JPH0373020B2 - - Google Patents

Description

[Detailed description of the invention]

[Technical Field of the Invention] The present invention relates to a data processing device that can efficiently perform a projection operation on sorted relations using a relational operation based on a merge/sort process. [Technical background of the invention and its problems] With the advancement of large-capacity storage devices, the number of computer systems that construct large-scale databases on the large-capacity storage devices is increasing. However, as the database becomes larger, a problem arises in that the time required to search the database for necessary data becomes longer. As one method for solving this problem, various proposals have been made to execute the processing necessary for the search, that is, the relational calculations, using dedicated hardware. However, in conventional data processing devices that execute relational operations based on merge/sort processing, each record between two relations that has been sorted in advance is compared with each other, and the results of the comparison are Since the method is based on the operation of selectively outputting records from one of the two relations in response to the above, there is a problem that, for example, it is difficult to perform the projection calculation necessary for the database search. That is, this projection operation is, for example, relation {R1}

From [Table], remove records whose keys (attribute A, attribute B) have the same value and create a new relation {R2}.

[Purpose of the invention]

The present invention has been made in consideration of these circumstances, and its purpose is to provide a data processing device that can easily and quickly implement projection operations necessary for, for example, database search processing. There is a particular thing. [Summary of the Invention] According to the present invention, when a relation consisting of a plurality of records that has been sorted according to a predetermined rule is input, the record is input to the input part of a calculation unit that performs a relational operation on the relation. A buffer memory and a detection means (comparison circuit) for detecting whether the value of the attribute (key) of the record read from the record buffer memory is equal to the value of the attribute (key) of the current input record are provided. By detecting whether consecutively input records are equal or not, it is determined whether or not to output the record according to this detection result, and the output control unit is notified, and the record is output. The projection calculation can be executed by controlling the . Furthermore, the present invention has a function of checking whether the sorting process has been performed correctly. In other words, if a sorting inspection circuit is provided and it is detected that the key values of records do not have a magnitude relationship according to a predetermined rule (ascending order), the sorting inspection circuit instructs the sorting processing section to perform the sorting process on multiple records again. This is a data processing device characterized by the following. [Effects of the Invention] Thus, according to the present invention, it is detected whether or not records that are sequentially input in relations that have been subjected to a predetermined sorting process are equal to records that are input next, and if they are equal, the record is determined to be the same. Since the records are prevented from being output and the records are sequentially output only if they are not equal, the output record relation is the one in which the duplicate records in the input relation have been removed. Therefore, projection calculations that conventionally relied on software can now be performed easily and at high speed. Moreover, by simply providing a record buffer memory and a comparison circuit in the calculation section,
This has great practical effects, such as being able to easily construct the hardware. Furthermore, a sorting check circuit is installed to detect whether the sorting process has been performed correctly, and if it is not performed correctly, it is set to re-sort the records that were previously performed, resulting in an error in the sorting process. This prevents erroneous output and wasteful processing. [Embodiment of the Invention] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a schematic configuration diagram of a data processing device according to an embodiment of the present invention. The data processing device basically includes a calculation unit 1 that executes relational calculations based on merge/sort processing, and an output control unit 2 that controls the output of the results of the relational calculations. A plurality of records constituting one relation, which have been sorted according to a predetermined rule, are sequentially input to the calculation unit 1 of the data processing device. This arithmetic unit 1 includes an input register 11 for setting one word of a record that is input sequentially, and a first record buffer memory that has a storage capacity for one record and has a FIFO (first-in, first-out) function. 12, a comparison circuit 13 for comparing the key value (word) read from the first record buffer memory 12 and the key value (word) set in the input register 11;
A data register 14 that sets one word of the record read from the record buffer memory 12.
and a flag circuit 15 that stores the equal signal E of the comparison circuit 13 as a duplication flag F. Note that the comparison circuit 13 detects matching records word by word, and outputs an equal signal E according to the detection result.
is outputting. Thus, a record matches when all of the plurality of words that make up the record match. Therefore, the flag circuit 15 to which the equal signal E outputted from the comparison circuit 13 is input,
reading the last word of the record stored in the record buffer memory 12 and providing it for the comparison;
At the same time, when the last word is set in the data register 14, the duplication flag F is output for the first time. By the way, since each key value of a record is compared word by word in this way, the timing when it becomes clear that the record to be processed overlaps with the record that follows is the last word of the record. This is when the comparison result is obtained. For this reason, it is necessary to buffer the records that are sequentially output word by word via the data register 14 of the first record buffer memory 12. A second record buffer memory 16 provided in the output control section 2 is intended for such buffering, and stores records outputted from the calculation section 1 via the data register 14 in order. . The output control circuit 17 inputs the duplication flag F from the arithmetic unit 1 and determines whether the record stored in the second record buffer memory 16 overlaps with the next record following that record. We are determining whether Then, if there is no input of the duplication flag F (F=0) and the record does not overlap with the next record in the relation, the output control circuit 17 sends the second record through the output register 18. The records stored in the record buffer memory 16 are output. Further, when the duplication flag F (F=1) is input, the output control circuit 17 controls the second record buffer memory 16.
is preventing the output of records stored in . Therefore, a record that is determined to be equal to the next record is not output from the output control unit 2. Although not shown here because it is not directly related to the gist of the present invention, the calculation unit 1 includes a circuit that indicates whether or not a key field is currently being processed;
Needless to say, a circuit or the like is provided to indicate whether or not the last word of the record is being processed. As shown in FIG. 2, the arithmetic unit 1 is provided with a sort check circuit 19 that inputs the comparison results (<, =, >) of the comparison circuit 13, and the sort check circuit 19 inputs the comparison results (<, =, >) of the comparison circuit 13. Based on the comparison result between the input record before the cycle and the current input record set in the input register 11, it is checked whether the sort processing performed on the plurality of records by the sort processing unit 3 was performed correctly. You can do it like this. That is, when sorting is performed in ascending order, the key value of the current input record should be greater than or equal to the key value of the previously input record. Therefore, by determining such record conditions from the comparison results of the comparison circuit 13, it becomes possible to easily determine whether or not the sorting process has been performed correctly. If an error in the sorting process is detected, the projection calculation for the input relation is immediately stopped, thereby eliminating unnecessary processing. At the same time, by instructing the sort processing unit 3 to redo the sorting process on the plurality of records, it becomes possible to promptly restart the projection calculation. Next, the flow of projection calculation processing in this apparatus will be explained using a specific example. Projection calculations in this device are basically performed as follows. () First, the relation {Rx} is sorted, for example, in ascending order (smallest first) with respect to its key value. For example, attribute A of relation {R1}
and attribute B are regarded as one attribute and the sorting process is performed. This sorting process can be performed using software or dedicated hardware for sorting (not shown).
It is carried out by. As a result, the following relations are obtained, for example.

[Table] Each record of such a relation is inputted word by word sequentially, and the projection calculation by this apparatus is executed as follows. () In the calculation unit 1, a pointer is placed at the first record of the relation. () Next, if the duplication flag F is "0", the record indicated by the pointer is output, and if the duplication flag F is "1", the output of the record is blocked. The pointer is then advanced to the next record. () If there is a next record in the relation, the process returns to the process in () above, and if there is no next record in the relation, the process ends. The above-mentioned processes () to () in this process flow will be explained in more detail. Note that each record of relation {R1} in this embodiment is composed of a plurality of attributes. Therefore, here, the word length of each attribute is all one word length,
It is assumed that the record length of each record is given as a two-word length. It is assumed that the time to process one word is one phase, and the time to process one record is one cycle, that is, two phases are one cycle. Now, to perform the projection operation, only two consecutive records in the sorted relation are required. Then, every time two consecutive records are input, the execution is started. In other words, the first two records [1/
r] and [2/q] are input, the calculation unit 1
It is checked whether the two records are duplicated or not, and the check result is output as a duplication flag F to the output control unit 2 in synchronization with the record. This process is performed each time a record is input one by one and between records that were input earlier. That is, in the 0th cycle, record [1/r]
is input and stored in the first record buffer memory 12. Specifically, in the first phase, [1] is input after the beginning of the first record, and this is set in the input register 11. The data (words) are then stored in the first record buffer memory 12. In the next second phase, the next word [r] of the record is set in the input register 11 and stored in the first record buffer memory 12. In the above 0th cycle, the input of the first record is completed, and the record is stored in the first record buffer memory 12. In the subsequent first cycle, the following three processes are executed in parallel. One of them is the process of inputting the second record [2/q] and storing it in the first record buffer memory 12, and this process is performed in the same way as in the 0th cycle. In this case, the FIFO function of the record buffer memory 12 causes the records stored in the record buffer memory 12 in the previous cycle to be read word by word in synchronization with the storage of the records. In the second process, the first record read from the first record buffer memory 12 and the current input state are stored in the input register 11.
This is an operation to compare with the second record set in . The third process is an operation of transferring the first record read from the first record buffer memory 12 to the output control section 2 and storing it in the second record buffer memory 16. That is, in the first phase, [2], which is the first word of the second record, is input and set in the input register 12. This data is stored in the first record buffer memory 12, and at the same time, from the first record buffer memory 12 having a FIFO function, the record stored one cycle ago (=first record) is stored in the first record buffer memory 12. The first word is read. And this record buffer memory 12
The data read from the output controller 2, that is, the first word [1] of the first record, is
The data is transferred to the record buffer memory 16 of the record buffer memory 16, and is compared with the value set in the input register 11 ([2], which is the first word of the second record) in the comparison circuit 13. In the subsequent second phase, the second word [q] of the second input record is set in the input register 11 and stored in the first record buffer memory 12, and at the same time, the second word [q] of the second input record is set in the first record buffer memory 12. 2 of the first record read from
It is compared with the word [r]. This second phase is processing of the last word of the record, and since records [1/r] and [2/q] do not match, the equal signal E of the comparison circuit 13 is "0", and this In response, the flag circuit 15 sets the duplication flag F.
is set to “0”. With the above, the first cycle is completed, the second record is stored in the first record buffer memory 12, and the second record is stored in the second record buffer memory 16.
The first record is stored in . The output control circuit 17 stores flag information indicating whether or not the first record stored in the second record buffer memory 16 overlaps with the next record. In accordance with the information of this duplication flag F, the output control circuit 17 updates the record [1/1] stored in the second record buffer memory 16 in the next third cycle.
r]. In this case, the duplication flag F is "0", indicating that the record does not overlap with the next record, so the record [1/r] is output as is. In the next third cycle, in addition to the three processes described above, record output control according to the duplication flag F is performed. At this time, the above-mentioned three processes include processing for the third record that will be input next, comparison processing with the second record, and transfer of the second record to the second record buffer memory 16. Needless to say, the transfer will take place. However, in the third cycle, the duplication flag F is "0", indicating that the record does not overlap with the next record, so the record [1/r] is output as is. Then, in the fourth cycle, a match is detected by comparing the second record and the third record in the third cycle, and since the duplicate flag F is "1", output control is performed. The circuit 17 prevents the output of the second record [2/q] stored in the second record buffer memory 16. In other words, records that duplicate the next record are not output. Thereafter, similar processing is performed by repeating the above cycle until processing for all input records is completed. Thus, according to the present apparatus, by adding the duplication flag F to the second record buffer memory 16, the following relation is obtained.

[Table] Therefore, the duplicate flag F from this relation
By preventing the output of records for which is "1", that is, by removing duplicate records, it is possible to obtain the following new relation {R2}.

[Table] As described above, according to this device, the record stored in the first record buffer memory 12 is compared with the record input in the next cycle, and based on the result, whether or not to output the record is determined. Since the output of the record is controlled by determining the output of the record, the projection operation can be executed effectively and at high speed using the data processing device. Furthermore, the processing algorithm is simple and can be easily implemented in hardware. Note that the present invention is not limited to the embodiments described above. For example, the number of attributes constituting a record, its word length, etc. may be determined depending on the structure of the database to be processed. Also, the number of phases in one cycle may be determined in accordance with this. In addition, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of drawings]

FIG. 1 is a schematic diagram of an apparatus according to an embodiment of the present invention;
FIG. 2 is a schematic configuration diagram of an apparatus showing an application example of the present invention. 1... Arithmetic unit, 2... Output control unit, 3... Sort processing unit, 11... Input register, 12... First
record buffer memory, 13...comparison circuit,
14...Data register, 15...Flag circuit,
16...Second record buffer memory, 17...
...Output control circuit, 18...Output register, 19...
...Sort inspection circuit.

Claims (1)

[Scope of Claims] 1. A sort processing unit that sorts a plurality of records according to a predetermined rule; an input register that sequentially inputs and sets each record sorted in this sort processing unit; a first record buffer memory that temporarily stores the input record in record units; a first key value representing an attribute of the record stored in the first record buffer memory; and the input register. a comparison circuit that mutually compares the values of the second key representing the attributes of the record currently being input; a sort inspection circuit for inspecting whether or not there is a size relationship according to the rule;
If it is detected that the test results of the first and second key values in the sort test circuit match the second record buffer memory that temporarily stores records in the record buffer memory in record units;
In response to this inspection result, the output of records stored in and read from the second record buffer memory is prevented, or if it is detected that the magnitude relationship conforms to the predetermined rule, the second record buffer memory an output control circuit that outputs records read from the memory, and when it is detected that the values of the first and second keys in the sorting inspection circuit do not have a magnitude relationship according to the predetermined rule, the sorting inspection circuit The data processing device is characterized in that it instructs the sorting processing unit to perform sorting processing on the plurality of records again. 2 The first record buffer memory is a FIFO that operates with one record processing time as one cycle.
2. The data processing device according to claim 1, wherein the data processing device has a function of reading out the key value of the record stored one cycle before when writing the key value of the current input record. 3. The data processing device according to claim 1, wherein the comparison circuit includes a flag circuit that generates a record duplication flag according to the comparison result.