JPH0855050A - Dynamic index preparing device - Google Patents

Abstract

PURPOSE:To automatically decide the optimum number of hierarchies so as to enable more efficient retrieval processing by flexibly changing the preparation of a hierarchical index file for data retrieval corresponding to the number of data records. CONSTITUTION:This device is composed of an optimum index dimension deciding means 1 for judging the number of hierarchies as no index or one-dimension to n-dimension according to a power root calculating system corresponding to the increase/decrease of the number of data records, index file preparing means 2 for preparing an index file 5 equipped with the hierarchical structure for the decided dimension, and managing information file preparing means 3 for preparing a managing information file 6 storing a data file name, index file name and index dimension in order to relate a data file 4 and an index file 5.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the creation of an index for performing efficient data search in a data search system, and particularly to the search of a plurality of data files having different data amounts and the increase or decrease of the data amount even in the same file. The present invention relates to indexing of a data search system when searching for a data file that causes a problem.

[0002]

2. Description of the Related Art Conventionally, when a data search is performed using an index, particularly when there is a large amount of data, the search processing has been speeded up by hierarchizing the index. The hierarchical structure of this index was fixed and did not support the increase or decrease of data.

[0003]

Generally, when there is a small amount of data, it is more efficient to search without an index, but when there is a large amount of data, it is effective to hierarchize the index. Become. However, in the conventional technique, since the index is fixed, there is a problem that an effective index configuration cannot be always taken when increasing or decreasing the data amount. In order to solve this, it is necessary to change the index each time using a tool or the like in proportion to the amount of data, which requires a lot of man-hours.

[0004]

Optimum index order determining means for determining the index order of an index of a hierarchical index file by a power root calculation method according to increase or decrease of data files to be searched, and the optimum index order determining means. And an index file name for associating the index file with means for sequentially creating the index file having a hierarchy of the degree determined by the method, and a data file storing data and the index file. , And a management information file creating means for creating a management information file which is a sequentially organized file storing index orders.

[0005]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described with reference to the drawings.

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

According to the present invention, the optimum index order determining means 1 for determining the index order of the index of the hierarchical index file 5 by the power root calculation method according to the increase or decrease of the data file 4 to be searched, and the optimum index order determination. An index file creating means 2 for creating an index file 5 having a hierarchy for the degree determined by the means 1 by a sequential storage method, and a data file name for associating the data file 4 storing the data with the index file 5 And a management information file creating means 3 for creating a management information file 6 which is a sequential organization file storing the index file name and the index order.

The flow of the entire process of creating a dynamic index will be described with reference to the flowchart of FIG.

When the data file name and the index file name are given as parameters from the terminal of the data retrieval system (step 6-1), the optimum index order determining means 1 is started (step 6-2), and the record of the data file 4 is recorded. The number of cases is input, the index order is determined by the power root calculation method described later, and the result is notified to the index file creating means 2 (step 6-
3, 6-4).

The data file 4 is premised on the fact that the number of records can be acquired without reading all the data from the file identifier and the like, and that it has already been sorted by the search key.

Next, the index file creating means 2 creates the index file 5 having a hierarchical structure according to the index order notified by the sequential storage method described later (step 6-5).

Finally, the management information file creating means 3 is activated (step 6-6) to create the management information file 6 which is a sequential file in which the data file name 61, the index degree 62 and the index file name 63 are stored. (Step 6-7) and the process ends.

Before describing the power root calculation method, the structure of the index file 5 created in this embodiment will be described with reference to FIGS. 2 and 3.

Index file 5 as shown in FIG.
In, the index blocks are stored in an n-order hierarchical structure. That is, there are a plurality of (l in FIG. 3) secondary index blocks 52-1 to 52-l below the primary index block 51-1 and m tertiary index blocks exist below it. To do. Similarly, the n-th index block is 5n-1 to 5n-o.
It is shown that there are o pieces of.

The basic unit of the index file 5 is the index block 5n-p. Here, n is the order of the hierarchical structure, and o is the index block number in the order.

FIG. 2 shows the primary index block 51-.
1 is shown as an example. A plurality of index records 5n-o-p are stored in the index block 5n-o (p is the number of the index record in one index block), and one record contains the data file 4 to be searched. Search key 5n-op-1 and the index block 5n-op-2 of the next layer, that is, the record pointer are set.

Next, the hierarchy of the index file 5 will be described with reference to FIG.

FIG. 3 shows the theoretical structure of the index file 5 in the case of having a hierarchy up to the nth order.

Only one primary index block 51-1 is created, and the offset 51-1 of each record is created.
-P-2 indicates the beginning of each secondary index block 52-1.

A plurality of records are created after the secondary index block 52-1 and the offset 5n-op-2 of each record points to the head of the index block 5n-o of the next layer. .

In the same manner, the indexes are searched step by step in the same manner, and the n-th order index block 5n- of the final hierarchy is searched.
When it reaches o, the offset of each record is 5n-o
-P-2 indicates the head of the data record having the corresponding key, and the data to be searched can be acquired for the first time here.

As for the physical structure of the index file 5, a primary index block 51-1 is stored at the head, then secondary index blocks 52-1 to 52-l, and a final index block, as shown in FIG. Are stored in the order of the nth index block 5n-o.

The power root calculation method will be described.

The power root calculation method is a method of determining the optimum index order by the following arithmetic expression using the number of records in the data file 4 as an input value. That is, when R ≦ a ... (1), the index order is the 0th order. In cases other than (1), the smallest natural number n such that R ≦ b raised to the nth power (2) is obtained and the index order is set to the nth order.

Here, R: the number of records in the data file 4 a: the upper limit of the number of records in the data file 4 when the index file 5 is not created b: to a storage medium such as a magnetic disk or a CD-ROM The number of index records 5n-op that can be stored in the index block 5n-o having the block length c determined in consideration of access efficiency. Search key length d and offset length e
Is obtained from

B = c ÷ (d + e) (3) For example, if one block has 2048 bytes in a storage medium, it is assumed that the access efficiency is improved.
At this time, if the record length of the data record 4i is 512 bytes, 4 records can be read by one access, so that the index file 5 need not be created if the number of records in the data file 4 is 4 or less. . That is, a = 4 (example 1).

When the index file 5 is created when the number of data records is sufficiently large, the search key 5n-
16 bytes for op-1 and offset 5n-op-2
Is 16 bytes, primary index record 5n
The record length of -op is 32 bytes, and 64 records can be stored in the primary index block 5n-op. That is, b = 64 (example 2).

If the number of records in the data file 4 in Example 2 is 262,000, then 262,0
Since 00 <64 is the third power, the index order is determined to be 3 (Example 3).

Next, the sequential storage method will be described with reference to FIG. 5 with a specific example.

Assuming that b = 10 and the number of data records is 9
In the case of 9 cases, the index order becomes 2 from the expressions (1) and (2), and the index file creating means 2 creates the index file 5 by the following procedure.

As shown in FIG. 5, first, the data file 4
Of the first record 41 of the key 1 (72-1) to the secondary index record buffer 72-1-1 of the last layer.
-1-1) and offset 1 (72-1-1-2) indicating the beginning of the data record 41 are set (step 8).
1). Similarly, the data block 410 is sequentially read, and the index block buffer 72-1 becomes full when the key 10 (72-1-10-1) and the offset 10 (72-1-10-2) are set. Therefore, key 10 (72-1-10-1) and offset 1 that points to the beginning of the secondary index block buffer 72-1
0-1 (71-1-1-2) is set in the primary index block buffer 71-1 (step 82). The data records 411 are sequentially read again (step 83), and when the index block buffer 72-y is full, the primary index record buffer 71 is read.
Set -1-z. Repeat the same process below,
When the last data record 499 is read, the key 99 is stored in the secondary index record buffer 72-10-9.
After setting (72-10-9-1) and the offset 99 (72-10-9-2) of the data record (step 84), null padding of the remaining portion of the index block buffer is performed. Next, the key 99 (71-1-1-) is stored in the primary index record buffer 71-1-10.
1) and offset 99-1 (71-1-1-2) pointing to the beginning of the secondary index block buffer 72-10 are set (step 85). Finally, the primary index block buffer 71-1 (step 86), the secondary index block buffer 72-1 (step 87),
.... Secondary index block buffer 72
Index file 5 in the order of -10 (step 88)
To the index file 5, and the process of creating the index file 5 ends.

A process for searching data according to the present invention will be described.

Key 9 in data file 4 in FIG.
When searching 8, the search program refers to the management information file 6 (see FIG. 3) to obtain the index file name 63 and the index degree 62, and the primary index block 71 at the top of the index file 5 is acquired.
Access -1-z. Here, the current index record such that (key of previous index record) <(key to be searched) ≦ (key of index record) is searched. In this case, the index record 71-
1-10 corresponds to this. Next, offset 71-1-
Since 10-2 points to the beginning of the secondary index block 72-10, it is accessed and the key 9 to be searched is searched.
8 and the key 72-10-z-2 in the secondary index record 72-10-z. At this time, since the value of the index order 62 is 2, it is determined that the secondary index is the last layer, and the (current index record key) = (key to be searched) secondary index record 72-10-z Look for. In this case, the secondary index record 72-10
Since -8 corresponds, the offset 72-10-8-2 points to the beginning of the data record having the search target key 98 in the data file 4, and the data file 4 is accessed based on this to perform the search process. finish.

It is assumed that there is a data file 4 in which data update occurs every month and the number of records increases / decreases, and an index file 5 is created.

For example, when the data amount is relatively small in May and the index order is judged to be 1, the index file 5 of only one layer is created, and the index order 62 in the management information file 6 is set to 1. Is stored.
However, when the amount of data increases significantly in June of the next month and the index degree 62 is determined to be 3, the index file 5 of three layers is created and 3 is stored in the index degree 62 of the management information file 6. It The data search program refers to the index order 62 in the management information file 6 to determine the index block 5
It is possible to grasp how many times the noo is accessed to obtain the offset 5n-op-2 of the data record 4i to be searched, and it is possible to easily deal with the change in the hierarchical structure of the index file 5.

[0036]

As described above, according to the present invention, when a plurality of data files having different data amounts are searched, or when the same file is searched for a data file whose data amount frequently increases and decreases. Has the following effects. 1. If there is a large amount of data, create a hierarchical index,
When the amount of data is small, by creating an index of only one layer, it is easy to cache the upper index and make the index resident in memory. Also, when the data amount is particularly small, the index itself is not created. By such a measure, the speed of the data search process can be increased. 2. By creating a management information file on the index creation processing side and managing the file structure such as the index order, the data search processing side can easily handle changes in the dynamic index.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of the present invention.

FIG. 2 is a configuration diagram of an index block according to an embodiment of the present invention.

FIG. 3 is a logical configuration diagram of an index file according to an embodiment of the present invention.

FIG. 4 is a physical configuration diagram of an index file according to an embodiment of the present invention.

FIG. 5 is a block diagram showing a procedure for creating an index file according to the sequential storage method of the embodiment of the present invention.

FIG. 6 is a flowchart showing the flow of processing of the present invention.

[Explanation of symbols]

 1 Optimal Index Order Determining Means 2 Index File Creating Means 3 Management Information File Creating Means 4 Data File 5 Index File 6 Management Information File 4i Data Record 5n-o nth Index Block 5n-o-p nth Index Record 5n-o- p-1 search key 5n-op-2 offset 61 data file name 62 index order 63 index file name 7x-y x-order index block buffer 7x-y-z x-order index record buffer 7x-y-z-1 search Key buffer 7x-yz-2 offset buffer

Claims (2)

[Claims] 1. An optimum index order determining means for determining an index order of an index of a hierarchical index file by a power root calculation method according to increase or decrease of a data file to be searched in a data search system using a computer, and the optimum index order. Index file creating means for creating the index file having a hierarchy for the degree determined by the determining means by a sequential storage method, and a data file name and an index for associating the data file storing the data with the index file A dynamic index creating device, comprising: a management information file creating means for creating a management information file that is a sequentially organized file that stores a file name and an index order. 2. The index file is composed of a plurality of index blocks, and a plurality of index records are stored in the index block, and a search key of the data file to be searched is stored in the index record. , The index block of the next layer, that is, the record pointer is set, and only one primary index block is created,
The offset of each index record points to the beginning of each secondary index block. After the secondary index block, a plurality of index records are created, and the offset of each index record is the index block of the next hierarchy. A dynamic index creation device, which points to the beginning and stores the information of the beginning of the data record of the corresponding search key at the offset of each record in the n-th index block of the final hierarchy.