JPH07210563A - Index processing method - Google Patents

Abstract

PURPOSE:To accelerate retrieval processing concerning the index processing method using a B tree for retrieving or updating data at a data base managing system. CONSTITUTION:A direct access node setting processing part 111 decides a direct access node for enabling direct access without retrieval from a route node and stores the information in a direct access node information storage area 117. At the time of retrieval, a direct access node access processing part 112 decides the direct access node closest to retrieving object data based on the information in the direct access node information storage part 117, and a B tree search processing part 113 advances the retrieval from the direct access node, which is closest to the retrieving object data, decided by the direct access node access processing part 112 to the target retrieving object data.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an index processing method used for searching or updating data in a database management system.

[0002]

2. Description of the Related Art B-trees have been used as a method for realizing data storage and index processing in conventional database management systems (monthly "Interface", July 1991, p207 to p208). The B-tree is a set of information storage nodes called nodes. A node is composed of a pair of a segment value storage area for storing segment values and a pointer. A pointer located to the left of the partition value points to another node with a partition value or key value less than the partition value. In this way, the node group starts from the root node (hereinafter referred to as the root node), which is the highest node, to the leaf node (hereinafter,
Called a leaf node) is hierarchically linked.
Like the upper node, the leaf node is composed of a key storage area and a pointer, and the pointer points to the address of the data record having the key value.

[0003]

In the conventional B-tree indexing process, when data is searched or added or deleted, the root node is used as a starting point and the key value of the data to be searched, added or deleted, Follow the links between nodes while comparing the magnitude relation with the segment value or key value stored in the node, move to the lowest hierarchy node that stores the desired key value, and finally arrive at the target data. It is a search method.

Since the number of comparisons with the key value increases as the number of layers in the hierarchy increases, the processing time to reach the target node increases proportionally. Even for frequently accessed data, it was necessary to start the search from the root node each time a search or change was made. Even in this case, a processing time proportional to the number of layers in the hierarchy is required.

A first object of the present invention is to reduce the processing time required to reach a target node.

Further, when searching the data satisfying a given condition from all the data, the key value stored in all the leaf nodes is compared with the contents of the data record pointed to by the pointer and the condition, starting from the root node. Had to do. All data is retrieved by sequentially extracting records from the beginning of the key value stored in each leaf node and making a condition determination up to the last data. Since it is performed for all the data registered in the B tree, the processing time increases in proportion to the number of data items.

A second object of the present invention is to reduce the processing time required for searching related data.

[0008]

In order to achieve the first object, the index processing method of the present invention comprises a direct access node setting section, a direct access node access processing section, and a B-tree search processing section.

The direct access node setting unit sets a node for storing frequently accessed data or data designated by a database operator at the time of or after the B tree is created.
Register directly to the access node group. Alternatively, the direct access node determination unit determines an appropriate direct access node from the total number of data items, the number of node layers, and the number of data items stored in each node, and registers the appropriate direct access node in the direct access node group. The node designated as the direct access node may be a leaf node or a node other than the leaf node.

At the time of search, the direct access node access processing section detects the data position from the access node information and directly accesses the data if the search target data is registered in the direct access node list. If the search data is not registered in the direct access node list, the "closest" direct access node in the B-tree is detected from the direct access node information.

When the search target data is not registered in the direct access node list during the search, the B-tree search processing unit departs from the "closest" direct access node in the B-tree detected by the direct access node access processing unit. As a point, the B-tree is searched up to the node that stores the target data.

In order to achieve the second object, the index processing method of the present invention comprises a key value sublink setting processing section and a key value sublink search processing section.

The key value sublink setting processing unit detects data that satisfies the sublink setting condition specified by the database operator during or after the B tree is created, and the key provided for each data correspondence of the node. Set the pointer to the next data in the value sublink area. The next data refers to the data that satisfies the sublink setting condition next to the own data when the key values are arranged in ascending order.

If the search condition is specified as a sublink setting condition at the time of search, the key value sublink search processing unit detects the head data of the sublink and is set in the sublink area of each data thereafter. The search is advanced along the pointer.

[0015]

The direct access node determination unit registers the node storing the frequently accessed data or the data specified by the database operator in the direct access node group at the time of or after the B tree is created, and the related information is used as the direct access node information. Store in the storage area. Alternatively, an appropriate direct access node is determined from the total number of data items, the number of stages of the node hierarchy, and the number of key values stored in each node, that is, the order, and registered in the direct access node group, and the related information is directly accessed. It is stored in the storage area as information. The direct access node determination unit provides a direct access node that can be directly accessed from the root node without comparing the magnitude relationships of all layers in the B-tree.

If the data to be searched is registered in the direct access node list at the time of search, the direct access node access processing unit detects the data position from the direct access node information stored by the direct access node determination unit, and directly detects that node or data. to access. If the search data is not directly registered in the access node list,
The "closest in the B-tree" direct access node or data is detected from the direct access node information. Here, "B
The direct access node closest to the search target data in the tree "
Means that the number of tier crossings is minimum, including the number of returns to the upper tier, until the node that stores the key value of the search target data is reached from the direct access node group as a starting point. A direct access node in a node. The direct access node access processing unit provides a means for accessing the direct access node from the root node without comparing the magnitude relationships of all layers in the B-tree.

When the search target data is not registered in the direct access node list at the time of search, the B-tree search processing unit finds the "closest" direct access node or data in the B-tree detected by the direct access node access processing unit. B as a starting point to the node that stores the target data
Search through the tree. The B-tree search processing unit provides means for performing a search process from the root node to a target node without comparing the magnitude relationships of all layers in the B-tree.

The key value sub-link setting processing unit detects data which meets the sub-link setting condition specified by the database operator during or after the B-tree is created, and the key value sub-link of each data storage area of the node is detected. In the link area, when the key values are arranged in ascending order, the pointer to the data that satisfies the sublink setting condition next to the own data is set. The key value sublink setting processing unit can retrieve only the data satisfying the specified condition without searching all the data.

During the search, the key value sublink search processing unit detects the top data of the sublink if the search condition is designated as the sublink setting condition, and is set in the sublink area of each data thereafter. Advance the search according to the pointer. The key value sublink search processing unit provides means for picking up only the data satisfying a specified condition along the chain of data without searching all the data.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Prior to the description of the embodiments of the present invention, an outline of the overall structure of an index processing method according to the present invention will be described first with reference to the structure block diagram of FIG.

In FIG. 1, reference numeral 100 designates a processor.
Reference numeral 10 represents a storage device, and 120 represents an input / output device. A B-tree main body is stored in the B-tree storage area 116 in the storage device. When a direct access node designation is input via the input / output device 122, the access node setting processing unit 111 generates a direct access node or data and stores the information in the direct access node information storage area 117. When the sublink condition designation is input 123, the sublink setting processing unit 114 creates a sublink in the data in the B tree and stores the information in the sublink information storage area 118. When the search is started, the direct access node access processing unit 112 causes the direct access node information storage area 117
Based on, the direct access node or data closest to the search target data is selected, and the B-tree search processing unit 113 performs search processing up to the search target data. If the search condition is stored as the sublink condition in the sublink information storage area 118, the sublink search processing unit 115 performs the search process.

The first feature of the present invention is that the direct access node setting unit 111 and the direct access node access processing unit 1 are provided.
12 and the B-tree search processing unit 113.

The direct access node setting unit 111 stores a direct access node or data that can be directly accessed without comparing the size relation between the segment value of the node for all layers in the B-tree and the target key value from the root node. provide. Direct access node access processing unit 112
Provides a means for directly accessing the access node or the data from the root node without comparing the size relationship between the partition values of the nodes for all layers in the B-tree and the target key value. The B-tree search processing unit 113 starts from the root node to B
Provided is a means for advancing a search process from a nearest direct access node or data to target data without comparing the magnitude relationship between the segmental values of nodes for all layers in the tree and the target key value.

From the above, it is possible to directly access the data or node in the B-tree without comparing the section values of the nodes of all layers in the B-tree and the target key value from the root node. Will be possible.

The second feature of the present invention is that the key value sublink setting processing unit 114 and the key value sublink search processing unit 115.
And to prepare. Key value sublink setting processing unit 1
14 provides a chain of data in which only data satisfying a specified condition can be picked up without searching all data. Key value sublink search processing unit 115
Provides a means of picking up only the data that meets the specified conditions according to the chain of data without searching the entire data. As described above, it is possible to pick up only the data that meets the specified condition without searching all the data in the B tree.

Next, the node configuration of the B-tree will be described with reference to FIG. FIG. 2A shows the configuration of nodes other than the leaf node of the conventional B tree. Node address storage areas 201, 201
Reference numerals 03, 205, and 207 are areas for storing node addresses of lower layers. Section value storage areas 202, 204, 206
Stores the upper limit of the key value stored in each node of the lower hierarchy in ascending order from the left. Node number value storage area (for example, 20
1) is a pointer to a node that stores a key value equal to or less than the segment value (for example, 202).

FIG. 2B shows the configuration of nodes other than the leaf node of the B tree in the present invention. It is composed of a pair of a node address storage area and a segment value storage area, and an upper layer node address storage area 208. The upper layer node address storage area stores the address of a node in the upper layer of the own node, and is used when starting a search starting from the direct access node. 2 (a) and (b),
The node address storage area stores the addresses of the nodes in the upper layer or the lower layer, but this may be the node number.

FIG. 2C shows the structure of a leaf node of a conventional B-tree. The leaf node is a key value storage area 211.21 that stores the key value of the data registered in the B tree.
3, 215 and associated information storage areas 212, 214, 216 for storing associated information.

On the other hand, the leaf node of the B tree in the present invention is shown in FIG. The leaf node of the present invention stores a key value storage area for storing a key value of data registered in the B-tree, an accompanying information storage area for storing its accompanying information, and sublink information storage areas 221, 222 for the data.
223 sets. When a sublink is set up, the sublink information storage area stores the data record address of the sublink destination from each data.

An outline of the structure and operation of the B-tree will be described with reference to FIG. FIG. 3A shows a conventional B-tree structure. The node 301 is a root node. Nodes 304 to 30
Reference numeral 9 is a leaf node. When searching for data having a key value of 24, first, the value 311 “30” in the root node is larger than the target key value “24”, so the link 312 on the left side of the key value 30 is followed. Move to the node 302 in the lower hierarchy. Furthermore, here the key value "1
The comparison is made with 0 ”313 and“ 20 ”314, and since both are larger than the both, the link 315 on the right side of the key value“ 20 ”314 is obtained. The reached node 306 is a profit / non-node. The data record address is obtained from the associated information storage area corresponding to the key value "24" in that area, and the data is accessed.

FIG. 3B shows the structure of the B tree in the present invention. The node 301 is a root node. Node 3
04 to 309 are leaf nodes. Similar to the conventional method, the target data can be searched from the root node according to this procedure. Also, in the B-tree construction of the present invention,
In addition, pointers 323 and 324 to direct access nodes and key value sublinks 325 to 331 are linked. At the time of search, if the key value to be searched is registered in the direct access node group 320 in the direct access node information storage area 117 (321), the corresponding pointer 322
To reach the target node or data by following links 323 and 324. For example, the key value for search is "21"
If so, since it is registered in the direct access node group (321), it is possible to access the data having the key value “21” in the leaf node 306 by following the pointers 322 and 324. If not registered, the pointers 322 corresponding to the “direct access node or data closest to the search target” determined by the direct access node access processing unit 112 are linked to links 323, 32.
Follow 4 to search for the desired data. For example, when the key value for search is “24”, it is not registered in the direct access node group, so “closest direct access node or data” is selected from the key value “21” (321),
After arriving at the leaf node 306 by following the pointer 324 and searching the inside of the node, it is possible to finally access the data having the key value “24”.

Sublinks 325 to 331 in FIG. 3 are
It is an example of a sub-link set between data such that the key value is an even number. When “search for even-numbered data among all data” is specified, the pointer 323 is traced from the first element here, the key value “2” 321, and the key value is even according to the sublink groups 325 to 331. To retrieve all data.

Next, an outline of the processing of each unit will be shown.

FIG. 4 is a flowchart showing the direct access node setting process.

First, when the direct access node creation is designated (401), the direct access node determination processing (40
Proceed to 2). When the direct access node key value is designated, the designated key value is set as the direct access node (403). If not specified, the number of nodes to be set as the direct access node is calculated from the total number of data, the order of the B-tree, and the number of stages (404), and the direct access nodes corresponding to the number of nodes are determined (405). A pair of the determined direct access node segment value or key value and its address is registered in the direct access node list (40
6) Then, the direct access node determination process ends.

Here, the total number of data is the total number of data stored in all leaf nodes in the B-tree, and the order is one half of the segment value (or key value) stored in each node. And the number of stages means the number of layers of nodes from the root node to the leaf node. As described above, when the direct access node key value is not specified, the number of nodes to be set as the direct access node is calculated from the total number of data items, the order of the B-tree, and the number of stages (404), and the direct access nodes for the number of nodes are calculated. create. Ideally, the data of the first element of the node is generated by the direct access node, one for each leaf node, and when the direct access node is accessed, "the maximum key in the direct access node having a key value that does not exceed the key value of the search target data" is generated. If the access node is directly determined under the condition of "node having a value", the search target data always exists in the accessed node, so that the search target data can be accessed without crossing the hierarchy. If the number of returns to the upper layer is less than or equal to one half of the number of stages of the B tree, the number of layer crossings when searching from the root node is not exceeded. When the number of direct access nodes is reduced to reduce the amount of information, if the number of stages is N and the order is M, (2M (N / 2) th power) direct access nodes may be created. The number of data items is evenly divided and generated as the data of the first element of each leaf node. According to the same access node determination method as in the previous example, even if there is a return to the upper layer, the search is performed from the root node. It is possible to reach the leaf node having the target data with a smaller number of crossovers than the above.

Finally, for the B-tree search processing, the node number of the upper layer is set in the upper layer node number storage area of each node (406). The B-tree search processing will be described with reference to FIG.

Direct access nodes are designated as data that is frequently accessed, or if data that is frequently accessed is "close in the B-tree" and that is representative of them.

FIG. 5 is a flow chart showing the direct access node access processing.

After the search is started (501), the direct access node to be accessed is determined based on the key value (502) of the designated data (503). If the determined direct access node is the designated data, the address of the direct access node is obtained and the data is accessed (505). If the determined direct access node is not the specified data itself, it is "closest" to the search target data in the B-tree.
It is a direct access node. The direct access node is used as a starting point to proceed with the search in the B-tree and finally arrive at the target data.

The "direct access node closest to the data to be searched in the B-tree" means the upper layer when the B-tree search is advanced from the direct access node to the data to be searched starting from the direct access node. The direct access node is located at a position where the total number of times across the hierarchy including the number of search returns to is minimum. In the method of determining the direct access node to access (503), as described above, “a node having the largest partition value among the direct access nodes having the partition value that does not exceed the key value of the search target data” is obtained by using a hash function. There is a determination method such as using a function to obtain a “direct access node having a segmental value with the smallest difference from the key value of the search target data”.

FIG. 6 is a flowchart showing the B-tree search process. The B-tree search processing unit operates when there is no data to be searched in the direct access node or its child or grandchild node determined by the direct access node access processing (601). First, the direct access node access processing unit obtains the node address of the upper layer from the upper layer node address storage area of the direct access node (602) and moves to the node of the upper layer (603).
If the search target data exists in the destination node or its child or grandchild node (604), the partition value in the partition value storage area of the node is compared with the key value of the search target data as in the conventional search processing. The node is moved to the target node according to the pointer (605), and the target data is accessed (606). If the search target data does not exist in the destination node or its child or grandchild node, the node is further moved to the node in the higher hierarchy (603) and the segment value in the node is compared with the search target key value.

The B-tree search processing unit does not compare the magnitude relationships of all the layers in the B-tree from the root node and proceed with the search to the target node. The search is started from the “closest” direct access node to the target node and data. If the number of search returns to the upper hierarchy increases, the effect of the direct access node decreases, and conversely the search processing time increases. Therefore, the direct access node setting unit determines the number of direct access nodes so that the number of search returns to the upper layer does not exceed the number of times of hierarchy traversal in the conventional B-tree search, and the direct access node access process determines the number of search returns to the upper layer. Including "," the "closest" direct access node in the B-tree is determined so as to minimize the number of crossing over all layers.

FIG. 7 is a flowchart showing the sublink setting process. When the sublink setting condition is designated (701), first, the sublink setting condition is registered in the sublink setting condition list in the sublink information storage area 711 (702). Next, it is determined from the top element of the B tree whether or not the condition is satisfied, and the key value of the data that first satisfies the condition or the data record address is set to the sublink start address of the sublink information storage area 711. (70
3) At the same time, the key value of this data or the data record address is registered in the preceding element securing area P (704). The B tree is continuously searched for data satisfying the sublink setting condition (705). When data satisfying the sublink setting condition is found (706), the previous data registered in the previous element securing area P is searched. The own data data record address is set in the sublink information storage area (221 to 223 in FIG. 2C) provided for the key value (707). At the same time, the key value of the own data or the data record address is newly rewritten in the preceding element securing area P (708). Similarly, the conditions are sequentially determined in the B-tree up to the last element (709), and sublinks are set between the data satisfying the conditions.

By this processing, a chain in which the data satisfying the specified sublink condition is arranged in ascending order according to the key value is created. Even if a plurality of key value sublink setting conditions are set, all the data in the B-tree may be simultaneously judged as the sublink condition for all the conditions.

FIG. 8 is a flowchart showing the sublink search process. When searching for data satisfying the conditions in all cases, first, if the search condition is registered in the sublink setting condition of the sublink information storage area 811
(802), the data head address is obtained from the sublink information storage area 811 (803), and the head data is accessed (804). From the data link information storage area (Fig. 2
(221 to 223 of (c)) The address of the next data is obtained (805), and the sub data is traced to access the next data (806). This process is continued until the last data of the sublink.

[0047]

By providing a direct access node that can be directly accessed from the root node of the B-tree without performing division value comparison processing of all layers and hierarchical movement processing, the processing speed can be increased by a maximum (number of layers). Can be changed.

[Brief description of drawings]

FIG. 1 is a block diagram of the entire system of the present invention.

FIG. 2 is an explanatory diagram of a node configuration of a B tree.

FIG. 3 is an explanatory diagram of a configuration of a B tree.

FIG. 4 is a flowchart showing a direct access node setting process.

FIG. 5 is a flowchart showing a direct access node access process.

FIG. 6 is a flowchart showing a B-tree search process.

FIG. 7 is a flowchart showing a key value sublink setting process.

FIG. 8 is a flowchart showing a key value sublink search process.

[Explanation of symbols]

111 ... Direct access node setting processing unit, 112 ... Direct access node access processing unit, 113 ... B-tree search processing unit, 117 ... Direct access node information storage area.

Claims (4)

[Claims] 1. A B stored in a storage device of a computer system.
The index search processing method using a tree comprises a direct access node that can be directly accessed from the root node of the B tree without searching the B tree. 2. The direct access node setting according to claim 1, wherein the direct access node that can be directly accessed from the root node of the B tree without searching the B tree is determined, and information about the direct access node is stored. Section, a direct access node access processing section for directly accessing the direct access node from the root node of the B tree during the search processing without searching the B tree, and when the search target is not set as the direct access node. An index processing method comprising a B-tree search processing unit that performs a search process from the accessed direct access node to target search target data. 3. B stored in a storage device of a computer system
An index search method using a tree, wherein each data storage area of a node is provided with a key value sublink. 4. The key value sublink setting unit for setting a key value sublink in each data storage area of the node according to claim 3, and when the search condition is set as the sublink setting condition, the search is performed according to the sublink. Index value processing method including a key value sublink search processing unit.