Find All Anagrams in a String

Use a window whose length equals the pattern length. Count the pattern characters and maintain counts for the current substring, or keep one difference array and a counter of how many character counts are nonzero. Slide the window one position at a time: add the right character, remove the left character once the window is too long, and report the start when all counts match. The invariant is that the tracked counts describe exactly the current length-m window, so equality with the pattern means that window is an anagram. For lowercase English letters an array of 26 is compact; for arbitrary characters use a map and define how Unicode is handled. Test a pattern longer than the text, repeated pattern characters, overlapping matches, and an empty pattern according to the API contract. Sorting every window is unnecessarily O(nm log m). The frequency-window solution is O(n plus alphabet) time and O(alphabet) space, normally O(n) time with fixed alphabet.