http://history.ioinformatics.org/oi/files/volume9.pdf#page=41

尝试整一些树状数组的花活，然后单点修改的这一段看不懂：

We make the following observation: when we generate the associated intervals of the nodes we pass by, we can cover the whole interval [ p + 1, y ] by starting from node p + 1 and climbing the first tree (Fig. 2.1). So instead of doing a query for every node we update, we compute the results of the queries on the fly by climbing the tree once. Analogously, we can update all the intervals of the form [ x, p – 1] by starting from node p – 1 and climbing the second tree (Fig. 2.2). The same algorithm is applied for updating both trees.

翻译无果，不知道是怎么爬树的时候更新的