JPH01234931A - Projection processing system for relational data base - Google Patents

Abstract

PURPOSE:To speed up projection processing by executing the projection processing by means of hardware based upon an exclusive circuit. CONSTITUTION:A processor 1, a memory 2 and an exclusive circuit 3 for projection processing are connected through various buses. The processor 1 controls the whole data processor and the memory 2 is provided with an input table 21 for a record group to be projected and an output table 22 for storing projected record group in its inside. The circuit 3 is provided with a projection column information buffer 31, an input address counting circuit 32 and an output address counting circuit 33 to execute the reading and writing of a projection column. Since the projecting function is based upon hardware, the projection processing can be rapidly executed.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention provides one of the operational functions of a relational database.
The present invention relates to a projection processing method, and particularly relates to a projection processing method for relational databases that is implemented in hardware to speed up processing.

[Conventional technology]

In recent years, with the spread of relational databases (hereinafter referred to as RDB), there has been a strong demand for higher performance processing. As a conventional technique for satisfying such a demand, for example, a technique described in Japanese Patent Application Laid-Open No. 57-30034 is known. This type of conventional technology is a technology related to converting the RDB operation function, which was conventionally performed by software, into hardware, and in particular, RDB
The search speed is improved by converting the search function, which is one of the DB operation functions, into hardware.

[Problem to be solved by the invention]

As mentioned above, the prior art relates to the hardware implementation of the search function among the functions of the RDB system.
There was a problem in that no consideration was given to hardware implementation of the projection function, which is generally executed after search processing in the RDB system, and the speed of the projection processing was not achieved.

SUMMARY OF THE INVENTION An object of the present invention is to provide a hardware-based RDB projection processing method in order to realize high-speed projection processing.

[Means to solve the problem]

According to the present invention, the purpose is to provide a dedicated buffer for storing projected column information in the order of column numbers, an input address count circuit that indicates the memory address of the input record to which the projection process is to be performed, and a memory to which the projected record is to be output. This is achieved by having an address count circuit that indicates an address and executing reading and writing of the projection column.

[For production]

The dedicated buffer stores projection column information such as the column number, column type, and data length (meaningful only for fixed-length columns; undefined for variable-length columns) in order of column number. The projection column information is read out in order from the smallest column number. Based on the read column number, the input address counting circuit counts up to the address of the corresponding column, and the corresponding column data is read from the memory based on this address.

This read column data is written to the memory address indicated by the output address count circuit. The aforementioned read/write operation is executed while counting up the input and output address count circuits by the amount of column data, thereby completing the memory movement of the corresponding column. Note that variable length columns are processed according to the data length at the beginning of the column. By executing such processing for the number of projection columns, a target projection record is generated at a predetermined position in the memory.

〔Example〕

Hereinafter, one embodiment of the RDB projection processing method according to the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of a data processing apparatus to which the present invention is applied. In FIG. 1, 1 is a processor, 2 is a memory, 3 is a projection processing dedicated circuit, 21 is an input table, 22 is an output table, 31 is a projection column information buffer, 32 is an input address count circuit, and 33 is an output address count circuit. It is.

As shown in FIG. 1, a data processing device to which the present invention is applied is constructed by connecting a processor 1, a memory 2, and a projection processing dedicated circuit 3 for the present invention through various buses. . A processor 1 controls the entire data processing device, and a memory 2 includes therein an input table 21 for a record group to be projected, and an output table 22 for storing a record group subjected to projection processing. ing.

Further, the projection processing dedicated circuit 3 includes a projection column information buffer 31, an input address count circuit 32, and an output address count circuit 33.

Each record stored in the input table 21 is composed of, for example, four columns having data regarding "product number", "product name", "price j", and "inventory quantity" in order from the beginning. shall be. Such an input table 21 is generated by a search process or the like, and in the records stored in the table, the address of the j-th column of the i-th record in the input table 21 is set as IA, . do. Then, projection processing is performed on the "product name" in column number 2 and the "inventory quantity" in column number 4 of each record stored in the input table 21, and the output table 22 is obtained. The operation of one embodiment will be explained.

First, as preprocessing, the processor 1 stores the column number of the column of each record in the input table 21 to be projected, and the column type (in the case of a fixed length column) in the projection column information buffer 31 in the projection processing dedicated circuit 3. "F", "■" for variable length columns) and data length are stored in order of column number.

Next, the processor 1 sets the start address IA of the first input record in the input table 21 to the input address register IAR in the input address count circuit 32, and sets the memory address to which the first output record is to be output. That is, the address in the output table 22 in the memory 2 is set in the output address register OAR in the output address count circuit 33, and the projection processing dedicated circuit 3 is activated.

Upon receiving this activation, the projection processing dedicated circuit 3 first retrieves the first column number, column type and data length corresponding to the column number from the projection column information buffer 31, and inputs the column number register CNR, column type register CTR, and data. Load long register DLR. After data is loaded into these registers, the input address count circuit 32 counts up the input address register IAR until it reaches the memory address of the column number set in the column address register CNR. In this operation example, the projection process is performed on the second and fourth columns of each record, so the value of the input address register IAR becomes the address IA+z of the second column of the first record. At this time, this address IA, □ is output as a memory read address, and a memory read for the memory 2 is executed. As a result, the data at the beginning of the second column of the first record in the input table of memory 2 is read out and fetched into memory data register MDR. This data is written to the memory address of the memory 2 indicated by the output address register OAR in the output address count circuit 33. Thereafter, the contents of the input address register IAR and the output address register OAR are counted up by JlIN times, and each time the above-mentioned read and write processing to the memory 2 is performed. When the above-mentioned read and write processes are repeated for the data length indicated in the data length register DLR, all the data in the second column of the first record in the input table 21 is transferred to the predetermined position in the output table 22. It has been transferred to . In this way, when the processing of one column is completed, the projection processing dedicated circuit 3 transfers the next projection column information in the projection column information buffer 31 to the register C.
The data is loaded into NR, CTR, and DLR, and the specified column data is moved in the same manner as described above.

When the projection processing dedicated circuit 3 finishes the projection processing for one record as described above, it reports the completion of the processing for one record to the processor 1. The processor 1 calculates the next input record address and output record address using a count circuit 32.3 in the projection processing dedicated circuit 3.
3 input address registers IAR and OAR,
The projection processing dedicated circuit 3 is activated to execute processing of the next record. The above-described processing is performed on all records of the input table 21 in the memory 2, and as a result, an output table 22 that has been subjected to the projection processing is obtained in the memory 2. In the above explanation, it was assumed that the column length is fixed, but when processing variable length columns, it is possible to read the data length at the beginning of the column and set it in the data length register. , projection processing can be performed in the same way as described above.

In the embodiment of the present invention described above, the projection processing can be executed by hardware using a dedicated circuit, so that the processing speed can be increased and the load on the processor can be reduced.

〔Effect of the invention〕

As described above, according to the present invention, in parallel with the search process, the hit record projection process can be performed at high speed using hardware with a simple configuration.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of a data processing apparatus to which the present invention is applied. DESCRIPTION OF SYMBOLS 1... Processor, 2... Memory, 3... Projection processing dedicated circuit, 21... Input table, 22... Output table, 31... Projection column information buffer, 32.
...Input address count circuit, 33...Output address count circuit, IAR...Input address register,
OAR...output address register, MDR...memory data register, CNR...column number register,
CTR: Column type register, DLR: Data length register.

Claims (1)

[Claims] 1. In a relational database projection processing method that extracts one or more columns of a predetermined length from an input record consisting of one or more columns of a predetermined length and generates an output record, the input of the column to be projected is a projected column information buffer that stores column numbers, column types, and data lengths indicating the arrangement order within a record in order of column numbers; and an address count circuit that counts addresses of input records in a memory in which the input records are stored; an address count circuit for counting the memory address to which the output record is to be written; the input address count circuit counts the address of the corresponding column in the input record according to the column number in the projection column information buffer; A relational database projection processing method, characterized in that data in a column is read and the read data is written to an address counted by the output address counting circuit.