Regular-Expresso Changeset

Changeset - 61a2b8f09823

Parent rev.

Child rev.

[Not reviewed]

default

0 3 0

Laman - 10 months ago 2024-06-24 13:47:20

refactoring: states represented by basic integers

3 files changed with 40 insertions and 51 deletions:

regexp.py

src/main.rs

src/regexp.rs

0 comments (0 inline, 0 general)

regexp.py

➞

Show inline comments

@@ @@ -153,220 +153,221 @@ class Chain(Token): @@
 def find_closing_parenthesis(pattern, k):
 	counter = 0
 	for (i, c) in enumerate(pattern[k:]):
 		if c == "(":
 			counter += 1
 		elif c == ")":
 			counter -= 1
 		if counter == 0:
 			return k+i
 	raise ParsingError(f'A closing parenthesis not found. Pattern: "{pattern}", position: {k}')
 def parse(pattern, offset=0):
 	res = []
 	is_alternative = False
 	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 == "*":
 			try:
 				token = res.pop()
 			except IndexError as e:
 				raise ParsingError(f'The asterisk operator is missing an argument. Pattern: "{pattern}", position {i}')
 			res.append(Asterisk(token))
 			i += 1
 		elif c == ")":
 			raise ParsingError(f'An opening parenthesis not found. Pattern: "{pattern}", position: {i}')
 		elif c == "|" or c == "+":
 			is_alternative = True
 			res.append(AlternativeSeparator())
 			i += 1
 		elif c == "_":
 			res.append(Lambda())
 			i += 1
 		else:
 			res.append(Symbol(i+offset, c))
 			i += 1
 	if is_alternative:
 		return Alternative(res)
 	else:
 		return Chain(res)
 class Regexp:
 	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()
 		end_states = {-1} if -1 in self.end_states else set()
 		index = {(-1,): -1}
 		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)
 				target_tup = tuple(sorted(target_set))
 				if target_tup not in index:
 					new_target = len(index)-1
 					index[target_tup] = new_target
 					stack.append(target_tup)
 				rules[(index[multistate], c)] = index[target_tup]
 				if any(st in self.end_states for st in target_set):
 					end_states.add(index[target_tup])
 		return (rules, end_states)
 class RegexpDFA:
 	def __init__(self, rules, end_states):
 		self.rules = rules
 		self.end_states = end_states
 	@classmethod
 	def create(cls, pattern):
 		r = Regexp(pattern)
 		(rules, end_states) = r.determinize()
 		return cls(rules, end_states)
 	def match(self, s):
-		st = 0
+		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
 	def reduce(self):
 		equivalents = self._find_equivalent_states()
 		(rules, end_states) = self._collapse_states(equivalents)
 		return RegexpDFA(rules, end_states)
 	def normalize(self):
 		index = {(-1,): 0}
 		queue = deque([(-1,)])
 		index = {-1: -1}
 		queue = deque([-1])
 		while queue:
 			state = queue.popleft()
 			edges = [((st, c), t) for ((st, c), t) in self.rules.items() if st == state]
 			edges.sort()
 			for ((st, c), t) in edges:
 				if t not in index:
 					index[t] = len(index)
+					index[t] = len(index)-1
 					queue.append(t)
 		rules = dict(((index[st], c), index[t]) for ((st, c), t) in self.rules.items())
 		end_states = {index[st] for st in self.end_states}
 		return RegexpDFA(rules, end_states)
 	def _find_equivalent_states(self):
-		state_set = [(-2,), (-1,)] + sorted(set(self.rules.values()))
 		state_set = [-2, -1] + sorted(set(self.rules.values()))
 		alphabet = {c for (st, c) in self.rules.keys()}
 		equivalents = {(s1, s2) for (i, s1) in enumerate(state_set) for s2 in state_set[i+1:]}
 		for (s1, s2) in equivalents.copy():
 			if (s1 in self.end_states and s2 not in self.end_states) or (s1 not in self.end_states and s2 in self.end_states):
 				equivalents.remove((s1, s2))
 		ctrl = True
 		while ctrl:
 			ctrl = False
 			for (s1, s2) in equivalents.copy():
 				for c in alphabet:
 					t1 = self.rules.get((s1, c), (-2,))
 					t2 = self.rules.get((s2, c), (-2,))
 					t1 = self.rules.get((s1, c), -2)
 					t2 = self.rules.get((s2, c), -2)
 					key = (min(t1, t2), max(t1, t2))
 					if t1 != t2 and key not in equivalents:
 						equivalents.remove((s1, s2))
 						ctrl = True
 						break
 		return equivalents
 	def _collapse_states(self, equivalents):
 		rules = self.rules.items()
 		end_states = self.end_states.copy()
 		for (s1, s2) in sorted(equivalents):
 			rules = map(
 				lambda item: (item[0], s1 if item[1] == s2 else item[1]),
 				filter(lambda item: item[0][0] != s2, rules)
+			)
 			end_states.discard(s2)
 		return (dict(rules), end_states)
 if __name__ == "__main__":
 	tests = ["", "a", "ab", "aabb", "abab", "abcd", "abcbcdbcd"]
 	for pattern in ["a(b|c)", "a*b*", "(ab)*", "a((bc)*d)*", "(a|b)*a(a|b)(a|b)(a|b)"]:
+	for pattern in ["a(b|c)", "a*b*", "(ab)*", "a((bc)*d)*", "(a|b)*a(a|b)(a|b)(a|b)", "abcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyz"]:
 		print("#", pattern)
 		try:
 			r = RegexpDFA.create(pattern).reduce().normalize()
 		except ParsingError as e:
 			print("Failed to parse the regexp:")
 			print(e)
 			continue
 		for t in tests:
 			print(t, r.match(t))
 		print()

src/main.rs

➞

Show inline comments

 use regexp::Regexp;
 fn main() {
 	let tests = ["", "a", "ab", "aabb", "abab", "abcd", "abcbcdbcd"];
 	for pattern in ["a(b|c)", "a*b*", "(ab)*", "a((bc)*d)*", "(a|b)*a(a|b)(a|b)(a|b)"] {
+	for pattern in ["a(b|c)", "a*b*", "(ab)*", "a((bc)*d)*", "(a|b)*a(a|b)(a|b)(a|b)", "abcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyz"] {
 		println!("# {pattern}");
 		let r = match Regexp::new(&pattern.to_string()) {
 			Ok(r1) => r1.determinize().reduce().normalize(),
 			Err(e) => {
 				println!("{e}");
 				continue;
+			}
 		};
 		for &t in tests.iter() {
 			println!("{t} {}", r.eval(t.to_string()));
+		}
 		println!();
+	}
+}

src/regexp.rs

➞

Show inline comments

 use std::collections::{HashMap, HashSet, VecDeque};
 mod token;
 pub use token::ParsingError;
 use token::parse;
 const START: i32 = -1;
 const FAIL: i32 = -2;
 fn encode_set(set: &HashSet<i32>) -> i32 {
 	let mut res = 0;
 	for x in set.iter() {
 		assert!(*x >= 0);
 		res ^= 1<<x;
+	}
 	return res;
+}
 fn decode_set(x: i32) -> HashSet<i32> {
 	if x == START {return HashSet::from([START]);}
 	let mut x = x;
 	let mut res: HashSet<i32> = HashSet::new();
 	while x > 0 {
 		let y = x.trailing_zeros();
 		res.insert(y as i32);
 		x ^= 1 << y;
+	}
 	return res;
 fn encode_set(set: &HashSet<i32>) -> String {
 	let mut v = Vec::from_iter(set.iter());
 	v.sort();
 	let res: Vec<String> = v.into_iter().map(|x| x.to_string()).collect();
 	return res.join(",");
+}
 #[derive(Debug)]
 pub struct Regexp {
 	rules: HashMap<(i32, char), HashSet<i32>>,
 	end_states: HashSet<i32>
+}
 impl Regexp {
 	pub fn new(pattern: &String) -> Result<Regexp, ParsingError> {
 		let r = parse(pattern, 0)?;
 		let pattern_chars = Vec::from_iter(pattern.chars());
 		let mut rules: HashMap<(i32, char), HashSet<i32>> = 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 as i32);},
 				None => {rules.insert(key, HashSet::from([i as i32]));}
 			};
+		}
 		for (i, j) in r.list_neighbours() {
 			let c = pattern_chars[j];
 			let key = (i as i32, c);
 			match rules.get_mut(&key) {
 				Some(set) => {set.insert(j as i32);},
 				None => {rules.insert(key, HashSet::from([j as i32]));}
 			};
+		}
 		let mut end_states = HashSet::from_iter(r.list_last().into_iter().map(|i| i as i32));
 		if r.is_skippable() {
 			end_states.insert(START);
+		}
 		return Ok(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));
+	}
 	pub fn determinize(&self) -> RegexpDFA {
 		let mut rules: HashMap<(i32, char), i32> = HashMap::new();
 		let mut end_states: HashSet<i32> = HashSet::new();
 		if self.end_states.contains(&START) {end_states.insert(START);}
 		let mut stack = Vec::from([START]);
 		let mut processed_states = HashSet::new();
 		let mut index_new = HashMap::from([(START.to_string(), START)]);
 		let mut index_multi = HashMap::from([(START.to_string(), HashSet::from([START]))]);
 		let mut stack = Vec::from([START.to_string()]);
 		while !stack.is_empty() {
 			let state = stack.pop().unwrap();
 			let multistate = decode_set(state);
 			let state_hash = stack.pop().unwrap();
 			let multistate = &index_multi[&state_hash];
 			let mut new_rules: HashMap<char, HashSet<i32>> = 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);
 				let target_hash = encode_set(&target_set);
 				let is_end = target_set.iter().any(|st| self.end_states.contains(st));
 				if !index_new.contains_key(&target_hash) {
 					let target_new = index_new.len() as i32;
 					index_new.insert(target_hash.clone(), target_new);
 					index_multi.insert(target_hash.clone(), target_set);
 					stack.push(target_hash.clone());
+				}
 				if !processed_states.contains(&encoded_target) {
 					stack.push(encoded_target);
 					processed_states.insert(encoded_target);
 				rules.insert((index_new[&state_hash], c), index_new[&target_hash]);
 				if is_end {
 					end_states.insert(index_new[&target_hash]);
+				}
+			}
+		}
 		return RegexpDFA{rules, end_states};
+	}
+}
 pub struct RegexpDFA {
 	rules: HashMap<(i32, char), i32>,
 	end_states: HashSet<i32>
+}
 impl RegexpDFA {
 	pub fn eval(&self, s: String) -> bool {
 		let mut state = START;
 		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);
+	}
 	pub fn reduce(&self) -> RegexpDFA {
 		let equivalents = self.find_equivalent_states();
 		return self.collapse_states(equivalents);
+	}
 	pub fn normalize(&self) -> RegexpDFA {
 		let mut index = HashMap::from([(START, START)]);
 		let mut queue = VecDeque::from([START]);
 		while !queue.is_empty() {
 			let state = queue.pop_front().unwrap();
 			let mut edges: Vec<((i32, char), i32)> = self.rules.iter()
 				.filter(|((st, c), t)| *st == state)
 				.map(|((st, c), t)| ((*st, *c), *t)).collect();
 			edges.sort();
 			for ((_st, _c), t) in edges {
 				if !index.contains_key(&t) {
 					index.insert(t, index.len() as i32);
 					queue.push_back(t);
+				}
+			}
+		}
 		let rules = self.rules.iter().map(|((st, c), t)| ((index[st], *c), index[t])).collect();
 		let end_states = self.end_states.iter().map(|st| index[st]).collect();
 		return RegexpDFA{rules, end_states};
+	}
 	fn find_equivalent_states(&self) -> Vec<(i32, i32)> {
 		let state_set: HashSet<i32> = HashSet::from_iter(self.rules.values().copied());
 		let mut state_vec: Vec<i32> = Vec::from_iter(state_set.into_iter());
 		state_vec.push(START);
 		state_vec.push(FAIL);
 		state_vec.sort();
 		let alphabet: HashSet<char> = self.rules.keys().map(|(_st, c)| c).copied().collect();
 		let mut equivalents = HashSet::new();
 		state_vec.iter().enumerate().for_each(|(i, s1)| {
 			equivalents.extend(
 				state_vec[i+1..].iter()
 				.filter(|s2| !(self.end_states.contains(s1)^self.end_states.contains(s2)))
 				.map(|s2| (*s1, *s2))
 			);
 		});
 		let mut n = usize::MAX;
 		while equivalents.len() < n {
 			n = equivalents.len();
 			equivalents = equivalents.iter().filter(|(s1, s2)| {
 				!alphabet.iter().any(|c| {
 					let t1 = self.rules.get(&(*s1, *c)).unwrap_or(&FAIL);
 					let t2 = self.rules.get(&(*s2, *c)).unwrap_or(&FAIL);
 					let key = (*t1.min(t2), *t1.max(t2));
 					return t1 != t2 && !equivalents.contains(&key);
 				})
 			}).copied().collect();
+		}
 		return Vec::from_iter(equivalents.into_iter());
+	}
 	fn collapse_states(&self, mut equivalents: Vec<(i32, i32)>) -> RegexpDFA {
 		let mut rules = self.rules.clone();
 		let mut end_states = self.end_states.clone();
 		equivalents.sort();
 		for (s1, s2) in equivalents.into_iter() {

0 comments (0 inline, 0 general)