Design of Index Schema based on Bit-Streams for XML Documents

Transcription

1 Design of Index Schema based on Bit-Streams for XML Documents Youngrok Song 1, Kyonam Choo 3 and Sangmin Lee 2 1 Institute for Information and Electronics Research, Inha University, Incheon, Korea 2 Department of Electronic Engineering, Inha University, Incheon, Korea 3 Computing and Information Center, Incheon University, Incheon, Korea gateway32@inha.ac.kr, kyonam@incheon.ac.kr, sanglee@inha.ac.kr Abstract In this paper, the structural information between tree nodes of XML documents is represented without any structural changes of the tree by converting that number information added to the tree into bit streams. It is also shown that other structural information can be retrieved and added to index schema in the process. In this process, we confirm that the response time can be minimized by conducting a bit operation between the bit streams with index schema in use, and accurate results can be reached by searching only with information of record sets by nodes in index files. Keywords: Index schema, XML, Database structure 1. Introduction As demand for using XML(eXtensible Markup Language) documents and the quantity of documents are increased, the amount of information needed to define the documents by XML standard is rapidly increasing as well. It has become more difficult to locate needed information, so methods to search and manage XML documents information more efficiently are necessary [1-4], and recent studies have been actively under way to store XML documents information in storage media such as databases[5-7]. These studies aim at supporting efficient route search of a single large volume XML document or several XML documents with the same structure. Thus, when using the aforementioned indexing techniques to find the wanted routes from several XML documents with different structures, we need to compose each index for each XML document and investigate every index. Though indexing methods have been suggested to complement the defects, such methods still have the drawback of lowered performance when the ancestordescendant relationship appears in the middle of the route, not in the root [8]. In this paper, query processing in diverse forms is given flexibility by adding the structural information field that is needed for the structure of the existing index files when additional structural information is added to XML documents. As a result, the given query analysis time is reduced, and thus response time to output return is reduced by using query schema based on index schema in order to improve the query processing efficiency. 2. Index Schema based on Bit-Streams This paper has gone through the following processes to achieve an index schema as in Figure 1. 1) Build up XML documents into trees by using DOM (document object model) trees. 2) Give sequential numbers to the built-up trees by the node of each level. 3) Trees given numbers are rebuilt-up. 4) Data to be stored in index files is acquired from the trees. 5) Make bit streams by using the number given to each node. 6) Store 131

2 the data gained from [4), 5)] in the index files. Figure 2 is a DTD(Document Type Definition) used for delivery between buyers and sellers. Figure 1. Index Schema Figure 2. DTD of Delivery XML Documents 2.1. XML Tree Numbering Technique In Figure 3, the node of each level is sequentially given a number, and the bits that can express each node are assigned in as the same number as that of the node of each level, among the numbers given to nodes of trees, those starting with bit 0 are excluded, and the numbers should be given from bit 1. In this way, the sequential numbers are given from the root node to the least significant nodes. After the operations are completed, the only bit stream of each route is generated when the assigned numbers are connected with the root node as the reference point. The value itself shows the super-sub relationship, the structural information of the entire node in the same tree such as parent-children, ancestor-descendant and sibling nodes from the root node up to the node where the route ends. Figure 3. Example Tree of Delivery XML Documents Given Bits 132

3 Bit stream per each node International Journal of Software Engineering and Its Applications 2.2. Technique of Converting Bit Stream of XML Structural Information Figure 3 shows that the bit stream value that is given to each node is gained by successively visiting each from the root node, and with this bit stream, the bit value by level is stored from the least significant bit in the assigned fixed bit space. The bit values from L 0 to L 7 expressed in table 1 show the bit stream value concerning the hierarchical relationship by node in Figure 3. The bit stream values correspond with all the nodes, one to one, and they are the only and unique values. Moreover, as this value is the unique value, if the bit value by each node name is known, then the whole XML document can be restored. The entire size of the bit stream is 64 bits, which means that the fixed space is assigned. Yet, if the level of an XML document goes up or the node number by level increases, then the value will exceed the fixed bit at 64 bit. In that case, it can be solved by assigning a space bigger than 64 bits. Table 1. Bit Stream Assigned to Each Node Level L 0 L 1 L 2 L 3 L 4 L 5 L 6 L i : i times repetition of bit Building up Index Schema based on Database Table In the index schema table, the names of each of the nodes, unique bit stream values, bit stream values of parent node and level values that are needed for query analysis when user's queries come up are stored. Table 2 shows the schema structure of the index files and an example of an index schema table that shows the attribute structural information of a node. Table 2. Table Field of Index Schema Table Field Roles Example N_name names of XML tree nodes Tel B_value bit stream values of each node B Tb_len certain node length of the whole bit stream 6 B_len bit stream length assigned temporarily to the present node 3 Level the level of the present node on XML tree 2 P_value bit stream value of the parent node of the present node Ctype shows if the present node is a primitive or an attribute 1 Position shows the order of the present node as a child node to the parent node 3 D_num the number of XML document 1 133

4 Data the value each node has Table 3 is a table that writes the basic document of Figure 3 based upon the DTD of Figure 2 and stores the information gained from the XML tree. Here, the record that is set by node stored in the database is the only value. Thus, the query processing method of using record sets, which is suggested in Table 3, is much more useful than the joining method based on the association between tables in which fragmented data are stored and the case of performing its additional operations in order to find out result sets. Now, it is necessary to find out how to use the structural information in Table 3. First, if a query is about a particular document, then the query can be more easily addressed by referring to the document information D_num field. Even when the query does not have any document information, the query forms that can be expressed by combinations of node information of different documents can be avoided in advance by looking at the returned results of referring to D_num field. For example, suppose the search query about (2) iname node in Table 3 is given, and looks at the result search processes. First, (2) iname node can be searched and located from table 3 by using the document information that is gained from the given query analysis(d_num field), primitives composing query and name of attributes(n_name field), values about attributes and primitives(data field) and level information(level field). At this point, B, bit stream value of iname node(b_value field) shows the location of iname node on the XML document trees. Again, the host nodes can also be identified through the bit stream(p_value field) of the parent node. Because the bit stream values of (1) item node in Table 3 are identical to B, the bit stream values of host nodes of (2) iname, it is possible to know that (1) item is the parent node of (2) iname. When such operations are repeated, any node from (2) iname to the root node can be searched. If the increase in the amount of documents may cause bit streams of the parent nodes to agree with each other, but it is possible to identify a parent node by comparing the level information, names of primitives and attributes and data. Moreover, if search query for sibling nodes of (2) iname is given, all node information that have the same value of the D_num field and Level field of (2) iname node makes sibling nodes, without particular join operations. If different structural information is added that cannot be stored in the index file structure suggested in documents, expansibility of index algorithm is guaranteed by adding a field in which new structural information can be stored without modifying the existing index file structures. Table 3. Example of Index Structure based on Conceptual Database Structure N_name B_value Tb_len B_len Level Data P_value Ctype Position D_num delivery none receiver none name B Leo date tel B ellipsis (1) item B none B iname 00000B6B xml B payment B true B (2) iname B ellipsis XML Guide B

5 3. Experiments and Results Experiment data was arranged and the performance was evaluated in order to determine the validity of index and query schema to process the XML documents presented so far. The accuracy test was conducted about the general query types by using 1,000 all different XML documents, and another test was carried out to compare the performances between the experiments of existing XRel [7] and INRIA [9] and the methods suggested in this paper. The index algorithm suggested by Chien[6] focuses on the index techniques centered on three points. That is, to effectively process XML queries, it takes 1) quickly testing the structural relationship of ancestor-descendant (parent-child) between the two given primitives, 2) quickly finding out the candidate list that satisfies the structural relationship in 1), and 3) effectively drawing the pairs that satisfy the structural relationship from the candidate list, the result of 2). This paper also suggested the algorithm concerning the new structure search of 1) and 2) mentioned above. However, Chien proved its efficiency with B+ trees in stage 3), but this paper presents a more efficient method of storing and searching the index algorithm that suits the processes of 1) and 2). As the queries are already optimized after the processes of 1) and 2), the result of one to one corresponding can be directly obtained from database in stage 3). However, in comparison with Chien's method, this paper shows a slight difference. As Chien concentrates on the structural relationship index, he does not consider the characteristics of the storage media such as databases. Rather, he simply stores and searches the structural information with B+ tree. However, there is a close relationship between the index structure and the storage structure in this paper, it is not appropriate to be compared with Chien's method the only indexes XML documents. The index schema of XML documents suggested in this paper was conducted 1,000 times by repeating the queries with the XPath expression depth of 10 by using experiment data, and the results are in figure 4. As shown in figure 4, the accuracy was more than 94%, though the number of primitives, attributes and texts in expressions increases. The error rate was less than 6%, the reason is that with the increased complexity of documents, the number of nodes by level in XML document trees increases. Thus, as the length of bit stream that can express structural information of each node exceeds 64 bits, query processing operations fail, otherwise, inappropriate results that were returned though query processing were successful. Figure 4. Accuracy of Retrieval Results for XML Queries 135

6 4. Conclusion This paper proposed an effective index and query schema to search large amounts of XML document structural information. The query accuracy test with 1,000 various XML documents and the query-response time measured in the experiment over Shakespeare plays whose documents are deep and complicated were conducted, and this proved the efficiency of the suggested index and query methods. Therefore, it can be said that using the index and query schema suggested in this paper shows that the accuracy of drawing results of the query search in the query accuracy performance test is more than 94%. The query expressions were made of queries in the forms with general and complicated structures that can occur in all XML document structures. One of the factors affecting the query accuracy performance test experiment is the bit stream overflow that the lengths of bit streams expressing each node by level exceed 64 bits as the complexity of documents increases. To solve this problem, the 64 bit bit-stream length should be expanded. Acknowledgements This work was supported by Key Research Institute Program through the National Research Foundation of Korea(NRF) funded by the Ministry of Education, Science and Technology( ) References [1] T. Dao, An indexing model for structured document to support queries on content, structure and attributes, Proc. of IEEE ADL, (1998), pp [2] T. Milo and D. Suciu, Index structures for path expressions, Proc. of 7th International Conference on Database Theory, (1999), pp [3] B. F. Cooper, N. Sample, M. J. Franklin, G. R. Hjaltason and M. Shadmon, A fast index for semistructured data, Proc. of the 27th VLDB Conference, (2001), pp [4] Y. R. Song, K. N. Choo, Y. S. Woo, H. K. Min and W. U. Lee, Multi-indexing system for news stories based on XML documents, Lecture Notes in Computer Science, vol. 3815, (2005), pp [5] C. Zhang, J. Naughton, D. DeWitt, Q. Luo and G. Lohman, On supporting containment queries in relational database management systems, Proc. of the 2001 ACM SIGMOD International Conference on Management of Data, vol. 30, (2001), pp [6] S. Y. Chien, Z. Vagena, D. Zhang, V. J. Tsotras and C. Zaniolo, Efficient structural joins on indexed XML documents, Proc. of the 28th VLDB Conference, (2002), pp [7] M. Yoshikawa, T. Amagasa, T. Shimura and S. Uemura, XRel: a path-based approach to storage and retrieval of XML documents using relational databases, ACM Transaction on Internet Technology, vol. 1, (2001), pp [8] C. W. Chung, J. K. Min and K. S. Shim, APEX: an adaptive path index for XML data, Proc. of the 2002 ACM SIGMOD International Conference on Management of Data, (2002), pp [9] D. Florescu and D. Kossmann, A performance evaluation of alternative mapping schemes for storing XML data in a relational database, INRIA Technical Report 3680, (1999). 136