JPH0518148B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a system for simultaneously generating indexes using the values of the respective fields as keys for a plurality of fields of a data record.

[Conventional technology]

Conventionally, in order to generate an index for one field, the index generation process was performed by reading all data records on a data file once and sorting them in the order of the values of the field that is the key of the index. Even when creating indexes for fields at the same time, the above process is repeated multiple times.

[Problem that the invention seeks to solve]

In the conventional index generation method described above, it is necessary to repeatedly read data records from a data file for the number of indexes to be generated at the same time, and when a large number of indexes are to be generated at the same time, it is necessary to read a very large amount of data records. .

The time required to read data records plays a large role in index generation, and repeatedly reading the same data record several times to generate indexes at the same time greatly affects the efficiency of index generation. There is a problem with high speed.

Therefore, the present invention provides a multiple index generation method for the purpose of generating indexes at high speed by reading the data file only once when multiple indexes are generated for the same data file at the same time. It is.

[Means for solving problems]

The multiple index generation method of the present invention includes means for reading data records one by one from a data file, extracting a plurality of fields for generating indexes from the data records, and adding a field identifier and a field identifier to the value of each field on the data file. means for generating a sort record by adding an address to the field; a first storage means for storing the sort record; a means for generating field attribute information for a plurality of fields for which an index is to be generated; and a second storage means for storing the sort record stored in the first storage means, the field identifier being the first key, and the sort record stored in the first storage means being the field identifier as the first key. It is constructed to include a sorting means for sorting values using a second key, and a means for reading a sorted record from the first storage means and generating an index.

〔Example〕

Next, an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention, in which a data record reading means 2 reads a data record from a data file 1, and a field identifier and an address on the data file are assigned to the value of the field that is the key of the index. Sort record generation means 3 that generates sort records by adding
and field attribute information generation means 4 that generates attribute information of a field that is a key of the index.
, a sorting means 6 for sorting according to the field attribute information with the field identifier as the first key and the field value as the second key, an index generating means 8 for generating an index for reading sorted records after sorting, and a field attribute information storage. It is composed of means 5, sort record storage means 7, and a generated index file 9.

The sort record storage means 7 is used for sorting sort records generated from data records, and the information stored in the field attribute information storage means 5 is necessary for sorting the sort records. Also index file 9
is a file that stores an index generated from data records.

As shown in FIG. 2, the data record format of this embodiment consists of four fields, A10, B11, C12, and D13.
An attempt is made to simultaneously generate indexes for three fields: 2 and D13.

The data record of this example is shown in FIG. The data records consist of 10 items. The suffix for each field indicates the order of the list.

As shown in FIG. 4, the sort record format consists of a field identifier 14 corresponding to each field, a field value 15, and an address 16 indicating the position of the data record on the data file 1.

As attribute information for each field, a correspondence table is created using pairs of field identifiers 17 and field attributes 18 as shown in FIG.

The procedure for generating an index using this method is as follows.

(1) The data record reading means 2 reads data records one by one from the data record group shown in FIG. In the example, fields A:01, C:02, D:
03), and using that number as the field identifier 14, perform sorting using the sort record format shown in Figure 4 from the field value 15 of the field to be indexed and the address 16 indicating the position of the data record on the data file 1. By performing this processing on all data records to be generated and stored in the sort record storage means 7, a sort record group shown in FIG. 7 is created on the sort record storage means 7.

(2) The correspondence table between the field identifier 17 and the field attribute 18 in FIG. : "C" (letter), C: "U" (numeric value), D: "N" (Japanese) (as shown in FIG. 6) are created on the field attribute information storage means 5.

(3) For the sort record group on the sort record storage means 7 generated by the sort record generation means 3, the sort means 6 stores field attribute information with the field identifier 14 as the first key and the field value 15 as the second key. Sort according to field attributes 18 on means 5. The 8th one is sorted using field identifier 14 as a key.
In addition, FIG. 9 shows the field values 15 rearranged according to the field attributes 18.

(4) The sorted records on the sorted record storage means 7 that have been rearranged by the sorting means 6 are read in sequence, and the index generating means 8 generates an index on the index file 9. The index is generated on the same index file 9 while the field identifier 14 is the same by reading the sort record.

Field identifier: 01 index is the first
0, the index for field identifier: 02 is generated as shown in FIG. 11, and the index for field identifier: 03 is generated as shown in FIG. 12.

〔Effect of the invention〕

The present invention enables data records in a data file to be sorted by adding field identifiers corresponding to fields to the sort records to be sorted in the process of sorting keys, and by creating a correspondence table between field identifiers and field attributes. The number of reads can be reduced to one, which has the effect of improving performance when multiple indexes are generated simultaneously.

FIG. 13 shows the time required to read a data record when the number of data records is n, the number of indexes to be generated at the same time is m, and the time to read all data records once is t1.

The required reading time in the conventional index generation method is as shown by the straight line 19,
Since all data records are read repeatedly for the number of indexes to be generated at the same time, the time required to read the data records is mt1.

In this method, as shown by the straight line 20, all data records are read once regardless of the number of indexes generated at the same time, and the time required to read the data records is t1.

Next, FIG. 14 shows the time required to rearrange the sorted records, where t2 is the time required to rearrange the n sorted records.

The required sorting time in the conventional method is a straight line 2
As shown in 1, it takes mt2 time to sort n sorted records the number of times the indexes are generated at the same time.

This method is as shown in the straight line 22, mn
The sorted records will be sorted using two keys. In other words, the time required for sorting by the field value, which is the second key, is approximately the same as the conventional method, mt2, and the time required for sorting by the field identifier, which is the first key, is longer than in the conventional method. Let this increment be t3.

From the above, the time required to generate an index is Conventional method: mt1+mt2 This method: t1+mt2+t3 becomes.

Generally, the time required to read data records occupies a larger proportion of the time required for index generation than the time required for sorting, and t1>>t3. Therefore, the efficiency of this method can be expected to improve as the number of indexes that are generated simultaneously increases.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the multiple index generation method of the present invention, and FIGS. 2 to 14 are diagrams for explaining the embodiment, of which FIG. Figure 3 shows the data record format, Figure 4 shows the sort record format, Figure 5 shows the correspondence table between identifiers and attributes,
FIG. 6 is a correspondence table of identifiers and attributes in the embodiment,
FIG. 7 shows a group of sort records in the embodiment;
The figure shows a sorted record group after sorting by the first key, Fig. 9 shows a sorted record group after sorting by the second key, Fig. 10 shows the index for field A, Fig. 11 shows the index for field C, and Fig. 12 shows the sorted record group after sorting by the second key. The figure shows an index for field D, FIG. 13 shows a comparison of reading times, and FIG. 14 shows a comparison of sorting times. Explanation of symbols: 1 is a data file, 2 is a data record reading means, 3 is a sort record generation means, 4 is a field attribute information generation means, 5 is a field attribute information storage means, 6 is a sorting means, 7
8 represents a sort record storage means, 8 represents an index generation means, 9 represents an index file, 19 and 20 represent the required reading time, and 21
and 22 represent the required sorting time, respectively.

Claims (1)

[Scope of Claims] 1. Means for reading data records one by one from a data file, extracting a plurality of fields for generating an index from the data record, and assigning a field identifier and an address on the data file to the value of each field. a first storage means for storing the sort record; a means for generating field attribute information for a plurality of fields for which an index is to be generated; and a means for storing the field attribute information. a second storage means for storing the sort records stored in the first storage means, and storing field values using field identifiers as first keys according to field attribute information stored in the second storage means; A method for generating multiple indexes, comprising a sorting means for sorting as a second key, and a means for reading a sorted record after sorting from the first storage means and generating an index.