| Abstract: | Proposals have been made for fast string search algorithms that can search a string for a given pattern without having to examine every character of the text passed over. These algorithms are effective only if search patterns are in practice long enough, and enough text is traversed in a search to justify the costs of initializing the sophisticated search algorithm. We report observations of routine uses of an editor instrumented to determine the characteristics of use. It appears that many searches are indeed for patterns of 3,5 or more characters, and that these searches cover substantial amounts of text. However, searches for patterns of a single character are very common, while only moving a short distance on average. The implementation of a sophisticated search algorithm must take care to deal with short single-character searches efficiently. |
