khenaidoo | ab1f7bd | 2019-11-14 14:00:27 -0500 | [diff] [blame] | 1 | // Copyright 2014 Google Inc. |
| 2 | // |
| 3 | // Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | // you may not use this file except in compliance with the License. |
| 5 | // You may obtain a copy of the License at |
| 6 | // |
| 7 | // http://www.apache.org/licenses/LICENSE-2.0 |
| 8 | // |
| 9 | // Unless required by applicable law or agreed to in writing, software |
| 10 | // distributed under the License is distributed on an "AS IS" BASIS, |
| 11 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 12 | // See the License for the specific language governing permissions and |
| 13 | // limitations under the License. |
| 14 | |
| 15 | // Package btree implements in-memory B-Trees of arbitrary degree. |
| 16 | // |
| 17 | // btree implements an in-memory B-Tree for use as an ordered data structure. |
| 18 | // It is not meant for persistent storage solutions. |
| 19 | // |
| 20 | // It has a flatter structure than an equivalent red-black or other binary tree, |
| 21 | // which in some cases yields better memory usage and/or performance. |
| 22 | // See some discussion on the matter here: |
| 23 | // http://google-opensource.blogspot.com/2013/01/c-containers-that-save-memory-and-time.html |
| 24 | // Note, though, that this project is in no way related to the C++ B-Tree |
| 25 | // implementation written about there. |
| 26 | // |
| 27 | // Within this tree, each node contains a slice of items and a (possibly nil) |
| 28 | // slice of children. For basic numeric values or raw structs, this can cause |
| 29 | // efficiency differences when compared to equivalent C++ template code that |
| 30 | // stores values in arrays within the node: |
| 31 | // * Due to the overhead of storing values as interfaces (each |
| 32 | // value needs to be stored as the value itself, then 2 words for the |
| 33 | // interface pointing to that value and its type), resulting in higher |
| 34 | // memory use. |
| 35 | // * Since interfaces can point to values anywhere in memory, values are |
| 36 | // most likely not stored in contiguous blocks, resulting in a higher |
| 37 | // number of cache misses. |
| 38 | // These issues don't tend to matter, though, when working with strings or other |
| 39 | // heap-allocated structures, since C++-equivalent structures also must store |
| 40 | // pointers and also distribute their values across the heap. |
| 41 | // |
| 42 | // This implementation is designed to be a drop-in replacement to gollrb.LLRB |
| 43 | // trees, (http://github.com/petar/gollrb), an excellent and probably the most |
| 44 | // widely used ordered tree implementation in the Go ecosystem currently. |
| 45 | // Its functions, therefore, exactly mirror those of |
| 46 | // llrb.LLRB where possible. Unlike gollrb, though, we currently don't |
| 47 | // support storing multiple equivalent values. |
| 48 | package btree |
| 49 | |
| 50 | import ( |
| 51 | "fmt" |
| 52 | "io" |
| 53 | "sort" |
| 54 | "strings" |
| 55 | "sync" |
| 56 | ) |
| 57 | |
| 58 | // Item represents a single object in the tree. |
| 59 | type Item interface { |
| 60 | // Less tests whether the current item is less than the given argument. |
| 61 | // |
| 62 | // This must provide a strict weak ordering. |
| 63 | // If !a.Less(b) && !b.Less(a), we treat this to mean a == b (i.e. we can only |
| 64 | // hold one of either a or b in the tree). |
| 65 | Less(than Item) bool |
| 66 | } |
| 67 | |
| 68 | const ( |
| 69 | DefaultFreeListSize = 32 |
| 70 | ) |
| 71 | |
| 72 | var ( |
| 73 | nilItems = make(items, 16) |
| 74 | nilChildren = make(children, 16) |
| 75 | ) |
| 76 | |
| 77 | // FreeList represents a free list of btree nodes. By default each |
| 78 | // BTree has its own FreeList, but multiple BTrees can share the same |
| 79 | // FreeList. |
| 80 | // Two Btrees using the same freelist are safe for concurrent write access. |
| 81 | type FreeList struct { |
| 82 | mu sync.Mutex |
| 83 | freelist []*node |
| 84 | } |
| 85 | |
| 86 | // NewFreeList creates a new free list. |
| 87 | // size is the maximum size of the returned free list. |
| 88 | func NewFreeList(size int) *FreeList { |
| 89 | return &FreeList{freelist: make([]*node, 0, size)} |
| 90 | } |
| 91 | |
| 92 | func (f *FreeList) newNode() (n *node) { |
| 93 | f.mu.Lock() |
| 94 | index := len(f.freelist) - 1 |
| 95 | if index < 0 { |
| 96 | f.mu.Unlock() |
| 97 | return new(node) |
| 98 | } |
| 99 | n = f.freelist[index] |
| 100 | f.freelist[index] = nil |
| 101 | f.freelist = f.freelist[:index] |
| 102 | f.mu.Unlock() |
| 103 | return |
| 104 | } |
| 105 | |
| 106 | // freeNode adds the given node to the list, returning true if it was added |
| 107 | // and false if it was discarded. |
| 108 | func (f *FreeList) freeNode(n *node) (out bool) { |
| 109 | f.mu.Lock() |
| 110 | if len(f.freelist) < cap(f.freelist) { |
| 111 | f.freelist = append(f.freelist, n) |
| 112 | out = true |
| 113 | } |
| 114 | f.mu.Unlock() |
| 115 | return |
| 116 | } |
| 117 | |
| 118 | // ItemIterator allows callers of Ascend* to iterate in-order over portions of |
| 119 | // the tree. When this function returns false, iteration will stop and the |
| 120 | // associated Ascend* function will immediately return. |
| 121 | type ItemIterator func(i Item) bool |
| 122 | |
| 123 | // New creates a new B-Tree with the given degree. |
| 124 | // |
| 125 | // New(2), for example, will create a 2-3-4 tree (each node contains 1-3 items |
| 126 | // and 2-4 children). |
| 127 | func New(degree int) *BTree { |
| 128 | return NewWithFreeList(degree, NewFreeList(DefaultFreeListSize)) |
| 129 | } |
| 130 | |
| 131 | // NewWithFreeList creates a new B-Tree that uses the given node free list. |
| 132 | func NewWithFreeList(degree int, f *FreeList) *BTree { |
| 133 | if degree <= 1 { |
| 134 | panic("bad degree") |
| 135 | } |
| 136 | return &BTree{ |
| 137 | degree: degree, |
| 138 | cow: ©OnWriteContext{freelist: f}, |
| 139 | } |
| 140 | } |
| 141 | |
| 142 | // items stores items in a node. |
| 143 | type items []Item |
| 144 | |
| 145 | // insertAt inserts a value into the given index, pushing all subsequent values |
| 146 | // forward. |
| 147 | func (s *items) insertAt(index int, item Item) { |
| 148 | *s = append(*s, nil) |
| 149 | if index < len(*s) { |
| 150 | copy((*s)[index+1:], (*s)[index:]) |
| 151 | } |
| 152 | (*s)[index] = item |
| 153 | } |
| 154 | |
| 155 | // removeAt removes a value at a given index, pulling all subsequent values |
| 156 | // back. |
| 157 | func (s *items) removeAt(index int) Item { |
| 158 | item := (*s)[index] |
| 159 | copy((*s)[index:], (*s)[index+1:]) |
| 160 | (*s)[len(*s)-1] = nil |
| 161 | *s = (*s)[:len(*s)-1] |
| 162 | return item |
| 163 | } |
| 164 | |
| 165 | // pop removes and returns the last element in the list. |
| 166 | func (s *items) pop() (out Item) { |
| 167 | index := len(*s) - 1 |
| 168 | out = (*s)[index] |
| 169 | (*s)[index] = nil |
| 170 | *s = (*s)[:index] |
| 171 | return |
| 172 | } |
| 173 | |
| 174 | // truncate truncates this instance at index so that it contains only the |
| 175 | // first index items. index must be less than or equal to length. |
| 176 | func (s *items) truncate(index int) { |
| 177 | var toClear items |
| 178 | *s, toClear = (*s)[:index], (*s)[index:] |
| 179 | for len(toClear) > 0 { |
| 180 | toClear = toClear[copy(toClear, nilItems):] |
| 181 | } |
| 182 | } |
| 183 | |
| 184 | // find returns the index where the given item should be inserted into this |
| 185 | // list. 'found' is true if the item already exists in the list at the given |
| 186 | // index. |
| 187 | func (s items) find(item Item) (index int, found bool) { |
| 188 | i := sort.Search(len(s), func(i int) bool { |
| 189 | return item.Less(s[i]) |
| 190 | }) |
| 191 | if i > 0 && !s[i-1].Less(item) { |
| 192 | return i - 1, true |
| 193 | } |
| 194 | return i, false |
| 195 | } |
| 196 | |
| 197 | // children stores child nodes in a node. |
| 198 | type children []*node |
| 199 | |
| 200 | // insertAt inserts a value into the given index, pushing all subsequent values |
| 201 | // forward. |
| 202 | func (s *children) insertAt(index int, n *node) { |
| 203 | *s = append(*s, nil) |
| 204 | if index < len(*s) { |
| 205 | copy((*s)[index+1:], (*s)[index:]) |
| 206 | } |
| 207 | (*s)[index] = n |
| 208 | } |
| 209 | |
| 210 | // removeAt removes a value at a given index, pulling all subsequent values |
| 211 | // back. |
| 212 | func (s *children) removeAt(index int) *node { |
| 213 | n := (*s)[index] |
| 214 | copy((*s)[index:], (*s)[index+1:]) |
| 215 | (*s)[len(*s)-1] = nil |
| 216 | *s = (*s)[:len(*s)-1] |
| 217 | return n |
| 218 | } |
| 219 | |
| 220 | // pop removes and returns the last element in the list. |
| 221 | func (s *children) pop() (out *node) { |
| 222 | index := len(*s) - 1 |
| 223 | out = (*s)[index] |
| 224 | (*s)[index] = nil |
| 225 | *s = (*s)[:index] |
| 226 | return |
| 227 | } |
| 228 | |
| 229 | // truncate truncates this instance at index so that it contains only the |
| 230 | // first index children. index must be less than or equal to length. |
| 231 | func (s *children) truncate(index int) { |
| 232 | var toClear children |
| 233 | *s, toClear = (*s)[:index], (*s)[index:] |
| 234 | for len(toClear) > 0 { |
| 235 | toClear = toClear[copy(toClear, nilChildren):] |
| 236 | } |
| 237 | } |
| 238 | |
| 239 | // node is an internal node in a tree. |
| 240 | // |
| 241 | // It must at all times maintain the invariant that either |
| 242 | // * len(children) == 0, len(items) unconstrained |
| 243 | // * len(children) == len(items) + 1 |
| 244 | type node struct { |
| 245 | items items |
| 246 | children children |
| 247 | cow *copyOnWriteContext |
| 248 | } |
| 249 | |
| 250 | func (n *node) mutableFor(cow *copyOnWriteContext) *node { |
| 251 | if n.cow == cow { |
| 252 | return n |
| 253 | } |
| 254 | out := cow.newNode() |
| 255 | if cap(out.items) >= len(n.items) { |
| 256 | out.items = out.items[:len(n.items)] |
| 257 | } else { |
| 258 | out.items = make(items, len(n.items), cap(n.items)) |
| 259 | } |
| 260 | copy(out.items, n.items) |
| 261 | // Copy children |
| 262 | if cap(out.children) >= len(n.children) { |
| 263 | out.children = out.children[:len(n.children)] |
| 264 | } else { |
| 265 | out.children = make(children, len(n.children), cap(n.children)) |
| 266 | } |
| 267 | copy(out.children, n.children) |
| 268 | return out |
| 269 | } |
| 270 | |
| 271 | func (n *node) mutableChild(i int) *node { |
| 272 | c := n.children[i].mutableFor(n.cow) |
| 273 | n.children[i] = c |
| 274 | return c |
| 275 | } |
| 276 | |
| 277 | // split splits the given node at the given index. The current node shrinks, |
| 278 | // and this function returns the item that existed at that index and a new node |
| 279 | // containing all items/children after it. |
| 280 | func (n *node) split(i int) (Item, *node) { |
| 281 | item := n.items[i] |
| 282 | next := n.cow.newNode() |
| 283 | next.items = append(next.items, n.items[i+1:]...) |
| 284 | n.items.truncate(i) |
| 285 | if len(n.children) > 0 { |
| 286 | next.children = append(next.children, n.children[i+1:]...) |
| 287 | n.children.truncate(i + 1) |
| 288 | } |
| 289 | return item, next |
| 290 | } |
| 291 | |
| 292 | // maybeSplitChild checks if a child should be split, and if so splits it. |
| 293 | // Returns whether or not a split occurred. |
| 294 | func (n *node) maybeSplitChild(i, maxItems int) bool { |
| 295 | if len(n.children[i].items) < maxItems { |
| 296 | return false |
| 297 | } |
| 298 | first := n.mutableChild(i) |
| 299 | item, second := first.split(maxItems / 2) |
| 300 | n.items.insertAt(i, item) |
| 301 | n.children.insertAt(i+1, second) |
| 302 | return true |
| 303 | } |
| 304 | |
| 305 | // insert inserts an item into the subtree rooted at this node, making sure |
| 306 | // no nodes in the subtree exceed maxItems items. Should an equivalent item be |
| 307 | // be found/replaced by insert, it will be returned. |
| 308 | func (n *node) insert(item Item, maxItems int) Item { |
| 309 | i, found := n.items.find(item) |
| 310 | if found { |
| 311 | out := n.items[i] |
| 312 | n.items[i] = item |
| 313 | return out |
| 314 | } |
| 315 | if len(n.children) == 0 { |
| 316 | n.items.insertAt(i, item) |
| 317 | return nil |
| 318 | } |
| 319 | if n.maybeSplitChild(i, maxItems) { |
| 320 | inTree := n.items[i] |
| 321 | switch { |
| 322 | case item.Less(inTree): |
| 323 | // no change, we want first split node |
| 324 | case inTree.Less(item): |
| 325 | i++ // we want second split node |
| 326 | default: |
| 327 | out := n.items[i] |
| 328 | n.items[i] = item |
| 329 | return out |
| 330 | } |
| 331 | } |
| 332 | return n.mutableChild(i).insert(item, maxItems) |
| 333 | } |
| 334 | |
| 335 | // get finds the given key in the subtree and returns it. |
| 336 | func (n *node) get(key Item) Item { |
| 337 | i, found := n.items.find(key) |
| 338 | if found { |
| 339 | return n.items[i] |
| 340 | } else if len(n.children) > 0 { |
| 341 | return n.children[i].get(key) |
| 342 | } |
| 343 | return nil |
| 344 | } |
| 345 | |
| 346 | // min returns the first item in the subtree. |
| 347 | func min(n *node) Item { |
| 348 | if n == nil { |
| 349 | return nil |
| 350 | } |
| 351 | for len(n.children) > 0 { |
| 352 | n = n.children[0] |
| 353 | } |
| 354 | if len(n.items) == 0 { |
| 355 | return nil |
| 356 | } |
| 357 | return n.items[0] |
| 358 | } |
| 359 | |
| 360 | // max returns the last item in the subtree. |
| 361 | func max(n *node) Item { |
| 362 | if n == nil { |
| 363 | return nil |
| 364 | } |
| 365 | for len(n.children) > 0 { |
| 366 | n = n.children[len(n.children)-1] |
| 367 | } |
| 368 | if len(n.items) == 0 { |
| 369 | return nil |
| 370 | } |
| 371 | return n.items[len(n.items)-1] |
| 372 | } |
| 373 | |
| 374 | // toRemove details what item to remove in a node.remove call. |
| 375 | type toRemove int |
| 376 | |
| 377 | const ( |
| 378 | removeItem toRemove = iota // removes the given item |
| 379 | removeMin // removes smallest item in the subtree |
| 380 | removeMax // removes largest item in the subtree |
| 381 | ) |
| 382 | |
| 383 | // remove removes an item from the subtree rooted at this node. |
| 384 | func (n *node) remove(item Item, minItems int, typ toRemove) Item { |
| 385 | var i int |
| 386 | var found bool |
| 387 | switch typ { |
| 388 | case removeMax: |
| 389 | if len(n.children) == 0 { |
| 390 | return n.items.pop() |
| 391 | } |
| 392 | i = len(n.items) |
| 393 | case removeMin: |
| 394 | if len(n.children) == 0 { |
| 395 | return n.items.removeAt(0) |
| 396 | } |
| 397 | i = 0 |
| 398 | case removeItem: |
| 399 | i, found = n.items.find(item) |
| 400 | if len(n.children) == 0 { |
| 401 | if found { |
| 402 | return n.items.removeAt(i) |
| 403 | } |
| 404 | return nil |
| 405 | } |
| 406 | default: |
| 407 | panic("invalid type") |
| 408 | } |
| 409 | // If we get to here, we have children. |
| 410 | if len(n.children[i].items) <= minItems { |
| 411 | return n.growChildAndRemove(i, item, minItems, typ) |
| 412 | } |
| 413 | child := n.mutableChild(i) |
| 414 | // Either we had enough items to begin with, or we've done some |
| 415 | // merging/stealing, because we've got enough now and we're ready to return |
| 416 | // stuff. |
| 417 | if found { |
| 418 | // The item exists at index 'i', and the child we've selected can give us a |
| 419 | // predecessor, since if we've gotten here it's got > minItems items in it. |
| 420 | out := n.items[i] |
| 421 | // We use our special-case 'remove' call with typ=maxItem to pull the |
| 422 | // predecessor of item i (the rightmost leaf of our immediate left child) |
| 423 | // and set it into where we pulled the item from. |
| 424 | n.items[i] = child.remove(nil, minItems, removeMax) |
| 425 | return out |
| 426 | } |
| 427 | // Final recursive call. Once we're here, we know that the item isn't in this |
| 428 | // node and that the child is big enough to remove from. |
| 429 | return child.remove(item, minItems, typ) |
| 430 | } |
| 431 | |
| 432 | // growChildAndRemove grows child 'i' to make sure it's possible to remove an |
| 433 | // item from it while keeping it at minItems, then calls remove to actually |
| 434 | // remove it. |
| 435 | // |
| 436 | // Most documentation says we have to do two sets of special casing: |
| 437 | // 1) item is in this node |
| 438 | // 2) item is in child |
| 439 | // In both cases, we need to handle the two subcases: |
| 440 | // A) node has enough values that it can spare one |
| 441 | // B) node doesn't have enough values |
| 442 | // For the latter, we have to check: |
| 443 | // a) left sibling has node to spare |
| 444 | // b) right sibling has node to spare |
| 445 | // c) we must merge |
| 446 | // To simplify our code here, we handle cases #1 and #2 the same: |
| 447 | // If a node doesn't have enough items, we make sure it does (using a,b,c). |
| 448 | // We then simply redo our remove call, and the second time (regardless of |
| 449 | // whether we're in case 1 or 2), we'll have enough items and can guarantee |
| 450 | // that we hit case A. |
| 451 | func (n *node) growChildAndRemove(i int, item Item, minItems int, typ toRemove) Item { |
| 452 | if i > 0 && len(n.children[i-1].items) > minItems { |
| 453 | // Steal from left child |
| 454 | child := n.mutableChild(i) |
| 455 | stealFrom := n.mutableChild(i - 1) |
| 456 | stolenItem := stealFrom.items.pop() |
| 457 | child.items.insertAt(0, n.items[i-1]) |
| 458 | n.items[i-1] = stolenItem |
| 459 | if len(stealFrom.children) > 0 { |
| 460 | child.children.insertAt(0, stealFrom.children.pop()) |
| 461 | } |
| 462 | } else if i < len(n.items) && len(n.children[i+1].items) > minItems { |
| 463 | // steal from right child |
| 464 | child := n.mutableChild(i) |
| 465 | stealFrom := n.mutableChild(i + 1) |
| 466 | stolenItem := stealFrom.items.removeAt(0) |
| 467 | child.items = append(child.items, n.items[i]) |
| 468 | n.items[i] = stolenItem |
| 469 | if len(stealFrom.children) > 0 { |
| 470 | child.children = append(child.children, stealFrom.children.removeAt(0)) |
| 471 | } |
| 472 | } else { |
| 473 | if i >= len(n.items) { |
| 474 | i-- |
| 475 | } |
| 476 | child := n.mutableChild(i) |
| 477 | // merge with right child |
| 478 | mergeItem := n.items.removeAt(i) |
| 479 | mergeChild := n.children.removeAt(i + 1) |
| 480 | child.items = append(child.items, mergeItem) |
| 481 | child.items = append(child.items, mergeChild.items...) |
| 482 | child.children = append(child.children, mergeChild.children...) |
| 483 | n.cow.freeNode(mergeChild) |
| 484 | } |
| 485 | return n.remove(item, minItems, typ) |
| 486 | } |
| 487 | |
| 488 | type direction int |
| 489 | |
| 490 | const ( |
| 491 | descend = direction(-1) |
| 492 | ascend = direction(+1) |
| 493 | ) |
| 494 | |
| 495 | // iterate provides a simple method for iterating over elements in the tree. |
| 496 | // |
| 497 | // When ascending, the 'start' should be less than 'stop' and when descending, |
| 498 | // the 'start' should be greater than 'stop'. Setting 'includeStart' to true |
| 499 | // will force the iterator to include the first item when it equals 'start', |
| 500 | // thus creating a "greaterOrEqual" or "lessThanEqual" rather than just a |
| 501 | // "greaterThan" or "lessThan" queries. |
| 502 | func (n *node) iterate(dir direction, start, stop Item, includeStart bool, hit bool, iter ItemIterator) (bool, bool) { |
| 503 | var ok, found bool |
| 504 | var index int |
| 505 | switch dir { |
| 506 | case ascend: |
| 507 | if start != nil { |
| 508 | index, _ = n.items.find(start) |
| 509 | } |
| 510 | for i := index; i < len(n.items); i++ { |
| 511 | if len(n.children) > 0 { |
| 512 | if hit, ok = n.children[i].iterate(dir, start, stop, includeStart, hit, iter); !ok { |
| 513 | return hit, false |
| 514 | } |
| 515 | } |
| 516 | if !includeStart && !hit && start != nil && !start.Less(n.items[i]) { |
| 517 | hit = true |
| 518 | continue |
| 519 | } |
| 520 | hit = true |
| 521 | if stop != nil && !n.items[i].Less(stop) { |
| 522 | return hit, false |
| 523 | } |
| 524 | if !iter(n.items[i]) { |
| 525 | return hit, false |
| 526 | } |
| 527 | } |
| 528 | if len(n.children) > 0 { |
| 529 | if hit, ok = n.children[len(n.children)-1].iterate(dir, start, stop, includeStart, hit, iter); !ok { |
| 530 | return hit, false |
| 531 | } |
| 532 | } |
| 533 | case descend: |
| 534 | if start != nil { |
| 535 | index, found = n.items.find(start) |
| 536 | if !found { |
| 537 | index = index - 1 |
| 538 | } |
| 539 | } else { |
| 540 | index = len(n.items) - 1 |
| 541 | } |
| 542 | for i := index; i >= 0; i-- { |
| 543 | if start != nil && !n.items[i].Less(start) { |
| 544 | if !includeStart || hit || start.Less(n.items[i]) { |
| 545 | continue |
| 546 | } |
| 547 | } |
| 548 | if len(n.children) > 0 { |
| 549 | if hit, ok = n.children[i+1].iterate(dir, start, stop, includeStart, hit, iter); !ok { |
| 550 | return hit, false |
| 551 | } |
| 552 | } |
| 553 | if stop != nil && !stop.Less(n.items[i]) { |
| 554 | return hit, false // continue |
| 555 | } |
| 556 | hit = true |
| 557 | if !iter(n.items[i]) { |
| 558 | return hit, false |
| 559 | } |
| 560 | } |
| 561 | if len(n.children) > 0 { |
| 562 | if hit, ok = n.children[0].iterate(dir, start, stop, includeStart, hit, iter); !ok { |
| 563 | return hit, false |
| 564 | } |
| 565 | } |
| 566 | } |
| 567 | return hit, true |
| 568 | } |
| 569 | |
| 570 | // Used for testing/debugging purposes. |
| 571 | func (n *node) print(w io.Writer, level int) { |
| 572 | fmt.Fprintf(w, "%sNODE:%v\n", strings.Repeat(" ", level), n.items) |
| 573 | for _, c := range n.children { |
| 574 | c.print(w, level+1) |
| 575 | } |
| 576 | } |
| 577 | |
| 578 | // BTree is an implementation of a B-Tree. |
| 579 | // |
| 580 | // BTree stores Item instances in an ordered structure, allowing easy insertion, |
| 581 | // removal, and iteration. |
| 582 | // |
| 583 | // Write operations are not safe for concurrent mutation by multiple |
| 584 | // goroutines, but Read operations are. |
| 585 | type BTree struct { |
| 586 | degree int |
| 587 | length int |
| 588 | root *node |
| 589 | cow *copyOnWriteContext |
| 590 | } |
| 591 | |
| 592 | // copyOnWriteContext pointers determine node ownership... a tree with a write |
| 593 | // context equivalent to a node's write context is allowed to modify that node. |
| 594 | // A tree whose write context does not match a node's is not allowed to modify |
| 595 | // it, and must create a new, writable copy (IE: it's a Clone). |
| 596 | // |
| 597 | // When doing any write operation, we maintain the invariant that the current |
| 598 | // node's context is equal to the context of the tree that requested the write. |
| 599 | // We do this by, before we descend into any node, creating a copy with the |
| 600 | // correct context if the contexts don't match. |
| 601 | // |
| 602 | // Since the node we're currently visiting on any write has the requesting |
| 603 | // tree's context, that node is modifiable in place. Children of that node may |
| 604 | // not share context, but before we descend into them, we'll make a mutable |
| 605 | // copy. |
| 606 | type copyOnWriteContext struct { |
| 607 | freelist *FreeList |
| 608 | } |
| 609 | |
| 610 | // Clone clones the btree, lazily. Clone should not be called concurrently, |
| 611 | // but the original tree (t) and the new tree (t2) can be used concurrently |
| 612 | // once the Clone call completes. |
| 613 | // |
| 614 | // The internal tree structure of b is marked read-only and shared between t and |
| 615 | // t2. Writes to both t and t2 use copy-on-write logic, creating new nodes |
| 616 | // whenever one of b's original nodes would have been modified. Read operations |
| 617 | // should have no performance degredation. Write operations for both t and t2 |
| 618 | // will initially experience minor slow-downs caused by additional allocs and |
| 619 | // copies due to the aforementioned copy-on-write logic, but should converge to |
| 620 | // the original performance characteristics of the original tree. |
| 621 | func (t *BTree) Clone() (t2 *BTree) { |
| 622 | // Create two entirely new copy-on-write contexts. |
| 623 | // This operation effectively creates three trees: |
| 624 | // the original, shared nodes (old b.cow) |
| 625 | // the new b.cow nodes |
| 626 | // the new out.cow nodes |
| 627 | cow1, cow2 := *t.cow, *t.cow |
| 628 | out := *t |
| 629 | t.cow = &cow1 |
| 630 | out.cow = &cow2 |
| 631 | return &out |
| 632 | } |
| 633 | |
| 634 | // maxItems returns the max number of items to allow per node. |
| 635 | func (t *BTree) maxItems() int { |
| 636 | return t.degree*2 - 1 |
| 637 | } |
| 638 | |
| 639 | // minItems returns the min number of items to allow per node (ignored for the |
| 640 | // root node). |
| 641 | func (t *BTree) minItems() int { |
| 642 | return t.degree - 1 |
| 643 | } |
| 644 | |
| 645 | func (c *copyOnWriteContext) newNode() (n *node) { |
| 646 | n = c.freelist.newNode() |
| 647 | n.cow = c |
| 648 | return |
| 649 | } |
| 650 | |
| 651 | type freeType int |
| 652 | |
| 653 | const ( |
| 654 | ftFreelistFull freeType = iota // node was freed (available for GC, not stored in freelist) |
| 655 | ftStored // node was stored in the freelist for later use |
| 656 | ftNotOwned // node was ignored by COW, since it's owned by another one |
| 657 | ) |
| 658 | |
| 659 | // freeNode frees a node within a given COW context, if it's owned by that |
| 660 | // context. It returns what happened to the node (see freeType const |
| 661 | // documentation). |
| 662 | func (c *copyOnWriteContext) freeNode(n *node) freeType { |
| 663 | if n.cow == c { |
| 664 | // clear to allow GC |
| 665 | n.items.truncate(0) |
| 666 | n.children.truncate(0) |
| 667 | n.cow = nil |
| 668 | if c.freelist.freeNode(n) { |
| 669 | return ftStored |
| 670 | } else { |
| 671 | return ftFreelistFull |
| 672 | } |
| 673 | } else { |
| 674 | return ftNotOwned |
| 675 | } |
| 676 | } |
| 677 | |
| 678 | // ReplaceOrInsert adds the given item to the tree. If an item in the tree |
| 679 | // already equals the given one, it is removed from the tree and returned. |
| 680 | // Otherwise, nil is returned. |
| 681 | // |
| 682 | // nil cannot be added to the tree (will panic). |
| 683 | func (t *BTree) ReplaceOrInsert(item Item) Item { |
| 684 | if item == nil { |
| 685 | panic("nil item being added to BTree") |
| 686 | } |
| 687 | if t.root == nil { |
| 688 | t.root = t.cow.newNode() |
| 689 | t.root.items = append(t.root.items, item) |
| 690 | t.length++ |
| 691 | return nil |
| 692 | } else { |
| 693 | t.root = t.root.mutableFor(t.cow) |
| 694 | if len(t.root.items) >= t.maxItems() { |
| 695 | item2, second := t.root.split(t.maxItems() / 2) |
| 696 | oldroot := t.root |
| 697 | t.root = t.cow.newNode() |
| 698 | t.root.items = append(t.root.items, item2) |
| 699 | t.root.children = append(t.root.children, oldroot, second) |
| 700 | } |
| 701 | } |
| 702 | out := t.root.insert(item, t.maxItems()) |
| 703 | if out == nil { |
| 704 | t.length++ |
| 705 | } |
| 706 | return out |
| 707 | } |
| 708 | |
| 709 | // Delete removes an item equal to the passed in item from the tree, returning |
| 710 | // it. If no such item exists, returns nil. |
| 711 | func (t *BTree) Delete(item Item) Item { |
| 712 | return t.deleteItem(item, removeItem) |
| 713 | } |
| 714 | |
| 715 | // DeleteMin removes the smallest item in the tree and returns it. |
| 716 | // If no such item exists, returns nil. |
| 717 | func (t *BTree) DeleteMin() Item { |
| 718 | return t.deleteItem(nil, removeMin) |
| 719 | } |
| 720 | |
| 721 | // DeleteMax removes the largest item in the tree and returns it. |
| 722 | // If no such item exists, returns nil. |
| 723 | func (t *BTree) DeleteMax() Item { |
| 724 | return t.deleteItem(nil, removeMax) |
| 725 | } |
| 726 | |
| 727 | func (t *BTree) deleteItem(item Item, typ toRemove) Item { |
| 728 | if t.root == nil || len(t.root.items) == 0 { |
| 729 | return nil |
| 730 | } |
| 731 | t.root = t.root.mutableFor(t.cow) |
| 732 | out := t.root.remove(item, t.minItems(), typ) |
| 733 | if len(t.root.items) == 0 && len(t.root.children) > 0 { |
| 734 | oldroot := t.root |
| 735 | t.root = t.root.children[0] |
| 736 | t.cow.freeNode(oldroot) |
| 737 | } |
| 738 | if out != nil { |
| 739 | t.length-- |
| 740 | } |
| 741 | return out |
| 742 | } |
| 743 | |
| 744 | // AscendRange calls the iterator for every value in the tree within the range |
| 745 | // [greaterOrEqual, lessThan), until iterator returns false. |
| 746 | func (t *BTree) AscendRange(greaterOrEqual, lessThan Item, iterator ItemIterator) { |
| 747 | if t.root == nil { |
| 748 | return |
| 749 | } |
| 750 | t.root.iterate(ascend, greaterOrEqual, lessThan, true, false, iterator) |
| 751 | } |
| 752 | |
| 753 | // AscendLessThan calls the iterator for every value in the tree within the range |
| 754 | // [first, pivot), until iterator returns false. |
| 755 | func (t *BTree) AscendLessThan(pivot Item, iterator ItemIterator) { |
| 756 | if t.root == nil { |
| 757 | return |
| 758 | } |
| 759 | t.root.iterate(ascend, nil, pivot, false, false, iterator) |
| 760 | } |
| 761 | |
| 762 | // AscendGreaterOrEqual calls the iterator for every value in the tree within |
| 763 | // the range [pivot, last], until iterator returns false. |
| 764 | func (t *BTree) AscendGreaterOrEqual(pivot Item, iterator ItemIterator) { |
| 765 | if t.root == nil { |
| 766 | return |
| 767 | } |
| 768 | t.root.iterate(ascend, pivot, nil, true, false, iterator) |
| 769 | } |
| 770 | |
| 771 | // Ascend calls the iterator for every value in the tree within the range |
| 772 | // [first, last], until iterator returns false. |
| 773 | func (t *BTree) Ascend(iterator ItemIterator) { |
| 774 | if t.root == nil { |
| 775 | return |
| 776 | } |
| 777 | t.root.iterate(ascend, nil, nil, false, false, iterator) |
| 778 | } |
| 779 | |
| 780 | // DescendRange calls the iterator for every value in the tree within the range |
| 781 | // [lessOrEqual, greaterThan), until iterator returns false. |
| 782 | func (t *BTree) DescendRange(lessOrEqual, greaterThan Item, iterator ItemIterator) { |
| 783 | if t.root == nil { |
| 784 | return |
| 785 | } |
| 786 | t.root.iterate(descend, lessOrEqual, greaterThan, true, false, iterator) |
| 787 | } |
| 788 | |
| 789 | // DescendLessOrEqual calls the iterator for every value in the tree within the range |
| 790 | // [pivot, first], until iterator returns false. |
| 791 | func (t *BTree) DescendLessOrEqual(pivot Item, iterator ItemIterator) { |
| 792 | if t.root == nil { |
| 793 | return |
| 794 | } |
| 795 | t.root.iterate(descend, pivot, nil, true, false, iterator) |
| 796 | } |
| 797 | |
| 798 | // DescendGreaterThan calls the iterator for every value in the tree within |
| 799 | // the range (pivot, last], until iterator returns false. |
| 800 | func (t *BTree) DescendGreaterThan(pivot Item, iterator ItemIterator) { |
| 801 | if t.root == nil { |
| 802 | return |
| 803 | } |
| 804 | t.root.iterate(descend, nil, pivot, false, false, iterator) |
| 805 | } |
| 806 | |
| 807 | // Descend calls the iterator for every value in the tree within the range |
| 808 | // [last, first], until iterator returns false. |
| 809 | func (t *BTree) Descend(iterator ItemIterator) { |
| 810 | if t.root == nil { |
| 811 | return |
| 812 | } |
| 813 | t.root.iterate(descend, nil, nil, false, false, iterator) |
| 814 | } |
| 815 | |
| 816 | // Get looks for the key item in the tree, returning it. It returns nil if |
| 817 | // unable to find that item. |
| 818 | func (t *BTree) Get(key Item) Item { |
| 819 | if t.root == nil { |
| 820 | return nil |
| 821 | } |
| 822 | return t.root.get(key) |
| 823 | } |
| 824 | |
| 825 | // Min returns the smallest item in the tree, or nil if the tree is empty. |
| 826 | func (t *BTree) Min() Item { |
| 827 | return min(t.root) |
| 828 | } |
| 829 | |
| 830 | // Max returns the largest item in the tree, or nil if the tree is empty. |
| 831 | func (t *BTree) Max() Item { |
| 832 | return max(t.root) |
| 833 | } |
| 834 | |
| 835 | // Has returns true if the given key is in the tree. |
| 836 | func (t *BTree) Has(key Item) bool { |
| 837 | return t.Get(key) != nil |
| 838 | } |
| 839 | |
| 840 | // Len returns the number of items currently in the tree. |
| 841 | func (t *BTree) Len() int { |
| 842 | return t.length |
| 843 | } |
| 844 | |
| 845 | // Clear removes all items from the btree. If addNodesToFreelist is true, |
| 846 | // t's nodes are added to its freelist as part of this call, until the freelist |
| 847 | // is full. Otherwise, the root node is simply dereferenced and the subtree |
| 848 | // left to Go's normal GC processes. |
| 849 | // |
| 850 | // This can be much faster |
| 851 | // than calling Delete on all elements, because that requires finding/removing |
| 852 | // each element in the tree and updating the tree accordingly. It also is |
| 853 | // somewhat faster than creating a new tree to replace the old one, because |
| 854 | // nodes from the old tree are reclaimed into the freelist for use by the new |
| 855 | // one, instead of being lost to the garbage collector. |
| 856 | // |
| 857 | // This call takes: |
| 858 | // O(1): when addNodesToFreelist is false, this is a single operation. |
| 859 | // O(1): when the freelist is already full, it breaks out immediately |
| 860 | // O(freelist size): when the freelist is empty and the nodes are all owned |
| 861 | // by this tree, nodes are added to the freelist until full. |
| 862 | // O(tree size): when all nodes are owned by another tree, all nodes are |
| 863 | // iterated over looking for nodes to add to the freelist, and due to |
| 864 | // ownership, none are. |
| 865 | func (t *BTree) Clear(addNodesToFreelist bool) { |
| 866 | if t.root != nil && addNodesToFreelist { |
| 867 | t.root.reset(t.cow) |
| 868 | } |
| 869 | t.root, t.length = nil, 0 |
| 870 | } |
| 871 | |
| 872 | // reset returns a subtree to the freelist. It breaks out immediately if the |
| 873 | // freelist is full, since the only benefit of iterating is to fill that |
| 874 | // freelist up. Returns true if parent reset call should continue. |
| 875 | func (n *node) reset(c *copyOnWriteContext) bool { |
| 876 | for _, child := range n.children { |
| 877 | if !child.reset(c) { |
| 878 | return false |
| 879 | } |
| 880 | } |
| 881 | return c.freeNode(n) != ftFreelistFull |
| 882 | } |
| 883 | |
| 884 | // Int implements the Item interface for integers. |
| 885 | type Int int |
| 886 | |
| 887 | // Less returns true if int(a) < int(b). |
| 888 | func (a Int) Less(b Item) bool { |
| 889 | return a < b.(Int) |
| 890 | } |