s: String) -> bool {
 		let mut state = START as u64;
 		for c in s.chars() {
 			if let Some(x) = self.rules.get(&(state, c)) {
 				state = *x;
 			} else {
 				return false;
+			}
+		}
 		return self.end_states.contains(&state);
+	}
+}
 fn main() {
 	let tests = ["", "a", "ab", "aabb", "abab", "abcd", "abcbcdbcd"];
 	for pattern in ["a*b*", "a+b+", "(ab)*", "(ab)+", "a((bc)*d)*"] {
 		println!("# {pattern}");
 		let r = Regexp::new(&pattern.to_string());
+		let r = Regexp::new(&pattern.to_string()).determinize();
 		for &t in tests.iter() {
 			println!("{t} {}", r.eval(t.to_string()));
+		}
 		println!();
+	}
+}

0 comments (0 inline, 0 general)