Processing Text Files as Is: Pattern Matching over Compressed Texts, Multi-byte Character Texts, and Semi-structured Texts

Abstract

Techniques in processing text files “as is” are presented, in which given text files are processed without modification. The compressed pattern matching problem, first defined by Amir and Benson (1992), is a good example of the “as-is” principle. Another example is string matching over multi-byte character texts, which is a significant problem common to oriental languages such as Japanese, Korean, Chinese, and Taiwanese. A text file from such languages is a mixture of single-byte characters and multi-byte characters. Naive solution would be (1) to convert a given text into a fixed length encoded one and then apply any string matching routine to it; or (2) to directly search the text file byte after byte for (the encoding of) a pattern in which an extra work is needed for synchronization to avoid false detection. Both the solutions, however, sacrifice the searching speed. Our algorithm runs on such a multi-byte character text file at the same speed as on an ordinary ASCII text file, without false detection. The technique is applicable to any prefix code such as the Huffman code and variants of Unicode. We also generalize the technique so as to handle structured texts such as XML documents. Using this technique, we can avoid false detection of keyword even if it is a substring of a tag name or of an attribute description, without any sacrifice of searching speed.