JPH0652231A - Relational data base operation mechanism - Google Patents

Abstract

PURPOSE:To constitute the operation mechanism so that in the case there are plural items in which a range of a value is narrow in a relational data base, a matrix index obtained by integrating them is generated, and retrieval at the time when a condition is designated to these items is executed at a high speed. CONSTITUTION:At the time of defining a table, a value which can be taken by at item in which a range of the value is narrow is registered in a data base definition file 6. At the time of adding and updating data, a matrix index 72 consisting of a matrix table and a data record address list is generated in a data base file 7 by a matrix index generating and updating means 12. At the time of retrieval, in the case a condition is designated to plural items in which the range of the value is narrow, retrieval is executed at a high speed by using the matrix index 72 by a matrix index retrieving means 15.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a relational database operating system for adding / updating data to / from a relational database and retrieving data according to conditions on a computer system.

[0002]

2. Description of the Related Art In a conventional relational database search system, the following method is generally used when a plurality of conditions are specified for an item having a narrow value range.

That is, an index is searched for one of a plurality of conditions to obtain a set of corresponding records, this is performed for all the conditions, and then the set is logically operated. Get the set of records you want.

Alternatively, a virtual item in which a plurality of items are combined is created in advance, and an index is created for that item, so that when all the conditions of a plurality of basic items are specified, the combination is performed. A set of desired records is obtained by searching the index of the created virtual item.

[0005]

In the conventional method, in which the set of records is created by searching under the condition of each item and the logical operation is performed, the search time of the index and the creation time of the set of records are -Overhead such as the time for logical operation of the set is large and the processing time is long.

Further, in the method of creating a virtual item index in which items are combined, the search speed becomes very high when conditions are specified for all corresponding items, but this index is used for individual items. Since it has to be created separately from the index, it takes extra time to create it and consumes that much space in the database. In addition, this index also has to be updated when a record is added or updated, resulting in a large overhead. Also, if there are many combinations of conditions to be used for searching, it will be necessary to create several types of indexes according to the combinations. If you create an index that includes them without doing so, the value will be partially undefined depending on the search conditions, and an extra index search will be required, resulting in slower processing speed. .

It is an object of the present invention to provide a relational database operating mechanism which solves the problems in the prior art as described above and speeds up the search speed.

[0008]

A relational database operation mechanism of the present invention is a database processing system for searching / updating a relational database.
In the case of a table format in which one side is an item of each record and the other side is an array of records, when there are a plurality of items in which the range of values that can be retrieved / updated is narrow, An n-dimensional matrix table in which the number of types of possible values of each item is the number of sides is created in the database, and the address list that points to the corresponding data record is pointed to from each element of the matrix table. The matrix index forming / updating means is provided which can speed up the search speed when the index is created in the matrix table format and a plurality of conditions having a narrow range of values are specified to search the data.

In addition, even if the range of items is not narrow in the above configuration, if there is a bias in appearance and it is possible to classify by some values and other values, a value with a range is also one element of the matrix table. And the element without the range is the same as 1 means,
For elements with a range, point to a tree-like index to search for values for that range from that element, create a matrix table format index, and specify multiple values even if the value range is wide. It has a matrix index creating / updating unit that can speed up the search speed when searching for data.

Further, in the system having the matrix table format index created by the above configuration, when a search request is input from a database user, the values and conditions of each side of the matrix of the matrix table format index created. To determine which pointer row on the matrix should be read, and if multiple pointer rows must be read, read them in parallel to minimize database access during search. It has a matrix index search means.

[0011]

According to the above construction, the matrix index creating / updating means sets the number of types of possible values of each item to 1 when the table to be searched / updated has a plurality of items having a narrow possible range of values. An n-dimensional matrix table having the number of sides is created on a database, and an address list pointing to a corresponding data record is created from each element of the matrix table, and an index in a matrix table format is created.

Alternatively, even if the range of the value of the item is not narrow, if the appearance is biased and can be classified by some value and other value, the value with range is also one element of the matrix table, Create a matrix tabular index that points to a tree-like index to search for values in the range. Further, when a search request is input, the matrix index search means compares the value of each side of the matrix of the matrix table format index with the condition to determine which pointer column on the matrix should be read, and When reading multiple columns, processing is performed so that they are read in parallel at the same time, so the search speed when specifying multiple conditions with a narrow range of values, or when searching for data by specifying multiple conditions even with a wide range of values. Can be speeded up.

[0013]

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

FIG. 1 is a block diagram of a relational database operating mechanism according to an embodiment of the present invention.

Referring to FIG. 1, in the embodiment, a host computer 1, a workstation 2, an application program 3, a database definition data file 4, a database operation mechanism 5, a database definition file 6, a database file 7, a search work file 8, Table definition means 9, user interface means 10, database update means 11, matrix index creation and update means 12,
It comprises normal index creation / update means 13, database search means 14, matrix index search means 15, and normal index search means 16, the database definition file 6 comprises definition records 61, the database file 7 comprises data records 71, matrix index 72. The search work file 8 is composed of a record set 81.

FIG. 2 is a diagram showing the structure of the matrix index shown in FIG. 1. Referring to FIG. 2, the matrix index 72 includes a matrix table 721, a data record address list 722, and a tree-like index 7.
23, and the matrix table 721 has a column αm721.
1, the column βn7212, the element αmβn7213, the data record address list 722 is composed of the data record pointer 7221, and the tree-like index 723 is composed of the key value 7231 and the data record pointer 7232.

FIG. 3 shows an example of a table having matrix indexes described in the embodiment. The table is composed of four items, and item A has a range of possible values (a
1, a2, a3, a4), the matrix index 72 is created. The item B has a wide range, but the appearances are biased, and almost all (b1, b2, b3) occupy most of them, so the matrix index 72 is created in the form of (b1, b2, b3, else). Since the item C is a non-overlapping item and its value range is too wide, a normal index 73 is created. Item D is not used as a search condition, and thus is not indexed. Here, since the number of matrix indexes and the number of items are set to the above values for the sake of simplicity of explanation, there is no particular limitation, and it is possible to create an n-dimensional matrix index. is there.

Next, the operation will be described.

In the computer system equipped with the present invention, the user first requests the database operation mechanism 5 to register the table definition information from the workstation 2 or the database definition data file 4. Upon receipt of the request, the database operation mechanism 5 receives the definition information from the workstation 2 in the table defining means 9 or inputs the definition data in the database definition data file 4 to create a definition record 61, which is stored in the database definition file 6. register. The definition record 61 is composed of information such as the item length and attributes of each item in the table, but regarding the item for which the matrix index 72 is created, the matrix index 72
A mark for creating the item and a possible value of the item are set. That is, regarding item A, possible values (a1, a2, a
3, a4) is set.

Next, when the user makes an additional update request for data through the workstation 2 or an additional update request is issued from the application program 3 to the table for which the definition has been completed, the database operation mechanism 5 causes the user interface means. At 10 the request and the additional update data are received and passed to the database updating means 11. The database updating means 11 reads the corresponding definition record 61 of the database definition file 6 in advance, and upon receiving the additional update record from the user interface means 10, first adds a data record to the database file 7 or creates a corresponding data record 71. To update. Then, the normal index 73 of the additional update record
The value of the item that should be created is normally updated by index creation 13
And the value of the item for which the matrix index 72 is to be created is passed to the matrix index creation / update means 12. In the embodiment, the value of the item C is normally passed to the index creating / updating means 13, and the value of the items A and B is passed to the matrix index creating / updating means 12.

Upon receiving the value of the item, the normal index creation / update means 13 additionally updates the normal index 73 of the database file 7 according to the definition record 61 of the database definition file 6.

Matrix index creating / updating means 12
In (1), when the value of the item is received, the matrix index 72 of the database file 7 is additionally updated according to the definition record 61 of the database definition file 6 as follows.

When the request is the first record addition, the matrix table 721 is first stored in the database file 7.
To create. At this time, one column αm7211 of the matrix table 721 corresponds to the value of item A. Ie item A
The values (a1, a2, a3, a4) of are the columns (α1, α2, α
3, α4). Similarly, the other row βn721
2 corresponds to the value of item B. That is, the value of item B (b
1, b2, b3, else) are columns (β1, β2, β3,
Corresponds to β4). Then, a null value is set in each element αmβn7213 of the matrix table 721. Next, the data record address list 722 is created, the pointer to the corresponding data record 71 is set in the data record pointer 7221 of the data record address list 722, and the element αmβn7213 corresponding to the values of the items A and B is set.
A pointer to the data record address list 722 is set to. That is, for example, the values of the items A and B are A = a
If 2, B = b3, the pointer is set to the element α2β3.

For the second and subsequent additions, the element αmβn721 of the matrix table 721 corresponding to the values of the items A and B is added.
3, the data record address list 722 is created and the data record pointer 7221 is set, and the element αmβn721 of the matrix table 721 is checked.
A pointer to the data record address list 722 is set to 3. If it is not a null value, the data record pointer 7221 is added to the data record address list 722 pointed to by the value in ascending order.

Further, the element αmβn of the matrix table 721
If 7213 corresponds to a wide value, the column β4 in which the item B corresponds to a value other than b1, b2, and b3 in the embodiment.
In the case of the element αmβ4 corresponding to, the value of item A is fixed,
A tree-shaped index 723 in which the value of the item B is other than b1, b2, b3 is created, and the tree-shaped index 723 is pointed to from the element αmβ4. In the tree-like index 723, the data record 71 can be pointed by the key value 7231 of the value of the item B and the data record pointer 7232 to the data record 71. Element α
If mβn7213 corresponds to the values of multiple wide items, the corresponding number of tree-like indices 723
From the element αmβn7213.

When updating the data, nothing is done unless the values of the items A and B change. If they change, the data record pointer 7221 of the data record address list 722 pointing to the data record 71 before the update is set. Delete and add a new pointer in the same way as when adding data.

Next, when the user issues a search request through the workstation 2 or the application program 3 issues a search request to the database operation mechanism 5 equipped with this method, the database operation mechanism 5 is executed. 5, the user interface means 10 first accepts the request and passes the search condition to the database search means 14.

The database search means 14 that has received the search conditions divides the conditions into those for the items having the matrix index and those for the items having the normal index, and the matrix index searching means 1 respectively.
5 and the normal index search means 16 are notified.

When the normal index search means 16 receives the condition, the normal index 7 of the database file 7 is searched.
3 is searched to obtain a set of corresponding records, and the set is output to the record set 81 of the search work file 8.

Upon receiving the condition, the matrix index search means 15 searches the matrix index 72 of the database file 7 in the following manner to obtain a set of corresponding records and outputs it to the record set 81 of the search work file 8.

When all the items corresponding to the matrix are specified and the range is not specified, the data record address list 722 is found from the element αmβn7213 of the matrix index 72 of the database file 7 corresponding to the value of the item, and the data record address list 7 is found.
The record set 81 may be created in the search work file 8 by sequentially reading the 22 data record pointers 7221.

When all the items corresponding to the matrix are not specified or the range is specified, a plurality of elements must be read. First, the condition is verified to determine which element should be read. Then, a plurality of elements αmβn of the matrix table 721 of the matrix index 72
Data record address list 72 corresponding to 7213
2, the data record address list 722 is arranged in ascending order, so a plurality of data record address lists 722 are alternately read at the same time, and the record set 81 of the search work file 8 is created at one time. It can be so.

When the element αmβn7213 in the matrix table 721 of the matrix index 72 corresponds to a wide value, the tree-like index 723 pointed to by the element αmβn7213 is searched to create the record set 81 in the search work file 8. In this case, it becomes the same as a normal index search, but the number of records corresponding to the index to be searched is reduced because other items are narrowed down.

When a record set 81 is created in the search work file 8 by the matrix index search means 15 and the normal index search means 16, the database search means 14 logically operates the record set 81 to create a resulting record set 81. User interface means 1
The result is notified to the user on the workstation 2 and the application program 3 through 0. Of course, if there is no condition for a normal index item, the record set 81 output by the matrix index search means 15 is notified as it is.

As described above, additional updating / searching of data can be performed in the database operation mechanism provided with the matrix index 72. As described above, since an extra index is not created in the additional update, the extra space is small and the overhead at the time of the additional update is also small. Further, at the time of searching, the amount of indexes to be searched can be minimized, and the search speed can be increased.

[0036]

As described above, according to the present invention,
It is possible to speed up the search speed when performing a database search by specifying multiple conditions for items with a small range of values, and even if the range of values is wide, there is a bias in appearance, If it can be expressed as "other", the search speed can be increased in the database search in the same manner as described above. In addition, since the auxiliary index is not created separately, no extra area is used in the database and the auxiliary index is not created separately, so the overhead when adding and updating data is reduced, etc. Has the effect of.

[Brief description of drawings]

FIG. 1 is a diagram showing an overall configuration of an embodiment of the present invention.

FIG. 2 is a diagram showing a structure of a matrix index in FIG.

FIG. 3 is a diagram showing an example of a table including matrix indexes according to an embodiment.

[Explanation of symbols]

 1 host computer 2 workstation 3 application program 4 database definition data file 5 database operation mechanism 6 database definition file 7 database file 8 search work file 9 table definition means 10 user interface means 11 database update means 12 matrix index creation update means 13 normal index Creation / update means 14 Database search means 15 Matrix index search means 16 Normal index search means

Claims (3)

[Claims] 1. A database processing system for searching / updating a relational database, wherein a set of records having the same record format has a table format in which one side is an item of each record and the other side is a row of records. When there are a plurality of items whose range of values is narrow in the table to be searched / updated, the number of types of possible values of each item is defined as the number of sides.
A matrix table format index is created by creating a matrix table of dimensions on a database, and pointing to an address list that points to the corresponding data record from each element of the matrix table. A relational database operation mechanism having a matrix index creating / updating unit capable of increasing the search speed in the case of searching data by designating a plurality of items. 2. The relational database operation mechanism according to claim 1, wherein the appearance is biased even if the value range of the item is not narrow, and it is possible to classify by some values and other values, A value with a range is also one element of the matrix table, and an element without a range is similar to the processing according to claim 1, and an element with a range has a tree-like shape for searching a value corresponding to the range of the element from the element. A matrix index creation / update means that creates an index in matrix table format by pointing to the index, and makes it possible to speed up the search speed when searching for data by specifying multiple even in the condition of a wide range of values. A relational database operation mechanism characterized by having. 3. In a system with matrix tabular index created in claim 1 and claim 2,
When a search request is submitted by a database user,
It is necessary to read a plurality of pointer columns by comparing the values of each side of the matrix of the matrix table format index created in claim 1 and claim 2 and conditions to determine which pointer column on the matrix should be read. A relational database operation mechanism characterized by having a matrix index search means capable of minimizing database access at the time of search by reading it in parallel when it is not necessary.