JPS62287350A - Index integrally updating system - Google Patents

Abstract

PURPOSE:To decrease a block read request and a write request by holding the updating information of an index accompanied to the updating of a data record, as a sort record, in case of updating the data record, and updating an index file in the order of a key value integrally after updating the data record. CONSTITUTION:A sort record group on an index record updating information storage means 5 rearranged by a sorting means 4 is read successively by an index record reading means 6, and an index on an index file 8 is updated by an index updating means 7. The index updating means 7 updates the index record existing on the index file 8 in accordance with an updating processing identifier 65, while an index identifier 61 has the same value.

Description

Detailed Description of the Invention 3. Detailed Description of the Invention (Field of Industrial Application) The present invention relates to a data file update method provided with access means using an index.

(Prior Art) Conventionally, in a data file equipped with an index file for improving data file search efficiency, 1
When updating a data record, if each field of the data record is updated, the associated index file is also updated at the same time.

Assuming an index file where several index records are grouped together into one block and input/output processing is performed in units of blocks, updating of unit index records can be explained using Figure 9 as shown in the following five steps. It becomes like this.

In the first step, the upper level block 91 is read.

In the second step, if the update process of the index record to be updated is "delete" or "change", the management record 92 of the lower level block containing the key value is found; on the other hand, if the update process is "add" For example, it finds the management record 92 of the lower level block that stores the index record.

In the third step, the lower level block management record 9
Read the lowest level block 93 indicated by 2. In the fourth step, if the update process is "delete", the index record 9G of the lowest level block 93 is deleted, and if the update process is "change", the lowest level block 93 is read. The index record 94 of the level block 93 is changed, and if the update process is "add", the index record 95 is added to the lowest level block 93.

In the fifth step, the lowest level block 93 is written.

In updating the index record, two block read requests and one block write request occur.

This method of updating index records does not utilize the property that index records stored in an index file are arranged in the order of key values in order to maintain index records in response to data records outside the update layer. .

(Problem to be Solved by the Invention) In the conventional index update method, when only the field code of a data record is updated, even if there is only one index that corresponds to the updated field, it is necessary to update the data record twice. A block read request and one block write request occur.

If the index record to be updated first is stored in the currently processed block, or is guaranteed to be stored in the currently processed block, then in updating the index record to be updated next, the above index record will be updated. Of the methods, only the fourth step is required.

In this case, neither a block read request nor a block write request occurs.

An object of the present invention is to eliminate the above-mentioned drawbacks by retaining the accompanying index update information as a sort record when updating a data record, and updating the index file in the order of key values all at once after updating the data record. An object of the present invention is to provide an index batch update method configured to reduce block read requests and write requests.

(Means for Solving the Problems) The index batch update method according to the present invention includes a data record update processing means, an index record update information generation means, a sort record storage means, a sorting means, and an index record update information reading means. , and index updating means.

The data record update processing means is for updating data records on a data file having one or more indexes.

The index record update information generation means uses the updated field value of the updated data record as a key in order to update one or more indexes for at least a plurality of data records updated by the data record update processing means. For an index record whose pointer value is the address on the data file of the updated data record, the index identifier corresponding to each field of the data record, the update order of the updated data record, and the index record's address. This is for generating index record update information to which an update method identifier is added.

The sort record storage means is for storing the index record update information generated by the index record update means as a sort record.

The sorting means uses the index identifier added by the index record update information generation means as a first key and the key value of the index record as a second key for the sort records stored in the sort record storage means. This is used to rearrange the update order of data records added and updated by the index record update information generating means as a third key.

How to read index record update information.

This is for reading the sorted records rearranged by the sorting means from the sorting record storage means and fetching index record update information.

The index update means is for collectively updating the index files indicated by the index identifiers based on the index record update information read by the index record update information reading means.

(Example) Next, an example of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the index batch update method according to the present invention.

In FIG. 1, 1 is a data file, 2 is a data record update processing means, 3 is an index record update information generation means, 4 is a sorting means, 5 is a sort record storage means, 6 is an index record update information reading means, and 7
8 is an index update means, and 8 is an index file.

FIG. 2 is an explanatory diagram showing the data record format in the embodiment shown in FIG.

The data record format of this embodiment is A as shown in Figure 2.
It consists of four fields: field 10, B field 11, C field 12, and D field 13, and two fields, B field 11 and D field 13, each have an index.

FIG. 3 shows a data record group before updating according to this embodiment, and FIG. 4 shows a data record group after updating.

The suffix of each field indicates the order of arrangement.

FIG. 5 shows a data record update process in which errors are added to the data record group of FIG. 3.

Update Experience The first data record update process is to delete the data record at address P1.

Update order The second data record update process is for address P.
Change the value of field B in data record 2 from "B2" to "B5", and change the value of field D from "B2" to "
B5".

Update order No. 3 has the value of field A rA4J,
The value of field B is "B4", the value of field C is "C"
4", and the value of field D is "D4". This is the process of adding a new data record.

FIG. 6 shows the sort record format to be created by the index record update information generating means 3 in conjunction with the data record update process.

An index identifier 61 and a key value 62 corresponding to the field with the index of the data record, a pointer value 63 determined corresponding to the address indicating the storage position of the data record on the data file 1, and the update order of the data record. and an update order 64 for associating the update order of the index record with the update order of the index record, and an update process identifier 65 for associating the update method of the data record with the update method of the index record.

Figure 7: Index record update information generation means 3
This is a group of index record update information generated by. In FIG. 7, 71 to 78 are index record update information, respectively.

Next, referring back to FIG. 1, an overview of each element will be explained.

Data record updating means 2 updates data records on data file 1 having one or more indexes.

The index record update information generation means 3 generates information only when the field serving as the key value of the index of at least a plurality of data records updated by the data record update processing means 2 is updated in accordance with the update of the data records. Set the field value to key value 62,
Generate an index record with the address on the data file 1 of the data record as a pointer 63, and add an index identifier 61 corresponding to each field of the data record to the generated index record,
The update order 64 of the data record, if the update process is to delete a data record, it is a deletion process identifier, if a data record is added, it is an addition process identifier, and if a data record is changed, it is a deletion process identifier and an addition process identifier. The index record update information to which the update processing identifier 65 determined by dividing the index record is added is generated.

The sorting means 4 uses the index identifier 61 as a first key, the key value 62 of the index record as a second key, and the update order 64 of the data record as a third key, which are stored in the sort record storage means 5. Sort the existing sort records.

The sort record storage means 5 is a working storage means for rearranging the sort records generated from the data records based on the index record update information.

The index record update information reading means 6 reads the sorted records from the sort record storage means 5 and takes in the index record update information.

The index update means 7 collectively updates the index records in the index file 8 indicated by the index identifier 61 based on the index record update information read by the index record update information reading means 6 in order of key value according to the index record update information. Update.

The index file 8 is a file that stores an index generated from data records.

The above is an overview of each element shown in FIG.

Assuming that the update process shown in Figure 5 is applied to the data record group shown in Figure 3, the index update procedure according to this method from the first process to the third process is as follows. .

First, generation of index update information will be described.

Update order: The first update process is a deletion process.

The index identifier "index B" added to field B is assumed to be the index identifier 61 of the index record update information.

The value of field B, "B1", is assumed to be the key value 62 of the index record update information.

Let "Pl", which is the address of the data record, be the pointer value 63 of the index record update information.

The update order 64 is set to "1", and the update process identifier 65 is set to "delete".

As a result, index record update information 71 is generated.

This index record update information is stored in the sort record storage means 5 as a sort record.

By applying the same index record update information generating means 3 to index D, new index record update information 72 is generated.

Update order: The second update process is a change process.

"Index B", which is the index identifier added to field B, is assumed to be the index identifier 61 of the index record update information.

"B2", which is the value of field B before the change, is set as the key value 62 of the index record update information.

"B2", which is the address of the data record, is set as the pointer value 63 of the index record update information.

The update order 64 is set to "2", and the update process identifier 65 is set to "delete".

As a result, index record update information 73 is generated.

Next, "index B", which is the index identifier added to field B, is set as the index identifier 61 of the index record update information.

The changed value of field B, "B5", is set as the key [62] of the index record update information.

"B2", which is the address of the data record, is set as the pointer value 63 of the index record update information.

The update order 64 is set to "2", and the update process identifier 65 is set to "add".

As a result, index record update information 74 is generated.

By applying the same index record update information generating means 3 to index D, new index record update information 75 and 76 are generated.

Update order: The third update process is an additional process.

"Index B", which is the index identifier added to field B, is assumed to be the index identifier 61 of the index record update information.

"B", which is the value of field B of the data record to be added.
4'' is the key value 62 of the index record update information.

"B4", which is the address of the data record, is set as the pointer value 63 of the index record update information.

The update order 64 is set to "3" and the update process identifier 65 is set to "add".

As a result, index record update information 77 is generated.

By applying the same index record update information generating means 3 to index D, new index record update information 78 is generated.

Second, sorting of index record update information will be described.

The sorting means 4 updates the sorted records on the index record update information storage means 5 generated by the index record update information generation means 3 with the index identifier 61 as the first key and the key value 62 as the second key. Sort with rank 64 as the third key.

The state after the rearrangement is completed is shown in FIG. 8(a).

Third, updating of the index file will be described.

The sorted records on the index record update information storage means 5 that have been sorted by the sorting means 4 are sequentially read by the index record reading means 6, and the index on the index file 8 is updated by the index updating means 7.

The index update means 7 updates index records existing on the same index file 8 according to the update processing identifier 65 while the index identifier 61 has the same value.

Within the same index file, when a key value on index record update information changes, it is determined whether the new key value can be processed in the block currently being processed.

If the new key value on the index record update information is smaller than the maximum key value of the lowest block currently being processed, it is possible to process the block currently being processed.

If processing is not possible, the block currently being processed is eliminated and a new block to be processed is determined by searching the index file starting from the upper blocks.

FIG. 8(b) shows the index B before updating, and FIG. 8<c) shows the index B after updating according to the procedure from the first processing description to the third processing description.
It shows.

(Effects of the Invention) The present invention replaces the index update process associated with data record update with the creation of a sort record to which an index identifier, an update method identifier, and an update order identifier are added in accordance with the field. Since the index update information is sorted in the order of key values when the index is updated at once, it is possible to reduce the number of read blocks and the number of write blocks.

The effects of the present invention can be summarized as follows according to a mathematical formula.

Let T be the sum of the number of block read requests when searching an index from the highest block to the lowest block and the number of block write requests when writing the lowest block.

In the conventional method, updating one index record always generates T block read requests and T write requests, so if Nu is the number of all updated index records, the total number of reads/writes is TXNu. The number of blocks is obtained.

According to the method of the present invention, T occurs only as many times as the lowest index block, so the sum of the number of read blocks and the number of write blocks is the number of index records stored in one block. Let f be
Let N be the total number of index records.

+ It is at most Tx (N/f). Here, ten [ ] represents rounding up.

For example, assuming that the total number of index records is 1 million, the number of index records stored in one block is 100, and that 100,000 index records, which is 10% of the total, have been updated, the read block The ratio of the sum of the number of blocks and the number of write blocks is + Nu: (N/f) = 100000: 10 between the conventional method and the method of the present invention.
000=10:1.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the index batch update method according to the present invention. FIG. 2 is an explanatory diagram showing the data record format in the embodiment shown in FIG. FIG. 3 is an explanatory diagram showing the data record format before update. FIG. 4 is an explanatory diagram showing the updated data record format. FIG. 5 is an explanatory diagram showing updates made to the data file. FIG. 6 is an explanatory diagram showing the sort record format. FIG. 7 is an explanatory diagram showing a sort record group in the embodiment shown in FIG. FIG. 8(a) is an explanatory diagram showing the sorted record group after sorting, and FIG. 8(b) is an explanatory diagram showing the index B before updating.
FIG. 8(C) is an explanatory diagram showing the updated index B. FIG. FIG. 9 is an explanatory diagram showing an example of an index structure. 1...Data file 2...Data record update processing means 3...Index record update information generation means 4...
Sorting means 5... Index record storage means 6... Index record update information reading means 7... Index updating means 8... Index files 10 to 13... Field 61... Index identifier 62...・Key value 63...Pointer value 6
4...Update order 65...Update process identifier

Claims (1)

[Claims] a data record update processing means for updating a data record on a data file having one or more indexes; and one or more indexes for at least the plurality of data records updated by the data record update processing means. In order to update the index record, the updated field value of the updated data record is used as the key value, and the address on the data file of the updated data record is used as the pointer value. Index record update information generation means for generating index record update information to which an index identifier corresponding to a field, an update order of an updated data record, and an index record update method identifier are added; and the index record update means. a sort record storage means for storing the index record update information generated by the index record storage means as a sort record; and an index identifier added by the index record update information generation means to the sort records stored in the sort record storage means. a first key, a key value of the index record as a second key, and a sorting means for sorting the update order of data records added and updated by the index record update information generating means as a third key. and index record update information reading means for reading the sort records rearranged by the sorting means from the sort record storage means and importing index record update information;
An index comprising an index updating means for collectively updating index files indicated by index identifiers based on the index record update information read by the index record update information reading means. Batch update method.