Thanks for the explanation. After using \registerctxluafile{sort-hanzi}{}, the bumping message did not appear. In terms of daily practical use, I really don't need so many characters. I just don't have the energy to pick out those thousands of commonly used Chinese characters from these 40,000 or 50,000 characters (In China, for example, there are only about 6,000-8,000 characters actually used on a daily basis. In Japanese, you may only need about 1000-3000 characters) There are only two commonly used sorts for these characters: (Sorting has nothing to do with unicode sorting) 1 according to the actual pronunciation of the characters and 2 according to the order in which the characters are written (strokes). (The situation in Japanese should probably be mostly sorted by actual pronunciation based on kana, but the pronunciation of kanji in Japanese is much more complicated than in Chinese.) But sorting by strokes, I don't have the ability to achieve it at the moment. So the three indexes I designed are sorted according to the actual pronunciation of the Chinese characters. The difference between them is only in the entries. 1 Sort in the order of a, b, c d, and use these letters as entries.(mostly used) 2 Sort in the order of a ai ao an ...... , and these pronunciations are used as entries. 3 Sort Chinese characters directly by their pronunciation and use them as entries. Because I know almost nothing about lua myself, just referring to sort-lang (just applying templates) For the sorting of Japanese, the sorting I see on latex so far is also directly marked out of the pronunciation, and then sorted by the pronunciation of the kana. (Because kanji in Japanese may have more than one pronunciation, and maybe even as many as 5). Unless there is a tool that can simultaneously phonetize the Chinese characters in the index at compile time.