1 | use alloc::vec::Vec; |
2 | |
3 | use crate::util::primitives::StateID; |
4 | |
5 | /// Remappable is a tightly coupled abstraction that facilitates remapping |
6 | /// state identifiers in DFAs. |
7 | /// |
8 | /// The main idea behind remapping state IDs is that DFAs often need to check |
9 | /// if a certain state is a "special" state of some kind (like a match state) |
10 | /// during a search. Since this is extremely perf critical code, we want this |
11 | /// check to be as fast as possible. Partitioning state IDs into, for example, |
12 | /// into "non-match" and "match" states means one can tell if a state is a |
13 | /// match state via a simple comparison of the state ID. |
14 | /// |
15 | /// The issue is that during the DFA construction process, it's not |
16 | /// particularly easy to partition the states. Instead, the simplest thing is |
17 | /// to often just do a pass over all of the states and shuffle them into their |
18 | /// desired partitionings. To do that, we need a mechanism for swapping states. |
19 | /// Hence, this abstraction. |
20 | /// |
21 | /// Normally, for such little code, I would just duplicate it. But this is a |
22 | /// key optimization and the implementation is a bit subtle. So the abstraction |
23 | /// is basically a ham-fisted attempt at DRY. The only place we use this is in |
24 | /// the dense and one-pass DFAs. |
25 | /// |
26 | /// See also src/dfa/special.rs for a more detailed explanation of how dense |
27 | /// DFAs are partitioned. |
28 | pub(super) trait Remappable: core::fmt::Debug { |
29 | /// Return the total number of states. |
30 | fn state_len(&self) -> usize; |
31 | /// Return the power-of-2 exponent that yields the stride. The pertinent |
32 | /// laws here are, where N=stride2: 2^N=stride and len(alphabet) <= stride. |
33 | fn stride2(&self) -> usize; |
34 | /// Swap the states pointed to by the given IDs. The underlying finite |
35 | /// state machine should be mutated such that all of the transitions in |
36 | /// `id1` are now in the memory region where the transitions for `id2` |
37 | /// were, and all of the transitions in `id2` are now in the memory region |
38 | /// where the transitions for `id1` were. |
39 | /// |
40 | /// Essentially, this "moves" `id1` to `id2` and `id2` to `id1`. |
41 | /// |
42 | /// It is expected that, after calling this, the underlying value will be |
43 | /// left in an inconsistent state, since any other transitions pointing to, |
44 | /// e.g., `id1` need to be updated to point to `id2`, since that's where |
45 | /// `id1` moved to. |
46 | /// |
47 | /// In order to "fix" the underlying inconsistent state, a `Remapper` |
48 | /// should be used to guarantee that `remap` is called at the appropriate |
49 | /// time. |
50 | fn swap_states(&mut self, id1: StateID, id2: StateID); |
51 | /// This must remap every single state ID in the underlying value according |
52 | /// to the function given. For example, in a DFA, this should remap every |
53 | /// transition and every starting state ID. |
54 | fn remap(&mut self, map: impl Fn(StateID) -> StateID); |
55 | } |
56 | |
57 | /// Remapper is an abstraction the manages the remapping of state IDs in a |
58 | /// finite state machine. This is useful when one wants to shuffle states into |
59 | /// different positions in the machine. |
60 | /// |
61 | /// One of the key complexities this manages is the ability to correctly move |
62 | /// one state multiple times. |
63 | /// |
64 | /// Once shuffling is complete, `remap` must be called, which will rewrite |
65 | /// all pertinent transitions to updated state IDs. Neglecting to call `remap` |
66 | /// will almost certainly result in a corrupt machine. |
67 | #[derive(Debug)] |
68 | pub(super) struct Remapper { |
69 | /// A map from the index of a state to its pre-multiplied identifier. |
70 | /// |
71 | /// When a state is swapped with another, then their corresponding |
72 | /// locations in this map are also swapped. Thus, its new position will |
73 | /// still point to its old pre-multiplied StateID. |
74 | /// |
75 | /// While there is a bit more to it, this then allows us to rewrite the |
76 | /// state IDs in a DFA's transition table in a single pass. This is done |
77 | /// by iterating over every ID in this map, then iterating over each |
78 | /// transition for the state at that ID and re-mapping the transition from |
79 | /// `old_id` to `map[dfa.to_index(old_id)]`. That is, we find the position |
80 | /// in this map where `old_id` *started*, and set it to where it ended up |
81 | /// after all swaps have been completed. |
82 | map: Vec<StateID>, |
83 | /// A mapper from state index to state ID (and back). |
84 | idxmap: IndexMapper, |
85 | } |
86 | |
87 | impl Remapper { |
88 | /// Create a new remapper from the given remappable implementation. The |
89 | /// remapper can then be used to swap states. The remappable value given |
90 | /// here must the same one given to `swap` and `remap`. |
91 | pub(super) fn new(r: &impl Remappable) -> Remapper { |
92 | let idxmap = IndexMapper { stride2: r.stride2() }; |
93 | let map = (0..r.state_len()).map(|i| idxmap.to_state_id(i)).collect(); |
94 | Remapper { map, idxmap } |
95 | } |
96 | |
97 | /// Swap two states. Once this is called, callers must follow through to |
98 | /// call `remap`, or else it's possible for the underlying remappable |
99 | /// value to be in a corrupt state. |
100 | pub(super) fn swap( |
101 | &mut self, |
102 | r: &mut impl Remappable, |
103 | id1: StateID, |
104 | id2: StateID, |
105 | ) { |
106 | if id1 == id2 { |
107 | return; |
108 | } |
109 | r.swap_states(id1, id2); |
110 | self.map.swap(self.idxmap.to_index(id1), self.idxmap.to_index(id2)); |
111 | } |
112 | |
113 | /// Complete the remapping process by rewriting all state IDs in the |
114 | /// remappable value according to the swaps performed. |
115 | pub(super) fn remap(mut self, r: &mut impl Remappable) { |
116 | // Update the map to account for states that have been swapped |
117 | // multiple times. For example, if (A, C) and (C, G) are swapped, then |
118 | // transitions previously pointing to A should now point to G. But if |
119 | // we don't update our map, they will erroneously be set to C. All we |
120 | // do is follow the swaps in our map until we see our original state |
121 | // ID. |
122 | // |
123 | // The intuition here is to think about how changes are made to the |
124 | // map: only through pairwise swaps. That means that starting at any |
125 | // given state, it is always possible to find the loop back to that |
126 | // state by following the swaps represented in the map (which might be |
127 | // 0 swaps). |
128 | // |
129 | // We are also careful to clone the map before starting in order to |
130 | // freeze it. We use the frozen map to find our loops, since we need to |
131 | // update our map as well. Without freezing it, our updates could break |
132 | // the loops referenced above and produce incorrect results. |
133 | let oldmap = self.map.clone(); |
134 | for i in 0..r.state_len() { |
135 | let cur_id = self.idxmap.to_state_id(i); |
136 | let mut new_id = oldmap[i]; |
137 | if cur_id == new_id { |
138 | continue; |
139 | } |
140 | loop { |
141 | let id = oldmap[self.idxmap.to_index(new_id)]; |
142 | if cur_id == id { |
143 | self.map[i] = new_id; |
144 | break; |
145 | } |
146 | new_id = id; |
147 | } |
148 | } |
149 | r.remap(|next| self.map[self.idxmap.to_index(next)]); |
150 | } |
151 | } |
152 | |
153 | /// A simple type for mapping between state indices and state IDs. |
154 | /// |
155 | /// The reason why this exists is because state IDs are "premultiplied." That |
156 | /// is, in order to get to the transitions for a particular state, one need |
157 | /// only use the state ID as-is, instead of having to multiple it by transition |
158 | /// table's stride. |
159 | /// |
160 | /// The downside of this is that it's inconvenient to map between state IDs |
161 | /// using a dense map, e.g., Vec<StateID>. That's because state IDs look like |
162 | /// `0`, `0+stride`, `0+2*stride`, `0+3*stride`, etc., instead of `0`, `1`, |
163 | /// `2`, `3`, etc. |
164 | /// |
165 | /// Since our state IDs are premultiplied, we can convert back-and-forth |
166 | /// between IDs and indices by simply unmultiplying the IDs and multiplying the |
167 | /// indices. |
168 | #[derive(Debug)] |
169 | struct IndexMapper { |
170 | /// The power of 2 corresponding to the stride of the corresponding |
171 | /// transition table. 'id >> stride2' de-multiplies an ID while 'index << |
172 | /// stride2' pre-multiplies an index to an ID. |
173 | stride2: usize, |
174 | } |
175 | |
176 | impl IndexMapper { |
177 | /// Convert a state ID to a state index. |
178 | fn to_index(&self, id: StateID) -> usize { |
179 | id.as_usize() >> self.stride2 |
180 | } |
181 | |
182 | /// Convert a state index to a state ID. |
183 | fn to_state_id(&self, index: usize) -> StateID { |
184 | // CORRECTNESS: If the given index is not valid, then it is not |
185 | // required for this to panic or return a valid state ID. We'll "just" |
186 | // wind up with panics or silent logic errors at some other point. |
187 | StateID::new_unchecked(index << self.stride2) |
188 | } |
189 | } |
190 | |
191 | #[cfg (feature = "dfa-build" )] |
192 | mod dense { |
193 | use crate::{dfa::dense::OwnedDFA, util::primitives::StateID}; |
194 | |
195 | use super::Remappable; |
196 | |
197 | impl Remappable for OwnedDFA { |
198 | fn state_len(&self) -> usize { |
199 | OwnedDFA::state_len(self) |
200 | } |
201 | |
202 | fn stride2(&self) -> usize { |
203 | OwnedDFA::stride2(self) |
204 | } |
205 | |
206 | fn swap_states(&mut self, id1: StateID, id2: StateID) { |
207 | OwnedDFA::swap_states(self, id1, id2) |
208 | } |
209 | |
210 | fn remap(&mut self, map: impl Fn(StateID) -> StateID) { |
211 | OwnedDFA::remap(self, map) |
212 | } |
213 | } |
214 | } |
215 | |
216 | #[cfg (feature = "dfa-onepass" )] |
217 | mod onepass { |
218 | use crate::{dfa::onepass::DFA, util::primitives::StateID}; |
219 | |
220 | use super::Remappable; |
221 | |
222 | impl Remappable for DFA { |
223 | fn state_len(&self) -> usize { |
224 | DFA::state_len(self) |
225 | } |
226 | |
227 | fn stride2(&self) -> usize { |
228 | // We don't do pre-multiplication for the one-pass DFA, so |
229 | // returning 0 has the effect of making state IDs and state indices |
230 | // equivalent. |
231 | 0 |
232 | } |
233 | |
234 | fn swap_states(&mut self, id1: StateID, id2: StateID) { |
235 | DFA::swap_states(self, id1, id2) |
236 | } |
237 | |
238 | fn remap(&mut self, map: impl Fn(StateID) -> StateID) { |
239 | DFA::remap(self, map) |
240 | } |
241 | } |
242 | } |
243 | |