JPH11232283A - Information retrieving method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To compile a file to be suited to a purpose of retrieval when the file is retrieved by dividing each item of records to constitute a file into a plurality of ranks depending on a range of values capable of being taken by each item and providing it with a string of ranks. SOLUTION: A groups of records is judged whether it is divided or not by regarding the entire records of a retrieving object file 11 as a root of a look-up tree about a rank string, starting the retrieval from the root of the look-up tree and referring to a parameter 14 by a look-up tree generating part 12. When the group of records is divided, it is divided into the group of records corresponding to each rank of unprocessed item with top priority by referring to a data item comparison table 15. The group of records not to be divided is successively stored in a file 18 and the corresponding rank string is stored in a look-up tree table 17. A table 20 is created by providing each rank string of the table 17 with an address and a compilation file 21 for indexing sequence formed by storing each group of records of the file 18 in an area of its address is created by a file compiling part 19.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for retrieving information using a computer, and in particular, each data item of a record constituting a file is divided into a plurality of ranks according to a range that the data item value can take. Such a file is searched by a column of a given rank.

[0002]

2. Description of the Related Art Generally, a file or database, which is a set of data records, is shared by a plurality of applications or a plurality of departments. However, the organization of the files and databases is not always suitable for the search purpose of each application or department, so reorganize the original files and databases so that they are optimal for each search purpose. Is done. For example, depending on the application, it is considered that each data item of a record constituting a file is divided into a plurality of ranks according to a range that the data item value can take, and a record corresponding to a target rank sequence is obtained from a file by giving a rank sequence. Search methods that extract groups may be useful. In this case, each data item is classified into ranks according to the search purpose, and file search is generally performed using a plurality of keys.

As described in "Database Specialist Text" (Central Information Processing Educational Research Institute, 1994), edited by the Japan Information Processing Development Association, speeding up retrieval of index-sequenced files using a plurality of keys. for,
A method is known in which an index corresponding to each key is made resident in a main storage area (memory), and a target record or record group is searched using the plurality of sets of indexes. However, this method generally requires a large amount of memory because a plurality of indexes are made to be memory-resident, and when extracting a target record by giving a rank sequence as described above, each index is used one by one. It takes a lot of time to search for the target records while searching, and it is not the most suitable search method for such a search purpose.

A method of directly accessing an organization file by using a hash function is described in the above-mentioned "database specialist text". This method uses a hash function that takes the data item value of the data item selected as a search key as an argument, finds the address of the storage area that stores the target record, and then directly enters the record at the found address in the organization file. Access,
There is no need to provide an index table on the memory, and a target record can be searched at high speed. However, since the search key is limited to one, it is not suitable for a file search using a plurality of keys as in the above-described rank sequence.

[0005] Also, "Oracle 7.3 New Function Course"
As described in (Japan Oracle Corporation, 1996), among data items constituting a record, a data item whose item value can be classified into a rank, a bitmap index for expressing each rank by 1-bit information. And a method of quickly searching for a record group corresponding to a given rank sequence by referring to the bitmap index. Although this method can reduce the required memory capacity compared to the above-mentioned plural sets of indexes even if the bitmap index is made memory-resident, in order to extract a record group corresponding to a given rank column, it is necessary to specify each specified data. This is a method of taking the logical sum of all bitmaps that match the search condition for the item, taking the logical product of all the logical sums of the obtained data items, and extracting the record group that hits from the file to be searched. There is no concept of reorganizing search target files to be optimal for individual search purposes.

[0006]

As described above, according to the prior art, when performing a file search using a plurality of keys, particularly a search using a rank sequence designating a rank over a plurality of data items, individual search purposes are performed. There was no idea to restructure both the index and the search target file to be optimal for.

[0007] An object of the present invention is to perform a search using a rank sequence specifying a rank over a plurality of data items,
It consists in reorganizing both the index and the file to be searched so as to be optimal for the respective search purpose.

[0008]

According to the present invention, the entire record group included in a file is regarded as the root of a search tree relating to a rank sequence, and the record group is divided according to a division condition set starting from the record group corresponding to the root of the search tree. It is determined whether or not to further divide, and if the record group is regarded as the root or node of the search tree and further divided, a record group corresponding to each rank of the data item having the highest priority among the data items not yet involved in the division When the record group is regarded as a leaf of the search tree and is not further divided, a rank sequence corresponding to the record group is sequentially stored in the search target file and the search tree table, respectively. In this way, the search tree table stores search tree information in order from the rank sequence having the highest priority according to the search purpose, and the search target file stores records corresponding to the search tree information in the search tree table in order.

To create an index sequential file,
A storage area is sequentially allocated to a record group corresponding to each rank column of the search tree table according to the storage area allocation method of the index sequential organization file, and a head address thereof is added to the rank column to create a search tree index table. The records are sequentially stored in a storage area starting from the determined address on the search target file. A rank sequence is given as a search condition, and when searching this index-sequential file, the search tree index table is searched for an address corresponding to the given rank sequence, and the target address in the index-sequential file is obtained from this address. Access records.
In this way, a record group that matches the search condition can be extracted by one sequential access, and a file search suitable for the search purpose can be efficiently performed.

When a direct organization file is created, a search tree index table is created by adding a sequence number of a record group corresponding to each rank sequence contained in the search tree table, and this sequence number is used as an argument. Then, the address of the storage area accommodating the record group is calculated by a hash function, and each record group accommodated in the search target file is directly stored in the storage area of the determined address of the organized file. When a rank sequence is given as a search condition and this direct organization file is searched, a search tree index table is searched to find a sequence number corresponding to the given rank sequence, and a hash function is used to find a corresponding record group from this sequence number. Calculate the address and directly access the record group corresponding to the specified rank column in the organization file.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below in detail with reference to the drawings.

FIG. 1 shows a processing procedure of a computer for creating search tree information suitable for a search purpose using an electronic computer, organizing an index sequential file based on the search tree information, and performing information search. FIG. The search target file 11 is
The original file to be searched. The data item ranking table 13 is a table that ranks the values of the data items included in the records of the search target file 11 by setting a range between an upper limit value and a lower limit value. The division condition parameter 14 is a parameter for setting a condition for determining whether to further divide a record group having a specific rank sequence. Here, the rank column refers to the rank of item 1 and the rank of item 2 for the data items included in the record.
・ It is something which limits. Data item comparison table 15
Is a table that is referred to when a data item to be the next priority item is selected when the record group is further divided. The work file 16 is a work file used when dividing the record group of the search target file 11, and includes a work file 16-1 and a work file 16-.
And 2. The work file 16-1 and the work file 16-2 are used alternately as an input file and an output file. The search tree table 17 is a table that stores a rank sequence corresponding to a record group as search tree information. The search target file 18 is the search target file 1
1 is a file in which files 1 are rearranged so as to correspond to the search tree information in the search tree table 17. Search target file 11,
Data item ranking table 13, division condition parameter 14, data item comparison table 15, work file 16, search tree table 17, and search target file 18
Is stored in the storage device of the computer.

The search tree creating unit 12 is a program stored in the main storage of the computer. The search tree creation unit 12
The records of the search target file 11 are sequentially read and output to one of the work files 16 to output the search target file 11.
Create a record group for the whole. Next, the record group of the work file 16 is read, the data items are ranked according to the data item values included in the records with reference to the data item ranking table 13, and the number of records is counted for each item and rank. It is stored in the data item comparison table 15. Next, a priority item is determined by calculating a standard deviation based on the reference rank of the reference item from the number of records counted for each item and rank in the data item comparison table 15, and a record group is determined by referring to the division condition parameter 14. It is determined whether or not to divide further. If the record group is further divided, the record group is divided according to the rank of the priority item and the other work file 1
6 is output. If the record group is not further divided, the rank sequence corresponding to the record group is output to the search tree table 17 as search tree information, and the entities of the record group are sequentially output to the search target file 18. In this way, the record group is subdivided according to the rank column,
When the division of all the record groups is completed, the processing ends. Eventually, a search tree of a record group is created, search tree information of a rank sequence is created in the search tree table 17 in order from the upper layer of the search tree, and a search target file 18 corresponds to each search tree information. Records are sequentially stored.

The search tree index table 20 is obtained by adding an address for storing a record group corresponding to each piece of search tree information of the search tree table 17. Index sequential file 2
1 is a file that stores each record group on the search target file 18 at an address corresponding to the search tree information in the search tree index table 20. The search tree index table 20 and the index sequential organization file 21 are stored in a storage device of the computer. The file organization unit 19 is a program stored in the main storage device of the computer, reads each search tree information of the search tree table 17, and stores the corresponding record group in the search target file 18 in the index order organization file 21. The start address at the time is determined, the address is added to the search tree information, and the search tree information is stored in the search tree index table 20. The corresponding record group is read from the search target file 18 and starts from the determined address on the index sequential organization file 21. Store in area.

23 is an input device for inputting a rank sequence to be searched, 25 is a search result stored in a storage device,
26 is a display device for displaying the search results, and 27 is a printer for outputting the search results. The information search unit 24 is a program stored in the main storage device of the computer.
The search tree index table 20 is searched by using the rank sequence input from 3 as a key to obtain a corresponding address, a record group is extracted from a storage area starting from the obtained address on the index sequential file 21, and a search result 25 is obtained. The output is displayed on the display device 26 or output to the printer 27.

An information processing device such as a personal computer, a workstation, or a mainframe is used as computer hardware, and the above-described files, parameters, and tables are stored in a storage device of the computer.
2. An information search system in which the programs of the file organization unit 19 and the information search unit 24 are stored in the main storage device and executed by the processing device can be constructed. Further, these programs can be stored in a computer-readable storage medium and stored in a main storage device via a driving device connected to the computer or by transmission and executed.

FIG. 2 is a conceptual diagram showing the structure of a search tree of an index sequential file. The entirety of the data record to be searched is positioned at the root 31 of the search tree. The root 31 is divided into nodes 32 and leaves 33 of the search tree. In FIG. 2, the dotted frame indicates the root 31 or the node 32, and indicates that the data record group is a branching node. The thick line frame indicates the leaf 33, which indicates that the data record group is a branch stop node. The upper part of the frame indicating the node is a rank column indicating the rank of each of the three data items included in the record. The rank is a classification made based on the value range of the data item. * Indicates that the item is not classified. For example, a node whose rank sequence is * A * is 2
The second item is a set of records corresponding to rank A, and a node whose rank column is aA2 is the first item having rank a, the second item having rank A, and the third item having rank 2
Is a set of records corresponding to. The middle part of the frame expresses the address of the first record of the node in the file by a relative address such as a relative byte address or a relative block address. The root 31 and the node 32 have no relative address because no record entity is assigned. The lower part of the frame indicates the number of records included in that node. In this example, the total number of records in the file is 500. In this example, a node having a rank column of * B * has a record count of 90, and is not further divided into nodes according to the limit on the record count in the record division condition, and the branch is stopped.

FIG. 3 shows a data item ranking table 1.
FIG. 3 is a diagram illustrating a data configuration example of No. 3; The item name is the name of the data item to be ranked included in the record, and the data type indicates whether the value of the data item is a numeric string or a character string, or the type of data. The number of ranks indicates the number of ranks of the data item, and the upper and lower limits of each rank indicate the range of the data item value for that rank. However, when the data type is a character string, there is no range since the data item value is determined by each rank, and the data item value is set as the lower limit value for convenience.

FIG. 4 is a diagram showing an example of the data structure of the data item comparison table 15. The item name is the name of the ranked data item excluding the reference item included in the record, the rank name is the name of each rank of the data item, and the number of records is the number of records belonging to that rank. The number of records corresponding to the reference rank r of the reference item f is the number of records having the attribute of the reference rank r of the reference item f among the records of the corresponding rank. Here, the reference item is a data item suitable for a search purpose. The intermediate value of the standard deviation calculation is an intermediate value calculated for each rank of each item, and the standard deviation is a standard deviation of each item calculated from this intermediate value. The processing end flag is set to ON when the corresponding item becomes a priority item so as to exclude the item from the calculation of the standard deviation later as an item that has already been involved in the record division. Numerical values from the number of records to the standard deviation except for the item name, rank name, and processing end flag vary for each record group corresponding to a node in the search tree. The data item comparison table 15 is created corresponding to a record group corresponding to the root 31, the node 32, or the leaf 33. The layer added to the table indicates the layer number to which the data item comparison table 15 belongs, and the rank column indicates the rank of each item of the record group corresponding to the table. In the initial state, only one data item comparison table 15 corresponding to the root 31 is prepared. The hierarchy is 1 and the rank column is the rank column of the root 31. In addition, fixed data is set for item names and rank names,
Other items have been initialized.

FIG. 5 is a diagram showing an example of data of the division condition parameter 14. Number of records in record group ≥ 140
The case indicates that the record group is further divided if the number of records corresponding to the search tree node is equal to or greater than the specified number. If the upper limit value ≧ the ratio of the records corresponding to the reference rank r of the reference item f ≧ the lower limit value, the record group is set if the ratio of the records corresponding to the reference rank r of the reference item f is within the specified range. Is further divided. A condition that targets only the upper limit or the lower limit may be used. Alternatively, a division condition such as “divide if the number of records having 90 or more reference items is less than 3 records” may be used. The data capacity of the record group ≧ the storage page capacity indicates that the data is divided if the data capacity of the target record group is equal to or larger than the page capacity (for example, 4 KB). When the target record group satisfies all the conditions, the record group is further divided. If any of the conditions is not satisfied, the division is stopped.

FIG. 6 is a diagram showing an example of the data structure of the search tree table 17. The search tree table 17 stores search tree information corresponding to each record group corresponding to the leaves 33 of the search tree. The search tree information is arranged in order from the record group with the highest priority, that is, from the record group with the highest hierarchy in the search tree. Each piece of search tree information is data for one row of the table, and includes the rank of each data item and the number of records. The search tree index table 20 is the search tree table 1
7 is obtained by adding an address to each row. The address indicates the start address of the corresponding record group. The search tree index table 28 described later is obtained by adding a sequence number to each row of the search tree table 17. The sequence number is a number indicating the order of the corresponding record group in the direct organization file.

FIGS. 7A and 7B are flowcharts showing the processing flow of the search tree creating unit 12. The search tree creating unit 12 sequentially inputs records from the search target file 11 and outputs the records to one work file 16 (step 4).
1). This work file 16 is used as a work file on the input side. Next, it is determined whether the division processing has been completed for all records of the work file 16 (step 4).
2). The point in time when there is no valid record in both the input side and output side work files 16 is the end point of the division process. If the processing has not been completed, it is determined whether or not the processing of the hierarchy in the search tree has been completed (step 43). The point in time when the data item comparison table 15 corresponding to the current layer disappears is the point at which the processing of the layer is completed. If the processing of the layer has not been completed, the record is positioned in the record group on the work file 16 on the input side to be processed next and the corresponding data item comparison table 15 (step 4).
4). Next, it is determined whether or not there is an item for which the processing end flag of the target data item comparison table 15 is off, that is, there is an unprocessed item (step 45). If there is an unprocessed item (step 45 YES), each record of the target record group of the work file 16 on the input side is sequentially read, and the target data item comparison table 15 is read by referring to the data item ranking table 13. The rank is determined for each item for which the processing end flag is off, the number of records of the corresponding rank is counted by one, the number of records corresponding to the reference rank r of the reference item f is counted, and stored in the data item comparison table 15. (Step 46). Next, a standard deviation expressing the degree of influence of the reference rank r of the reference item f is calculated for each item for which the processing end flag is off (step 47). Next, if there are two or more unprocessed items, the standard deviation of each item is compared, and the item having the largest standard deviation is set as the priority item (step 48). Here, the item having the largest standard deviation is the item having the highest influence on the reference item. If there is only one unprocessed item, step 48 is skipped. Next, data item comparison table 1
The processing end flag corresponding to the priority item No. 5 is set on (step 49).

7B, the target data item comparison table 15 is compared with the division condition parameter 14 to determine whether or not the record group is to be further divided (step 50). If the record group is to be divided (step 50 YES), the data item comparison table 15 is generated for the number of ranks of the priority items in the data item comparison table 15 (step 51). The hierarchy of each generated data item comparison table 15 is the current hierarchy + 1, and the rank column is obtained by updating the rank of the item corresponding to the priority item in the current rank column to each rank name. In addition, for items for which the processing end flag is off, numerical items from the number of records to the standard deviation are initialized. Further, the data item comparison table 15 that is the source of the generation is deleted.

Next, the process proceeds to a process of dividing the record group into record groups for each rank of the priority item. If the division processing of all ranks has not been completed (step 52 NO),
Positioned at the head of the record group on the input work file 16, read only the records corresponding to the rank to be processed next among the ranks of the priority items in order, and written before the output work file 16 The extracted record groups are sequentially output to the area following the record group (step 53). At this time, the rank division information of the data item ranking table 13 is applied to each read record,
Only records corresponding to the rank of the priority item are extracted and output to the work file 16 on the output side. Also, the head address of the record group of the rank is stored for positioning the subsequent record group. Next, the rank is raised to the next rank of the priority item (step 54), and the process returns to step 52.
When the division processing has been completed for all ranks of the priority item (YES in step 52), determination and switching of input / output switching of the work file 16 are performed (step 55). That is, the input-side work file 16 for which the input processing has been completed.
Are determined as invalid records, and it is determined whether or not valid records remain in the work file 16 on the input side. If there is a valid record in the work file 16 on the input side, the process immediately proceeds to step 43. When there are no more valid records in the work file 16 on the input side, the work file 16 on the output side is changed to the input side, and the work file 1 on the input side is used.
After switching 6 to the output side, go to step 43. If there is no unprocessed item (NO at step 45) or if the record group is not to be further divided (step 50N)
O), the target record group of the work file 16 on the input side is sequentially read and output to the area following the record group written before the search target file 18 (step 5).
6). The number of records output at this time is counted. Next, search tree information of the record group is created (step 5).
7). The rank sequence of the search tree information stores the rank sequence of the data item comparison table 15 corresponding to the record group.
Also, the number of records counted as the number of records is stored. Next, the created search tree information is output as the next row of the search tree table 17 (step 58), and the procedure goes to step 55.

When the processing has been completed for all the record groups of the hierarchy (step 43YE)
S) Up only one current layer (step 5)
9) Return to step 42. When the division processing has been completed for all the record groups on the work file 16 (step 42 YES), the processing of the search tree creating unit 12 ends.

FIG. 8 is a flowchart showing the flow of the standard deviation calculation process in step 47. Search tree creation unit 1
2 refers to the data item comparison table 15 of the target record group (step 71), and determines whether or not the number of items for which the processing end flag is off, that is, the number of unprocessed items is one (step 72). If there are two or more unprocessed items (NO in step 72), the data item comparison table 15
(Step 73). Next, it is determined whether or not processing of all items has been completed (step 74).
If processing of all items has not been completed (step 74N
O), it is determined whether or not the processing end flag of the item to be currently processed is on (step 75). If the processing end flag is on, that is, if processing has been completed (step 75YE
S) Go to step 77. If the processing end flag is off (NO in step 75), the standard deviation is calculated for the item, and the result is stored in the column of the standard deviation of the item in the data item comparison table 15 (step 76).

In calculating the standard deviation, first,

[0028]

(Equation 1)

The standard deviation calculation intermediate value is calculated for each rank of the item according to the calculation formula, and is stored in the column of the standard deviation calculation intermediate value of the data item comparison table 15. When the standard deviation calculation intermediate value is calculated for all ranks of the item in this way,

[0030]

(Equation 2)

The standard deviation of the item is calculated in accordance with the calculation formula and stored in the column of the standard deviation of the data item comparison table 15. Here, n is the number of ranks of the item, and Σ indicates that the values from the intermediate value 1 to the intermediate value n are totaled.

When the process of step 76 is completed in this way, the next item is positioned (step 77), and the process returns to step 74. When the processing of all the items is completed (YES in step 74), the process proceeds to step 48. If there is only one unprocessed item (YES at step 72), the unprocessed item is set as a priority item (step 78), and the process goes to step 48.

FIG. 9 is a flowchart showing details of the processing in step 50. The search tree creation unit 12 refers to the division condition parameter 14 (step 81), and when the total number of records of each rank for the priority item determined in step 48 is equal to or greater than the number of records set in the division condition parameter 14, It is determined whether or not there is (step 82). If the number of records of the priority item is equal to or greater than the number of records of the division condition (step 82 YES), it is determined whether the data capacity occupied by the record group of the number of records of the item is equal to or greater than the page capacity set in the division condition parameter 14 (step 82). Step 83). If the data capacity is equal to or larger than the page capacity (step 83 YES), whether the ratio (record group average value) of the records corresponding to the reference rank r of the reference item f is within the range of the division condition set in the division condition parameter 14 (Step 8
4). The average value of a record group is calculated by the following formula. Average value = number of records corresponding to reference rank r of reference item f / total number of records of this data item The number of records corresponding to reference rank r of reference item f is
This is the total number of records for all ranks of the data item. If the average value is within the range of the division condition (step 84 YES), the process goes to step 51 as the division. If the number of records of the record group is less than the number of records of the division condition (step 82 NO), if the data capacity of the record group is less than the page capacity of the division condition (step 83 NO), or if the average value of the record group is If the condition is not satisfied (step 84N
O), it is determined that the record group is not to be divided, and the flow proceeds to step 56.

The file organizing section 19 searches the search tree table 1
7, the search tree information for one row is read, and the number of records is acquired. Next, the records of the number of records acquired from the search target file 18 are sequentially read, sequentially stored in a predetermined storage area of the index sequential file 21 according to the index sequential organization method of the file, and the relative address at the head of the storage area is read. It is stored in the search tree index table 20 in addition to the search tree information. When the address information is added to all the search tree information on the search tree table 17 and stored in the search tree index table 20 in this manner, and all the records on the search target file 18 are stored in the index sequential organization file 21 The processing of the file organization unit 19 ends.

When a search key consisting of a rank sequence is input via the input device 23, the information search unit 24 searches the search tree index table 20 and adds it to search tree information corresponding to the specified rank sequence. The obtained address is accessed, the indexed sequential file 21 is accessed to read the corresponding record group, written as a search result 25 on an external storage device, displayed on a display device 26, or output to a form on a printer 27. I do. The search tree index table 20 may be resident on the main memory (memory) in order to speed up the index search.

FIG. 10 shows an example in which a directly organized file is created based on the search tree table 17 and the search target file 18.
FIG. 9 is a diagram showing a processing procedure for retrieving information on the directly organized file. The processing of the search tree creating unit 12 is as described above,
The search tree creating unit 12 outputs a search tree table 17 and a search target file 18 as the processing result. The search tree index table 28 and the direct organization file 29 are files stored in the storage device of the computer. File organization unit 2
2 and the information retrieval unit 30 are programs stored in the main storage device of the computer.

The file organization unit 22 searches the search tree table 1
The search tree information for one row is read, and a sequence number starting from 0 is added to the search tree information and stored in the search tree index table 28. Next, the relative address at the head of the storage area for storing the corresponding record group is calculated using the hash function. The relative address is calculated using, for example, the following hash function. Relative address = hash function = capacity of file to be searched × sequence number of record group ÷ total number of leaves Here, the sequence number is an argument of the hash function. The file capacity can be calculated from the total number of records obtained by summing the number of records in each row of the search tree table 17 for all record groups. The total number of leaves is equal to the number of rows of search tree information stored in the search tree table 17. If the file capacity is the number of data bytes, the relative address is a relative byte address. If the file capacity is the number of data blocks, the relative address is a relative block address. Next, search target file 1 for the record group
From eight, records of the number of records are sequentially read and sequentially stored in the storage area starting from the calculated relative address on the direct organization file 29. In this way, the sequence number is incremented by one, all the search tree information on the search tree table 17 is added with the sequence number and stored in the search tree index table 28, and all records on the search target file 18 are directly organized. File organization unit 2 when stored in file 29
The process of No. 2 ends.

When a search key consisting of a rank sequence is input via the input device 23, the information search unit 30 searches the search tree index table 28 and adds it to search tree information corresponding to the specified rank sequence. The obtained sequence number is obtained, the relative address corresponding to the specified rank sequence is calculated using the above hash function, and the corresponding record group is read by directly accessing the organization file 29 to obtain the search result 2.
5 is written on the external storage device, displayed on the display device 26, or output to a form on the printer 27. The search tree index table 28 may be resident on the main memory (memory) in order to speed up the index search.

The structure of the search tree for the direct organization file is the same as that shown in FIG. However, in the direct organization file, the position of the record group in the file is represented by a sequence number instead of the relative address shown in FIG.

FIG. 11 shows another data item comparison table 1.
It is a figure which shows the example of 5 '. The item name is the name of a data item included in the record. Generally, the data item comparison table 15 'also includes reference items. The search frequency is the number of times that the item extracted from the past search history is used as a search key. The processing end flag is set to ON when the corresponding item becomes a priority item.

The search tree creating unit 12 performs steps 46 to
At 48, the number of records is counted for the unprocessed item with reference to the data item comparison table 15, the standard deviation is calculated, and the priority item is determined by comparing the standard deviation of the unprocessed items. Instead of steps 46 to 48, the search frequency may be compared for unprocessed items with reference to the data item comparison table 15 ', and the item with the highest search frequency may be set as the priority item.

FIG. 12 is a flowchart showing the processing flow of a rank division program for calculating the upper limit value and the lower limit value of each rank for each item of the record in the search target file 11 and creating the data item ranking table 13. is there. The rank division program displays the item names of the data items constituting the record on the display device, and prompts the user to input the number of ranks set for each item and the upper / lower limit specified by the user. Data items for which only the lower limit is set are specified by the user. When the number of ranks of each item is input via the input device and the upper limit / lower limit is input for the rank specified by the user (step 91), the item is positioned at the first item (step 92). If the processing for all the items has not been completed (NO in step 93), it is determined whether or not the upper limit / lower limit specified by the user is set for the item (step 94). If there is an upper limit / lower limit set (YES in step 94), the designated upper limit / lower limit is stored in each rank of the corresponding item in the data item ranking table 13 (step 95), and the process proceeds to step 99. If no upper limit / lower limit is specified (step 94N
O) The records are sequentially read from the search target file 11, the items are sorted in ascending order of the item values, and output to the work file (step 96). Next, the sequence number of the record corresponding to the upper / lower limit of each rank is calculated (step 97). The ranking of the records included in the search target file 11 is performed by dividing the number of records of the search target file 11 almost equally by the rank number. Assuming that the designated rank number of the item is n, the sequence number of the record that is the lower limit of the rank m (1 ≦ m ≦ n) is obtained by the following equation.

[0043]

(Equation 3)

Here, the upward arrow indicates that the numerical value after the decimal point of the calculation result is rounded up to an integer value by rounding up. The sequence number of the record, which is the upper limit of the rank m, is obtained by the following equation.

[0045]

(Equation 4)

Next, the records are sequentially read from the work file while counting the number of records, and the item value of the item of the record corresponding to the calculated sequence number is obtained.
It is stored in the upper limit value / lower limit value of the corresponding rank of the corresponding item in the data item ranking table 13 (step 9).
8). Next, it is positioned at the next item (step 99), and the procedure goes to step 93. When the processing for all items is completed in this way (step 93 YES), the processing of this program is completed.

In the method shown in FIG. 12, there is a case where the upper limit of rank m and the lower limit of rank (m + 1) match. In such a case, the same data item value is incorporated into the upper limit value of rank m, and a different data item value appearing by sequentially reading records of the work file is set as the lower limit value of rank (m + 1).

Note that the data item comparison table 15 'shown in FIG. 11 sets a priority specified by the user according to the search purpose instead of the search frequency, and determines a priority item for dividing the next record group according to the priority. May be.

The search tree index table 20 in which an address is added to the search tree information and the data having the data structure of the index sequential file 21 are stored in a computer-readable storage medium, and can be used for information search as described above. it can. Further, data having the data structure of the search tree index table 28 obtained by adding a sequence number to the search tree information and the direct organization file 29 are stored in a computer-readable storage medium, and can be used for information search as described above.

[0050]

As described above, according to the present invention, the search tree table is narrowed down to only the rank sequence suitable for the search purpose, so that the index search time is reduced as compared with the conventional index table. The records on the search target file are arranged in the order of the search tree information according to the search purpose. If a rank sequence is specified, the corresponding records can be extracted by one sequential access, and the time of the file search can be reduced. Be shortened.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a processing procedure of an information search method according to an embodiment.

FIG. 2 is a diagram illustrating a configuration example of a search tree of an index sequential organization file.

FIG. 3 is a data item ranking table 13 according to the embodiment;
FIG. 3 is a diagram showing a data configuration of FIG.

FIG. 4 is a diagram showing a data configuration of a data item comparison table 15 of the embodiment.

FIG. 5 is a diagram showing a data example of a division condition parameter 14;

FIG. 6 is a diagram illustrating a data configuration of a search tree table 17 according to the embodiment.

FIG. 7A is a flowchart illustrating a processing flow of a search tree creating unit 12 according to the embodiment.

FIG. 7B is a flowchart (continued) showing a processing flow of the search tree creating unit 12 of the embodiment.

FIG. 8 is a flowchart further illustrating step 47 of the embodiment.

FIG. 9 is a flowchart further illustrating step 50 of the embodiment.

FIG. 10 is a diagram illustrating a processing procedure of an information search method according to another embodiment.

FIG. 11 is a diagram illustrating a data configuration of a data item comparison table 15 ′ according to the embodiment.

FIG. 12 is a flowchart illustrating a flow of processing of a rank division program according to the embodiment.

[Explanation of symbols]

12: Search tree creation unit, 13: Data item ranking table, 14: Division condition parameter, 15: Data item comparison table, 17: Search tree table, 18: Search target file, 19: File organization unit, 20, 28 ... Search tree index table, 21: index sequential organization file, 24: information retrieval unit, 29: direct organization file, 30: information retrieval unit

Claims (5)

[Claims] 1. A rank provided with a file containing a plurality of records in which each record is composed of a plurality of data items and each data item is divided into a plurality of ranks according to a range that the data item value can take. In an information retrieval method using a computer, a search is performed by using a column of a search sequence. It is determined whether or not the group is to be further divided, and when the record group is regarded as a root or a node of the search tree and is further divided, it corresponds to each rank of the data item having the highest priority among the data items not yet involved in the division. If the record group is divided into record groups and the record group is regarded as a leaf of the search tree, The rank sequence corresponding to the group is sequentially stored in the search target file and the search tree table, respectively, and the address of the storage area accommodating the record group corresponding to each rank sequence stored in the search tree table is added to the search tree index. A table is created, and each record group contained in the search target file is stored in a storage area of the determined address to create an index sequential file. When a rank sequence is given, the search tree index table is created. And searching for a corresponding address, and accessing a record group corresponding to a specified rank sequence of the indexed sequential file. 2. A data item that matches a search purpose among a plurality of data items is set as a reference item, and a data item having the greatest influence on the reference item among data items not yet involved in division is set as a priority item. 2. The information search method according to claim 1, wherein: 3. The information search method according to claim 1, wherein among the data items not yet involved in the division, the data item having the highest search frequency based on the past search history is set as the priority item. 4. A search tree by adding a sequence number of a record group corresponding to each rank sequence contained in the search tree table, instead of creating the index sequential file and searching the index sequential file. While creating an index table, calculate the address of the storage area accommodating the record group by a hash function with the sequence number as an argument,
Each record group contained in the search target file is stored in the storage area of the determined address to directly create an organization file, and when a rank sequence is given, the search tree index table is searched and a corresponding sequence is created. 2. The information retrieval method according to claim 1, wherein a number is obtained, an address of a corresponding record group is calculated by the hash function, and a record group corresponding to a specified rank sequence of the directly organized file is accessed. 5. A computer program embodied in a computer-readable storage medium, wherein each program is composed of a plurality of data items, and each data item is composed of a plurality of data items according to a possible range of the data item value. A program for retrieving a file in which a plurality of records are classified according to a given rank by a given rank column, and includes the following steps: (a) The entire record group contained in the file is related to the rank column. When the record group is regarded as the root of the search tree and it is determined that the record group is to be further divided based on the division condition set starting from the record group corresponding to the root of the search tree, For each rank of the highest priority data item that is not yet involved in the split, If the record group is regarded as a leaf of the search tree and is not further divided, the record group and the corresponding rank sequence are sequentially stored in the search target file and the search tree table, respectively.
(B) A search tree index table is created by adding an address of a storage area accommodating a record group corresponding to each rank sequence accommodated in the search tree table, and each record group accommodated in the search target file Is stored in the storage area of the determined address to create an index sequential file,
(C) When a rank sequence is given, the search tree index table is searched for a corresponding address, and a record group corresponding to the specified rank sequence in the index sequential file is accessed.