Cartesian tree: Difference between revisions

Revision as of 03:15, 12 August 2023

Latest revision as of 21:45, 11 July 2025

(37 intermediate revisions by 8 users not shown)

Line 1:

{{good article}}

{{Short description|Binary tree derived from a sequence of numbers}}

[[File:Cartesian tree.svg|thumb|240px|A sequence of numbers and the Cartesian tree derived from them.]]

Line 14 ⟶ 15:

#The tree has the [[Binary heap|min-heap property]]: the parent of any non-root node has a smaller value than the node itself.<ref name=minmax/>

These two definitions are equivalent: the tree defined recursively as described above is the unique tree that has the properties listed above. If a sequence of numbers contains repetitions, a Cartesian tree can be determined for it by following a consistent tie-breaking rule before applying the above construction. For instance, the first of two equal elements can be treated as the smaller of the two.{{sfnp|Vuillemin|1980}}

==History==

Cartesian trees were introduced and named by {{harvtxt|Vuillemin|1980}}, who used them as an example of the interaction between [[geometric combinatorics]] and the design and analysis of [[data structure]]s. In particular, Vuillemin used these structures to analyze the [[average-case complexity]] of concatenation and splitting operations on [[binary search tree]]s. The name is derived from the [[Cartesian coordinate]] system for the plane: in one version of this structure, as in the two-dimensional range searching application discussed abovebelow, a Cartesian tree for a point set has the sorted order of the points by their <math>x</math>-coordinates as its symmetric traversal order, and it has the heap property according to the <math>y</math>-coordinates of the points. {{harvtxt|Vuillemin|1980}} described both this geometric version of the structure, and the definition here in which a Cartesian tree is defined from a sequence. Using sequences instead of point coordinates provides a more general setting that allows the Cartesian tree to be applied to non-geometric problems as well.{{sfnp|Vuillemin|1980}}

==Efficient construction==

A Cartesian tree can be constructed in [[linear time]] from its input sequence.

One method is to simply process the sequence values in left-to-right order, maintaining the Cartesian tree of the nodes processed so far, in a structure that allows both upwards and downwards traversal of the tree. To process each new value <math>a</math>, start at the node representing the value prior to <math>a</math> in the sequence and follow the [[Path (graph theory)|path]] from this node to the root of the tree until finding a value <math>b</math> smaller than <math>a</math>. The node <math>a</math> becomes the right child of <math>b</math>, and the previous right child of <math>b</math> becomes the new left child of <math>a</math>. The total time for this procedure is linear, because the time spent searching for the parent <math>b</math> of each new node <math>a</math> can be [[Potential method|charged]] against the number of nodes that are removed from the rightmost path in the tree.{{sfnp|Gabow|Bentley|Tarjan|1984}}

An alternative linear-time construction algorithm is based on the [[all nearest smaller values]] problem. In the input sequence, define the ''left neighbor'' of a value <math>a</math> to be the value that occurs prior to <math>a</math>, is smaller than <math>a</math>, and is closer in position to <math>a</math> than any other smaller value. The ''right neighbor'' is defined symmetrically. The sequence of left neighbors can be found by an algorithm that maintains a [[stack (data structure)|stack]] containing a subsequence of the input. For each new sequence value <math>a</math>, the stack is popped until it is empty or its top element is smaller than <math>a</math>, and then <math>a</math> is pushed onto the stack. The left neighbor of <math>a</math> is the top element at the time <math>a</math> is pushed. The right neighbors can be found by applying the same stack algorithm to the reverse of the sequence. The parent of <math>a</math> in the Cartesian tree is either the left neighbor of <math>a</math> or the right neighbor of <math>a</math>, whichever exists and has a larger value. The left and right neighbors can also be constructed efficiently by [[parallel algorithm]]s, making this formulation useful in efficient parallel algorithms for Cartesian tree construction.<ref>{{harvtxt|Berkman|Schieber|Vishkin|1993}}.</ref>

Another linear-time algorithm for Cartesian tree construction is based on [[Divide-and-conquer algorithm|divide-and-conquer]]. The algorithm recursively constructs the tree on each half of the input, and then merges the two trees. byThe merger process involves takingonly the nodes on the left and right spine''spines'' of these trees: the left treespine andis the path obtained by following left spinechild ofedges from the root until reaching a node with no left child, and the right treespine (bothis ofdefined whichsymmetrically. areAs pathswith whoseany root-topath in a min-leafheap, orderboth spines sortshave their values in sorted order, from the smallest value at their root to their largest) andvalue performsat athe standardend [[Mergeof algorithm#Mergingthe path. To merge the two lists|mergingtrees, apply a [[merge algorithm]] operationto the right spine of the left tree and the left spine of the right tree, replacing these two paths in two trees by a single path that contains the same nodes. In the merged path, the successor in the sorted order of each node from the left tree is placed in its right child, and the successor of each node from the right tree is placed in its left child, the same position that was previously used for its successor in the spine. The left children of nodes from the left tree and right children of nodes from the right tree are leftremain unchanged. The algorithm is also parallelizable since on each level of recursion, each of the two sub-problems can be computed in parallel, and the merging operation can be [[Merge algorithm#Parallel merge|efficiently parallelized]] as well.{{sfnp|Shun|Blelloch|2014}}

Yet another linear-time algorithm, using a linked list representation of the input sequence, is based on ''locally maximum linking'': the algorithm repeatedly identifies a ''local maximum'' element, i.e., one that is larger than both its neighbors (or than its only neighbor, in case it is the first or last element in the list). This element is then removed from the list, and attached as the right child of its left neighbor, or the left child of its right neighbor, depending on which of the two neighbors has a larger value, breaking ties arbitrarily. This process can be implemented in a single left-to-right pass of the input, and it is easy to see that each element can gain at most one left-, and at most one right child, and that the resulting binary tree is a Cartesian tree of the input sequence. {{sfnp|Kozma|Saranurak|2020}} {{sfnp|Hartmann|Kozma|Sinnamon|Tarjan|2021}}

It is possible to maintain the Cartesian tree of a dynamic input, subject to insertions of elements and [[lazy deletion]] of elements, in logarithmic [[amortized time]] per operation. Here, lazy deletion means that upa todeletion aoperation constantis fractionperformed ofby themarking elementsan element in the current tree mayas bebeing markeda asdeleted deletedelement, ratherbut thannot actually removedremoving it from the tree. When toothe largenumber of marked elements reaches a constant fraction of elementsthe aresize marked,of the whole tree, it is rebuilt, keeping only its unmarked elements.{{sfnp|Bialynicka-Birula|Grossi|2006}}

==Applications==

===Range searching and lowest common ancestors===

[[File:Cartesian tree range searching.svg|thumb|300px|Two-dimensional range-searching using a Cartesian tree: the bottom point (red in the figure) within a three-sided region with two vertical sides and one horizontal side (if the region is nonempty) can be found as the nearest common ancestor of the leftmost and rightmost points (the blue points in the figure) within the slab defined by the vertical region boundaries. The remaining points in the three-sided region can be found by splitting it by a vertical line through the bottom point and recursing.]]

Cartesian trees form part of an efficient [[data structure]] for [[Range Minimum Query|range minimum queries]],. aAn [[rangeinput searching]]to problemthis involvingkind queriesof thatquery askspecifies fora contiguous subsequence of the minimumoriginal valuesequence; inthe aquery contiguousoutput subsequenceshould ofbe the originalminimum sequencevalue in this subsequence.<ref>{{harvtxt|Gabow|Bentley|Tarjan|1984}}; {{harvtxt|Bender|Farach-Colton|2000}}.</ref> In a Cartesian tree, this minimum value can be found at the [[lowest common ancestor]] of the leftmost and rightmost values in the subsequence. For instance, in the subsequence (12,10,20,15,18) of the example sequence shown in the first illustration, the minimum value of the subsequence (10) forms the lowest common ancestor of the leftmost and rightmost values (12 and 1518). Because lowest common ancestors can be found in constant time per query, using a data structure that takes linear space to store and can be constructed in linear time, the same bounds hold for the range minimization problem.<ref>{{harvtxt|Harel|Tarjan|1984}}; {{harvtxt|Schieber|Vishkin|1988}}.</ref>

{{harvtxt|Bender|Farach-Colton|2000}} reversed this relationship between the two data structure problems by showing that data structures for range minimization could also be used for finding lowest common ancestors. Their data structure associates with each node of the tree its distance from the root, and constructs a sequence of these distances in the order of an [[Euler tour]] of the (edge-doubled) tree. It then constructs a range minimization data structure for the resulting sequence. The lowest common ancestor of any two vertices in the given tree can be found as the minimum distance appearing in the interval between the initial positions of these two vertices in the sequence. Bender and Farach-Colton also provide a method for range minimization that can be used for the sequences resulting from this transformation, which have the special property that adjacent sequence values differ by ±1one. As they describe, for range minimization in sequences that do not have this form, it is possible to use Cartesian trees to reduce the range minimization problem to lowest common ancestors, and then to use Euler tours to reduce lowest common ancestors to a range minimization problem with this special form.{{sfnp|Bender|Farach-Colton|2000}}

The same range minimization problem may also be given an alternative interpretation in terms of two dimensional range searching. A collection of finitely many points in the [[Cartesian plane]] can be used to form a Cartesian tree, by sorting the points by their <math>x</math>-coordinates and using the <math>y</math>-coordinates in this order as the sequence of values from which this tree is formed. If <math>S</math> is the subset of the input points within some vertical slab defined by the inequalities <math>L\le x\le R</math>, <math>p</math> is the leftmost point in <math>S</math> (the one with minimum <math>x</math>-coordinate), and <math>q</math> is the rightmost point in <math>S</math> (the one with maximum <math>x</math>-coordinate) then the lowest common ancestor of <math>p</math> and <math>q</math> in the Cartesian tree is the bottommost point in the slab. A three-sided range query, in which the task is to list all points within a region bounded by the three inequalities <math>L\le x\le R</math> and <math>y\le T</math>, can be answered by finding this bottommost point <math>b</math>, comparing its <math>y</math>-coordinate to <math>T</math>, and (if the point lies within the three-sided region) continuing recursively in the two slabs bounded between <math>p</math> and <math>b</math> and between <math>b</math> and <math>q</math>. In this way, after the leftmost and rightmost points in the slab are identified, all points within the three-sided region can be listed in constant time per point.{{sfnp|Gabow|Bentley|Tarjan|1984}}

Line 38 ⟶ 44:

===As a binary search tree===

{{main article|Treap}}

Because a Cartesian tree is a binary tree, it is natural to use it as a [[binary search tree]] for an ordered sequence of values. However, defining a Cartesian tree based on the same values that form the search keys of a binary search tree does not work well: theThe Cartesian tree of a sorted sequence is just a [[path graph|path]], rooted at its leftmost endpoint,. and binaryBinary searching in this tree degenerates to [[sequential search]] in the path. However, ita isdifferent possibleconstruction uses Cartesian trees to generate more-balanced[[binary search treestree]]s byof generatinglogarithmic ''priority''depth valuesfrom forsorted eachsequences searchof keyvalues. thatThis arecan differentbe thandone theby keygenerating itself,''priority'' sortingnumbers thefor inputseach by their key valuesvalue, and using the corresponding sequence of priorities to generate a Cartesian tree. This construction may equivalently be viewed in the geometric framework described above, in which the <math>x</math>-coordinates of a set of points are the searchvalues keysin a sorted sequence and the <math>y</math>-coordinates are thetheir priorities.{{sfnp|Seidel|Aragon|1996}}

This idea was applied by {{harvtxt|Seidel|Aragon|1996}}, who suggested the use of random numbers as priorities. The data[[self-balancing structurebinary search tree]] resulting from this random choice is called a [[treap]], due to its combination of binary search tree and binary min-heap features. An insertion into a treap can be performed by inserting the new key as a leaf of an existing tree, choosing a priority for it, and then performing [[tree rotation]] operations along a path from the node to the root of the tree to repair any violations of the heap property caused by this insertion; a deletion can similarly be performed by a constant amount of change to the tree followed by a sequence of rotations along a single path in the tree.{{sfnp|Seidel|Aragon|1996}} A variation on this data structure called a zip tree uses the same idea of random priorities, but simplifies the random generation of the priorities, and performs insertions and deletions in a different way, by splitting the sequence and its associated Cartesian tree into two subsequences and two trees and then recombining them.{{sfnp|Tarjan|Levy|Timmel|2021}}

If the priorities of each key are chosen randomly and independently once whenever the key is inserted into the tree, the resulting Cartesian tree will have the same properties as a [[random binary search tree]], a tree computed by inserting the keys in a randomly chosen [[permutation]] starting from an empty tree, with each insertion leaving the previous tree structure unchanged and inserting the new node as a leaf of the tree. Random binary search trees hadhave been studied for much longer than treaps, and are known to behave well as search trees. (theyThe have[[expected value|expected]] length of the search path to any given value is at most <math>2\ln n</math>, and the whole tree has [[logarithm]]ic depth (its maximum root-to-leaf distance) [[with high probability);]]. More formally, there exists a constant <math>C</math> such that the depth is <math>\le C\ln n</math> with probability tending to one as the number of nodes tends to infinity. The same good behavior carries over to treaps. It is also possible, as suggested by Aragon and Seidel, to reprioritize frequently-accessed nodes, causing them to move towards the root of the treap and speeding up future accesses for the same keys.{{sfnp|Seidel|Aragon|1996}}

===In sorting===

Line 56 ⟶ 62:

#* Add the Cartesian tree children of the removed value to the priority queue

As Levcopoulos and Petersson show, for input sequences that are already nearly sorted, the size of the priority queue will remain small, allowing this method to take advantage of the nearly-sorted input and run more quickly. Specifically, the worst-case running time of this algorithm is <math>O(n\log k)</math>, where <math>n</math> is the sequence length and <math>k</math> is the average, over all values in the sequence, of the number of consecutive pairs of sequence values that bracket the given value (meaning that the given value is between the two sequence values). They also prove a lower bound stating that, for any <math>n</math> and (non-constant) <math>k</math>, any comparison-based sorting algorithm must use <math>\Omega(n\log k)</math> comparisons for some inputs.{{sfnp|Levcopoulos|Petersson|1989}}

===In pattern matching===

The problem of ''Cartesian tree matching'' has been defined as a generalized form of [[string matching]] in which one seeks a [[substring]] (or in some cases, a [[subsequence]]) of a given string that has a Cartesian tree of the same form as a given pattern. Fast algorithms for variations of the problem with a single pattern or multiple patterns have been developed, as well as data structures analogous to the [[suffix tree]] and other text indexing structures.<ref>{{harvtxt|Park|Amir|Landau|Park|2019}}; {{harvtxt|Park|Bataa|Amir|Landau|2020}}; {{harvtxt|Song|Gu|Ryu|Faro|2021}}; {{harvtxt|Kim|Cho|2021}}; {{harvtxt|Nishimoto|Fujisato|Nakashima|Inenaga|2021}}; {{harvtxt|Oizumi|Kai|Mieno|Inenaga|2022}}</ref>

==History==

Cartesian trees were introduced and named by {{harvtxt|Vuillemin|1980}}. The name is derived from the [[Cartesian coordinate]] system for the plane: in one version of this structure, as in the two-dimensional range searching application discussed above, a Cartesian tree for a point set has the sorted order of the points by their <math>x</math>-coordinates as its symmetric traversal order, and it has the heap property according to the <math>y</math>-coordinates of the points. {{harvtxt|Vuillemin|1980}} described both this geometric version of the structure, and the definition here in which a Cartesian tree is defined from a sequence. Using sequences instead of point coordinates provides a more general setting that allows the Cartesian tree to be applied to non-geometric problems as well.{{sfnp|Vuillemin|1980}}

==Notes==

Line 101 ⟶ 104:

| title = STACS 2006, 23rd Annual Symposium on Theoretical Aspects of Computer Science, Marseille, France, February 23-25, 2006, Proceedings

| volume = 3884

| year = 2006}}| isbn = 978-3-540-32301-3

}}

*{{citation|contribution=On Cartesian trees and range minimum queries|first1=Erik D.|last1=Demaine|author1-link=Erik Demaine|first2=Gad M.|last2=Landau|first3=Oren|last3=Weimann|series=Lecture Notes in Computer Science|volume=5555|year=2009|pages=341–353|doi=10.1007/978-3-642-02927-1_29|title=Automata, Languages and Programming, 36th International Colloquium, ICALP 2009, Rhodes, Greece, July 5-12, 2009|isbn=978-3-642-02926-4|hdl=1721.1/61963|hdl-access=free}}

*{{citation

Line 150 ⟶ 154:

| volume = 13

| year = 1984}}

*{{citation

*{{citation|title=The maximum capacity route problem|first=T. C.|last=Hu|journal=Operations Research|volume=9|issue=6|year=1961|pages=898–900|jstor=167055|doi=10.1287/opre.9.6.898|doi-access=free}}

| last1 = Hartmann | first1 = Maria

| last2 = Kozma | first2 = László

| last3 = Sinnamon | first3 = Corwin

| last4 = Tarjan | first4 = Robert E. | author4-link = Robert Tarjan

| doi =10.4230/LIPIcs.ICALP.2021.79

| journal = [[ICALP]]

| pages = 79:1–79:20

| title = Analysis of Smooth Heaps and Slim Heaps

| series = Leibniz International Proceedings in Informatics (LIPIcs)

| year = 2021| volume = 198

| doi-access = free

| isbn = 978-3-95977-195-5

}}

*{{citation|title=The maximum capacity route problem|first=T. C.|last=Hu|author-link=T. C. Hu|journal=Operations Research|volume=9|issue=6|year=1961|pages=898–900|jstor=167055|doi=10.1287/opre.9.6.898|doi-access=free}}

*{{citation

| last1 = Kim | first1 = Sung-Hwan

Line 163 ⟶ 181:

| title = 32nd Annual Symposium on Combinatorial Pattern Matching, CPM 2021, July 5-7, 2021, Wrocław, Poland

| volume = 191

| year = 2021| doi-access = free
| isbn = 9783959771863

}}

*{{citation

| last1 = Kozma | first1 = László

| last2 = Saranurak | first2 = Thatchaphol

| doi = 10.1137/18M1195188

| number = 5

| journal = [[SIAM J. Comput.]]

| publisher = SIAM

| title = Smooth Heaps and a Dual View of Self-Adjusting Data Structures

| volume = 49

| year = 2020

| arxiv = 1802.05471

}}

*{{citation

Line 185 ⟶ 216:

| title = WADS '89: Proceedings of the Workshop on Algorithms and Data Structures

| volume = 382

| year = 1989}}| isbn = 978-3-540-51542-5

}}

*{{citation

| last1 = Nishimoto | first1 = Akio

Line 201 ⟶ 233:

| title = String Processing and Information Retrieval - 28th International Symposium, SPIRE 2021, Lille, France, October 4-6, 2021, Proceedings

| volume = 12944

| year = 2021| s2cidisbn = 235313506978-3-030-86691-4

| s2cid = 235313506

}}

*{{citation

Line 218 ⟶ 251:

| title = 33rd Annual Symposium on Combinatorial Pattern Matching, CPM 2022, June 27-29, 2022, Prague, Czech Republic

| volume = 223

| year = 2022| doi-access = free
| isbn = 9783959772341

| s2cid = 246679910

}}

Line 236 ⟶ 270:

| title = 30th Annual Symposium on Combinatorial Pattern Matching, CPM 2019, June 18-20, 2019, Pisa, Italy

| volume = 128

| year = 2019| doi-access = free
| isbn = 9783959771030

| s2cid = 162168587

}}

Line 262 ⟶ 297:

| url = http://citeseer.ist.psu.edu/seidel96randomized.html

| volume = 16

| year = 1996| doi-broken-date = 1 July 2025 }}

*{{citation

| last1 = Schieber | first1 = Baruch