3 //! Implements a map from integer indices to data.
4 //! Rather than storing data for every index, internally, this maps entire ranges to the data.
5 //! To this end, the APIs all work on ranges, not on individual integers. Ranges are split as
6 //! necessary (e.g., when [0,5) is first associated with X, and then [1,2) is mutated).
7 //! Users must not depend on whether a range is coalesced or not, even though this is observable
8 //! via the iteration APIs.
11 use std::num::NonZeroU64;
13 use rustc::ty::layout::Size;
15 #[derive(Clone, Debug)]
17 /// The range covered by this element; never empty.
18 range: ops::Range<u64>,
19 /// The data stored for this element.
22 #[derive(Clone, Debug)]
23 pub struct RangeMap<T> {
28 /// Creates a new `RangeMap` for the given size, and with the given initial value used for
31 pub fn new(size: Size, init: T) -> RangeMap<T> {
32 let size = size.bytes();
33 let mut map = RangeMap { v: Vec::new() };
43 /// Finds the index containing the given offset.
44 fn find_offset(&self, offset: u64) -> usize {
45 // We do a binary search.
46 let mut left = 0usize; // inclusive
47 let mut right = self.v.len(); // exclusive
49 debug_assert!(left < right, "find_offset: offset {} is out-of-bounds", offset);
50 let candidate = left.checked_add(right).unwrap() / 2;
51 let elem = &self.v[candidate];
52 if offset < elem.range.start {
53 // We are too far right (offset is further left).
54 debug_assert!(candidate < right); // we are making progress
56 } else if offset >= elem.range.end {
57 // We are too far left (offset is further right).
58 debug_assert!(candidate >= left); // we are making progress
67 /// Provides read-only iteration over everything in the given range. This does
68 /// *not* split items if they overlap with the edges. Do not use this to mutate
69 /// through interior mutability.
70 pub fn iter<'a>(&'a self, offset: Size, len: Size) -> impl Iterator<Item = &'a T> + 'a {
71 let offset = offset.bytes();
72 let len = len.bytes();
73 // Compute a slice starting with the elements we care about.
74 let slice: &[Elem<T>] = if len == 0 {
75 // We just need any empty iterator. We don't even want to
76 // yield the element that surrounds this position.
79 let first_idx = self.find_offset(offset);
82 // The first offset that is not included any more.
83 let end = offset + len;
85 .take_while(move |elem| elem.range.start < end)
86 .map(|elem| &elem.data)
89 pub fn iter_mut_all<'a>(&'a mut self) -> impl Iterator<Item = &'a mut T> + 'a {
90 self.v.iter_mut().map(|elem| &mut elem.data)
93 // Splits the element situated at the given `index`, such that the 2nd one starts at offset
94 // `split_offset`. Do nothing if the element already starts there.
95 // Returns whether a split was necessary.
96 fn split_index(&mut self, index: usize, split_offset: u64) -> bool
100 let elem = &mut self.v[index];
101 if split_offset == elem.range.start || split_offset == elem.range.end {
105 debug_assert!(elem.range.contains(&split_offset),
106 "the `split_offset` is not in the element to be split");
108 // Now we really have to split. Reduce length of first element.
109 let second_range = split_offset..elem.range.end;
110 elem.range.end = split_offset;
111 // Copy the data, and insert second element.
114 data: elem.data.clone(),
116 self.v.insert(index+1, second);
120 /// Provides mutable iteration over everything in the given range. As a side-effect,
121 /// this will split entries in the map that are only partially hit by the given range,
122 /// to make sure that when they are mutated, the effect is constrained to the given range.
123 /// Moreover, this will opportunistically merge neighbouring equal blocks.
128 ) -> impl Iterator<Item = &'a mut T> + 'a
130 T: Clone + PartialEq,
132 let offset = offset.bytes();
133 let len = len.bytes();
134 // Compute a slice containing exactly the elements we care about
135 let slice: &mut [Elem<T>] = if len == 0 {
136 // We just need any empty iterator. We don't even want to
137 // yield the element that surrounds this position, nor do
141 // Make sure we got a clear beginning
142 let mut first_idx = self.find_offset(offset);
143 if self.split_index(first_idx, offset) {
144 // The newly created 2nd element is ours
147 let first_idx = first_idx; // no more mutation
148 // Find our end. Linear scan, but that's ok because the iteration
149 // is doing the same linear scan anyway -- no increase in complexity.
150 // We combine this scan with a scan for duplicates that we can merge, to reduce
151 // the number of elements.
152 // We stop searching after the first "block" of size 1, to avoid spending excessive
153 // amounts of time on the merging.
154 let mut equal_since_idx = first_idx;
155 // Once we see too many non-mergeable blocks, we stop.
156 // The initial value is chosen via... magic. Benchmarking and magic.
157 let mut successful_merge_count = 3usize;
158 let mut end_idx = first_idx; // when the loop is done, this is the first excluded element.
160 // Compute if `end` is the last element we need to look at.
161 let done = (self.v[end_idx].range.end >= offset+len);
162 // We definitely need to include `end`, so move the index.
164 debug_assert!(done || end_idx < self.v.len(), "iter_mut: end-offset {} is out-of-bounds", offset+len);
165 // see if we want to merge everything in `equal_since..end` (exclusive at the end!)
166 if successful_merge_count > 0 {
167 if done || self.v[end_idx].data != self.v[equal_since_idx].data {
168 // Everything in `equal_since..end` was equal. Make them just one element covering
170 let removed_elems = end_idx - equal_since_idx - 1; // number of elements that we would remove
171 if removed_elems > 0 {
172 // Adjust the range of the first element to cover all of them.
173 let equal_until = self.v[end_idx - 1].range.end; // end of range of last of the equal elements
174 self.v[equal_since_idx].range.end = equal_until;
175 // Delete the rest of them.
176 self.v.splice(equal_since_idx+1..end_idx, std::iter::empty());
177 // Adjust `end_idx` because we made the list shorter.
178 end_idx -= removed_elems;
179 // Adjust the count for the cutoff.
180 successful_merge_count += removed_elems;
182 // Adjust the count for the cutoff.
183 successful_merge_count -= 1;
185 // Go on scanning for the next block starting here.
186 equal_since_idx = end_idx;
189 // Leave loop if this is the last element.
194 // Move to last included instead of first excluded index.
195 let end_idx = end_idx-1;
196 // We need to split the end as well. Even if this performs a
197 // split, we don't have to adjust our index as we only care about
198 // the first part of the split.
199 self.split_index(end_idx, offset+len);
200 // Now we yield the slice. `end` is inclusive.
201 &mut self.v[first_idx..=end_idx]
203 slice.iter_mut().map(|elem| &mut elem.data)
211 /// Query the map at every offset in the range and collect the results.
212 fn to_vec<T: Copy>(map: &RangeMap<T>, offset: u64, len: u64) -> Vec<T> {
213 (offset..offset + len)
216 .iter(Size::from_bytes(i), Size::from_bytes(1))
226 let mut map = RangeMap::<i32>::new(Size::from_bytes(20), -1);
228 for x in map.iter_mut(Size::from_bytes(10), Size::from_bytes(1)) {
232 assert_eq!(to_vec(&map, 10, 1), vec![42]);
233 assert_eq!(map.v.len(), 3);
235 // Insert with size 0.
236 for x in map.iter_mut(Size::from_bytes(10), Size::from_bytes(0)) {
239 for x in map.iter_mut(Size::from_bytes(11), Size::from_bytes(0)) {
242 assert_eq!(to_vec(&map, 10, 2), vec![42, -1]);
243 assert_eq!(map.v.len(), 3);
248 let mut map = RangeMap::<i32>::new(Size::from_bytes(20), -1);
249 for x in map.iter_mut(Size::from_bytes(11), Size::from_bytes(1)) {
252 for x in map.iter_mut(Size::from_bytes(15), Size::from_bytes(1)) {
255 assert_eq!(map.v.len(), 5);
257 to_vec(&map, 10, 10),
258 vec![-1, 42, -1, -1, -1, 43, -1, -1, -1, -1]
261 for x in map.iter_mut(Size::from_bytes(10), Size::from_bytes(10)) {
266 assert_eq!(map.v.len(), 6);
268 to_vec(&map, 10, 10),
269 vec![23, 42, 23, 23, 23, 43, 23, 23, 23, 23]
271 assert_eq!(to_vec(&map, 13, 5), vec![23, 23, 43, 23, 23]);
274 for x in map.iter_mut(Size::from_bytes(15), Size::from_bytes(5)) {
277 assert_eq!(map.v.len(), 6);
279 to_vec(&map, 10, 10),
280 vec![23, 42, 23, 23, 23, 19, 19, 19, 19, 19]
282 // Should be seeing two blocks with 19.
283 assert_eq!(map.iter(Size::from_bytes(15), Size::from_bytes(2))
284 .map(|&t| t).collect::<Vec<_>>(), vec![19, 19]);
286 // A NOP `iter_mut` should trigger merging.
287 for x in map.iter_mut(Size::from_bytes(15), Size::from_bytes(5)) { }
288 assert_eq!(map.v.len(), 5);
290 to_vec(&map, 10, 10),
291 vec![23, 42, 23, 23, 23, 19, 19, 19, 19, 19]