Sliding Window

Sliding window is a technique, not a single algorithm. It applies to contiguous ranges when adding a right element and removing a left element update the relevant state cheaply. State an invariant first: for example, a window has sum at most K, contains at most K distinct values, or has no duplicate character. Move right forward, update counts or sum, and while the invariant is violated move left forward, reversing that element's contribution. Record lengths, counts, or best values only at the correct validity point. For fixed-size windows, initialize the first k elements and then add one/remove one per step. For variable-size windows, both pointers only advance, so the usual complexity is O(n) time and O(alphabet) or O(n) state. It does not automatically work for negative-number sum constraints because shrinking may increase or decrease the sum unpredictably.