KR100358348B1 - Multi-Path Index Method for The Efficient Retrieval of XML Data - Google Patents

Abstract

The present invention relates to a multipath index method for efficiently searching XML data, which assigns a unique path identifier to each single path included in an extended path expression, stores an index key value and a path identifier in combination, The present invention provides a multiple index method that simultaneously supports searching for a plurality of paths and index function for a single path.

The present invention improves the search performance of a query represented by an extended path expression in a multipath index, and index management is facilitated by integrally providing an index for a plurality of paths. In addition, it is possible to provide a search function for each path as well as an extension path. In addition, the method using the path identifier can be easily applied to the existing single path indexing technique, and the B + It is excellent in practicality.

Description

{Multi-Path Index Method for Efficient Retrieval of XML Data}

The present invention relates to a multi-path index method for efficiently searching XML data, and more particularly, to a method for efficiently processing an extended path query by briefly showing a plurality of paths in a database, that is, a multi-path.

Many studies have been conducted recently to efficiently store and retrieve XML data on the Internet. Particularly, researches for storing and retrieving XML data more efficiently using a relational database management system (RDBMS) or an object-oriented database management system (OODBMS) are active. The object-oriented database is attracting attention as an XML storage device because the data model can naturally reflect the characteristics of XML and can support the existing object query language by extending the query language to XML.

Conventionally, when the number of individual paths corresponding to the extended path increases, the total number of indexes increases. As a result, the index storage space is increased and search performance is deteriorated. Particularly, in the case of a point query, there is a problem that performance is further deteriorated.

In addition, the database administrator must determine the individual single path corresponding to the extended path, and set the index for each, and when deleting the index, the administrator should know which index is set for the related extended path. In this process, the index may not exist for some single paths in the extended path, so that it is difficult to perform consistent index management and the gain using the index is decreased.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and it is an object of the present invention to provide a path identifier unique to each single path included in the extended path expression, The present invention provides a multiple index method that simultaneously supports searching for a plurality of paths and index function for a single path.

1 is a conventional XML DTD (document type definition) program.

Figure 2 is a schema of an object database for storing data of an XML DTD in an object database.

3 is a flow diagram of a multipath indexing method in accordance with the present invention.

4 is a flowchart illustrating a method of allocating a path identifier in a multi-path index method according to the present invention.

5 is an algorithm for allocating path identifiers according to the present invention.

6 is a record structure diagram of a terminal node in a multipath index according to the present invention;

FIG. 7 illustrates an example of storing an XML document according to the present invention in an object database. FIG.

8 is a structural view of a path used in the laboratory of the present invention.

9 is a graph showing performance ratios of point queries on the number of paths in the lab time of the present invention.

10 is a graph showing performance ratios of the area queries to the number of paths in the laboratory of the present invention;

According to the present invention, there is provided an index method for searching XML data,

Finding a single path corresponding to a given extended path expression and assigning a path identifier to each of the single paths; Collecting index information collection by searching actual data in the database for each single path; And storing the path identifier and the collected index information in a multi-path index.

The method of allocating the path identifier comprises: a first step of inputting a mitigation path expression into an algorithm for allocating a path identifier; A second step of searching for a path between a trailing element of a foreground path and a trailing element of a trailing path in the relaxation path expression by calling a search-path function in an algorithm for allocating the path identifier; A third step of inserting all the paths found in the second step and the corresponding sub-elements into a path extents set; And allocating a path identifier for each single path found when the extent is obtained.

The algorithm for allocating the path identifier receives the mitigation path expression and searches the database for all paths between the backmost element of the foreground path and the foremost element of the backward path of the mitigation path expression, And assigns a path identifier to the path identifier.

The record structure of the multipath index includes header information, an index key value, a number of different paths, a path identifier of the path, a number of object identifiers corresponding to the path, and an array of object identifiers of the object corresponding to the path .

A method for searching and updating a multipath index includes: obtaining all path extents for a search path and finding a path identifier for each single path in the path extent; Searching for an intermediate node and an end terminal node of the database as a key value; Searching for only an object identifier whose value of the key identifier matches the value of the path identifier.

The conditions for using the multi-path index according to the present invention are as follows.

The names of the elements and attributes in the XML DTD must be unique within the DTD, and they should be different, even if they have the same name in different DTDs. In the case of elements of the same name existing in different XML DTDs, it is assumed that the classes in the object database system generated corresponding to the elements are also different. And the proposed multi - path index method can not set the index for the path where the cycle exists at the DTD level. It is also assumed that if an attribute of the same name is used for multiple elements, their data types are all the same.

The multipath index method can not use the query path when referring to another XML object through IDREF that represents ID property value in XML or when pointing to another object in the same XML object. In the case of IDREF, it is not possible to know what kind of object it will actually refer to because it can not express the reference relationship graph in the schema of the object database system, and thus the path index can not be used.

There is only one top object for all objects. In other words, the same object is not shared by multiple objects. However, there can be multiple subobjects for one topobject. The lower object also maintains an object identifier (OID) for the upper object. However, references to the upper object are maintained only within the system and are not exposed at the application level. Considering that the reference relation by IDREF is not supported by the reference relationship at the level of the object database system, as assumed above, the schema of the object database corresponding to the XML DTD will be a DAG type, Objects in the object database can be represented in tree form. At this time, the top object of the objects in the object database representing the XML object is defined as a 'root object'. The top-level object in the tree is the object of the object database system class with the DTD type name. For every element of XML that is transformed into an object database system, an Extent object for each element (or class) is created, which maintains a pointer to all instances of the class, do. The difference from the general data in storing XML data is that partial path instanciation starting from the root object may exist but the path is not generated from non-root object.

For example, when a '1' represents a root object, a reference path of the form 1-2-3-4-5 or 1-2-3 may exist, but the form 3-4-5 without 1-2-3 It is not. This is because all objects except root must have a reference relationship between the upper object and the lower object.

Term Definition

Since the existing path expressions used in the object query language represent only one path, when a single path expression is used, a path expression represented by the following form by combining '*' characters with an existing path expression is referred to as a relaxed path expression, Quot;

Existing path expressions such as A.B.C.D. and A.B.C.D.D1 are simple path expressions, and path expressions such as A. * .D or A.B. * .D.D1 are mitigation path expressions, all of which are extension path expressions. A. * .D and A.B. * D.D1 are A and A.B respectively, and the back path is D and D.D1 respectively.

A DTD element in XML or a path expression by the name of a class that appears in the schema of an object database system is called a 'schema path', and it is defined as a 'query path' when it is used in a query. A set of objects stored in the database according to a schema path is defined as a 'path instance' or 'object path'.

In an algorithm for allocating path identifiers

Pi - The index path.

Di - The direction of the index, forward (to match the direction of the path) or backward (reverse of the path).

KAi - The name of the attribute to be keyed.

The start class of the Hi-path corresponds to the forward element (head (post_path (P)) of the backward path in the case of a forward index and the tail (post_path (P)) of the foreground path in the reverse direction .

(The tail (post_path (P)) of the foreground path in the case of a forward index, and the head (post_path (P)) of the backward path in the reverse direction do.

Ei - all searchable paths through index i = .

BTi - The object identifier of the B + - tree object storing the index data.

For each single path: in,

- path expression,

- the number of path instances corresponding to the path,

PID - Path identifier.

Hereinafter, the present invention will be described in detail with reference to the drawings. However, these drawings are for illustrative purposes only and the present invention is not limited thereto.

1 shows an example of a conventional XML DTD (document type definition). FIG. 2 shows an example of a schema of an object-oriented database for storing data of the XML DTD of FIG. 1 in an object-oriented database. In FIG. 2, a square and an italic string show an example of attributes of classes and classes, respectively will be.

3 is a flow chart of a method for constructing a multipath index according to the present invention.

A method for constructing a multi-path index, finding individual paths corresponding to a given extended path expression, and allocating path identifiers for each path (S100), searching actual data in the database for each single path, (S200), storing the path identifier and the collected index information in a multi-path index (S300).

4 is a flowchart illustrating a method of assigning a path identifier in the multipath index method according to the present invention.

A method of allocating a path identifier includes a first step (S110) of inputting a mitigation path expression to an algorithm for allocating a path identifier, a search-path function is called in the algorithm for allocating the path identifier, A second step S120 of searching for all paths between the trailing element of the path expression and the frontmost element of the trailing path in step S120, a step of combining all the traversal traversed in the second step S120 and the corresponding lower element, A third step S130 of inserting the path identifier into the single path, and a fourth step S140 of allocating a path identifier to each single path found when the extent is obtained.

5 is an algorithm for allocating path identifiers according to the present invention.

Given an XML DTD and an extended path expression P to set the index, all single-path expressions belonging to P satisfy the given XML DTD and are defined as `path extents' for that path expression Respectively. The algorithm of FIG. 5 shows an algorithm for obtaining a path extent for a given extended path expression and assigning a path identifier. When the XML DTD is converted into the object database schema, the path extents can be found in the original XML DTD by using the upper element and the lower element or the element and attribute relation in the original XML DTD.

The head (P) used in the algorithm of FIG. 5 represents the frontmost element of the path P, and tail (P) represents the rearmost element or attribute of the path P. And child ( E ) for element E represents the set of attributes of E or E that constitute E.

The algorithm of FIG. 5 assumes path extents only for a given path expression if the given path expression is a single path expression that does not include '*'. If the mitigation path expression is given, all paths between the trailing element ( tail (pre_path (P))) of the foreground path and the frontmost element or attribute ( head (post_path (P))) Call the ' search_path ' function.

The search_path function searches for the element (head (post_path (P)) that is the forefront element of the backward path among the elements (c) of the element ( e ) If there is the same element, search path information stored in the stack and its corresponding sub-element (or attribute) are combined and inserted into the path extent set. If there is a lower element in the lower Elient burst (c), the search is continued by recursive call of the search_path function.

This search process continues until a leaf element or attribute without a sub-element is visited. Once extents are obtained, they are assigned path identifiers for each single path found.

For example, a path identifier is assigned to a single path belonging to the path extent of the extended path expression book.chapter. *. Title using the algorithm of FIG. 5 for the XML DTD of FIG. 1 as follows.

book.chapter.title - path identifier (PID): 1

book.chapter.section.title - Path identifier (PID): 2

book.chapter.section.subsection.title - Path identifier (PID): 3

FIG. 6 is a record structure diagram of a terminal node in a multipath index according to the present invention.

The index information of the multipath index is stored using the B + - tree in which the structure of the terminal record is changed. When each part of the record structure of the terminal node in the multipath index is described,

record header - header information that records the size of the record,

key value - the index key value,

# paths - the number of different paths,

PID - Path identifier of the path,

# OIDs - the number of object identifiers corresponding to the path,

OIDs - an array of object identifiers for the object corresponding to the path.

Where the path identifier is given by the path identifier assignment algorithm. The intermediate node structure of the multipath index is the same as the intermediate node structure of a general B + - tree.

FIG. 7 illustrates an XML document stored in an object database according to an exemplary embodiment of the present invention. Referring to FIG.

FIG. 7 is a diagram showing a part of a database in which XML documents conforming to the XML DTD schema shown in FIG. 1 are stored in an object database. In FIG. 7, a circle means an object, and a number in a circle means an object identifier of the object. The number next to the circle is the value of mdate. For this database, i is a multipath index for the path book.chapter. *. Mdate, i is implemented as follows:

Pi - book.chapter. *. Mdate

Di - backward

KAi - mdate

Hi-book

Ti - mdate

Ei - book.chapter.body.mdate

book.chapter.subsection.body.mdate

book.chapter.section.subsection.body.mdate

Information by route

book.chapter.body.mdate - PID: 1, :2

book.chapter.subsection.body.mdate - PID: 2, :2

book.chapter.section.subsection.body.mdate - PID: 3, : 7

Referring to FIG. 7, the terminal records in the index are as follows (record header information is omitted)

"1999: 11: 03", 3, (1,1, 0), (2,1,

"2000: 01: 01", 3, (1,1, o1), (2,1, o2), (3,4, o1, o2,

"2000: 02: 01", 1, (3,2, o1, o2))

A method for searching and updating a multi-path index,

The search process for a multipath index is the same as that of a general B + - tree, except that only the search of the terminal records is different. In other words, the index tree is searched in the same manner as the B + - tree, but only the object identifier satisfying the given query path is selected only in the terminal node. Therefore, we obtain all the path extents for the desired search path and find the path identifiers for each single path in the path extent. Then, the intermediate node and the end terminal node are searched by the given key value, and only the object identifier whose key value matches the path identifier value is searched.

For example, as shown in FIG. 6,

The path exponent of book.chapter.section. *. mdate, the path to select b from b in Books where b.chapter.section. *. mdate = "1999: 11:

With book.chater.section.body.mdate

book.chater.section.subsection.body.mdate and its corresponding path identifier (PID) is 2 and 3. Therefore, when the index is searched for "1999: 11: 03", and the key identifier of the terminal node is matched and the object identifier having the PID of 2 or 3 is searched, o1 can be found.

The path extents of b.chapter. *. mdate, which is the path to select b from b in Books where b.chapter. *. mdate = "2000: 02:

book.chater.section.subsection.body.mdate and the corresponding path identifier (PID) is 3. Therefore, when the index is searched for "2000: 02: 01", and the key identifier of the terminal node matches and the PID of the object identifier is 3, o1 and 02 can be found.

A multipath index is a method of maintaining the key value and the object identifier of the object corresponding to the end of the path for a specified path, such as a nested index. Therefore, when a reference reference between objects is inserted, deleted or changed, the change information for the path should be obtained by searching both ends of the object path where insertion or deletion occurred, as in the case of the overlap index, and reflecting the change information in the index.

Laboratory example

We compared the performance of indexing method using multipath index and the method using several single path indexes for arbitrary extended path expressions. Since there are many existing index methods for a single path, it is known that among the existing index methods, the search performance of the overlapped index is the best. In this experiment, a superposition index is used as a single path index method. Respectively.

FIG. 8 is a structural diagram of a path used in an experimental example according to the present invention, in which a path expression is expressed in the form of a schema of an object database (OODBMS). The class (element) that may exist in the middle of each path is omitted, and the start class of the path And the class that represents the end of the path Lt; / RTI > Here, N represents the number of different paths and is equal to the number of end classes of the path shown in FIG.

Table 1 shows the parameters used in the experiment. In the case of a nested index, one index is installed for each path, so the index is used as many as the number of paths. On the other hand, in the case of a multipath index, only one index is used regardless of the number of paths.

The per-fan-out (f) From class A class that is connected from an object to any single path to The number of instances of The number of objects belonging to " Thus if N = 8 and f = 10 There are 8 different single paths for one object and 10 objects are connected for each single path, meaning that 80 objects are connected to each other. kl is the size of the key, and kr is the number of duplicate keys.

Experiments were divided into point query and domain query. The point query is to search only key values that match exactly the given key value, and the domain query is to search for objects that correspond to a range of key values. Performance comparison is based on query processing cost to process a given query for each index technique and query processing cost is defined as the number of expected index pages to be accessed in the process of query processing.

9 is a graph showing the performance ratio of the area query to the number of paths in the experimental example according to the present invention.

FIG. 9 shows the query processing cost of the multipath index for the point query divided by the query processing cost of the overlap index. In the figure, 'F- *' and 'B- *' represent the performance comparison for the forward index and the performance comparison for the reverse index, respectively. The omnidirectional index is the key value of the index. Quot; < / RTI > Therefore, in this case, the index has a key value to be searched Linked to an instance of The instance of. In contrast, the reverse index ≪ / RTI > located in an instance of < RTI ID = 0.0 > Linked to an instance of Search for instances of. 1/10/100 connected after F (or B) represents the value of kr . And Respectively. Even if the values of f , c , and kl are different, the performance ratio is similar to that shown in FIG.

In FIG. 9, it can be seen that the search performance of the multipath index is significantly improved as the number of paths increases for all cases. In the case of a nested index, there is one index for each path, so all indexes must be searched to process a given query.

Point queries are particularly affected by the height of the index. When a single-path index is used, the cost of query processing is linearly proportional to the number of indexes because the indexes are visited. On the other hand, since the multipath index uses one index, the height of the index does not greatly differ even when the number of paths increases, so there is no difference in the cost of processing the point query according to the number of paths. Therefore, as the number of paths increases, the search performance of the multi-path index becomes much better.

10 is a graph showing the performance ratio of the point query to the number of paths in the experimental example according to the present invention.

FIG. 10 is a graph illustrating the performance ratio of the area query performed in the same manner as FIG. 9 And the performance ratio is similar to that of Fig. 10 even when the values of f , c , and kl are different.

FIG. 10 shows that as the number of paths increases for the area query, the multipath index performs better than the overlap index. In the case of the area query, the number of end nodes (or pages) of the index has a major effect on performance. As the number of paths increases in both the multipath index and the nested index, the number of end nodes increases, The number of nodes is greatly increased. In the case of the multipath index, as the number of paths increases, only the number of PIDs and object identifiers included in the same key value increases, so that the size of the terminal records increases slightly, whereas in the case of the overlap index, Since the number of path indexes increases, the sum of the number of terminal nodes of all single path indexes increases more rapidly than that of the multi-path indexes.

This point also explains the case where the multipath index in the case of B-1 falls by about 5% compared with the overlap index regardless of the number of paths. That is, B-1 uses a reverse index and the value of the key does not overlap. Even if the number of paths increases, a new key value is added continuously. Therefore, in the case of the multi-path index, The relative performance is not improved. Also, since the size of the terminal record increases due to the field such as the additional path identifier (PID) used, the number of terminal nodes is larger than that of the overlap index, so that the performance is somewhat lower than the overlap index.

As described above, according to the present invention,

The multi-path index improves the retrieval performance of the query represented by the extended path expression, and index management is facilitated by integrally providing indexes for a plurality of paths. In addition, it is possible to provide a search function for each path as well as an extension path. In addition, the method using the path identifier can be easily applied to the existing single path indexing technique, and the B + There is an effect that the practicality is excellent.

Claims (5)

Finding a single path corresponding to a given extended path expression and assigning a path identifier to each of the single paths; Collecting index information by searching actual data in a database for each single path; And storing the path identifier and the collected index information in a multi-path index. The method according to claim 1, A method of allocating a path identifier is as follows. A first step of inputting a mitigation path expression into an algorithm for allocating path identifiers; A second step of searching for a path between a trailing element of a foreground path and a trailing element of a trailing path in the relaxation path expression by calling a search-path function in an algorithm for allocating the path identifier; A third step of inserting all the paths found in the second step and the corresponding sub-elements into a path extents set; And a fourth step of allocating a path identifier for each single path found when the extent is obtained. The method according to claim 2, wherein The algorithm for allocating the path identifier comprises: And a path identifier is assigned to each single path that is searched for. The path identifier is a path identifier that is used to search for a path between the backward path of the foreground path and the path of the forward path. A multi-path index method for effective retrieval of XML data. 4. The method according to any one of claims 1 to 3, The record structure of the multipath index includes header information, an index key value, a number of different paths, a path identifier of the path, a number of object identifiers corresponding to the path, and an array of object identifiers of the object corresponding to the path A multi-path index method for effective retrieval of XML data. 4. The method according to any one of claims 1 to 3, The search and update method for the multi- Obtaining all the path extents for the search path and finding a path identifier for each single path in the path extent; Searching for an intermediate node and an end terminal node of the database as a key value; And searching for only an object identifier whose value of the key identifier matches the value of the path identifier.