JPS60181849A - Index control system - Google Patents

Abstract

PURPOSE:To reduce the record capacity of an index itself and at the same time to have a high-speed access to a desired record occurrence, by deciding the type of the index record occurrence in consideration of the key value of the record occurrence and the mutual relations among addresses having the record occurrence stored in the data base. CONSTITUTION:At the outset, the 1st index record type is defined with the record occurrence having the hundred million key value. Then the 2nd index record type is defined as long as the relation satisfies an equation I among relative code numbers of the record occurrence having the overlap key value. Here just a bit map item suffices if the displacement of N units of record relative numbers is within an (LBX8) range. Then plural bit map items are needed if said displacement exceeds the (LBX8) range. Otherwise the LB length can be increased. The 1st index record type is used in case the equation I is not satisfied despite the pressure of the overlap key value. As a result, the independent characteristics can be provided for each index to increase greatly the using efficiency of the index.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an index management method for realizing high-speed content search processing on a file device in a database management system.

(Prior Art) Conventionally, a system called B+ tree has often been adopted for indexing. In the B-tree, there is always one highest-level node (called the root), and each node has the maximum key value contained in the lower nodes, and
A pair of record occurrences constituted by pointers to the nodes, ie, a plurality of index record occurrences, are included, and the index record occurrences are nodes arranged in the order of key values. When a pointer to all index record occurrences that belong to a node points to a record occurrence that holds that key value, this node is called a 'f: R5' subordinate node, and The number of index record occurrences and the number of access target record occurrences match.

For indexes with B″ tree structure, even when adding a record occurrence with a new key value,
Or, even when adding a record occurrence with a duplicate key value, create an index record occurrence containing that key value at the corresponding location of the lowest node.
, already 4=ytakes record occurrence and logical order are guaranteed. Therefore, in a conventional index with a B+ tree structure, the number of nodes that make up the index increases in proportion to the number of records in the population, and even if the number of records with duplicate key values increases, the number of index occurrences decreases. The drawback was that it did not decrease.

(Objective of the Invention) The object of the present invention is to determine the format of an index record occurrence by taking into account the relationship between the key value of the record occurrence and the address where the record occurrence is stored in the database. It is an object of the present invention to provide an index management system configured to eliminate the drawbacks, reduce the recording capacity of the index itself, and enable high-speed access to a target record occurrence.

(Structure of the Invention) The index management method according to the present invention includes a plurality of keys 1
b, and 1 each representing one of the plurality of key values.
In an index management method that generates and maintains an index record occurrence that includes addresses of a plurality of record occurrences, the index generation means and the file means are provided.

The index maintenance means is also provided.

The index generating means includes a first type of index record format consisting of a pair of key values and an associated address, a plurality of key values, and a plurality of record occurrences each including one of the plurality of key values. a second type of index record format consisting of one or more bitmap items for indicating an address; If the number is within the standard value, the first type index record format is selected, and if it exceeds the standard value, the Asahi 2 type index record format is selected. It is useless to generate an index.

The file means is for storing the generated index in a randomly accessible manner.

The index maintenance means changes the index record occurrence related to the above processing in synchronization with the above image processing when changing the key value in accordance with record occurrence update processing or addition processing for the generated index, This is to pass control to the process of comparison and selection with respect to the reference value by the index generation means.

(Example) Next, the present invention will be explained in detail with reference to all the drawings.

In a database management system that has a large amount of data, it is important to be able to quickly search for a target record occurrence in the database. For example, end-user languages that include powerful functions are increasingly being provided to end-users who use them in an online real-time environment. The query processing created by the end user includes up to one case of simple conditional force and one complex condition, and is mainly atypical processing performed by trial and error. The capacity is almost nil (up to a few dozen at most).Therefore, if there were more cases than this, it would have given a strong moxibustion value, and from another point of view. Iterate again while giving empty items.When such a process is assumed and if access efficiency is to be improved by using an index, ir is essential, and the indexing method of the present invention that takes data characteristics into account is used. The effect of a rainbow can be even more impressive.

The explanation will be given below with reference to Figure Shun 1 and Figure f4L2.

In FIG. 1, when creating an index, the bitmap item length (LB) of the second type index record format in FIG. 2 is given in units of all bytes. For example, if LB=32, it is possible to manage a maximum of 256 (32×8) record occurrences with 11 focus map items (32 bytes). LB
Since the value of can be changed for each index, it is possible to reflect the characteristics of each target key value.

Next, in order to create an index occurrence, [key value + record relative number] is collected, sorted in order of key value, and given. While checking the key values using these input data, select the first type shown in Figure 2, or the
Create and hold a relative file on the file device using an Intex body of type . Since multiple indexes can be created for the same record type, it is possible to prevent a power check from being performed for the first time. using numbers. The reason for this is that when solving a complex condition such as the one described above, the principle operation using data on the index is effectively replaced with a bit operation.

In this method, the criteria for separating the first type of index record format and the second type of index record format are as follows.

■ A record occurrence with a key value of - is the first
The index record format is of type .

■ If the relationship between the relative record numbers of record occurrences with duplicate key values satisfies the contents of Figure 3, the second
The index record format is of type .

Here, the displacement of N record relative numbers is 1, Bx s
If it's within that range, just one P/Nodomap item <, LB
If it is X8 or more, it is possible to use a plurality of pit map items or to increase the LB length.

If the index record has a multiple key value but does not satisfy the contents shown in FIG. 3, it is set to the first type of index record format.

Even when the key value is changed by adding and updating record occurrences after the index is created, conversion maintenance between type I and type ■ is carried out according to the above criteria.

In FIG. 1, the broken line indicates the case where a key value is given and compared with the given 5 key values to obtain a record relative number that satisfies the matching condition.

In the above method, each index can have independent characteristics, so the usage effect is very large.

(Effects of the Invention) As explained above, the present invention can significantly reduce the index recording capacity by using an index with duplicate key values, and reduce the number of times (physical and logical input/output times) for accessing index record occurrences during content search processing. This has the effect of improving processing efficiency because the angular force (contrast) is reduced.

[Brief explanation of drawings]

FIG. 1 is a schematic diagram showing an overview of the index management method. FIG. 2 is a diagram schematically showing index occurrences generated and maintained by the index management method of FIG. 1. 1-1...Bitmap item length 1-2...Key value and 1-code relative number Patent applicant NEC Corporation representative Patent attorney Inoro ′ Juga 1 diagram (222== ): Input/output value

Claims (1)

[Claims] In an index management system configured to generate and maintain index record occurrences consisting of multiple key values and addresses of multiple record occurrences each containing one of the multiple key values, a first type of index record format consisting of a key value, an associated address, and a PUR; a plurality of e record occurrences, each of which includes one of the plurality of key values, and one of the plurality of key values; and a second type of index record format consisting of one or more focus map items (to indicate the address of the index), and the number of index record occurrences having the same key value at the time of index generation, Kl, If the number is within the reference value, the first type of index record format is selected; if the number exceeds the reference value, the second type is selected. 2. An index generating means for selecting an index record format to generate an index; 2. A file means for storing the generated index in a randomly accessible manner; and 2. File means for storing the generated index in a randomly accessible manner; When the key value is changed in accordance with additional processing, an index record occurrence related to the processing is changed in synchronization with the image processing, and the comparison with the reference value by the index generation means and the selection are performed. How do I pass control to index maintenance? An index management method characterized by comprising: