Study Of The Native XML Database Query And Optimization

Extensible Markup Language (XML), which is as a criterion of data exchanging,emphasizes on the uniform data format but not provides the database features. So in theapplication of XML, DBMS is still the main mean of data management. Sometraditional database vendors such as IBM, Microsoft and Oracle step into Native XMLDatabase (NXD) field recently. XQuery is the latest XML query languagerecommended by W3C. It borrows advantages from many XML query languages orsemi-structure languages. Implementing XQuery in NXD and query optimization arevery important in the study of NXD. The XQuery implemention in our system mainly divides into such parts: lexicanalysis, semantic analysis, XQuery normalization and query computation etc. It candistinguish different status and memorize the present context while analyzing andchoose different lexic analysis function according to different context state. It also cansolve the important problem of lexic analysis that we need analyze different lexis indifferent contexts. Lexic analysis constructs query semantic tree according to thesemantic unit which lexic analyzer provides. In semantic analysis, it deals with externalfunction (user-defined function) and internal function differently. That can avoid theproblems there are recursions in the external function and external function is still notdefined when it being referenced. It normalizes the query tree which has processedlexic analysis. The normalization uses Intermediate Language to facilitate machinereplant and migrate. Inspired by the traditional RDBMS query optimization technology, we separatethe XML document into structure and content, meanwhile we optimize themindividually according to the XML language's semi-structure feature. Extendingnumbering is taken to quantitate the nodes which present XML path information. Wetake Path Oriented Extensible Model POEM model on the basis of the numberingmechanism to choose the most optimum path in order to improve query effeciency.XML nodes and contents are stored separately as node sets. The node's content andstore can be optimized through the extending B+_tree index.