JPH05143644A - Index part management system - Google Patents

Abstract

PURPOSE:To shorten the wait time due to block exclusive control in circumstances of simultaneous execution of plural processes by increasing the processing speed for deletion of data records. CONSTITUTION:With respect, to a file 14 where one or more key items are provided in a data record and an index of a tree structure consisting of plural index blocks is provided for each key item, an index part non-changing means 12 of a file access control part 6 is controlled to not change key values of index records so that the change of the chain indicating order relations of index blocks is not spread to higher-order index blocks if this chain is changed by deletion of data records.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a file index management system.

[0002]

2. Description of the Related Art In a file holding a large number of data records, a predetermined item in the data record is used as a key item in order to facilitate the retrieval of the data record, and a tree structure is formed based on the key value. Has an index. In general, a data record has a plurality of key items, and an index is provided for each key item.

FIG. 4 shows an example of an index for one key item formed in the index storage area 17. Note that FIG. 4 is also used in the description of the operation of the embodiment of the invention.

In FIG. 4, B1 to B19 are index blocks, and here, the respective index blocks B1 to B1.
9 can store 3 index records. It should be noted that the numbers in the index record indicate the key value, and the index record for which the key value is not indicated is empty. Index blocks B1, B2, B4, B1
The MIN at 0 indicates the minimum value. Further, in the index block, index records are arranged in order of key value.

A thin line with an arrow indicates a pointer between the upper and lower index blocks, and each index record in the upper index block points to the index record of the minimum value which is the head of the lower index block. ing. In addition,
The index blocks B10 to B19 are the lowest index blocks, and the index records contained in these index blocks B10 to B19 point to data records having corresponding key values.

Now, for example, when it is desired to access the data record having the key value "310", first, the highest index block B1 is referred to, and the target key value "310" is the same or smaller than the target key value "310". The pointer of the index record having the key value “270” is traced as an index record having a close key value, and the index block B3 is referred to. Similarly, the index blocks B7 and B14 are referred to, the index record of the key value "310" is referred to, and a desired data record is accessed from the pointer.

When a data record is deleted, added, or updated, the index is updated or reorganized to reflect the state in the index.

[0008]

As described above, in the conventional file having the index, when the data record is deleted, the index is updated / reorganized each time. Due to this, the processing speed becomes slow, the efficiency of individual processing is inferior, and in an environment where multiple processes are executing simultaneously, other processes trying to access the same file wait for a long time due to block exclusive control. There was an inconvenience that

That is, when a data record is deleted and the index record corresponding to the data record has the minimum value in the index block, the index record of the upper index block is also affected. Further, if the index record is the minimum value in the index block, it is transmitted to a higher rank. Since the index unit is generally provided together with the data unit on the auxiliary storage device having a slow access speed, the processing speed increases significantly due to the increase in the number of input / output times.

For example, in FIG. 4, the key value "270"
Index block B1 when the data record of
It becomes necessary to update the index blocks from the lowest rank to the highest rank, such as 3, B7, B3, B1. Such a situation repeatedly occurs when a large number of data records are deleted in ascending order of the key value, and the processing time required for index part management cannot be ignored.

Further, when the number of data records in the file is large or when the key values do not overlap so much, the index block becomes multistage and the processing time further increases due to the increase in the number of input / output. ..

The present invention has been proposed in view of the above points, and an object of the present invention is to improve the processing speed by not updating the index part as much as possible when deleting a data record. It is to provide an index part management method that can do this.

[0013]

In order to achieve the above object, the present invention has a tree structure index having one or more key items in a data record, each index item having a plurality of index blocks. In the index management method for files with files, if the deletion of data records causes a change in the chain that indicates the vertical relationship of index blocks, the key value of the index record will be prevented so that the chain change will not be transmitted to the upper index blocks. The indexing section non-changing means for controlling so as not to change is provided.

[0014]

According to the index management system of the present invention, a file having one or more key items in a data record and having a tree-structured index composed of a plurality of index blocks for each key item is provided. In the above, when the data record deletion causes a change in the chain that indicates the vertical relationship of the index blocks, the index part no change means changes the key value of the index record so that the change in the chain is not transmitted to the upper index blocks. Control not to.

[0015]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the index management system of the present invention.

In FIG. 1, the functional units of this embodiment are a data unit updating unit 7, an index unit updating unit 8 and an index unit searching unit 9 in a file access control unit 6 realized as a process on the main memory 1. They are a comparison search means 10, an index part reorganization means 11, and an index part no-change means 12.

A file 14 is stored in the auxiliary storage device 13.
Is stored, and a data section 15 and an index section 16 are provided in the file 14.

Then, the file access control section 6 makes a predetermined request to the data section 15 and the index section 16 of the file 14 in response to a data record read request, an add request, a delete request, or a rewrite request made from the processes 2 to 5. Perform processing.

FIG. 2 shows the structure of the data part 15 and the index part 16 of the file 14. The data part 15 stores a plurality of data records R, and some items in the data record R are indexes. Is specified as a key item for.

The index section 16 is provided with an index section storage area 17 for each key item, and each index section storage area 17 is provided.
In, an index having a tree structure of a plurality of index blocks B is formed.

The operation of the above embodiment will be described below in different cases.

(1) At the time of deleting a data record In FIG. 1, the data record is deleted by the data part updating means 7 of the file access control part 6 in response to a deletion request from the process 4, and the change of the data part 15 is indexed. Part 16
In order to match the contents of the index part updating means 8, the index part updating means 8, the index part searching means 9, the comparison searching means 10, the index part reorganizing means 11,
The index part unchanging means 12 is appropriately used to update / reorganize the index. Below, the index part storage area 17 of the index part 16
The processing will be specifically described by focusing on the change of the internal state.

First, assume that the state of the index in the index storage area 17 at a certain point of time is as shown in FIG.

In FIG. 4, B1 to B19 are index blocks, and here, respective index blocks B1 to B1.
9 can store 3 index records. It should be noted that the numbers in the index record indicate the key value, and the index record for which the key value is not indicated is empty. Index blocks B1, B2, B4, B1
The MIN at 0 indicates the minimum value. In addition, index records are arranged in the order of key values in the index block.

A thin line with an arrow indicates a pointer between the upper and lower index blocks, and each index record in the upper index block points to the smallest index record at the head of the lower index block. ing. In addition,
The index blocks B10 to B19 are the lowest index blocks, and the index records contained in these index blocks B10 to B19 point to data records having corresponding key values.

Now, in the state of FIG. 4, the index block B
If the data record (not shown) corresponding to the key value “270” of No. 13 is deleted, the index record of the key value “270” of the index block B13 is deleted or invalidated. As a result, the substantial minimum key value of the index block B13 changes to "280".

Conventionally, since the minimum key value of the index block B13 has changed, the higher index block B7
Although the key value of the index record of the key value "270" of the above is also updated, in the present invention, the minimum key value of the upper index block B7 is set so that the update is not chained to the upper index block. The index record of "270" is not updated.
This control is performed by the index section non-changing means 12.

Then, the key value "2" of the index block B13
If the data record corresponding to "80" is deleted, only the index record having the key value "280" of the index block B13 is deleted.

Then, the key value "2" of the index block B13
If the data record corresponding to "90" is deleted, the index record of the key value "290" of the index block B13 is deleted. As a result, the index block B13 has no valid index record, and the index block B13 itself is deleted. Also in this case, the minimum key value "270" of the upper index block B7 is not changed.

However, if this condition is left as it is, the pointer of the index record of the key value "270" of the index block B7 causes inconvenience. Therefore, the pointer is changed so as to point to the subsequent index block B14, and the index block B14 is pointed until then. Index block B7 key value "300"
Delete the index record of.

Next, the key value "3" of the index block B14
If the data records corresponding to "00", "310", and "320" are deleted, the processing is the same as that described above, but since the index block B14 itself is deleted, there is nothing to point to in the index block B7. The index block B7 itself is also deleted. Further, along with the deletion of the index block B7, the pointer of the index block B3 is changed and the index record is deleted.

FIG. 5 shows this state. The hatched index blocks B7, B13, B14 and the index record of the index block B3 indicate that they have been deleted.

(2) At the time of adding a data record In FIG. 1, the addition of a data record is performed by the data section updating means 7 of the file access control section 6 in response to an addition request from the process 3, and the change of the data section 15 is indexed. Part 16
In order to match the contents of the above, the index part updating means 8 appropriately updates and reorganizes the index by using the index part searching means 9, the comparison searching means 10, and the index part reorganizing means 11. Hereinafter, the processing will be described focusing on the change in the state in the index storage area 17 of the index unit 16.

First, assume that the state of the index in the index storage area 17 at a certain point of time is as shown in FIG.

Now, for example, if a data record with the key value "355" is added, it is necessary to add a corresponding new index record, but since there is no free area in the index block B15, the index block is divided. Perform processing.

FIG. 3 shows a flow chart of the index block division processing. The above example will be described with reference to FIG.

First, the division point is obtained and the index block is divided (step S1). In this example, index block B1
The index record of the key value "360" of 5 and the key value of "37"
It is assumed that the data is divided into the index record of "0" and the key value "3
The index record of "60" is moved to the index block B before the division.
Index records of key values “370” and “380” are left in 15. At this point, the index record corresponding to the new data record is not added to the newly generated index block B20, but is added after the division process is completed.

Next, an index record pointing to itself of the index block above the index block before division is obtained (step S2). In the present example, the index record of the key value "360" of the index block B8 is specified.

Next, it is judged whether or not the upper index record of the index block before division is the minimum value in the block (step S3). In the present example, it is determined that the index record of the key value "360" of the index block B8 is the minimum value in the block.

Next, when it is determined that the value is the minimum value in the block, the key value of the upper index record is compared with the minimum value of the index block before the division at that time (step S4). .. In this example, the key value is "360"
And the key value “370” are compared, and it is determined that the upper index record side is smaller.

Next, when it is judged that the upper index record side is smaller, the key value of the upper index record is changed to the minimum value at that time of the index block before division (step S5). In the present example, the key value “360” of the index block B8 is changed to “370”.

Then, an index record pointing to the new index block after division is added to the upper index block (step S6). In this example, index block B8
Key value "3" that points to index block B20 at the beginning of the
60 ”index record is added.

Then, a new index block B2 after division
The key value “355” of the added data record is added to 0.

FIG. 6 shows this state.

In FIG. 3, steps S3 and S4 are performed.
When branching to the end side in, the processing is the same as normal addition processing, the upper index record of the index block before division is not changed, and the index record pointing to the newly generated index block is added. It will be.

(3) At the time of retrieving a data record In FIG. 1, the data record is retrieved in response to a read request from the process 2 (a search is required for addition, deletion, rewriting, etc.). The index section search means 9 and the comparison search means 10 of FIG. Hereinafter, the processing will be described focusing on the change in the state in the index storage area 17 of the index unit 16.

First, assume that the state of the index in the index storage area 17 at a certain point of time is as shown in FIG.

Now, for example, when searching the index to obtain the data record of the key value "355", the highest index block B1 is followed by the index blocks B3 and B8, but the minimum value of the index block B8. Is "360", which is larger than the target key value "355", so that a search according to the principle causes a dead end.

Therefore, in this case, the current index record which is the object of judgment at the time of the dead end is traced. More generally speaking, the principle search rule does not find the index record with the target key value The current index record is the smallest value in the index block The key value of the current index record is the target key value If the three conditions that are larger than are satisfied, the current index record is traced.

In the example of FIG. 6, since the index record of the key value "360" of the index block B8 becomes the current index record, the index block B20 is traced to refer to the index record of the key value "355". The record can be found and the data record can be accessed from the pointer.

[0052]

As described above, the index part management system of the present invention has the following effects.

The deletion of the data record causes the upper index block to be updated in the conventional case, but the key value of the index record is not changed so as not to spread to the upper position as much as possible. Since it is controlled as described above, the number of times of input / output with the index unit on the auxiliary storage device can be significantly reduced, and the time required for processing can be shortened.

In an environment in which a plurality of processes are simultaneously executed, it is possible to increase the efficiency of the system by reducing the number of cases where the access to the same file is made to wait by block exclusive control.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an embodiment of an index management system of the present invention.

FIG. 2 is a detailed view of the inside of the auxiliary storage device in FIG.

FIG. 3 is a flowchart showing an index block dividing process.

FIG. 4 is a diagram showing an example of an index composed of index blocks.

FIG. 5 is a diagram showing an example of an index composed of index blocks.

FIG. 6 is a diagram showing an example of an index composed of index blocks.

[Explanation of symbols]

 1 ………………………… Main memory 2-5 ……………… Process 6 …………………… File access controller 7 …………………… Data updater 8 ... ……………… Index section update means 9 …………………… Index section search means 10 ………………… Comparison search means 11 ………………… Index section reorganization means 12…. ……………… Index section unchangeable means 13 …………………… Auxiliary storage device 14 …………………… File 15 ………………… Data section 16 …………………… Index Part 17 …………………… Index storage area R …………………… Data record B, B1 to B20… Index block

Claims (1)

[Claims] 1. An index management method for a file having one or more key items in a data record, wherein each key item is provided with a tree-structured index composed of a plurality of index blocks. When a chain that indicates the hierarchical relationship of index blocks changes due to deletion, a non-change part of the index part that controls not to change the key value of the index record so that the change of the chain is not transmitted to the upper index block. Index part management method characterized by having.