JP2000076106A - Management method for indexed sequential file - Google Patents

Abstract

PROBLEM TO BE SOLVED: To shorten time required for the updating and elimination processing of plural indexed sequential files and to improve throughput. SOLUTION: An index read processing means 5 reads an index block, traces the lowest order level entry and checks whether or not the index block is extended. When it is extended, it is judged that an index tree structure is formed based on a record address and read is performed. Also, for a non-extended index block, it is judged as an index tree based on a normal key and a conventional read processing is performed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to management of an index sequential file in an electronic computer system, and more particularly to a method of managing a plurality of index sequential files having a plurality of indexes.

[0002]

2. Description of the Related Art As shown in FIG. 2, an indexed sequential file has a data section and an index section so that a record can be randomly accessed using a specific item in a record as a key. Records are continuously stored in the data section. The index section stores, for each record, a key of the record and a pointer indicating where the corresponding record is stored in the data section. The combination of the key and the record pointer is called an entry.
FIG. 3 shows the configuration of the entry. By referring to this index part, it is possible to directly access a record by specifying a key.

Each entry in the index part has an index tree structure in units of index blocks as shown in FIG. 4, thereby speeding up the index search. The index block is made up of a plurality of entries as shown in FIG. At the beginning of the index block,
A level S (Status) indicating whether the block is the lowest index block or the highest index block is provided. Further, a pointer P.3 pointing to an index block having the next lowest key value at the same level as this index block. P (P
rior-Pointer) and a pointer N.I. pointing to the index block with the next largest key value. P (Next-P
pointer) is set.

A pointer of a corresponding record is set as a pointer in an entry constituting an index block of the lowest level, and a record pointer in an entry of an index block other than the lowest level has a lower level than the index block. The address indicating the index block is set, and the key is set to the same content as the largest key in the index block lower than the index block.

On the other hand, as shown in FIG. 6, a multiple index sequential organization file is configured such that an index part and a data part are divided into individual files and a plurality of indexes can be provided.

The index portion of the multiple index sequential organization file is shown in FIG.
As shown in (1), the index tree structure is based on index blocks as in the index portion of the index sequential organization file. The point that the corresponding record address of the data section is set in the pointer set in the entry of the lowest-level index block differs from the index sequential file. Also in this multiple index sequential organization file, the key search of the index part is speeded up by the index tree structure.

[0007] The index sequential file has a format in which a record pointer is added to a record having the same key in the data section as shown in FIG. 8 when the key duplication is valid. A chain is formed by records having the same key in the data section. When updating a record, the efficiency of the process is improved by searching the key in the index part and tracing the record chain for the corresponding key in the data part.

[0008] A multiple index sequential file can have a plurality of indexes, but cannot have a configuration of a record chain for a single record unlike the above-mentioned index sequential file. When the duplicate key is valid in the organization file, as shown in FIG. 6, the key of each index part has a record pointer of the direct data part.

[0009]

In a multiple index sequential file, if duplicate keys are valid and there are many identical keys, a record chain cannot be created, and the time required for updating and deleting records is significantly increased.

Further, when processing an entry having a large number of identical keys, an updated entry cannot be searched only by the key.
Further, since it is necessary to check each entry from the address of the update record, the index block must be read out at any time before the entry is found,
This also contributes to an increase in processing time.

An object of the present invention is to solve such a problem, shorten the time required for updating and deleting a plurality of index sequential files, and improve the throughput.

[0012]

According to the present invention, there is provided a method for managing an indexed sequential file, wherein an index portion and a data portion are divided into individual files, and when there is no area for storing an entry in the data portion, the entry is extended. The index section is stored in a data section, and the index section forms an index tree in which a plurality of entries are linked to each other in a logical hierarchy. When a new index is additionally created in the index portion in the multiple index sequential organization file searched with reference to, the keys of a plurality of entries in the index block in which the index is created are compared, and if all the keys are the same, For example, the index tree of the hierarchy below the index block is converted from the key base expressed by the key correlation to the record address of the data section. And converting the index tree to a record address-based index tree represented by the correlation of: and determining whether the index tree is key-based or record address-based. In the process of directly reading the entry of the data section indicated by the key and simultaneously updating / deleting the relevant key at the time of entry update / deletion, if the judgment result of the index tree is a record address base, the relevant key is searched from the record address Update / deletion processing.

[0013]

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

FIG. 1 is a conceptual diagram showing the configuration of the present invention, wherein 1 is a management program, 2 is an operating system incorporated in an electronic computer system, and 3 is a storage device. The operating system 2 has an index addition processing unit 4, an index read processing unit 5, an index update processing unit 6, and an index deletion processing unit that manage the index sequential organization file management of the present invention.

Next, the processing of the embodiment will be described.

When all the keys in the index block become identical by adding an entry in the index block of a certain level, the index adding processing means 4 starts with the index block as a base point and sets the lower level index. The index tree structure composed of blocks is changed to an index tree structure based on the record address set in the pointer in the lowest-level entry.

When making this change, in order to distinguish between those based on a normal key and those based on a record address, the expanded index block after the change is expanded to its level area. A mark is set to indicate that

In the index read processing, the index read processing means 5 reads the index block and traces the lowest level entry. At this time, it is checked whether or not the index block has been expanded based on the presence or absence of a mark in the level area. In the case of an extended one, it is determined that an index tree structure is formed based on the record address, and reading is performed. For an index block that has not been extended, it is determined that the index tree structure is based on a normal key, and a conventional read process is performed.

When the corresponding entry of the index part is updated at the same time as the record update processing, the index update processing means 6 checks whether or not the index block is expanded, as in the case of the index read processing means 5. If it is expanded, it is determined that the index tree structure is configured based on the record address, the relevant entry is searched, the key of the relevant entry is changed to a new one, and then the index is moved to the index block indicated by the index part Let it. Then delete the original entry.

When the corresponding entry of the index part is deleted simultaneously with the record deletion processing, the index deletion processing means 7 performs an extension check of the index block similarly to the index reading processing means 5, and determines which index tree structure is used. Then, the corresponding entry is searched for, and this entry is deleted.

By performing these processes, when a large number of entries having the same key occur in the index section, a part of the index tree structure in the index section including the entry is expanded to the record address base. Correspondence "can be performed. As a result, the entry search processing for the changed / deleted record can be performed not based on the same key but based on the record pointer to be searched, thereby reducing unnecessary access to the storage device. As a result, the search time can be reduced.

Next, the operation of the index block expansion processing of the present invention will be described with reference to a diagram showing a more specific situation.

In one index block at the lowest level,
One entry for key “G” is added (FIG. 10
(A)). At this time, the lowest level index block in which the keys are all the same "G", starting from the corresponding entry of the next higher intermediate level index block, is configured (FIG. 10B).

When the keys of the index block at the lowest level are all the same, the entries are rearranged in ascending order by the record address of the data section set in each pointer. A mark indicating that the index tree structure has been expanded to the record address base is set (FIG. 10C).

Next, processing in an intermediate level index block will be described. As shown in FIG. 11A, when one entry of the key “G” is added to the intermediate-level index block, an intermediate-level index block in which all keys have the same “G” is configured. At this time, the key contents of the next lower index block (in this figure, the lowest index block) connected to the intermediate index block are all "G".

Thereafter, as shown in FIG. 11 (c), the entries of the respective lower index blocks to be linked are rearranged in ascending order by the record address set in each pointer, and all the intermediate keys having keys "G" are rearranged. The key content of the index block is
Set the value of the largest record address in the next lower index block. In addition, a mark is set in the level area of the index block, indicating that the index tree structure has been expanded to the record address base.

As described above, in each intermediate level index block, when the keys of the entries in the index block are all the same, the same processing is performed, whereby the index tree structure in which the keys are based on the record address is obtained. Can be changed.

In order to identify which key of the extended intermediate-level index block has an extended index tree structure, the contents of the key of the upper entry serving as the base point of the tree are required. Therefore, if an extension mark is set in the level area of the lower-level index block, the contents of this key are retained, and a search is performed for the entry in the lowest level with the record address set there. .

When the contents of the keys set in the keys of the intermediate level index block are all the same, the key area must be at least larger than the record address in order to set the record address of the data section to the key. Must have a head. That is, if the record address is represented by 4 bytes, it is necessary to secure at least 4 bytes or more in the key area even if the data length of the initially stored key is 1 byte.

By performing the above processing, the record address of the data part to be processed can be grasped beforehand when deleting or updating a record. When searching for the lowest entry of a duplicated key, first search by key, and once you reach the record address based index tree structure, switch to search by record address from there, Can be searched for.

[0031]

As described above, by using the record address formed in the index tree structure as a search key, all the entries from the first entry are used as in the conventional method of managing the index sequential file. There is no need to search for entries, and the number of times of reading unnecessary index blocks is reduced. Inevitably, the time required for updating and deleting records of the entire file is reduced, and the throughput is improved.

This is particularly effective in the case of a multi-index sequential file in a situation where there are many index parts, duplicate keys are valid, and there are many entries with the same contents.

[Brief description of the drawings]

FIG. 1 is a system configuration diagram of a method for managing an indexed sequential file according to the present invention.

FIG. 2 is a diagram showing a structure of a conventional index sequential organization file.

FIG. 3 is a diagram showing a configuration of an entry in an index sequential organization file.

FIG. 4 is a diagram showing a relationship between a logical configuration in an index portion and a data portion of a conventional index sequential organization file.

FIG. 5 is a diagram showing a configuration of an index block.

FIG. 6 is a diagram showing a structure of a multiple index sequential organization file of the present invention.

FIG. 7 is a diagram showing an index part of a multiple index sequential organization file and a logical configuration in the index part.

FIG. 8 is a diagram illustrating a data file search method in a conventional index sequential organization file when a duplicate key is valid.

FIG. 9 is a diagram showing a method of searching for a data file when a duplicate key is valid in a multiple index sequential organization file of the present invention.

FIG. 10 is a diagram showing processing when an entry is added to the lowest-level index block in the present invention;
(1) is a state before the entry is added, (2) is a state after the entry is added, and (3) is a state after the entries are rearranged.

FIG. 11 is a diagram showing processing when an entry is added to an intermediate-level index block in the present invention;
(1) is a state before an entry is added, and (2) is a state after an entry is added.

FIG. 12 is a diagram illustrating a process when an entry is added to an intermediate-level index block according to the present invention, in a state after rearranging the entries.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Management program 2 Operating system 3 Storage device 4 Index addition processing means 5 Index reading processing means 6 Index update processing means 7 Index deletion processing means

Claims (1)

[Claims] 1. An index section and a data section are divided into individual files, and when there is no area for storing an entry in the data section, the entry is stored in an extension data section, and the index section includes a plurality of entries. Form an index tree that links the linked index blocks in a logical hierarchy,
When reading out the entry, in a multiple index sequential file searched by referring to an index tree formed in the index part, when newly creating an index in the index part, the index block in which the index is created The keys of a plurality of entries are compared, and if the keys are all the same, the index tree of the hierarchy below the index block is represented by the correlation of the record address of the data section from the key base represented by the correlation of the key. Processing to change to the index tree based on the record address based and the index tree determines whether the record is based on a key or a record address. The process of reading the entry directly and updating the corresponding key simultaneously when updating / deleting the entry /
When deleting, if the result of the index tree determination is based on a record address, the method comprises a process of searching for a corresponding key from the record address and updating / deleting the same, characterized by that .