1 | use core::fmt; |
2 | |
3 | /// A representation of byte oriented equivalence classes. |
4 | /// |
5 | /// This is used in a DFA to reduce the size of the transition table. This can |
6 | /// have a particularly large impact not only on the total size of a dense DFA, |
7 | /// but also on compile times. |
8 | #[derive(Clone, Copy)] |
9 | pub struct ByteClasses([u8; 256]); |
10 | |
11 | impl ByteClasses { |
12 | /// Creates a new set of equivalence classes where all bytes are mapped to |
13 | /// the same class. |
14 | pub fn empty() -> ByteClasses { |
15 | ByteClasses([0; 256]) |
16 | } |
17 | |
18 | /// Creates a new set of equivalence classes where each byte belongs to |
19 | /// its own equivalence class. |
20 | pub fn singletons() -> ByteClasses { |
21 | let mut classes = ByteClasses::empty(); |
22 | for i in 0..256 { |
23 | classes.set(i as u8, i as u8); |
24 | } |
25 | classes |
26 | } |
27 | |
28 | /// Copies the byte classes given. The given slice must have length 0 or |
29 | /// length 256. Slices of length 0 are treated as singletons (every byte |
30 | /// is its own class). |
31 | pub fn from_slice(slice: &[u8]) -> ByteClasses { |
32 | assert!(slice.is_empty() || slice.len() == 256); |
33 | |
34 | if slice.is_empty() { |
35 | ByteClasses::singletons() |
36 | } else { |
37 | let mut classes = ByteClasses::empty(); |
38 | for (b, &class) in slice.iter().enumerate() { |
39 | classes.set(b as u8, class); |
40 | } |
41 | classes |
42 | } |
43 | } |
44 | |
45 | /// Set the equivalence class for the given byte. |
46 | #[inline ] |
47 | pub fn set(&mut self, byte: u8, class: u8) { |
48 | self.0[byte as usize] = class; |
49 | } |
50 | |
51 | /// Get the equivalence class for the given byte. |
52 | #[inline ] |
53 | pub fn get(&self, byte: u8) -> u8 { |
54 | self.0[byte as usize] |
55 | } |
56 | |
57 | /// Get the equivalence class for the given byte while forcefully |
58 | /// eliding bounds checks. |
59 | #[inline ] |
60 | pub unsafe fn get_unchecked(&self, byte: u8) -> u8 { |
61 | *self.0.get_unchecked(byte as usize) |
62 | } |
63 | |
64 | /// Return the total number of elements in the alphabet represented by |
65 | /// these equivalence classes. Equivalently, this returns the total number |
66 | /// of equivalence classes. |
67 | #[inline ] |
68 | pub fn alphabet_len(&self) -> usize { |
69 | self.0[255] as usize + 1 |
70 | } |
71 | |
72 | /// Returns true if and only if every byte in this class maps to its own |
73 | /// equivalence class. Equivalently, there are 256 equivalence classes |
74 | /// and each class contains exactly one byte. |
75 | #[inline ] |
76 | pub fn is_singleton(&self) -> bool { |
77 | self.alphabet_len() == 256 |
78 | } |
79 | |
80 | /// Returns an iterator over a sequence of representative bytes from each |
81 | /// equivalence class. Namely, this yields exactly N items, where N is |
82 | /// equivalent to the number of equivalence classes. Each item is an |
83 | /// arbitrary byte drawn from each equivalence class. |
84 | /// |
85 | /// This is useful when one is determinizing an NFA and the NFA's alphabet |
86 | /// hasn't been converted to equivalence classes yet. Picking an arbitrary |
87 | /// byte from each equivalence class then permits a full exploration of |
88 | /// the NFA instead of using every possible byte value. |
89 | #[cfg (feature = "std" )] |
90 | pub fn representatives(&self) -> ByteClassRepresentatives { |
91 | ByteClassRepresentatives { classes: self, byte: 0, last_class: None } |
92 | } |
93 | |
94 | /// Returns all of the bytes in the given equivalence class. |
95 | /// |
96 | /// The second element in the tuple indicates the number of elements in |
97 | /// the array. |
98 | fn elements(&self, equiv: u8) -> ([u8; 256], usize) { |
99 | let (mut array, mut len) = ([0; 256], 0); |
100 | for b in 0..256 { |
101 | if self.get(b as u8) == equiv { |
102 | array[len] = b as u8; |
103 | len += 1; |
104 | } |
105 | } |
106 | (array, len) |
107 | } |
108 | } |
109 | |
110 | impl fmt::Debug for ByteClasses { |
111 | fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { |
112 | if self.is_singleton() { |
113 | write!(f, "ByteClasses({{singletons}})" ) |
114 | } else { |
115 | write!(f, "ByteClasses(" )?; |
116 | for equiv in 0..self.alphabet_len() { |
117 | let (members, len) = self.elements(equiv as u8); |
118 | write!(f, "{} => {:?}" , equiv, &members[..len])?; |
119 | } |
120 | write!(f, ")" ) |
121 | } |
122 | } |
123 | } |
124 | |
125 | /// An iterator over representative bytes from each equivalence class. |
126 | #[cfg (feature = "std" )] |
127 | #[derive(Debug)] |
128 | pub struct ByteClassRepresentatives<'a> { |
129 | classes: &'a ByteClasses, |
130 | byte: usize, |
131 | last_class: Option<u8>, |
132 | } |
133 | |
134 | #[cfg (feature = "std" )] |
135 | impl<'a> Iterator for ByteClassRepresentatives<'a> { |
136 | type Item = u8; |
137 | |
138 | fn next(&mut self) -> Option<u8> { |
139 | while self.byte < 256 { |
140 | let byte = self.byte as u8; |
141 | let class = self.classes.get(byte); |
142 | self.byte += 1; |
143 | |
144 | if self.last_class != Some(class) { |
145 | self.last_class = Some(class); |
146 | return Some(byte); |
147 | } |
148 | } |
149 | None |
150 | } |
151 | } |
152 | |
153 | /// A byte class set keeps track of an *approximation* of equivalence classes |
154 | /// of bytes during NFA construction. That is, every byte in an equivalence |
155 | /// class cannot discriminate between a match and a non-match. |
156 | /// |
157 | /// For example, in the regex `[ab]+`, the bytes `a` and `b` would be in the |
158 | /// same equivalence class because it never matters whether an `a` or a `b` is |
159 | /// seen, and no combination of `a`s and `b`s in the text can discriminate |
160 | /// a match. |
161 | /// |
162 | /// Note though that this does not compute the minimal set of equivalence |
163 | /// classes. For example, in the regex `[ac]+`, both `a` and `c` are in the |
164 | /// same equivalence class for the same reason that `a` and `b` are in the |
165 | /// same equivalence class in the aforementioned regex. However, in this |
166 | /// implementation, `a` and `c` are put into distinct equivalence classes. |
167 | /// The reason for this is implementation complexity. In the future, we should |
168 | /// endeavor to compute the minimal equivalence classes since they can have a |
169 | /// rather large impact on the size of the DFA. |
170 | /// |
171 | /// The representation here is 256 booleans, all initially set to false. Each |
172 | /// boolean maps to its corresponding byte based on position. A `true` value |
173 | /// indicates the end of an equivalence class, where its corresponding byte |
174 | /// and all of the bytes corresponding to all previous contiguous `false` |
175 | /// values are in the same equivalence class. |
176 | /// |
177 | /// This particular representation only permits contiguous ranges of bytes to |
178 | /// be in the same equivalence class, which means that we can never discover |
179 | /// the true minimal set of equivalence classes. |
180 | #[cfg (feature = "std" )] |
181 | #[derive(Debug)] |
182 | pub struct ByteClassSet(Vec<bool>); |
183 | |
184 | #[cfg (feature = "std" )] |
185 | impl ByteClassSet { |
186 | /// Create a new set of byte classes where all bytes are part of the same |
187 | /// equivalence class. |
188 | pub fn new() -> Self { |
189 | ByteClassSet(vec![false; 256]) |
190 | } |
191 | |
192 | /// Indicate the the range of byte given (inclusive) can discriminate a |
193 | /// match between it and all other bytes outside of the range. |
194 | pub fn set_range(&mut self, start: u8, end: u8) { |
195 | debug_assert!(start <= end); |
196 | if start > 0 { |
197 | self.0[start as usize - 1] = true; |
198 | } |
199 | self.0[end as usize] = true; |
200 | } |
201 | |
202 | /// Convert this boolean set to a map that maps all byte values to their |
203 | /// corresponding equivalence class. The last mapping indicates the largest |
204 | /// equivalence class identifier (which is never bigger than 255). |
205 | pub fn byte_classes(&self) -> ByteClasses { |
206 | let mut classes = ByteClasses::empty(); |
207 | let mut class = 0u8; |
208 | let mut i = 0; |
209 | loop { |
210 | classes.set(i as u8, class as u8); |
211 | if i >= 255 { |
212 | break; |
213 | } |
214 | if self.0[i] { |
215 | class = class.checked_add(1).unwrap(); |
216 | } |
217 | i += 1; |
218 | } |
219 | classes |
220 | } |
221 | } |
222 | |
223 | #[cfg (test)] |
224 | mod tests { |
225 | #[cfg (feature = "std" )] |
226 | #[test] |
227 | fn byte_classes() { |
228 | use super::ByteClassSet; |
229 | |
230 | let mut set = ByteClassSet::new(); |
231 | set.set_range(b'a' , b'z' ); |
232 | |
233 | let classes = set.byte_classes(); |
234 | assert_eq!(classes.get(0), 0); |
235 | assert_eq!(classes.get(1), 0); |
236 | assert_eq!(classes.get(2), 0); |
237 | assert_eq!(classes.get(b'a' - 1), 0); |
238 | assert_eq!(classes.get(b'a' ), 1); |
239 | assert_eq!(classes.get(b'm' ), 1); |
240 | assert_eq!(classes.get(b'z' ), 1); |
241 | assert_eq!(classes.get(b'z' + 1), 2); |
242 | assert_eq!(classes.get(254), 2); |
243 | assert_eq!(classes.get(255), 2); |
244 | |
245 | let mut set = ByteClassSet::new(); |
246 | set.set_range(0, 2); |
247 | set.set_range(4, 6); |
248 | let classes = set.byte_classes(); |
249 | assert_eq!(classes.get(0), 0); |
250 | assert_eq!(classes.get(1), 0); |
251 | assert_eq!(classes.get(2), 0); |
252 | assert_eq!(classes.get(3), 1); |
253 | assert_eq!(classes.get(4), 2); |
254 | assert_eq!(classes.get(5), 2); |
255 | assert_eq!(classes.get(6), 2); |
256 | assert_eq!(classes.get(7), 3); |
257 | assert_eq!(classes.get(255), 3); |
258 | } |
259 | |
260 | #[cfg (feature = "std" )] |
261 | #[test] |
262 | fn full_byte_classes() { |
263 | use super::ByteClassSet; |
264 | |
265 | let mut set = ByteClassSet::new(); |
266 | for i in 0..256u16 { |
267 | set.set_range(i as u8, i as u8); |
268 | } |
269 | assert_eq!(set.byte_classes().alphabet_len(), 256); |
270 | } |
271 | } |
272 | |