Given some balls of three colors arranged in a line, rearrange them such that all the red balls go first, then green and then blue ones.

Do rearrange the balls in place. A solution that simply counts colors and overwrites the array is not the one we are looking for.

This is an important problem in search algorithms theory proposed by Dutch computer scientist Edsger Dijkstra. Dutch national flag has three colors (albeit different from ones used in this problem).

‍

Example

Input: [G, B, G, G, R, B, R, G]

Output: [R, R, G, G, G, G, B, B]

There are a total of 2 red, 4 green and 2 blue balls. In this order they appear in the correct output.

‍

Notes

Input Parameters: An array of characters named balls, consisting of letters from {‘R’, ‘G’, ‘B’}.

Output: Return type is void. Modify the input character array by rearranging the characters in-place.

Constraints:

• 1

• Do this in ONE pass over the array - NOT TWO passes, just one pass.

• Solution is only allowed to use constant extra memory.

‍

Solution

As mentioned in the notes: A naive solution to this problem, is to simply count how many balls of each color, and then overwrite the array with that many balls in the expected order of colors. This is not the solution we (and an interviewer would) look for here.

To solve the problem, taking one example will help. Try to play with:

[R, G, G, R, G, B, G, B, R, R, R, G]

[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]

Suppose our initial array is: [array to process]

----

Now "suppose" after processing some part of the given array, we are able to maintain its structure like:

[R, R, ..., R, R][G, G, ..., G, G][remaining array to process][B, B, ..., B, B]

Just assume that we are able to maintain this structure somehow, currently do not worry about how!

----

If we can maintain the same structure after processing the first ball of "remaining array to process" in O(1) time, then it means now we have to solve the same problem with the reduced size of "remaining array to process".

After we have processed the first ball, we repeat the process for the next ball; that also takes O(1) time and maintains the same structure! So at the end we will be able to maintain the same structure and that will be the sorted array! We are taking O(1) time to process each character hence the time complexity will also be n * O(1) that is O(n)! Let's think how can we maintain the same structure when first letter in the "remaining array to process" is:

----

'R', which means array is:

[R, R, ..., R, R]['G', G, ..., G, G]['R' + other remaining array to process][B, B, ..., B, B]

What should we do with 'R' to maintain the same structure? Swapping it with the leftmost 'G'?

Yes, that is possible and that will maintain the structure with only one swap, "assuming that we have the index of the leftmost G".

So now our array will look like:

[R, R, ..., R, R, 'R'][G, ..., G, G, 'G'][other remaining array to process][B, B, ..., B, B]

----

'G' which means array is:

[R, R, ..., R, R][G, G, ..., G, G]['G' + other remaining array to process][B, B, ..., B, B]

What should we do with 'G' to maintain the same structure?

Nothing to do, just go to the next character!

So now our array will look like:

[R, R, ..., R, R][G, G, ..., G, G, 'G'][other remaining array to process][B, B, ..., B, B]

----

'B' which means array is:

[R, R, ..., R, R][G, G, ..., G, G]['B' + other remaining array to process + 'last character'][B, B, ..., B, B]

What should we do with 'B' to maintain the same structure?

Swap 'B' with the 'last character' of the "remaining array to process"!  Yes that is possible and that will maintain the structure with only one swap "assuming that we have the index of last character". So now our array will look like:

[R, R, ..., R, R][G, G, ..., G, G]['last character' + other remaining array to process]['B', B, B, ..., B, B]

----

We Made Some Assumptions:

1) When the first character is 'R', we assumed that "we have the index of the leftmost G".

2) When the first character is 'B', we assumed that "we have the index of the last character".

So somehow we need to maintain these two indices and we will be able to solve the problem by following the steps above.

Also when we are starting we can take,

[R, R, ..., R, R], [G, G, ..., G, G] and [B, B, ..., B, B] parts = "". And then start our solution!

The idea sounds complex but the code can be very simple!

Time Complexity:

O(n).

Auxiliary Space Used:

O(1) as we are using only constant extra space.

Space Complexity:

O(n).