JPH0467660B2 - - Google Patents

Description

TECHNICAL FIELD The present invention relates to a chain search method, and more particularly to a chain search method for recognizing the order of records in a file having a chain structure stored in an auxiliary storage device in an electronic computer system.

Prior Art As a method for expressing the order of records recorded in an auxiliary storage device such as a magnetic disk, there is a method in which the +1-th record address is recorded as a pointer in the logically-th record from the starting point record. Hereinafter, an area in an auxiliary storage device in which records recorded using this recording method are stored will be referred to as a chain file, and a collection of records linked by pointers from one starting point record will be referred to as a chain.

The conventional method for recognizing the number of each record from the starting point record is as follows:
After reading the starting point record from the auxiliary storage to the main memory by some means, the address of the next record is obtained from the pointer recorded in each record, and the page to which the record belongs is transferred to the main memory by direct access means. There is a method of repeating the operation of reading data into the device and tracing it to the end point record.

Here, a page is a physical input/output unit between an auxiliary storage device and a main storage device, and hereinafter, an operation for inputting/outputting a page between an auxiliary storage device and a main storage device will be referred to as page IO.

FIG. 2 shows an example of a chain file consisting of four pages, and seven records constituting one chain are stored in the chain file. ai indicates the address of each record, a4 is a start point record, a5 is an end point record, and all others are intermediate records.

In the conventional method, if the entire chain corresponding to a certain start point record is closed and stored in one page, the page read into main memory can be searched, so one page IO can be used to reach the end point record. can be traced.

However, if a chain is recorded across multiple pages as shown in Figure 2, page IO is required for each record unless all read pages are stored. In Figure 2, there is only one start point record in the chain file, but there are many start point records in the chain file and chains starting from those records.
If each chain spans pages,
Tracing the entire chain requires a number of page IOs equal to the number of records per file.

Furthermore, as shown in FIG. 3, when there are multiple types of pointers that define ordering, page IOs corresponding to the product of the number of records per file and the number of pointers per record may be required.

OBJECTS OF THE INVENTION The present invention has been made to solve the problems of the prior art described above, and its purpose is to provide a chain search method that reduces page IO.

Structure of the Invention According to the present invention, there is provided a chain search method that recognizes the order of each record in a chain file, in which a first step of sequentially reading pages of the chain file into main memory from the beginning of the chain file. And for each record in the chain file, (A, B, C, N) from the set of the storage address of the record and the pointer in the record.
4 fields (A and C are fields that record the address and pointer according to each record at the start point, end point, and intermediate point; B records 0, 1, or 2 according to each record at the start point, end point, and intermediate point. field, N is a field indicating the order of records) and storing it in a classification work file, and sorting the field records in the classification work file in ascending order using the field as a key. The third step is to compare each adjacent field record in the classification work file after this sorting with each other according to a predetermined rule, and the field records that have a certain relationship as a result of the comparison are created as a result file. store,
a fourth step of storing the matching results of other field records in the matching work file; and, in response to the completion of the fourth step, copying the contents of the matching work file to the classification work file. From the third step to the fourth step
and a fourth step of repeating the above steps, and when no field record is output to the matching work file, the N field of the field record of the result file is used to output any field record in the chain file. A chain search method is obtained which is characterized in that information indicating how many chains it takes to reach a record from the first record in a chain file is obtained.

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

A block diagram of an embodiment of the invention is shown in FIG.
In FIG. 1, the auxiliary storage devices 11 to 41 do not necessarily have to be different devices, but may all or any combination of the chain file 11, result file 21, matching work file 31, and classification work file 41 be the same. However, in FIG. 1, these files are all shown as being located in different auxiliary storage devices.

The control means 5 uses the sequential access means 1 to read the chain file and outputs the records necessary for recognizing the orderliness to the classification work file 41 using the sequential access means 4, and then outputs the records to the classification work file 41 using the sequential access means 6 and the sequential access means. 3. While repeatedly using the sequential access means 4 to transfer records between the classification work file 41 and the matching work file 31 multiple times,
The records whose order has been determined are output to the result file 21 using the sequential access means 2.

The conditions of the chain file 11, which are the premise of the present invention, will be explained with reference to FIG.

(1) ai indicates a record address, and can be expressed in any format as long as it uniquely identifies each record within the chain file 11. However, it is necessary to be able to determine the AI value when the page is loaded.

(2) The first field of a record indicates a pointer, and the second field indicates whether the record is the start point of a chain (“S” in Figure 2), the end point (“E” in Figure 2), or an intermediate point. (blank in FIG. 2). The positions of the pointers and flags, and the values of the flags are arbitrary as long as they can be determined from the record itself or from external instructions.

New field records (A, B, C, N) each consisting of four fields are stored in the classification work file 41, the matching work file 31, and the result file 21. 1st and 3rd
In fields A and C, addresses and pointers are recorded according to the start point, end point, and intermediate records, and in the second field B, the start point, end point, and
Each value of 0, 1, or 2 is recorded according to each intermediate record, and an integer value indicating the order of the records is recorded in the fourth field N. Specifically, it is as follows.

(1) If B = 0 and N = 0, the record in the chain file is 2 to the G power (2 ^G , where G is the integer value register) following the chain from record A in the chain file. This is a value that represents the number of times the matching processing results have been repeatedly stored in the matching work file 31, which will be described later.

(2) If B=0 and N≠0, this indicates that record A in the chain file is the Nth record in the chain from the starting point record.

(3) If B=1 and N=0, this indicates that record A in the chain file is the 2<G>th record in the chain from record N.

(4) When B=2, it indicates that record A in the chain file is the end point record, and in this case, C=0 and N=0.

In the following explanation, the contents of each field record of the classification work file 41, the matching work file 31, and the result file 21 are described above (A,
For example, the first record in FIG. 4A is represented by (a6, 0, a1, 0).

The present invention can be easily applied to cases where there are multiple types of pointers that define order as shown in FIG. 3, or where there are multiple start point records in a file, so for the chain file 11 shown in FIG.
The details of the procedure will be explained based on FIG.

(1) The litigation control means 5 uses the sequential access means 1 to sequentially read pages from the beginning of the chain file 11 into the main memory, and
For all records, the following (1-1),
Perform either operation (1-2) or (1-3).

(1-1) If record ai is the starting record,
A field record [ai, 0, ai, 0] is generated and output to the result file, and a field record [ai pointer, 0, ai, 1] is generated and output to the classification work file 41.

(1-2) If record ai is the end record,
A field record [ai, 2, 0, 0] is generated and output to the classification work file 41.

(1-3) If record ai is an intermediate record, [ai
A field record of [ai, 1, pointer of ai, 0] and a field record of [ai, 1, pointer of ai, 0] are generated and output to the classification work file 41.

If all the records in the chain file 11 are start records, the process ends; if not, the value of the register G is set to 0, control is passed to the classification means 6, and the process goes to (2).

(2) The classification means 6 uses the classification main storage device 7 and sorts the records in the classification work file 41 in ascending order using the first field, second field, and third field as keys, and the control means 5 Return control to.

(3) The control means 5 reads records from the beginning of the classification work file 41 and repeats any one of the processes (31) to (36) while comparing the records before and after the classification work file 41 until the end of the classification work file 41 is reached. In (31) to (36), Ap, Bp, Cp,
Np indicates the values of the first, second, third, and fourth fields of the p-th record from the beginning of the classification work file 41.

(31) If Ap=Ap+1, Bp+1=1, Np=0, then [Cp+1,0,Cp,0] and [Cp,1,
Cp+1,0] is output to the matching work file 31, and p=p+2. In other words, record Ap in the chain file is 2 G from record Cp in the chain file, and record Cp+1 in the chain file is 2 G from Ap (=Ap+1).
Since it corresponds to the multiplicity, Cp+1 is G of 2 from Cp
This is the +1st power, and information indicating that the number from the beginning of the chain is undetermined is stored for the next G+1st processing.

(32) If Ap=Ap+1, Bp+1, Np≠0, then [Ap, 0, Cp, Np] and [Cp+1, 0,
Cp, Np + 2 to the G power] is output to the matching work file 31, and p=p
+2. That is, Ap is the Npth point along the chain from the starting point field, and
Cp+1 is 2 to the G power from Ap, and Cp+
Since 1 corresponds to the Gth power of Np + 2 from Cp,
Ap is the Npth record from the starting point record Cp, and Cp
+1 stores information indicating that it corresponds to the G-th power of Np+2 from Cp for the next G+1-th processing.

(33) If Ap=Ap+1, Bp+1=2, and Np=0, output the p+1st record to the matching work file 31 and set p=p+2.

(34) If Ap=Ap+1, Bp+1=2, Np≠0, output the pth record to the matching work file 31 and set p=p+2.

(35) Ap≠Ap+1, Bp=Bp+1=0, Np=
If it is 0, p=p+1.

(36) Ap≠Ap+1, Bp=Bp+1=0, Np≠
If it is 0, the pth record is saved in the result file 21.
, and p=p+1.

When the control means 5 reaches the end of the classification work file 41, it performs the following operations.

(37) If the fourth field of the final record is 0, go to (38); otherwise, output the final record to the result file 21 and go to (38).

(38) During the iterative process, the matching file 31
If no records are output to , the process ends, otherwise go to (4).

(4) The control means 5 adds 1 to the register G, copies the contents of the matching work file 31 to the classification file 41, passes control to the classification means 6, and goes to (2).

As described above, in the process of repeating the classification process, the result file 21 contains [own address, 0,
Records in the format [start point record address, relative number from the start point record] are accumulated, and the order of all records in the chain file 11 can be recognized.

The details of the procedure will be supplementarily explained using FIG. Enter the chain file 11 in Figure 2 and perform the above steps.
When (1) is executed, the contents of the classification work file 41 become as shown in FIG. 4A, and the record shown by C in FIG. 4 is output to the result file 21.

When the file A in FIG. 4 is classified according to step (2), the classification work file 41 becomes in the state shown in FIG. 4B. G =
0, when step (3) is executed, a matching work file 31 shown in FIG. 4D is created, and in step (4)
When classification is performed according to step (2) with =1, the classification work file 41 is in the state shown in FIG. 4E.

When G=1, when step (3) is executed, a matching work file 31 shown in FIG. 4F is created, and step
When classifying according to step (2) with G=2 in (4), the classification work file 41 becomes in the state shown in FIG. 4G.

When G=2, executing step (3) creates the matching work file 31 shown in FIG.
When the record shown in FIG. J is added to the result file 21 and classified according to step (2) with G=3 in step (4), the classification work file 41 becomes in the state shown in FIG. 4I.

When G=3, when step (3) is executed, a record shown in FIG. 4K is added to the result file 21, but since no record is output to the matching work file 31, the process is terminated.

By repeating the procedure as described above, the result file 21 contains Figure 4C, Figure 4J, and Figure 4K.
The records indicated by are accumulated. Therefore, the order of each record can be recognized based on the contents of the resulting file.

To summarize the above, all pages in the chain file 11 are read using the sequential access means 1,
By creating a work file consisting of the records necessary to recognize the order of each record, and repeating the classification process [log ₂ N] ⁺ +1 times for these records, all the records in the chain file are Regarding the chain, the order of each record within the chain is recognized.

Here, N is the maximum value obtained by subtracting 1 from the records that make up the chain for all chains included in the chain file, and [ ] ⁺ indicates the integer value rounded down to the decimal point. shall be taken as a thing.

In this way, even if the number of page IOs equal to the number of pages per chain file is the sum of the number of page IOs for the work file, the ordering of all chains in the chain file can be recognized, and [log ₂ N] ⁺ In the process of +1 classification operation, the number of records in the work file decreases rapidly.

In addition, the number of records per page of the work file can be maximized to the extent that the characteristics of the auxiliary storage device allow, reducing the page IO of the work file, and the main memory capacity for classification can be increased to reduce the page IO of the classification work file. It has the advantage of being easy to optimize. In particular, for chain files with many chains spanning pages, the overall page IO can be reduced compared to the conventional method.

Effects of the Invention As described above, the present invention is effective when recognizing the order of all records within the chain at once when a large number of records are stored in a chain file. This is particularly effective when there are many types of pointers that define different ordering within a chain file.

The present invention can be applied to database reorganization processing in which logical relationships between data are expressed in chains, and to inspection of chain connection states.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention, FIGS. 2 and 3 are diagrams showing an example of a chain file, and FIGS. FIG. 2 is a diagram showing the process of recognizing gender. Explanation of symbols of main parts: 1... sequential access means 1, 2... sequential access means 2, 3... sequential access means 3, 4... sequential access means 4, 5... control means, 6...
Classification means, 7...Main memory for classification, 11...Chain file, 21...Result file, 31...Work file for comparison, 41...Work file for classification.

Claims (1)

[Claims] 1 A chain search method that recognizes the order of each record in a chain file, which includes a first step of sequentially reading pages of the chain file into main memory from the beginning of the chain file, and , from the set of the storage address of the record and the pointer in the record, four fields (A, B, C, N) (A and C are the addresses and A field that records a pointer. B is a field that records 0, 1, or 2 depending on the start point, end point, and intermediate record. N
is a field indicating the order of records) and stores it in the classification work file, and the field records in the classification work file are rearranged in ascending order using each field as a key. In the third step, each of the adjacent field records in the classification work file after sorting is compared with each other according to a predetermined rule, and the field records that have a certain relationship as a result of the matching are stored in the result file. and a fourth step of storing the matching results of other field records in the matching work file, and in response to the completion of the fourth step, copying the contents of the matching work file to the classification work file. and a fifth step of repeating the third step to the fourth step, and a fifth step of repeating the third step to the fourth step; With the N field,
A chain search method characterized in that information indicating how many chains from the first record in a chain file can be reached to reach an arbitrary record in a chain file.