Suppose you are working in some audio streaming service. The service has (n) active users and (10^9) tracks users can listen to. Users can like tracks and, based on likes, the service should recommend them new tracks. Tracks are numbered from (1) to (10^9). It turned out that tracks the (i)-th user likes form a segment (l_i, r_i). Let's say that the user (j) is a predictor for user (i) ((j \neq i)) if user (j) likes all tracks the (i)-th user likes (and, possibly, some other tracks too). Also, let's say that a track is strongly recommended for user (i) if the track is not liked by the (i)-th user yet, but it is liked by every predictor for the (i)-th user. Calculate the number of strongly recommended tracks for each user (i). If a user doesn't have any predictors, then print (0) for that user. The first line contains one integer (t) ((1 \le t \le 10^4)) — the number of test cases. Next, (t) cases follow. The first line of each test case contains one integer (n) ((1 \le n \le 2 \cdot 10^5)) — the number of users. The next (n) lines contain two integers (l_i) and (r_i) per line ((1 \le l_i \le r_i \le 10^9)) — the segment of tracks the (i)-th user likes. Additional constraint on the input: the sum of (n) over all test cases doesn't exceed (2 \cdot 10^5). For each test case, print (n) integers, where the (i)-th integer is the number of strongly recommended tracks for the (i)-th user (or (0), if that user doesn't have any predictors). In the first test case: the first user has no predictors; the second user has no predictors; the third user has two predictors: users (1) and (2); only track (3) is liked by both of them and not liked by the third user. In the second test case, the second user is a predictor for the first user. Therefore, all tracks, except (42), are strongly recommended for the first user. In the third