Lecture Notes on

Analysis and Design of Algorithms

BCS401

Module-2 :Divide and Conquer

Contents
1. General method
2. Recurrence equation
3. Algorithm: Binary search
4. Algorithm: Finding the maximum and minimum
5. Algorithm: Merge sort
6. Algorithm: Quick sort
7. Advantages and Disadvantages
8. Decrease and Conquer Approach
9. Algorithm: Topological Sort
ADA BCS401

1. General method
Divide and Conquer is one of the best-known general algorithm design echnique. It works
according to the following general plan:
 Given a function to compute on ‘n’ inputs the divide-and-conquer strategy suggests
splitting the inputs into ‘k’ distinct subsets, 1<k<=n, yielding ‘k’ sub problems.
 These sub problems must be solved, and then a method must be found to combine sub
solutions into a solution of the whole.
 If the sub problems are still relatively large, then the divide-and-conquer strategy can
possibly be reapplied.
 Often the sub problems resulting from a divide-and-conquer design are of the same
type as the original problem.For those cases the reapplication of the divide-and-
conquer principle is naturally expressed by a recursive algorithm.

A typical case with k=2is diagrammatically shown below.

Problem of size n

Sub Problem of size n/2 Sub Problem of size n/2

Solution to sub problem 1 Solution to sub problem 2

Solution to the original problem

Control Abstraction for divide and conquer:

In the above specification,

 Initially DAndC(P) is invoked, where ‘P’ is the problem to be solved.
 Small (P) is a Boolean-valued function that determines whether the input size is small
enough that the answer can be computed without splitting. If this so, the function ‘S’
is invoked.Otherwise, the problem P is divided into smaller sub problems.These sub
problems P1, P2 …Pk are solved by recursive application of DAndC.
 Combine is a function that determines the solution to P using the solutions to the ‘k’
sub problems.

CSE, SVIT 1
ADA BCS401

2. Recurrence equation for Divide and Conquer

If the size of problem ‘p’ is n and the sizes of the ‘k’ sub problems are n1, n2….nk,
respectively, then the computing time of divide and conquer is described by the recurrence
relation

Where,
 T(n) is the time for divide and conquer method on any input of size n and
 g(n) is the time to compute answer directly for small inputs.
 The function f(n) is the time for dividing the problem ‘p’ and combining the solutions
to sub problems.
For divide and conquer based algorithms that produce sub problems of the same type as the
original problem, it is very natural to first describe them by using recursion.
More generally, an instance of size n can be divided into b instances of size n/b, with a of
them needing to be solved. (Here, a and b are constants; a>=1 and b > 1.). Assuming that
size n is a power of b(i.e. n=bk), to simplify our analysis, we get the following recurrence for
the running time T(n):
..... (1)
where f(n) is a function that accounts for the time spent on dividing the problem into smaller
ones and on combining their solutions.
The recurrence relation can be solved by i) substitution method or by using ii) master
theorem.
1. Substitution Method - This method repeatedly makes substitution for each
occurrence of the function T in the right-hand side until all such occurrences
disappear.
2. Master Theorem - The efficiency analysis of many divide-and-conquer algorithms is
greatly simplified by the master theorem. It states that, in recurrence equation T(n) =
aT(n/b) + f (n), If f (n)∈ Θ (nd) where d ≥ 0 then

Analogous results hold for the Ο and Ω notations, too.

For example, the recurrence for the number of additions A(n) made by thedivide-and-
conquer sum-computation algorithm (see above) on inputs of sizen = 2k is

Thus, for this example, a = 2, b = 2, and d = 0; hence, since a >bd,

CSE, SVIT 2
ADA BCS401

Problems on Substitution method & Master theorem to solve the recurrence relation

CSE, SVIT 3
ADA BCS401

CSE, SVIT 4
ADA BCS401

CSE, SVIT 5
ADA BCS401

CSE, SVIT 6
ADA BCS401

3. Binary Search
Problem definition:Let ai, 1 ≤ i ≤ n be a list of elements that are sorted in non-decreasing
order. The problem is tofind whether a given element x is present in the list or not. If x is
present we have to determine a value j (element’s position) such that aj=x. If x is not in the
list, then j is set to zero.
Solution:Let P = (n, ai…al , x) denote an arbitrary instance of search problem where n is the
number of elements in the list, ai…alis the list of elements andx is the key element to be
searched for in the given list.Binary search on the list is done as follows:
Step1: Pick an index q in the middle range [i, l] i.e. q= (𝑛 + 1)/2] and compare x with aq.
Step 2: if x = aqi.e key element is equal to mid element, the problem is immediately solved.
Step 3: if x <aqin this case x has to be searched for only in the sub-list ai, ai+1, ……, aq-
Therefore, problem reduces to (q-i, ai…aq-1, x).
Step 4: if x >aq,x has to be searched for only in the sub-list aq+1, ...,., al . Therefore problem
reduces to (l-i, aq+1…al, x).
For the above solution procedure, the Algorithm can be implemented as recursive or non-
recursive algorithm.

CSE, SVIT 7
ADA BCS401

Recursive binary search algorithm

Iterative binary search:

CSE, SVIT 8
ADA BCS401

Analysis
In binary search the basic operation is key comparison. Binary Search can be analyzed with
the best, worst, and average case number of comparisons.The numbers of comparisons for the
recursive and iterative versions of Binary Search are the same, if comparison counting is
relaxed slightly. For Recursive Binary Search, count each pass through the if-then-else block
as one comparison. For Iterative Binary Search, count each pass through the while block as
one comparison.Let us find out how many such key comparison does the algorithm make on
an array of n elements.
Best case –Θ(1) In the best case, the key is the middle in the array. A constant number of
comparisons (actually just 1) are required.
Worst case - Θ(log2 n) In the worst case, the key does not exist in the array at all. Through
each recursion or iteration of Binary Search, the size of the admissible range is halved. This
halving can be done ceiling(log2n ) times. Thus, ] log2 n ] comparisons are required.
Sometimes, in case of the successful search, it may take maximum number of comparisons.
] log2 n ]. So worst case complexity of successful binary search is Θ (log2 n).
Average case - Θ (log2n) To find the average case, take the sum of the product of number of
comparisons required to find each element and the probability of searching for that element.
To simplify the analysis, assume that no item which is not in array will be searched for, and
that the probabilities of searching for each element are uniform.

Space Complexity - The space requirements for the recursive and iterative versions of binary
search are different. Iterative Binary Search requires only a constant amount of space, while
Recursive Binary Search requires space proportional to the number of comparisons to
maintain the recursion stack.
Advantages: Efficient on very big list, Can be implemented iteratively/recursively.
Limitations:
 Interacts poorly with the memory hierarchy
 Requires sorted list as an input
 Due to random access of list element, needs arrays instead of linked list.

CSE, SVIT 9
ADA BCS401

4. Finding the maximum and minimum

Problem statement:Given a list of n elements, the problem is to find the maximum and
minimum items.
StraightMaxMin: A simple and straight forward algorithm to achieve this is given below.

Explanation:
 StraightMaxMin requires 2(n-1) comparisons in the best, average & worst cases.
 By realizing the comparison of a[i]>max is false, improvement in a algorithm can be
done. Hence we can replace the contents of the for loop by,
If(a[i]>Max) then Max = a[i]; Else if (a[i]<min) min=a[i]
 On the average a[i] is > max half the time.So, the avg. no. of comparison is 3n/2-1.

Algorithm based on Divide and Conquer strategy

Let P = (n, a [i],……,a [j]) denote an arbitrary instance of the problem. Here ‘n’ is the no. of
elements in the list (a[i],….,a[j]) and we are interested in finding the maximum and minimum
of the list.If the list has more than 2 elements, P has to be divided into smaller instances.
For example, we might divide ‘P’ into the 2 instances,
P1=( [n/2],a[1], a[n/2])
P2= ( n-[n/2], a[[n/2]+1], , a[n])
After having divided ‘P’ into 2 smaller sub problems, we can solve them
byrecursivelyinvoking the same divide-and-conquer algorithm.
Algorithm:

CSE, SVIT 10
ADA BCS401

Example:

of recursive calls of MaxMin is as follows

Analysis - Time Complexity

CSE, SVIT 11
ADA BCS401

Compared with the straight forward method (2n-2) this method saves 25% in comparisons.

Space Complexity
Compared to the straight forward method, the MaxMin method requires extra stack space for
i, j, max, min, max1 and min1. Given n elements there will be 𝑙o𝑔2𝑛] + 1 levels of
recursion and we need to save seven values for each recursive call. (6 + 1 for return address).

5. Merge Sort
Merge sort is a perfect example of a successful application of the divide-and conquer
technique. It sorts a given array A [O ... n - 1] by dividing it into two halves A [0 .. 𝑛/2]-1]
and A [ /2] .. n-1], sorting each of them recursively, and then merging the two smaller sorted
arrays into a single sorted one.

The merging of two sorted arrays can be done as follows.

 Two pointers (array indices) are initialized to point to the first elements of the arrays
being merged.
 The elements pointed to are compared, and the smaller of them is added to a new
array being constructed

CSE, SVIT 12
ADA BCS401

 After that, the index of the smaller element is incremented to point to its immediate
successor in the array it was copied from. This operation is repeated until one of the
two given arrays is exhausted, and then the remaining elements of the other array are
copied to the end of the new array.

Example:
The operation of the algorithm on the
list 8, 3, 2, 9, 7, 1, 5, 4 is illustrated in
the figure

Analysis

Here the basic operationis key comparison. As merge sort execution does not depend on the
order of the data, best case and average case runtime are the same as worst case runtime.

Worst case:During key comparison, neither of the two arrays becomes empty before the
other one contains just one element leads to the worst case of merge sort. Assuming for

CSE, SVIT 13
ADA BCS401

simplicity that total number of elementsn is a power of 2, therecurrence relation for the
number of key comparisons C(n) is

where, Cmerge(n) is the number of key comparisons made during the merging stage.
Let us analyze Cmerge(n), the number of key comparisons performed during themerging stage.
At each step, exactly one comparison is made, after which the totalnumber of elements in the
two arrays still needing to be processed is reducedby 1. In the worst case, neither of the two
arrays becomes empty before theother one contains just one element (e.g., smaller elements
may come from thealternating arrays).Therefore, for the worst case,Cmerge(n) = n –1.
Now,

Solving the recurrence equation using master theorem:

Here a = 2, b = 2, f (n) = n, d = 1. Therefore 2 = 21, case 2 holds in the master theorem
Cworst(n) = Θ (nd log n) = Θ (n1 log n) = Θ (n log n)Therefore Cworst(n) = Θ (n log n)

Advantages:
 Number of comparisons performed is nearly optimal.
 For large n, the num er of comparisons made by this algorithm in the average case
turns out to be about 0.25n less and hence is also in Θ(n log n).
 Mergesort will never degrade to O (n2)
 Anotheradvantage of mergesort over quicksort and heapsort is its stability. (A sorting
algorithm is said to be stable if two objects with equal keys appear in the same order
in sorted output as they appear in the input array to be sorted. )
Limitations:
 The principal shortcoming of mergesort is the linear amount [O(n) ] of extra storage
the algorithm requires. Though merging can be done in-place, the resulting algorithm
is quite complicated and of theoretical interest only.

Variations of merge sort

1. Thealgorithm can be implemented bottom up by merging pairs of the array’s
elements,then merging the sorted pairs, and so on. (If n is not a power of 2, only
slightbookkeeping complications arise.) This avoids the time and space overhead
ofusing a stack to handle recursive calls.
2. We can divide a list to be sortedin more than two parts, sort each recursively, and
then merge them tog ther. Thisscheme, which is particularly useful for sorting files
residing on secondary memorydevices, is called multiway mergesort.

CSE, SVIT 14
ADA BCS401

6. Quick sort
Quicksort is the other important sorting algorithm that is based on the divide-and-conquer
approach. Unlike mergesort, which divides its input elements according to their position in
the array, quicksort divides ( or partitions) them according to their value.
A partition is an arrangement of the array’s elements so that all the elements to the left of
some element A[s] are less than or equal to A[s], and all the elements to the right of A[s] are
greater than or equal to it:

Obviously, after a partition is achieved, A[s] will be in its final position in thesorted array,
and we can continue sorting the two subarrays to the left and to theright of A[s]
independently (e.g., by the same method).
In quick sort, the entire workhappens in the division stage, with no work required to combine
the solutions tothe subproblems.

Partitioning
We start by selecting a pivot—an element with respect to whose valuewe are going to divide
the subarray.There are several different strategies forselecting a pivot. We use
thesophisticated method suggested byC.A.R. Hoare, the prominent British computer scientist
who invented quicksort.
Select the subarray’s firstelement: p = A[l].Now scan the subarray from both ends,comparing
the subarray’s elements to the pivot.
 The left-to-right scan, denoted below by index pointer i, starts with the second
element. Since we want elements smaller than the pivot to be in the left part of the
subarray, this scan skips over elements that are smaller than the pivot and stops upon
encountering the first element greater than or equal to the pivot.
 The right-to-left scan, denoted below by index pointer j, starts with the last element of
the subarray. Since we want elements larger than the pivot to be in the right part of the

CSE, SVIT 15
ADA BCS401

subarray, this scan skips over elements that are larger than the pivot and stops on
encountering the first element smaller than or equal to the pivot.
After both scans stop, three situations may arise, depending on whether or notthe scanning
indices have crossed.
1. If scanning indices i and j have not crossed, i.e., i< j, we simply exchange A[i] and
A[j ] and resume the scans by incrementing I and decrementing j, respectively:

2. If the scanning indices have crossed over, i.e., i> j, we will have partitioned the
subarray after exchanging the pivot with A[j]:

3. If the scanning indices stop while pointing to the same element, i.e., i = j, the value
they are pointing to must be equal to p. Thus, we have the subarray partitioned, with
the split position s = i = j :

We can combine this with the case-2 byexchanging the pivot with A[j] whenever i≥j

ALGORITHM HoarePartition(A[l..r])
//Partitions a subarray by Hoare’s algorithm, using the first element as a pivot
//Input: Subarray of array A[0..n − 1], defined by its left and right indices l and r (l<r)
//Output: Partition of A[l..r], with the split position returned as this function’s value

Note that index i can go out of the subarray’s bounds in this pseudocode.

CSE, SVIT 16
ADA BCS401

Example: Example of quicksort operation. (a) Array’s transformations with pivotsshown in

bold. (b) Tree of recursive calls to Quicksort with input values land r of subarray bounds and
split position s of a partition obtained.

Analysis
Best Case -Here the basic operation is key comparison. Numberof key comparisons made
before a partition is achieved is n + 1 if the scanningindices cross over and n if they coincide.
If all the splits happen in themiddle of corresponding subarrays, we will have the best case.
The number of keycomparisons in the best case satisfies the recurrence,

According to the Master Theorem, Cbest(n) ∈Θ(n log2 n); solving it exactly forn = 2k yields
Cbest(n) = n log2 n.
Worst Case – In the worst case, all the splits will be skewed to the extreme: one of thetwo
subarrays will be empty, and the size of the other will be just 1 less than thesize of the
subarray being partitioned. This unfortunate situation will happen, inparticular, for increasing
arrays. Indeed, if A[0..n − 1] is a strictly increasing array and we use A[0] as thepivot, the

CSE, SVIT 17
ADA BCS401

left-to-right scan will stop on A[1] while the right-to-left scan will go allthe way to reach
A[0], indicating the split at position 0:So, after making n + 1 comparisons to get to this
partition and exchanging thepivot A[0] with itself, the algorithm will be left with the strictly
increasing arrayA[1..n − 1] to sort. This sorting of strictly increasing arrays of diminishing
sizes willcontinue until the last one A[n−2.. n−1] has been processed. The total numberof key
comparisons made will be equal to

Average Case - Let Cavg(n) be the average number of key comparisons made byquicksort on
a randomly ordered array of size n. A partition can happen in anyposition s (0 ≤ s ≤ n−1) after
n+1comparisons are made to achieve the partition.After the partition, the left and right
subarrays will have s and n − 1− s elements,respectively. Assuming that the partition split can
happen in each position s withthe same probability 1/n, we get the following recurrence
relation:

Its solution, which is much trickier than the worst- and best-case analyses, turnsout to be

Thus, on the average, quicksort makes only 39% more comparisons than in thebest case.
Moreover, its innermost loop is so efficient that it usually runs faster thanmergesort on
randomly ordered arrays of nontrivial sizes. This certainly justifies the namegiven to the
algorithm by its inventor.
Variations: Because of quicksort’s importance, there have been persistent efforts over
theyears to refine the basic algorithm. Among several improvements discovered
byresearchers are:
 Better pivot selection methods such as randomized quicksort that uses a random
element or the median-of-three method that uses the median of the leftmost,
rightmost, and the middle element of the array
 Switching to insertion sort on very small subarrays (between 5 and 15 elements for
most computer systems) or not sorting small subarrays at all and finishing the
algorithm with insertion sort applied to the entire nearly sorted array
 Modifications of the partitioning algorithm such as the three-way partition into
segments smaller than, equal to, and larger than the pivot
Limitations: 1. It is not stable. 2. It requires a stack to store parameters of subarrays that are
yet to be sorted. 3. WhilePerformance on randomly ordered arrays is known to be sensitive
not only to the implementation details of the algorithm but also to both computer architecture
and data type.

CSE, SVIT 18
ADA BCS401

7. Advantages and Disadvantages of Divide & Conquer

Advantages
1. Parallelism: Divide and conquer algorithms tend to have a lot of inherent parallelism.
Once the division phase is complete, the sub-problems are usually independent and
can therefore be solved in parallel. This approach typically generates more enough
concurrency to keep the machine busy and can be adapted for execution in multi-
processor machines.
2. Cache Performance: divide and conquer algorithms also tend to have good cache
performance. Once a sub-problem fits in the cache, the standard recursive solution
reuses the cached data until the sub-problem has been completely solved.
3. It allows solving difficult and often impossible looking problems like the Tower of
Hanoi. It reduces the degree of difficulty since it divides the problem into sub
problems that are easily solvable, and usually runs faster than other algorithms would.
4. Another advantage to this paradigm is that it often plays a part in finding other
efficient algorithms, and in fact it was the central role in finding the quick sort and
merge sort algorithms.
Disadvantages
5. One of the most common issues with this sort of algorithm is the fact that the
recursion is slow, which in some cases outweighs any advantages of this divide and
conquer process.
6. Another concern with it is the fact that sometimes it can become more complicated
than a basic iterative approach, especially in cases with a large n. In other words, if
someone wanted to add a large amount of numbers together, if they just create a
simple loop to add them together, it would turn out to be a much simpler approach
than it would be to divide the numbers up into two groups, add these groups
recursively, and then add the sums of the two groups together.
7. Another downfall is that sometimes once the problem is broken down into sub
problems, the same sub problem can occur many times. It is solved again. In cases
like these, it can often be easier to identify and save the solution to the repeated sub
problem, which is commonly referred to as memorization.

8. Decrease and Conquer Approach

Decrease-and-conquer is a general algorithm design technique, based on exploiting a
relationship between a solution to a given instance of a problem and a solution to a smaller
instance of the same problem. Once such a relationship is established, it can be exploited
either top down (usually recursively) or bottom up.There are three major variations of
decrease-and-conquer:
1. decrease-by-a-constant, most often by one (e.g., insertion sort)
2. decrease-by-a-constant-factor, most often by the factor of two (e.g., binary search)
3. variable-size-decrease (e.g., Euclid’s algorithm)

CSE, SVIT 19
ADA BCS401

In the decrease-by-a-constant variation, the size of an instance is reducedby the same

constant on each iteration of the algorithm. Typically, this constant is equal to one although
other constant size reductions do happenoccasionally.

Figure: Decrease-(by one)-and-conquer

technique
Example: an = an-1×a

The decrease-by-a-constant-factor technique suggests reducing a probleminstance by the

same constant factor on each teration of the algorithm. In mostapplications, this constant
factor is equal to two.

Figure: Decrease-(byhalf)-and-conquer
technique.

Example:

Finally, in the variable-size-decrease variety of decrease-and-conquer, thesize-reduction

pattern varies from one iteration of an algorithm to another.
Example: Euclid’salgorithm for computing the greatest common divisor. It is based on the
formula. gcd(m, n) = gcd(n, m mod n).
Though the value of the second argument is always smaller on the right-hand sidethan on the
left-hand side, it decreases neither by a constant nor by a constantfactor.

CSE, SVIT 20
Insertion sort

It is an application for decease-by-one technique to sort an array A[0…n-1]. Everytime the element is
inserted, it is stored in its position. So after every insertion the array remains sorted.

The number of key comparisons for such an input is

is the worst case.

For sorted array, the number of key comparisons is

And average case is Cavg(n) ᴝn2\4 € Θ(n2)

9. Topological Sort
Background: A directed graph, or digraph for short, is a graph with directions specified for
all its edges. The adjacency matrix and adjacency lists are the two principal means of
representing a digraph.Ther are only two notable differences between undirected and
directed graphs in representing them: (1) the adjacency matrix of a directed graph does not
have to be symmetric; (2) an edge in a directed graph has just one (not wo) corresponding
nodes in the digraph’s adjacency lists.
Depth-first search and breadth-first search are principal traversal algorithms for traversing
digraphs as well, but the structure of corresponding forests can be more complex than for
undirected graphs. Thus, even for the simple example of Figure, the depth-first search forest
(Figure b) exhibits all four types of edges possible in a DFS forest of a directed graph:
 tree edges (ab, bc, de),
 back edges (ba) from vertices to their ancestors,
 forward edges (ac) from vertices to their descendants in the tree other than their
children, and
 cross edges (dc), whichare none of the aforementioned types.

Note that a back edge in a DFS forest of a directed graph can connect a vertexto its parent.
Whether or not it is the case, the presence of a back edge indicatesthat the digraph has a
directed cycle. A directed cycle in a digraph is a sequenceof three or more of its vertices that
starts and ends with the same vertex and inwhich every vertex is connected to its immediate
predecessor by an edge directedfrom the predecessor to the successor. For example, a, b, a is
a directed cycle in the digraph in Figure given above. Conversely, if a DFS forest of a
digraph has no backedges, the digraph is a dag, an acronym for directed acyclic graph.

Motivation for topological sorting: Consider a set of five required

courses {C1, C2, C3, C4,C5} a part-time student has to take in some
degree program. The courses can betaken in any order as long as the
following course prerequisites are met: C1 andC2 have no prerequisites,
C3 requires C1 and C2, C4 requires C3, and C5 requiresC3 and C4. The
student can take only one course per term. In which order shouldthe student take the
ADA BCS401

courses?The situation can be modeled by a digraph in which vertices represent courses and
directed edges indicate prerequisite requirements.
In terms of this digraph, the uestion is whether we can list its vertices in such an order that
for every edge in the graph, the vertex where the edge starts is listed before the vertex where
the edge ends. In other words, can you find such an ordering of this digraph’s vertices? This
problem is called topological sorting.
Topological Sort: For topological sorting to be possible, a digraph in question must be a
DAG. i.e., if a digraph has nodirected cycles, the topological sorting problem for it has a
solution.
There are two efficient algorithms that both verify whether a digraph is a dag and, if it is,
produce an ordering of vertices that solves the topological sortingproblem.The first one is
based on depth-first search; the second is based on a direct application of the decrease-by-one
technique.
Topological Sorting based on DFS
Method
1. Perform a DFS traversal and note the order in which vertices become dead-ends
2. Reversing this order yields a solution to the topological sorting problem, provided, of
course, no back edge has been encountered during the traversal. If a back edge has
been encountered, the digraph is not a DAG, and topological sorting of its vertices is
impossible.
Illustration
a) Digraph for which the topological sorting problem needs to be solved.
b) DFS traversal stack w th the subscript numbers indicating the popping off order.
c) Solution to the problem. Here we have drawn the edges of the digraph, and they all
point from left to right as the problem’s statement requires. It is a convenient way to
check visually the correctness of a solution to an instance of the topological sorting
problem.

Topological Sorting using decrease-and-conquer technique

Method: The algorithm is based on a direct implementation of the decrease-(byone)-and-
conquer technique:

CSE, SVIT 23
ADA BCS401

1. Repeatedly, identify in a remaining digraph a source, which is a vertex with no

incoming edges, and delete it along with all the edges outgoing from it. (If there are
several sources, break the tie arbitrarily. If there are none, stop because the problem
cannot be solved.)
2. The order in which the vertices are deleted yields a solution to the topological sorting
problem.
Illustration - Illustration of the source-removal algorithm for the topological sorting problem
is given here. On each iteration, a vertex with no incoming edges is deleted from the digraph.

Note: The solution obtained by the source-removal algorithm is different from the one
obtained by the DFS-based algorithm. Both of them are correct, of course; the topological
sorting problem may have several alternative solutions.
Depth first search:

CSE, SVIT 24
Breadth first search:
Applications of Topological Sorting
 Instruction scheduling in program compilation
 Cell evaluation ordering in spreadsheet formulas,
 Resolving symbol dependencies in linkers.

***