Prefix Sum
The prefix sum is an incredibly powerful and straightforward technique. Its primary goal is to allow for constant time range sum queries on an array.
What is Prefix Sum?
The prefix sum of an array at index 'i' is the sum of all numbers from index '0' to 'i'. By computing and storing the prefix sum of an array, you can easily answer queries about the sum of any subarray in constant time.
1# Example for Prefix Sum in Python
2
3def prefix_sum_array(arr):
4 prefix_sum = [0]
5 for num in arr:
6 prefix_sum.append(prefix_sum[-1] + num)
7 return prefix_sum
8
9arr = [1, -20, -3, 30, 5, 4]
10print(prefix_sum_array(arr))
Prefix Sum vs. Sliding Window
Both prefix sum and sliding window are techniques to solve subarray-related problems. The choice between them depends on the nature of the problem:
- Prefix sum is generally used for problems that ask about the sum of subarrays, as it allows for constant-time range sum queries.
- The sliding window is typically utilized for problems related to contiguous subarrays of a certain size or condition, e.g., maximum sum of a subarray of size 'k'.
Dive Deeper into Subarray Sum Problems
Understanding how to manipulate subarrays is key to mastering many algorithms.
Subarray with Given Sum
The subarray sum problem, as showcased above, tries to find a contiguous subarray that sums up to a particular number.
Zero Sum Subarrays
A variant of the subarray sum problem is finding subarrays that sum to zero. This can be solved using the prefix sum approach and hashing.
Longest Subarray with Sum k
In some problems, instead of finding any subarray with a given sum, you might be tasked with finding the longest subarray with a specific sum. This is a slight twist but can be approached using similar methods.
Smallest Subarray with Sum Greater Than a Given Value
Another variant would be to find the smallest subarray whose sum is greater than a given value.
Maximum Sum of Subarrays of Size k
To find the maximum sum of all subarrays of a specific size 'k', you can utilize the sliding window technique.
Suffix Sum
Just as we have prefix sum that calculates the sum from the beginning to a specific index, we can also calculate suffix sum, which calculates the sum from a particular index to the end of the array.
Now let's practice one of such problems.
Practice Problem: Subarray Sum
Given an array of integers and an integer target, find a subarray that sums to target and return the start and end indices of the subarray.
Input:
arr: 1 -20 -3 30 5 4
target: 7
Output: 1 4
Explanation: -20 - 3 + 30 = 7. The indices for subarray [-20,-3,30] is 1 and 4 (right exclusive).
Try it yourself
Explanation
Intuition
The brute force way is to find the sum of each subarray and compare it with the target. Let N be the number of elements in the array, there are N subarrays with size 1, N-1 subarrays with size 2 .. and 1 subarray with size N. Time complexity is O(N^2).
A key observation is that the the sum of a subarray [i, j] is equal to the sum of [0, j] minus the sum of [0, i - 1].
The sum of elements from index 0 to i is called the prefix sum (prefix = from the beginning, prefix sum = sum from the beginning, i.e. index 0). If we have the prefix sum for each index, we can find the sum of any subarray in constant time.
With the knowledge of the prefix sum under our belt, the problem boils down to Two Sum Sorted. We keep a dictionary of prefix_sum: index while going through the array calculating prefix_sum. If at any point, prefix_sum - target is in the dictionary, we have found our subarray.
Time Complexity: O(n)
Space Complexity: O(n)
Solution
1from typing import List
2
3def subarray_sum(arr: List[int], target: int) -> List[int]:
4 # prefix_sum 0 happens when we have an empty array
5 prefix_sums = {0: 0}
6 cur_sum = 0
7 for i in range(len(arr)):
8 cur_sum += arr[i]
9 complement = cur_sum - target
10 if complement in prefix_sums:
11 return [prefix_sums[complement], i + 1]
12 prefix_sums[cur_sum] = i + 1
13
14if __name__ == '__main__':
15 arr = [int(x) for x in input().split()]
16 target = int(input())
17 res = subarray_sum(arr, target)
18 print(' '.join(map(str, res)))
191import java.util.Arrays;
2import java.util.HashMap;
3import java.util.List;
4import java.util.Scanner;
5import java.util.stream.Collectors;
6
7class Solution {
8 public static List<Integer> subarraySum(List<Integer> arr, int target) {
9 HashMap<Integer, Integer> prefixSums = new HashMap<>();
10 // prefix_sum 0 happens when we have an empty array
11 prefixSums.put(0, 0);
12 int curSum = 0;
13 for (int i = 0; i < arr.size(); i++) {
14 curSum += arr.get(i);
15 int complement = curSum - target;
16 if (prefixSums.containsKey(complement)) {
17 return List.of(prefixSums.get(complement), i + 1);
18 }
19 prefixSums.put(curSum, i + 1);
20 }
21 return null;
22 }
23
24 public static List<String> splitWords(String s) {
25 return s.isEmpty() ? List.of() : Arrays.asList(s.split(" "));
26 }
27
28 public static void main(String[] args) {
29 Scanner scanner = new Scanner(System.in);
30 List<Integer> arr = splitWords(scanner.nextLine()).stream().map(Integer::parseInt).collect(Collectors.toList());
31 int target = Integer.parseInt(scanner.nextLine());
32 scanner.close();
33 List<Integer> res = subarraySum(arr, target);
34 System.out.println(res.stream().map(String::valueOf).collect(Collectors.joining(" ")));
35 }
36}
371function subarraySum(arr, target) {
2 // prefix_sum 0 happens when we have an empty array
3 const prefixSums = new Map([[0, 0]]);
4 let curSum = 0;
5 for (let i = 0; i < arr.length; i++) {
6 curSum += arr[i];
7 const complement = curSum - target;
8 if (prefixSums.has(complement)) {
9 return [prefixSums.get(complement), i + 1];
10 }
11 prefixSums.set(curSum, i + 1);
12 }
13}
14
15function splitWords(s) {
16 return s == "" ? [] : s.split(' ');
17}
18
19function* main() {
20 const arr = splitWords(yield).map((v) => parseInt(v));
21 const target = parseInt(yield);
22 const res = subarraySum(arr, target);
23 console.log(res.join(' '));
24}
25
26class EOFError extends Error {}
27{
28 const gen = main();
29 const next = (line) => gen.next(line).done && process.exit();
30 let buf = '';
31 next();
32 process.stdin.setEncoding('utf8');
33 process.stdin.on('data', (data) => {
34 const lines = (buf + data).split('\n');
35 buf = lines.pop();
36 lines.forEach(next);
37 });
38 process.stdin.on('end', () => {
39 buf && next(buf);
40 gen.throw(new EOFError());
41 });
42}
431#include <algorithm> // copy
2#include <iostream> // boolalpha, cin, cout, streamsize
3#include <iterator> // back_inserter, istream_iterator, ostream_iterator, prev
4#include <limits> // numeric_limits
5#include <sstream> // istringstream
6#include <string> // getline, string
7#include <unordered_map> // unordered_map
8#include <vector> // vector
9
10std::vector<int> subarray_sum(std::vector<int> arr, int target) {
11 std::unordered_map<int, int> prefix_sums;
12 // an empty subarray has a sum of 0
13 prefix_sums[0] = 0;
14 int cur_sum = 0;
15 for (int i = 0 ; i < arr.size(); i++) {
16 cur_sum += arr[i];
17 int complement = cur_sum - target;
18 if (prefix_sums.count(complement)) {
19 return {prefix_sums[complement], i + 1};
20 }
21 prefix_sums[cur_sum] = i + 1;
22 }
23 // return no indices if arr is empty
24 return {};
25}
26
27template<typename T>
28std::vector<T> get_words() {
29 std::string line;
30 std::getline(std::cin, line);
31 std::istringstream ss{line};
32 ss >> std::boolalpha;
33 std::vector<T> v;
34 std::copy(std::istream_iterator<T>{ss}, std::istream_iterator<T>{}, std::back_inserter(v));
35 return v;
36}
37
38void ignore_line() {
39 std::cin.ignore(std::numeric_limits<std::streamsize>::max(), '\n');
40}
41
42template<typename T>
43void put_words(const std::vector<T>& v) {
44 if (!v.empty()) {
45 std::copy(v.begin(), std::prev(v.end()), std::ostream_iterator<T>{std::cout, " "});
46 std::cout << v.back();
47 }
48 std::cout << '\n';
49}
50
51int main() {
52 std::vector<int> arr = get_words<int>();
53 int target;
54 std::cin >> target;
55 ignore_line();
56 std::vector<int> res = subarray_sum(arr, target);
57 put_words(res);
58}
59Note that in the implementation, typically we use prefix_sum[i] to denote the sum of elements in 0...i - 1 (rightmost element i is not included in the sum). One good thing about this is prefix_sum[0] then means sum of array up to but not including the first element, i.e. empty array. The definition of empty array sum is useful when there exists a subarray starting from 0 that sums up to the target. Without the definition of empty array sum we would miss it because its complement 0 does not exist in the dictionary.
Follow up
Find the total number of subarrays that sums up to target.
Explanation
Since the new problem does not ask for index but total number instead, we can change our hashmap to "sum k: number of prefix sums that sums up to k".
1from typing import Counter, List
2
3def subarray_sum_total(arr: List[int], target: int) -> int:
4 prefix_sums = Counter()
5 prefix_sums[0] = 1 # since empty array's sum is 0
6 cur_sum = 0
7 count = 0
8 for i in range(len(arr)):
9 cur_sum += arr[i]
10 complement = cur_sum - target
11 if complement in prefix_sums:
12 count += prefix_sums[complement]
13 prefix_sums[cur_sum] += 1
14 return count
15
16if __name__ == '__main__':
17 arr = [int(x) for x in input().split()]
18 target = int(input())
19 res = subarray_sum_total(arr, target)
20 print(res)
211import java.util.Arrays;
2import java.util.HashMap;
3import java.util.List;
4import java.util.Scanner;
5import java.util.stream.Collectors;
6
7class Solution {
8 public static int subarraySumTotal(List<Integer> arr, int target) {
9 HashMap<Integer, Integer> prefixSums = new HashMap<>();
10 prefixSums.put(0, 1); // since empty array's sum is 0
11 int curSum = 0;
12 int count = 0;
13 for (int i = 0; i < arr.size(); i++) {
14 curSum += arr.get(i);
15 int complement = curSum - target;
16 if (prefixSums.containsKey(complement)) {
17 count += prefixSums.get(complement);
18 }
19 if (prefixSums.containsKey(curSum)) {
20 prefixSums.replace(curSum, prefixSums.get(curSum) + 1);
21 } else {
22 prefixSums.put(curSum, 1);
23 }
24 }
25 return count;
26 }
27
28 public static List<String> splitWords(String s) {
29 return s.isEmpty() ? List.of() : Arrays.asList(s.split(" "));
30 }
31
32 public static void main(String[] args) {
33 Scanner scanner = new Scanner(System.in);
34 List<Integer> arr = splitWords(scanner.nextLine()).stream().map(Integer::parseInt).collect(Collectors.toList());
35 int target = Integer.parseInt(scanner.nextLine());
36 scanner.close();
37 int res = subarraySumTotal(arr, target);
38 System.out.println(res);
39 }
40}
411function subarraySumTotal(arr, target) {
2 const prefixSums = new Map();
3 prefixSums.set(0, 1); // since empty array's sum is 0
4 let curSum = 0;
5 let count = 0;
6 for (let i = 0; i < arr.length; i++) {
7 curSum += arr[i];
8 const complement = curSum - target;
9 if (prefixSums.has(complement)) {
10 count += prefixSums.get(complement);
11 }
12 if (prefixSums.has(curSum)) {
13 prefixSums.set(curSum, prefixSums.get(curSum) + 1);
14 } else {
15 prefixSums.set(curSum, 1);
16 }
17 }
18 return count;
19}
20
21function splitWords(s) {
22 return s == "" ? [] : s.split(' ');
23}
24
25function* main() {
26 const arr = splitWords(yield).map((v) => parseInt(v));
27 const target = parseInt(yield);
28 const res = subarraySumTotal(arr, target);
29 console.log(res);
30}
31
32class EOFError extends Error {}
33{
34 const gen = main();
35 const next = (line) => gen.next(line).done && process.exit();
36 let buf = '';
37 next();
38 process.stdin.setEncoding('utf8');
39 process.stdin.on('data', (data) => {
40 const lines = (buf + data).split('\n');
41 buf = lines.pop();
42 lines.forEach(next);
43 });
44 process.stdin.on('end', () => {
45 buf && next(buf);
46 gen.throw(new EOFError());
47 });
48}
491#include <algorithm> // copy
2#include <iostream> // boolalpha, cin, cout, streamsize
3#include <iterator> // back_inserter, istream_iterator
4#include <limits> // numeric_limits
5#include <sstream> // istringstream
6#include <string> // getline, string
7#include <unordered_map> // unordered_map
8#include <vector> // vector
9
10int subarray_sum_total(std::vector<int> arr, int target) {
11 std::unordered_map<int, int> prefix_sums;
12 prefix_sums[0] = 1; // an empty array has a sum of 0
13 int cur_sum = 0, count = 0;
14 for (int i = 0; i < arr.size(); i++) {
15 cur_sum += arr[i];
16 int complement = cur_sum - target;
17 if (prefix_sums.count(complement)) {
18 count += prefix_sums[complement];
19 }
20 if (prefix_sums.count(cur_sum)) {
21 prefix_sums[cur_sum] += 1;
22 } else {
23 prefix_sums[cur_sum] = 1;
24 }
25 }
26 return count;
27}
28
29template<typename T>
30std::vector<T> get_words() {
31 std::string line;
32 std::getline(std::cin, line);
33 std::istringstream ss{line};
34 ss >> std::boolalpha;
35 std::vector<T> v;
36 std::copy(std::istream_iterator<T>{ss}, std::istream_iterator<T>{}, std::back_inserter(v));
37 return v;
38}
39
40void ignore_line() {
41 std::cin.ignore(std::numeric_limits<std::streamsize>::max(), '\n');
42}
43
44int main() {
45 std::vector<int> arr = get_words<int>();
46 int target;
47 std::cin >> target;
48 ignore_line();
49 int res = subarray_sum_total(arr, target);
50 std::cout << res << '\n';
51}
52Got a question? Ask the Teaching Assistant anything you don't understand.
Still not clear? Ask in the Forum, Discord or Submit the part you don't understand to our editors.