Related
Basically, I would like help designing an algorithm that takes a given number, and returns a random number that is unrelated to the first number. The stipulations being that a) the given output number will always be the same for a similar input number, and b) within a certain range (ex. 1-100), all output numbers are distinct. ie., no two different input numbers under 100 will give the same output number.
I know it's easy to do by creating an ordered list of numbers, shuffling them randomly, and then returning the input's index. But I want to know if it can be done without any caching at all. Perhaps with some kind of hashing algorithm? Mostly the reason for this is that if the range of possible outputs were much larger, say 10000000000, then it would be ludicrous to generate an entire range of numbers and then shuffle them randomly, if you were only going to get a few results out of it.
Doesn't matter what language it's done in, I just want to know if it's possible. I've been thinking about this problem for a long time and I can't think of a solution besides the one I've already come up with.
Edit: I just had another idea; it would be interesting to have another algorithm that returned the reverse of the first one. Whether or not that's possible would be an interesting challenge to explore.
This sounds like a non-repeating random number generator. There are several possible approaches to this.
As described in this article, we can generate them by selecting a prime number p and satisfies p % 4 = 3 that is large enough (greater than the maximum value in the output range) and generate them this way:
int randomNumberUnique(int range_len , int p , int x)
if(x * 2 < p)
return (x * x) % p
else
return p - (x * x) % p
This algorithm will cover all values in [0 , p) for an input in range [0 , p).
Here's an example in C#:
private void DoIt()
{
const long m = 101;
const long x = 387420489; // must be coprime to m
var multInv = MultiplicativeInverse(x, m);
var nums = new HashSet<long>();
for (long i = 0; i < 100; ++i)
{
var encoded = i*x%m;
var decoded = encoded*multInv%m;
Console.WriteLine("{0} => {1} => {2}", i, encoded, decoded);
if (!nums.Add(encoded))
{
Console.WriteLine("Duplicate");
}
}
}
private long MultiplicativeInverse(long x, long modulus)
{
return ExtendedEuclideanDivision(x, modulus).Item1%modulus;
}
private static Tuple<long, long> ExtendedEuclideanDivision(long a, long b)
{
if (a < 0)
{
var result = ExtendedEuclideanDivision(-a, b);
return Tuple.Create(-result.Item1, result.Item2);
}
if (b < 0)
{
var result = ExtendedEuclideanDivision(a, -b);
return Tuple.Create(result.Item1, -result.Item2);
}
if (b == 0)
{
return Tuple.Create(1L, 0L);
}
var q = a/b;
var r = a%b;
var rslt = ExtendedEuclideanDivision(b, r);
var s = rslt.Item1;
var t = rslt.Item2;
return Tuple.Create(t, s - q*t);
}
That generates numbers in the range 0-100, from input in the range 0-100. Each input results in a unique output.
It also shows how to reverse the process, using the multiplicative inverse.
You can extend the range by increasing the value of m. x must be coprime with m.
Code cribbed from Eric Lippert's article, A practical use of multiplicative inverses, and a few of the previous articles in that series.
You can not have completely unrelated (particularly if you want the reverse as well).
There is a concept of modulo inverse of a number, but this would work only if the range number is a prime, eg. 100 will not work, you would need 101 (a prime). This can provide you a pseudo random number if you want.
Here is the concept of modulo inverse:
If there are two numbers a and b, such that
(a * b) % p = 1
where p is any number, then
a and b are modular inverses of each other.
For this to be true, if we have to find the modular inverse of a wrt a number p, then a and p must be co-prime, ie. gcd(a,p) = 1
So, for all numbers in a range to have modular inverses, the range bound must be a prime number.
A few outputs for range bound 101 will be:
1 == 1
2 == 51
3 == 34
4 == 76
etc.
EDIT:
Hey...actually you know, you can use the combined approach of modulo inverse and the method as defined by #Paul. Since every pair will be unique and all numbers will be covered, your random number can be:
random(k) = randomUniqueNumber(ModuloInverse(k), p) //this is Paul's function
The question:
Given any string, add the least amount of characters possible to make it a palindrome in linear time.
I'm only able to come up with a O(N2) solution.
Can someone help me with an O(N) solution?
Revert the string
Use a modified Knuth-Morris-Pratt to find the latest match (simplest modification would be to just append the original string to the reverted string and ignore matches after len(string).
Append the unmatched rest of the reverted string to the original.
1 and 3 are obviously linear and 2 is linear beacause Knuth-Morris-Pratt is.
If only appending is allowed
A Scala solution:
def isPalindrome(s: String) = s.view.reverse == s.view
def makePalindrome(s: String) =
s + s.take((0 to s.length).find(i => isPalindrome(s.substring(i))).get).reverse
If you're allowed to insert characters anywhere
Every palindrome can be viewed as a set of nested letter pairs.
a n n a b o b
| | | | | * |
| -- | | |
--------- -----
If the palindrome length n is even, we'll have n/2 pairs. If it is odd, we'll have n/2 full pairs and one single letter in the middle (let's call it a degenerated pair).
Let's represent them by pairs of string indexes - the left index counted from the left end of the string, and the right index counted from the right end of the string, both ends starting with index 0.
Now let's write pairs starting from the outer to the inner. So in our example:
anna: (0, 0) (1, 1)
bob: (0, 0) (1, 1)
In order to make any string a palindrome, we will go from both ends of the string one character at a time, and with every step, we'll eventually add a character to produce a correct pair of identical characters.
Example:
Assume the input word is "blob"
Pair (0, 0) is (b, b) ok, nothing to do, this pair is fine. Let's increase the counter.
Pair (1, 1) is (l, o). Doesn't match. So let's add "o" at position 1 from the left. Now our word became "bolob".
Pair (2, 2). We don't need to look even at the characters, because we're pointing at the same index in the string. Done.
Wait a moment, but we have a problem here: in point 2. we arbitrarily chose to add a character on the left. But we could as well add a character "l" on the right. That would produce "blolb", also a valid palindrome. So does it matter? Unfortunately it does because the choice in earlier steps may affect how many pairs we'll have to fix and therefore how many characters we'll have to add in the future steps.
Easy algorithm: search all the possiblities. That would give us a O(2^n) algorithm.
Better algorithm: use Dynamic Programming approach and prune the search space.
In order to keep things simpler, now we decouple inserting of new characters from just finding the right sequence of nested pairs (outer to inner) and fixing their alignment later. So for the word "blob" we have the following possibilities, both ending with a degenerated pair:
(0, 0) (1, 2)
(0, 0) (2, 1)
The more such pairs we find, the less characters we will have to add to fix the original string. Every full pair found gives us two characters we can reuse. Every degenerated pair gives us one character to reuse.
The main loop of the algorithm will iteratively evaluate pair sequences in such a way, that in step 1 all valid pair sequences of length 1 are found. The next step will evaluate sequences of length 2, the third sequences of length 3 etc. When at some step we find no possibilities, this means the previous step contains the solution with the highest number of pairs.
After each step, we will remove the pareto-suboptimal sequences. A sequence is suboptimal compared to another sequence of the same length, if its last pair is dominated by the last pair of the other sequence. E.g. sequence (0, 0)(1, 3) is worse than (0, 0)(1, 2). The latter gives us more room to find nested pairs and we're guaranteed to find at least all the pairs that we'd find for the former. However sequence (0, 0)(1, 2) is neither worse nor better than (0, 0)(2, 1). The one minor detail we have to beware of is that a sequence ending with a degenerated pair is always worse than a sequence ending with a full pair.
After bringing it all together:
def makePalindrome(str: String): String = {
/** Finds the pareto-minimum subset of a set of points (here pair of indices).
* Could be done in linear time, without sorting, but O(n log n) is not that bad ;) */
def paretoMin(points: Iterable[(Int, Int)]): List[(Int, Int)] = {
val sorted = points.toSeq.sortBy(identity)
(List.empty[(Int, Int)] /: sorted) { (result, e) =>
if (result.isEmpty || e._2 <= result.head._2)
e :: result
else
result
}
}
/** Find all pairs directly nested within a given pair.
* For performance reasons tries to not include suboptimal pairs (pairs nested in any of the pairs also in the result)
* although it wouldn't break anything as prune takes care of this. */
def pairs(left: Int, right: Int): Iterable[(Int, Int)] = {
val builder = List.newBuilder[(Int, Int)]
var rightMax = str.length
for (i <- left until (str.length - right)) {
rightMax = math.min(str.length - left, rightMax)
val subPairs =
for (j <- right until rightMax if str(i) == str(str.length - j - 1)) yield (i, j)
subPairs.headOption match {
case Some((a, b)) => rightMax = b; builder += ((a, b))
case None =>
}
}
builder.result()
}
/** Builds sequences of size n+1 from sequence of size n */
def extend(path: List[(Int, Int)]): Iterable[List[(Int, Int)]] =
for (p <- pairs(path.head._1 + 1, path.head._2 + 1)) yield p :: path
/** Whether full or degenerated. Full-pairs save us 2 characters, degenerated save us only 1. */
def isFullPair(pair: (Int, Int)) =
pair._1 + pair._2 < str.length - 1
/** Removes pareto-suboptimal sequences */
def prune(sequences: List[List[(Int, Int)]]): List[List[(Int, Int)]] = {
val allowedHeads = paretoMin(sequences.map(_.head)).toSet
val containsFullPair = allowedHeads.exists(isFullPair)
sequences.filter(s => allowedHeads.contains(s.head) && (isFullPair(s.head) || !containsFullPair))
}
/** Dynamic-Programming step */
#tailrec
def search(sequences: List[List[(Int, Int)]]): List[List[(Int, Int)]] = {
val nextStage = prune(sequences.flatMap(extend))
nextStage match {
case List() => sequences
case x => search(nextStage)
}
}
/** Converts a sequence of nested pairs to a palindrome */
def sequenceToString(sequence: List[(Int, Int)]): String = {
val lStr = str
val rStr = str.reverse
val half =
(for (List(start, end) <- sequence.reverse.sliding(2)) yield
lStr.substring(start._1 + 1, end._1) + rStr.substring(start._2 + 1, end._2) + lStr(end._1)).mkString
if (isFullPair(sequence.head))
half + half.reverse
else
half + half.reverse.substring(1)
}
sequenceToString(search(List(List((-1, -1)))).head)
}
Note: The code does not list all the palindromes, but gives only one example, and it is guaranteed it has the minimum length. There usually are more palindromes possible with the same minimum length (O(2^n) worst case, so you probably don't want to enumerate them all).
O(n) time solution.
Algorithm:
Need to find the longest palindrome within the given string that contains the last character. Then add all the character that are not part of the palindrome to the back of the string in reverse order.
Key point:
In this problem, the longest palindrome in the given string MUST contain the last character.
ex:
input: abacac
output: abacacaba
Here the longest palindrome in the input that contains the last letter is "cac". Therefore add all the letter before "cac" to the back in reverse order to make the entire string a palindrome.
written in c# with a few test cases commented out
static public void makePalindrome()
{
//string word = "aababaa";
//string word = "abacbaa";
//string word = "abcbd";
//string word = "abacac";
//string word = "aBxyxBxBxyxB";
//string word = "Malayal";
string word = "abccadac";
int j = word.Length - 1;
int mark = j;
bool found = false;
for (int i = 0; i < j; i++)
{
char cI = word[i];
char cJ = word[j];
if (cI == cJ)
{
found = true;
j--;
if(mark > i)
mark = i;
}
else
{
if (found)
{
found = false;
i--;
}
j = word.Length - 1;
mark = j;
}
}
for (int i = mark-1; i >=0; i--)
word += word[i];
Console.Write(word);
}
}
Note that this code will give you the solution for least amount of letter to APPEND TO THE BACK to make the string a palindrome. If you want to append to the front, just have a 2nd loop that goes the other way. This will make the algorithm O(n) + O(n) = O(n). If you want a way to insert letters anywhere in the string to make it a palindrome, then this code will not work for that case.
I believe #Chronical's answer is wrong, as it seems to be for best case scenario, not worst case which is used to compute big-O complexity. I welcome the proof, but the "solution" doesn't actually describe a valid answer.
KMP finds a matching substring in O(n * 2k) time, where n is the length of the input string, and k substring we're searching for, but does not in O(n) time tell you what the longest palindrome in the input string is.
To solve this problem, we need to find the longest palindrome at the end of the string. If this longest suffix palindrome is of length x, the minimum number of characters to add is n - x. E.g. the string aaba's longest suffix substring is aba of length 3, thus our answer is 1. The algorithm to find out if a string is a palindrome takes O(n) time, whether using KMP or the more efficient and simple algorithm (O(n/2)):
Take two pointers, one at the first character and one at the last character
Compare the characters at the pointers, if they're equal, move each pointer inward, otherwise return false
When the pointers point to the same index (odd string length), or have overlapped (even string length), return true
Using the simple algorithm, we start from the entire string and check if it's a palindrome. If it is, we return 0, and if not, we check the string string[1...end], string[2...end] until we have reached a single character and return n - 1. This results in a runtime of O(n^2).
Splitting up the KMP algorithm into
Build table
Search for longest suffix palindrome
Building the table takes O(n) time, and then each check of "are you a palindrome" for each substring from string[0...end], string[1...end], ..., string[end - 2...end] each takes O(n) time. k in this case is the same factor of n that the simple algorithm takes to check each substring, because it starts as k = n, then goes through k = n - 1, k = n - 2... just the same as the simple algorithm did.
TL; DR:
KMP can tell you if a string is a palindrome in O(n) time, but that supply an answer to the question, because you have to check if all substrings string[0...end], string[1...end], ..., string[end - 2...end] are palindromes, resulting in the same (but actually worse) runtime as a simple palindrome-check algorithm.
#include<iostream>
#include<string>
using std::cout;
using std::endl;
using std::cin;
int main() {
std::string word, left("");
cin >> word;
size_t start, end;
for (start = 0, end = word.length()-1; start < end; end--) {
if (word[start] != word[end]) {
left.append(word.begin()+end, 1 + word.begin()+end);
continue;
}
left.append(word.begin()+start, 1 + word.begin()+start), start++;
}
cout << left << ( start == end ? std::string(word.begin()+end, 1 + word.begin()+end) : "" )
<< std::string(left.rbegin(), left.rend()) << endl;
return 0;
}
Don't know if it appends the minimum number, but it produces palindromes
Explained:
We will start at both ends of the given string and iterate inwards towards the center.
At each iteration, we check if each letter is the same, i.e. word[start] == word[end]?.
If they are the same, we append a copy of the variable word[start] to another string called left which as it name suggests will serve as the left hand side of the new palindrome string when iteration is complete. Then we increment both variables (start)++ and (end)-- towards the center
In the case that they are not the same, we append a copy of of the variable word[end] to the same string left
And this is the basics of the algorithm until the loop is done.
When the loop is finished, one last check is done to make sure that if we got an odd length palindrome, we append the middle character to the middle of the new palindrome formed.
Note that if you decide to append the oppoosite characters to the string left, the opposite about everything in the code becomes true; i.e. which index is incremented at each iteration and which is incremented when a match is found, order of printing the palindrome, etc. I don't want to have to go through it again but you can try it and see.
The running complexity of this code should be O(N) assuming that append method of the std::string class runs in constant time.
If some wants to solve this in ruby, The solution can be very simple
str = 'xcbc' # Any string that you want.
arr1 = str.split('')
arr2 = arr1.reverse
count = 0
while(str != str.reverse)
count += 1
arr1.insert(count-1, arr2[count-1])
str = arr1.join('')
end
puts str
puts str.length - arr2.count
I am assuming that you cannot replace or remove any existing characters?
A good start would be reversing one of the strings and finding the longest-common-substring (LCS) between the reversed string and the other string. Since it sounds like this is a homework or interview question, I'll leave the rest up to you.
Here see this solution
This is better than O(N^2)
Problem is sub divided in to many other sub problems
ex:
original "tostotor"
reversed "rototsot"
Here 2nd position is 'o' so dividing in to two problems by breaking in to "t" and "ostot" from the original string
For 't':solution is 1
For 'ostot':solution is 2 because LCS is "tot" and characters need to be added are "os"
so total is 2+1 = 3
def shortPalin( S):
k=0
lis=len(S)
for i in range(len(S)/2):
if S[i]==S[lis-1-i]:
k=k+1
else :break
S=S[k:lis-k]
lis=len(S)
prev=0
w=len(S)
tot=0
for i in range(len(S)):
if i>=w:
break;
elif S[i]==S[lis-1-i]:
tot=tot+lcs(S[prev:i])
prev=i
w=lis-1-i
tot=tot+lcs(S[prev:i])
return tot
def lcs( S):
if (len(S)==1):
return 1
li=len(S)
X=[0 for x in xrange(len(S)+1)]
Y=[0 for l in xrange(len(S)+1)]
for i in range(len(S)-1,-1,-1):
for j in range(len(S)-1,-1,-1):
if S[i]==S[li-1-j]:
X[j]=1+Y[j+1]
else:
X[j]=max(Y[j],X[j+1])
Y=X
return li-X[0]
print shortPalin("tostotor")
Using Recursion
#include <iostream>
using namespace std;
int length( char str[])
{ int l=0;
for( int i=0; str[i]!='\0'; i++, l++);
return l;
}
int palin(char str[],int len)
{ static int cnt;
int s=0;
int e=len-1;
while(s<e){
if(str[s]!=str[e]) {
cnt++;
return palin(str+1,len-1);}
else{
s++;
e--;
}
}
return cnt;
}
int main() {
char str[100];
cin.getline(str,100);
int len = length(str);
cout<<palin(str,len);
}
Solution with O(n) time complexity
public static void main(String[] args) {
String givenStr = "abtb";
String palindromeStr = covertToPalindrome(givenStr);
System.out.println(palindromeStr);
}
private static String covertToPalindrome(String str) {
char[] strArray = str.toCharArray();
int low = 0;
int high = strArray.length - 1;
int subStrIndex = -1;
while (low < high) {
if (strArray[low] == strArray[high]) {
high--;
} else {
high = strArray.length - 1;
subStrIndex = low;
}
low++;
}
return str + (new StringBuilder(str.substring(0, subStrIndex+1))).reverse().toString();
}
// string to append to convert it to a palindrome
public static void main(String args[])
{
String s=input();
System.out.println(min_operations(s));
}
static String min_operations(String str)
{
int i=0;
int j=str.length()-1;
String ans="";
while(i<j)
{
if(str.charAt(i)!=str.charAt(j))
{
ans=ans+str.charAt(i);
}
if(str.charAt(i)==str.charAt(j))
{
j--;
}
i++;
}
StringBuffer sd=new StringBuffer(ans);
sd.reverse();
return (sd.toString());
}
I'm reading the numbers 0, 1, ..., (N - 1) one by one in some order. My goal is to find the lexicography index of this given permutation, using only O(1) space.
This question was asked before, but all the algorithms I could find used O(N) space. I'm starting to think that it's not possible. But it would really help me a lot with reducing the number of allocations.
Considering the following data:
chars = [a, b, c, d]
perm = [c, d, a, b]
ids = get_indexes(perm, chars) = [2, 3, 0, 1]
A possible solution for permutation with repetitions goes as follows:
len = length(perm) (len = 4)
num_chars = length(chars) (len = 4)
base = num_chars ^ len (base = 4 ^ 4 = 256)
base = base / len (base = 256 / 4 = 64)
id = base * ids[0] (id = 64 * 2 = 128)
base = base / len (base = 64 / 4 = 16)
id = id + (base * ids[1]) (id = 128 + (16 * 3) = 176)
base = base / len (base = 16 / 4 = 4)
id = id + (base * ids[2]) (id = 176 + (4 * 0) = 176)
base = base / len (base = 4 / 4 = 1)
id = id + (base * ids[3]) (id = 176 + (1 * 1) = 177)
Reverse process:
id = 177
(id / (4 ^ 3)) % 4 = (177 / 64) % 4 = 2 % 4 = 2 -> chars[2] -> c
(id / (4 ^ 2)) % 4 = (177 / 16) % 4 = 11 % 4 = 3 -> chars[3] -> d
(id / (4 ^ 1)) % 4 = (177 / 4) % 4 = 44 % 4 = 0 -> chars[0] -> a
(id / (4 ^ 0)) % 4 = (177 / 1) % 4 = 177 % 4 = 1 -> chars[1] -> b
The number of possible permutations is given by num_chars ^ num_perm_digits, having num_chars as the number of possible characters, and num_perm_digits as the number of digits in a permutation.
This requires O(1) in space, considering the initial list as a constant cost; and it requires O(N) in time, considering N as the number of digits your permutation will have.
Based on the steps above, you can do:
function identify_permutation(perm, chars) {
for (i = 0; i < length(perm); i++) {
ids[i] = get_index(perm[i], chars);
}
len = length(perm);
num_chars = length(chars);
index = 0;
base = num_chars ^ len - 1;
base = base / len;
for (i = 0; i < length(perm); i++) {
index += base * ids[i];
base = base / len;
}
}
It's a pseudocode, but it's also quite easy to convert to any language (:
If you are looking for a way to obtain the lexicographic index or rank of a unique combination instead of a permutation, then your problem falls under the binomial coefficient. The binomial coefficient handles problems of choosing unique combinations in groups of K with a total of N items.
I have written a class in C# to handle common functions for working with the binomial coefficient. It performs the following tasks:
Outputs all the K-indexes in a nice format for any N choose K to a file. The K-indexes can be substituted with more descriptive strings or letters.
Converts the K-indexes to the proper lexicographic index or rank of an entry in the sorted binomial coefficient table. This technique is much faster than older published techniques that rely on iteration. It does this by using a mathematical property inherent in Pascal's Triangle and is very efficient compared to iterating over the set.
Converts the index in a sorted binomial coefficient table to the corresponding K-indexes. I believe it is also faster than older iterative solutions.
Uses Mark Dominus method to calculate the binomial coefficient, which is much less likely to overflow and works with larger numbers.
The class is written in .NET C# and provides a way to manage the objects related to the problem (if any) by using a generic list. The constructor of this class takes a bool value called InitTable that when true will create a generic list to hold the objects to be managed. If this value is false, then it will not create the table. The table does not need to be created in order to use the 4 above methods. Accessor methods are provided to access the table.
There is an associated test class which shows how to use the class and its methods. It has been extensively tested with 2 cases and there are no known bugs.
To read about this class and download the code, see Tablizing The Binomial Coeffieicent.
The following tested code will iterate through each unique combinations:
public void Test10Choose5()
{
String S;
int Loop;
int N = 10; // Total number of elements in the set.
int K = 5; // Total number of elements in each group.
// Create the bin coeff object required to get all
// the combos for this N choose K combination.
BinCoeff<int> BC = new BinCoeff<int>(N, K, false);
int NumCombos = BinCoeff<int>.GetBinCoeff(N, K);
// The Kindexes array specifies the indexes for a lexigraphic element.
int[] KIndexes = new int[K];
StringBuilder SB = new StringBuilder();
// Loop thru all the combinations for this N choose K case.
for (int Combo = 0; Combo < NumCombos; Combo++)
{
// Get the k-indexes for this combination.
BC.GetKIndexes(Combo, KIndexes);
// Verify that the Kindexes returned can be used to retrive the
// rank or lexigraphic order of the KIndexes in the table.
int Val = BC.GetIndex(true, KIndexes);
if (Val != Combo)
{
S = "Val of " + Val.ToString() + " != Combo Value of " + Combo.ToString();
Console.WriteLine(S);
}
SB.Remove(0, SB.Length);
for (Loop = 0; Loop < K; Loop++)
{
SB.Append(KIndexes[Loop].ToString());
if (Loop < K - 1)
SB.Append(" ");
}
S = "KIndexes = " + SB.ToString();
Console.WriteLine(S);
}
}
You should be able to port this class over fairly easily to the language of your choice. You probably will not have to port over the generic part of the class to accomplish your goals. Depending on the number of combinations you are working with, you might need to use a bigger word size than 4 byte ints.
There is a java solution to this problem on geekviewpoint. It has a good explanation for why it's true and the code is easy to follow. http://www.geekviewpoint.com/java/numbers/permutation_index. It also has a unit test that runs the code with different inputs.
There are N! permutations. To represent index you need at least N bits.
Here is a way to do it if you want to assume that arithmetic operations are constant time:
def permutationIndex(numbers):
n=len(numbers)
result=0
j=0
while j<n:
# Determine factor, which is the number of possible permutations of
# the remaining digits.
i=1
factor=1
while i<n-j:
factor*=i
i+=1
i=0
# Determine index, which is how many previous digits there were at
# the current position.
index=numbers[j]
while i<j:
# Only the digits that weren't used so far are valid choices, so
# the index gets reduced if the number at the current position
# is greater than one of the previous digits.
if numbers[i]<numbers[j]:
index-=1
i+=1
# Update the result.
result+=index*factor
j+=1
return result
I've purposefully written out certain calculations that could be done more simply using some Python built-in operations, but I wanted to make it more obvious that no extra non-constant amount of space was being used.
As maxim1000 noted, the number of bits required to represent the result will grow quickly as n increases, so eventually big integers will be required, which no longer have constant-time arithmetic, but I think this code addresses the spirit of your question.
Nothing really new in the idea but a fully matricial method with no explicit loop or recursion (using Numpy but easy to adapt):
import numpy as np
import math
vfact = np.vectorize(math.factorial, otypes='O')
def perm_index(p):
return np.dot( vfact(range(len(p)-1, -1, -1)),
p-np.sum(np.triu(p>np.vstack(p)), axis=0) )
I just wrote a code using Visual Basic and my program can directly calculate every index or every corresponding permutation to a given index up to 17 elements (this limit is due to the approximation of the scientific notation of numbers over 17! of my compiler).
If you are interested I can I can send the program or publish it somewhere for download.
It works fine and It can be useful for testing and paragon the output of your codes.
I used the method of James D. McCaffrey called factoradic and you can read about it here and something also here (in the discussion at the end of the page).
Suppose I have an alphabet of 'abcd' and a maximum string length of 3. This gives me 85 possible strings, including the empty string. What I would like to do is map an integer in the range [0,85) to a string in my string space without using a lookup table. Something like this:
0 => ''
1 => 'a'
...
4 => 'd'
5 => 'aa'
6 => 'ab'
...
84 => 'ddd'
This is simple enough to do if the string is fixed length using this pseudocode algorithm:
str = ''
for i in 0..maxLen do
str += alphabet[i % alphabet.length]
i /= alphabet.length
done
I can't figure out a good, efficient way of doing it though when the length of the string could be anywhere in the range [0,3). This is going to be running in a tight loop with random inputs so I would like to avoid any unnecessary branching or lookups.
Shift your index by one and ignore the empty string temporarily. So you'd map 0 -> "a", ..., 83 -> "ddd".
Then the mapping is
n -> base-4-encode(n - number of shorter strings)
With 26 symbols, that's the Excel-column-numbering scheme.
With s symbols, there are s + s^2 + ... + s^l nonempty strings of length at most l. Leaving aside the trivial case s = 1, that sum is (a partial sum of a geometric series) s*(s^l - 1)/(s-1).
So, given n, find the largest l such that s*(s^l - 1)/(s-1) <= n, i.e.
l = floor(log((s-1)*n/s + 1) / log(s))
Then let m = n - s*(s^l - 1)/(s-1) and encode m as an l+1-symbol string in base s ('a' ~> 0, 'b' ~> 1, ...).
For the problem including the empty string, map 0 to the empty string and for n > 0 encode n-1 as above.
In Haskell
encode cs n = reverse $ encode' n where
len = length cs
encode' 0 = ""
encode' n = (cs !! ((n-1) `mod` len)) : encode' ((n-1) `div` len)
Check:
*Main> map (encode "abcd") [0..84] ["","a","b","c","d","aa","ab","ac","ad","ba","bb","bc","bd","ca","cb","cc","cd","da","db","dc","dd","aaa","aab","aac","aad","aba","abb","abc","abd","aca","acb","acc","acd","ada","adb","adc","add","baa","bab","bac","bad","bba","bbb","bbc","bbd","bca","bcb","bcc","bcd","bda","bdb","bdc","bdd","caa","cab","cac","cad","cba","cbb","cbc","cbd","cca","ccb","ccc","ccd","cda","cdb","cdc","cdd","daa","dab","dac","dad","dba","dbb","dbc","dbd","dca","dcb","dcc","dcd","dda","ddb","ddc","ddd"]
Figure out the number of strings for each length: N0, N1, N2 & N3 (actually, you won't need N3). Then, use those values to partition your space of integers: 0..N0-1 are length 0, N0..N0+N1-1 are length 1, etc. Within each partition, you can use your fixed-length algorithm.
At worst, you've greatly reduced the size of your lookup table.
Here is a C# solution:
static string F(int x, int alphabetSize)
{
string ret = "";
while (x > 0)
{
x--;
ret = (char)('a' + (x % alphabetSize)) + ret;
x /= alphabetSize;
}
return ret;
}
If you want to optimize this further, you may want to do something to avoid the string concatenations. For example, you could store the result into a preallocated char[] array.
Suppose I have an int x = 54897, old digit index (0 based), and the new value for that digit. What's the fastest way to get the new value?
Example
x = 54897
index = 3
value = 2
y = f(x, index, value) // => 54827
Edit: by fastest, I definitely mean faster performance. No string processing.
In simplest case (considering the digits are numbered from LSB to MSB, the first one being 0) AND knowing the old digit, we could do as simple as that:
num += (new_digit - old_digit) * 10**pos;
For the real problem we would need:
1) the MSB-first version of the pos, that could cost you a log() or at most log10(MAX_INT) divisions by ten (could be improved using binary search).
2) the digit from that pos that would need at most 2 divisions (or zero, using results from step 1).
You could also use the special fpu instruction from x86 that is able to save a float in BCD (I have no idea how slow it is).
UPDATE: the first step could be done even faster, without any divisions, with a binary search like this:
int my_log10(unsigned short n){
// short: 0.. 64k -> 1.. 5 digits
if (n < 1000){ // 1..3
if (n < 10) return 1;
if (n < 100) return 2;
return 3;
} else { // 4..5
if (n < 10000) return 4;
return 5;
}
}
If your index started at the least significant digit, you could do something like
p = pow(10,index);
x = (x / (p*10) * (p*10) + value * p + x % p).
But since your index is backwards, a string is probably the way to go. It would also be more readable and maintainable.
Calculate the "mask" M: 10 raised to the power of index, where index is a zero-based index from the right. If you need to index from the left, recalculate index accordingly.
Calculate the "prefix" PRE = x / (M * 10) * (M * 10)
Calculate the "suffix" SUF = x % M
Calculate the new "middle part" MID = value * M
Generate the new number new_x = PRE + MID + POST.
P.S. ruslik's answer does it more elegantly :)
You need to start by figuring out how many digits are in your input. I can think of two ways of doing that, one with a loop and one with logarithms. Here's the loop version. This will fail for negative and zero inputs and when the index is out of bounds, probably other conditions too, but it's a starting point.
def f(x, index, value):
place = 1
residual = x
while residual > 0:
if index < 0:
place *= 10
index -= 1
residual /= 10
digit = (x / place) % 10
return x - (place * digit) + (place * value)
P.S. This is working Python code. The principle of something simple like this is easy to work out, but the details are so tricky that you really need to iterate it a bit. In this case I started with the principle that I wanted to subtract out the old digit and add the new one; from there it was a matter of getting the correct multiplier.
You gotta get specific with your compute platform if you're talking about performance.
I would approach this by converting the number into pairs of decimal digits, 4 bit each.
Then I would find and process the pair that needs modification as a byte.
Then I would put the number back together.
There are assemblers that do this very well.