JPH07101382B2 - Margin processing device - Google Patents

Description

TECHNICAL FIELD The present invention relates to a merge processing device that inputs a plurality of pre-sorted record strings and outputs it as one sorted record string.

[Conventional technology]

Generally, data handled in database processing is called a record, and the record is composed of a plurality of columns having different attributes. Therefore, in the database processing, sorting and merging processing is performed with the tabular data in which the columns are arranged horizontally and the records vertically. The tabular data itself is called a database.

In this kind of database processing, a database with a very large number of records and a database with a plurality of different information are targeted for processing. Therefore, sorting and merging with a general-purpose computer requires a huge amount of time. In many cases. For this reason, a dedicated processing device has been proposed in which part or all of the database processing is directly executed by hardware to increase the speed. Here, as an example thereof, the sort processing device disclosed in JP-A-61-42031 will be described.

FIG. 2 is a block diagram of a prior art sort processing device shown in Japanese Patent Laid-Open No. 61-42031. In the figure, 11 is a sort circuit, 12 is a large-capacity buffer memory, 50 is a control circuit, 51 is an input register, 52 is an output register, 53 is a bank address generation circuit, 54 is a sort input / output switching circuit, and 55 is a read circuit. Address generating circuit, 56 write address generating circuit, 57 address switching circuit, 58 data switching circuit, 59 switching control line (ST), 60 bank address line (BA), 61 I phase signal line (I), 62 is a II phase signal line (II), 63 is a state control line (IN
T), 64 is a sort input / output switching signal line (PUP), 65 is an input control line (PUSH), 66 is an output control line (POP), and 67 is a data input / output line (DIO).

The process of the sorting process in this sort processing device includes an initial sorting stage and a merging stage, which are designated by a switching control line (ST) 59. The initial sorting stage is data input / output line 67
It is a stage in which a large number of input records given from are divided and input to the sorting circuit 11, the decomposed records are sorted, and the sorted record string is stored in the buffer memory 12. The merging stage is a stage in which a plurality of record strings generated in the buffer memory 12 in the initial sorting stage are merged and output from the data input / output line 67. Since the operation of FIG. 2 is described in detail in Japanese Patent Laid-Open No. 61-42031, the description thereof will be omitted here.

[Problems to be solved by the invention]

In the prior art shown in FIG. 2, when reading a record from the buffer memory in the merge process, a bank address designating a record row to which the read record belongs,
It required a record address to specify the record position within the record string. For this reason, the calculation of the read address of the buffer memory becomes complicated, and the control hardware portion for performing the address calculation is inevitably large.
Therefore, there is a problem in constructing a compact and high-performance sort processing device. Further, since the record address is required for each record column, hardware that stores the same number of record addresses as the number of record columns to be merged is required, which is a factor that defines the upper limit of the number of record columns that can be merged at one time. There was also a problem.

Furthermore, in the process of merging, hardware is required to determine whether the output of all records in the record string has been completed. If all records have been output,
There is a problem that the control must be changed so that the output of the next record is not requested for the corresponding record string.
Detecting the completion of reading records from each record string,
It was desired to realize a mechanism that can automatically change the merge control with a small amount of hardware.

The present invention aims to eliminate the above-mentioned problems of the prior art, simplifies the control circuit when merging a plurality of record strings at one time, and reduces the amount of hardware.
It is to provide a smaller and higher-performance merge processing device.

[Means for solving problems]

The present invention, a record storage means for storing a plurality of record strings pre-sorted in ascending order or descending order, a sorting means for extracting the maximum or minimum record from the plurality of records and holding the remaining records, The address information of the next record in the record string to which the record belongs is added to the record read out from the record storage unit and input to the sort unit, and the address information added to the record extracted by the sort unit. Control means for reading the next record from the record storage means.

[Work]

First, the head record of each of the plurality of record strings is read from the record storage means, the address information of the next record in the record string to which the head record belongs is added to each read head record, and the result is input to the sorting means. . The sorting means extracts the maximum or minimum record from the plurality of input top records and retains the remaining records.

Next, one record is read from the record storage means by the address information added to the record extracted by the sorting means, and the read record is added with the address information of the next record in the record sequence to which the record belongs. Enter into the sorting means. The sorting means similarly extracts the maximum or minimum record from the input record and the held record, and holds the remaining records again.

Hereinafter, the process of reading a record from the record storage means based on the address information of the extracted record, adding the address information of the next record to the read record, and inputting it to the sorting means is stored in the record storage means. Merge multiple record columns. When the reading of all the records included in a certain record string is completed, the end record which is the maximum / minimum or more than all the records of the record string is input to the sorting means,
Do not read the next record from the record string.

As a result, it is possible to realize a more compact and high-performance merge processing device by eliminating the circuit and various counter circuits that were conventionally required to store the position of the read record in the record string.

〔Example〕

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

FIG. 1 is a block diagram of an embodiment of the present invention. In the figure, 11 is a sorting circuit for extracting the maximum or minimum input record, 12 is a buffer memory for storing sorted record strings, 13 is an address counter for calculating a read address of the buffer memory 12, and 14 is a read. Initial address input line for inputting initial value of address, 15 address line of buffer memory 12, 16 data output line of buffer memory 12, 17 record input line to sort circuit 11, 18 record from sort circuit 11 An output line, 19 is a record output line for outputting the merged result, and 20 is a read address input line for inputting the address of the record output from the sort circuit 11 to the address counter 13. Reference numeral 21 is a triangle that schematically shows the record string stored in the buffer memory 12, and the height of the triangle indicates the size of the record when comparing the records. Reference numeral 22 denotes a terminal record stored at the end of the record string, 23 denotes a record input to the sort circuit 11, and 24 denotes a configuration of an output record from the sort circuit 11.

The procedure of the merge process in this embodiment is shown below. Here, three record strings A, B, C sorted in descending order are stored in the buffer memory 12, and these are similarly merged into one record string in descending order.
End record 22 added to the end of each record string A, B, C
Is a record having a minimum value or less among the records to be merged, and as shown in the figure,
Expressed as ∞.

First, the three record strings stored in the buffer memory 12
The initial address for reading the first record of A, B, C is line 1
The address counter 13 is sequentially input from 4 and the three head records are sequentially read from the buffer memory 12.

Each record is read from the buffer memory 12 by dividing the number of times according to the record length, and the address at that time is calculated by the address counter 13. Therefore, for example, when the reading of the first record of the record string A is completed, the output of the address counter 13 indicates the read address of the next record in the same record string A as the first record. This is the same when the first record of the record strings B and C is read. Therefore, each time the first record of each of the three record strings A, B, and C is sequentially read from the buffer memory 12, the output value of the address counter 12 indicated by the address input line 15 at that time, that is, the same record string The read address of the next record is added to the corresponding top record and input to the sort circuit 11.

Reference numeral 23 in FIG. 1 shows a state in which the read address (next address) of the next record in the same record string as the record is added to the record (record body) read from the buffer memory 12. The position where the next address is added is a position that does not affect the comparison of the record bodies. Generally, since the comparison of records is performed with priority from the upper bit (MSB), the addition position of the next address should be on the lower bit (LSB) side.

The sort circuit 11 compares the sizes of the top records of the input record sequences A, B, and C with each other, and extracts the largest record. At this time, the record extracted by the sort circuit 11 is the largest record among all the records forming the three record strings A, B, and C to be merged. The record (main body) extracted by the sorting circuit 11 is output from the record output line 19 as a merged result, and the remaining two records are held in the sorting circuit 11. This completes the first merge process.

On the other hand, the next address of the record extracted by the sort circuit 11 and output from the record output line 19 is input to the address counter 13 via the read address input line 20. According to the address of the address counter 13, the buffer memory 12
The next record is read from, and the read address of the next record is added to the read record and input to the sort circuit 11 in the same manner as above. The sort circuit 11 compares the input record and the two held records with each other to extract the largest record. This extracted record consists of three record strings A, B,
It is the second largest record of all the records that make up C, and is output as the merged result through the record output line 19, and the remaining two records are held in the sorting circuit 11 again. This completes the second merge process.

Hereinafter, the next address of the merge result record output from the record output line 19 is input to the address counter 13, the next record is read from the buffer memory 12, and the next record read address is added to the read next record. By repeating the operation of inputting to the sort circuit 11, the merging process of the records forming the three record strings A, B and C is sequentially advanced.

In this way, when all the records in the target record string are read, merged by the sort circuit 11, and output, the next address having the final record in the record string output from the sort circuit 11 is , End record
22 read addresses. The end record 22 is also read from the buffer memory 12 in the same manner and input to the sorting circuit 11. However, since the end record 22 is −∞, which is the smallest value for all the records to be merged, it is not output from the sort circuit 11 until the merge process is completed. Therefore, the address of the next record added to the end record 22 is meaningless, and any value does not adversely affect the merge operation. When each end record 22 of the three record strings A, B, and C is held in the sort circuit 11, the merge process ends.

In this embodiment, the case where the three record strings A, B and C stored in the buffer memory 12 are merged and output is shown.
Even when the number of record strings stored in the buffer memory 12 is other than 3, the same expansion can be performed. Also,
In the present embodiment, the example of the merge process in which the record string sorted in the descending order is input and the record string sorted in the descending order is output is shown. However, when merging in the ascending order, the input record string is sorted in the ascending order. Then, it can be easily realized by setting the end record added to the end of the record string in the buffer memory as the record having the maximum value or more of the record to be processed, that is, + ∞. .

In addition, in the present embodiment, the method of pre-storing the end record at the end of the record string stored in the buffer memory 12 has been described, but the end record is generated externally by detecting the last record for each record string. It is also possible to control the merging operation by inputting it to the sorting circuit 11 in the same manner as in the present embodiment.

Further, in the embodiment, the example in which the buffer memory 12 capable of reading a record by designating an address is used as the record storage means is shown, but a magnetic tape device, a magnetic disk device, a magnetic bubble device or the like is used. A device capable of storing a large amount of records and reading consecutive records can also be used as the record storage means.

Further, when the records are stored in the record storage means, it is not necessary to store the records in physically continuous areas, and logically continuous records in the record sequence sorted in ascending or descending order, that is, sorted by sorting. It suffices if the pointer required for reading a record can be calculated from the storage positions of adjacent records. In this embodiment,
Since the record string is stored in the physically continuous area of the buffer memory 12, the record to be read next is determined by (storage address of record read immediately before) + (length of record read immediately before). The storage address of can be obtained.

〔The invention's effect〕

As is apparent from the above description, according to the present invention,
Control of merge processing is simplified, and circuits that were conventionally required to store the position of the read record in the record string and various counter circuits are unnecessary, and a smaller and higher-performance merge processing device is realized. it can.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment of the present invention, and FIG. 2 is a diagram showing a conventional configuration example. 11 …… sort circuit, 12 …… buffer memory, 13 …… address counter, 14 …… initial address input line, 22 …… end record, 23 …… input record to sort circuit, 24 …… output from sort circuit record.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Hiroshi Matsuo, Hiroshi Matsuo, 1356-2356 Take, Yokosuka City, Kanagawa Nippon Telegraph and Telephone Corporation, Information and Communication Processing Laboratory (56) Reference JP-A-61-72333 (JP, A)

Claims (2)

[Claims] 1. A record storage means for storing a plurality of record strings pre-sorted in ascending or descending order, and a sorting means for extracting a maximum or minimum record from the plurality of records and retaining the remaining records. The first time, the first record of each of the plurality of record strings is read from the record storage means, and the address information of the next record in the record string to which the first record belongs is added to each read first record and the sorting is performed. Input to the means, and from the second time onward, the record is read from the record storage means by the address information added to the record extracted by the sort means, and the next record in the record string to which the record belongs to the read record. And a control means for adding address information to the sorting means and inputting to the sorting means. Merge processing apparatus characterized by merging code string into one record sequence rearranged in ascending or descending order. 2. A record string stored in the record storage means is added to the end of each record string stored in the record storage means, as a terminal record, the maximum or minimum record or a record exceeding the minimum record record, and the respective record strings are merged. In the processing, if the record extracted from the sorting means is the last record of the corresponding record string, the corresponding end record following the last record is input to the sorting means. The merge processing device according to claim 1.