JPH0573623A - Learning type data base processing system - Google Patents

Abstract

PURPOSE:To provide a learning type data base processing system which can execute a data base processing only by means of describing the restriction relation of data and which can execute a high speed and secure processing. CONSTITUTION:A data base is stored in a data base holding means 11 and a restriction holding means 12 stores restriction expression. When the realization of the restriction relation is inspected and a restriction inspection means 13 indicates a unit to be corrected, a unit selection means 14 selects one unit and a sequence storage means 15 stores the unit to be corrected from the relation of restriction expression. A restriction holding means 12 is caused to indicate the unit to be corrected with a trigger as an opportunity. The name of the unit which the unit selection means 14 selects and the sufficient situation of a restriction rule are stored in the sequence storage means 15. When an operation similar to that of the trigger is executed again, a comparison means 16 compares them. The data base holding means 11 is corrected in accordance with a sequence stored in the sequence storage means 15 as long as the sufficient state of the restriction rule coincides.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a learning type database processing system, and more particularly, to a learning type database processing which enhances the correctness of the processing when the database processing is executed by describing only the mutual constraint relations of data. Regarding the system.

[0002]

The relational database model is widely used for managing large amounts of data. In the relational database model, all data is managed as a table called "relation".

FIG. 2 shows an example of a relation. In the figure, (A) shows an employee relation, (B) shows a single transfer notification relation, and (C) shows a support application relation. The vertical direction of the relation in the figure, for example, "employee code", "name", etc., is called an attribute. In addition, the horizontal direction of the relation, for example, one line corresponding to the employee code “0001” is called a “tuple”. There are 7 attributes and 3 tuples in the “employee” relation of FIG. The order of this tuple has no meaning. The tuple of the employee code “0001” is composed of seven pieces of data, which are called “attribute values”. Further, among the attributes of the relation, there is at least one attribute that uniquely determines the tuple in the relation by determining the value of the attribute, and this is called a "key attribute". The key attribute of the “employee” relation in (A) is “employee code”.

The information processing system executes processing such as adding a new tuple, deleting an old tuple, or updating data represented by a relational database model that rewrites attribute values in the tuple. .. Conventionally, in this type of information processing system, it is an application program side to consider which tuple should be updated, and the database management system has only a function of storing data.

In the conventional database management system, the decision as to which tuple should be updated is made by a program, that is, it is realized as a procedure according to an input from an external user. Therefore, if there is any modification to the program, it is necessary to review the entire procedure and it is necessary to generate all paths in advance, including rare cases. Has become.

In order to solve such a problem, Japanese Patent Application No.
-228983 uses a constraint expression in which the presence or absence of a tuple in the database and the attribute value in the tuple are unit variables,
Further, in order to speed up the processing, a declarative database processing method is disclosed in which the sequence of the previous processing is stored and used when the same processing is requested. The constraint here is a rule that describes a semantic relationship that data in the database must satisfy each other.

FIG. 4 shows an example in which the rule expression of constraints is shown as a table. The rule expression conforms to the format generally used in expert system construction support tools. In the same figure, (A) is a restriction on transfer assignment, (B) is a restriction on transfer status,
(C) is a rule expression regarding restrictions on dependent applications. According to the rules of the expert system construction support system, "?
A value added with “?” such as “x” or “? y” is called a “rule variable”, and the value of the data item is bound in the rule.

For example, the rule of the first line in the single transfer assignment restriction shown in FIG. 4A has the following meaning.
“A tuple exists in the employee relation (for the specified key attribute, ie employee code)
And there is a tuple in the single assignment report relation and the value of the "assignment" attribute of the tuple in the single assignment report relation = "address" of the tuple in the single relation and employee relationship
Attribute value =? y And the value of the "address" attribute of the tuple in the single assignment report relation =? x
And the value of the "address 1" attribute of the tuple in the employee relation =? The relationship of x may be established. The meaning of the above rule is that when there is an employee with the employee code in the employee relation, if the value of the attribute of "transfer" is "single", the tuple of the employee exists in the single transfer report relation. The value of the attribute of the "homecoming destination" of the tuple in the single assignment report relation (hereinafter referred to as the "single assignment report tuple") is the tuple in the employee relation The value of the “address” attribute of the single assignment report tuple is equal to the value of the “address” of the employee tuple. Here, "address 1" means the place where the family lives. As described above, the constraint rule may be any rule as long as it represents the mutual relation of attribute values.

In the above-mentioned "learning type database processing system" (Japanese Patent Application No. 3-228983), the constraint relation of FIG. 4 is corrected when the constraint relation with the database is checked and the constraint is not satisfied. It is verified by the constraint checking means having the function of presenting the unit to be processed. As the value of each attribute given to the constraint relation of FIG. 4, the value in the database is normally used as it is, but the constraint may not be satisfied if the value in the database is left as it is. It is necessary to temporarily change the attributes. Therefore, the unit is prepared to store the attribute value given to the constraint. FIG. 3 shows an example of the unit. The "unit" is a term widely used in the constraint satisfaction problem as a meaning of a variable that determines a value so as to satisfy the constraint relationship.

In FIG. 3, "employee.tuple" is a unit for displaying whether or not a tuple exists in the "employee" relation under a given key. The “employee.address” is a unit that displays whether or not tuples exist in the employee relation. Also, "employee.address"
Corresponds to the value of the "address" attribute of a tuple in the employee relation. Typically, these units correspond to database values, but some of the unit values may change. What is described as “possible value” in FIG. 3 indicates a range of values that may be substituted in the unit. Further, “(input value)” means that a value candidate cannot be specified in advance and must be acquired from the user of the system and input. "Nul
"L" means that no value is explicitly declared and cannot be entered. For example, "employee.name1" = Null
Means that the employee does not have a family equivalent to dependents.

FIG. 5 shows the unit value in the initial state. For example, it is assumed that the unit value shown in FIG. 3 is specifically the value shown in FIG. The values in FIG. 5 correspond to the waiting state in which the system is not performing any operation. In this case, "employee.tuple""single transfer notification. Tuple""support application. Tuple"
Is all "none". Further, in FIG. 4, there are three restrictions: (A) single transfer notification restriction, (B) transfer status restriction, and (C) dependent application restriction.
Although the kind restrictions are described, "Employee.tuple" in Figure 4
Since “single assignment report. Tuple” and “support application. Tuple” are “none”, the constraint rules at the bottom of FIGS. 4A, 4B, and 4C are satisfied. Each constraint in FIG. 4 has only to be satisfied in the horizontal direction (row), so that the unit values in FIG. 5 all satisfy the constraint.

FIG. 6 shows the unit value after the employee tuple is acquired. Here, it is assumed that an employee tuple is added as shown in FIG. In this case, for example, when the assignment status constraint of FIG. 4B is checked, the bottom line does not satisfy the constraint. In the remaining three lines, it is necessary to check which constraint is satisfied after acquiring values such as "employee.address".

Further, FIG. 7 shows unit values after data acquisition. If the unit value of FIG. 5 changes to the unit value of FIG. 7, the constraint is satisfied. The role of the constraint checking means is to check whether or not the constraint is satisfied corresponding to the unit value.

Another important function is added to the conventional constraint checking means. That is, it is known as a value that can be taken in advance by deciding a unit value whose value should be decided next when the result of the check of the constraint checking means shows that the constraint is unsatisfied or the constraint cannot be satisfied. There is a function to select an appropriate value from the ones. Furthermore, there is a function to change some unit values and search for a state where the constraint is satisfied. In the case where the value of the unit is not determined in advance and the input value from the outside (represented as “(input value)” in FIG. 3), the constraint checking means may be the constraint checking means itself. There is a function to instruct the system user to acquire the value, even if the module is requested to acquire the value.

If some of the unit values have unacquired data and the constraint inspecting means cannot execute the inspection, the constraint inspecting means instructs acquisition of the uninputted unit value and the unacquired unit. Determine the value. In this way, the constraint checking means presents a plurality of units to be modified or to newly acquire data.

In the example of the assignment status constraint of FIG. 4B,
In order to satisfy this constraint, in the unit value after the employee tuple acquisition shown in FIG. 6, write back “Employee.tuple” to “none”: “Employee.address” “employee.address1” values one after another And determine the value of “employee.transfer”: The following two are derived from the constraints in Figure 4. In this case, a person immediately determines that it is correct, but this cannot be determined only from the constraints shown in FIG. Furthermore, when “employee.transfer” becomes “single” by the above data acquisition, the single transfer notification restriction in FIG. 4A becomes unsatisfactory. To satisfy this constraint, rewrite "Employee.tuple" to "No" in Fig. 6: Rewrite "single transfer. Tuple" to "Yes": Rewrite "employee.transfer" to "accompanied" : There are three ways. However, it is not possible to judge which of the selections is semantically correct only from the constraint table of FIG. Therefore, it is necessary for a human to instruct the operation of rewriting “single assignment report. Tuple” to “present”. As a result, the "single assignment report restriction" will be satisfied. Figure 4
(C) Description of "restriction application restrictions" is omitted.

In the conventional "learning type database processing system" (Japanese Patent Application No. 3-228983), a unit selecting means is provided in order to allow the user to select one of these plural modifiable units. The selected unit name is stored in the sequence storage means. By storing the sequence in this way, when processing for exactly the same tuple is required at a later date, the processing can be speeded up by using the stored sequence.

[0018]

However, in the conventional "learning type database processing system", when the sequence stored in the sequence storing means is reused,
There is a problem that the exact same value as the last time is not always input by the user.

An example will be described. There is a “TEST condition” in (b) “assignment status constraint” in FIG. 4 and the value of the “address” attribute of the “employee” tuple and “address 1” of the “employee” tuple.
Depending on whether the values of the attributes match or do not match, the constraint rules that are satisfied (each constraint in the horizontal direction is called a constraint rule) will differ. Here, it is assumed that “? X” and “? Y” match at the time of the previous processing. In this case, the value of the “assignment” attribute of the “employee” tuple is “accompanied” due to restrictions.
Becomes In the previous processing, the value of “single transfer report. Tuple” is finally “none” due to the restriction on single transfer report in (A) of FIG.
It is assumed that the processing has ended. The sequence of acquisition of these data is stored in the sequence storage means.

Here, the same processing is triggered for another employee, and as a result, the sequence storage means in which the value of the attribute of “employee.address1” of “employee.tuple” is stored is stored. Suppose that it will be acquired according to the previous sequence. But this time, unlike last time,
Assuming that the target person is a single employee, the rule variable "? X"
And "? Y" have different values. Therefore, since the value of the “assignment” attribute of the “employee” tuple is “single”, the value of the “single assignment report. Tuple” is finally set to “single assignment report tuple” in FIG. 4A. Need to be treated as "Yes". However, in reality, this is not the case, and since the previously stored sequence is used as is,
The value of “single assignment report. Tuple” is set to “none”, and the validity of the process is not guaranteed. This does not mean that the same employee as the previous time is being processed, so even if the value of the acquired data may be different, the same processing can be performed without performing the constraint check. This is because

As a measure to cope with the problem of the above-mentioned conventional learning type database processing system, all the previously acquired data are stored, and as long as the values match the previous values, the sequence storing the processing is stored. If the matching state collapses, the constraint checking means is fully operated, and thereafter, which unit is corrected to switch the operation to select which unit satisfies the constraint by the unit selecting means. It However, this method cannot tolerate the diversity of the actual society in which each person has a different address, and has a problem of impairing the range of use of the learning database processing system having the learning effect of speeding up the processing.

The present invention has been made in view of the above points,
By declaratively describing only the mutual relations of data,
It is an object of the present invention to provide a learning type database processing system capable of performing database processing, operating at high speed, and performing reliable processing.

[0023]

A database holding means for storing a database based on a relational database model, the presence / absence of tuples stored in the database holding means, and a constraint expression in which an attribute value for each tuple is a unit variable. Check whether the relation between the constraint holding means to be stored and the constraint expression for the database is satisfied,
When the relation of the constraint expression is unsatisfied, the constraint checking means for presenting one or more units to be corrected and the plurality of units to be corrected designated by the constraint checking means Corresponding to the unit selection means for selecting one unit and the trigger for adding or deleting the first tuple that first transits the unsatisfied database state satisfying the constraint expression relation to the database holding means. , A unit to be corrected until the relation of the constraint expressions is finally satisfied, a sequence storage unit that stores the satisfaction state of each constraint rule in the constraint as a sequence, and the satisfaction state of the constraint rule stored in the sequence storage unit And a comparing means for comparing the satisfaction state of the constraint rule presented by the constraint checking means, and the constraint checking is triggered by the trigger. The unit to be modified is presented by means, and when there is one unit to be modified, the unit is selected, and when there are a plurality of units to be modified, the unit selected by the unit selection means is sequenced together with the satisfaction state of each constraint rule. When the same operation as the trigger is performed again when stored in the storage means, and the comparison results match by the comparison means, the database holding means is modified according to the unit sequence stored in the sequence storage means.

[0024]

In the learning-type database processing system of the present invention, the constraint that is satisfied is changed to correspond to the trigger of the operation of adding / deleting the first tuple that is changed to the unsatisfied state, and finally the constraint is satisfied. In addition to the conventional learning type database processing of storing a unit sequence to be corrected until it is satisfied again and using this stored sequence for processing again when the same type of trigger occurs again, What is stored in the sequence storage means is not only the unit name to be corrected, but also the satisfaction status of the constraint rule satisfied by the modification is stored together.
Even if the stored sequence is used for processing by the same trigger, the constraint relation is always checked, and the comparison means compares whether the satisfaction state of the constraint rule is the same as the stored satisfaction state. However, as long as the satisfaction states of the constraint rules match, the processing is performed according to the sequence stored in the sequence storage means.

As a result, the amount of calculation is small if only the checking is executed by the constraint checking means, but it is extremely large calculation to check which unit is modified to make a transition to satisfy the constraint. Need quantity. Even if the stored sequence is used, the operation of the present invention allows the constraint checking means to check whether the satisfaction state of the constraint rule is the same as the last time the sequence was stored, so Even if different data is input, the operation can be performed within the range where the processing is semantically guaranteed by the constraint. Further, by abandoning the sequence stored at the time when the transition of the constraint satisfaction state is different from the previous time, it is guaranteed that when the sequence finally ends, the constraint always stops in the satisfied state.

[0026]

1 is a block diagram showing the structure of a learning type database processing system according to an embodiment of the present invention. The learning type database processing system 1 is a database holding means 1
1, a constraint holding unit 12, a constraint checking unit 13, a unit selecting unit 14, a sequence storing unit 15, and a comparing unit 16.

The database holding means 11 holds data according to a relational model as shown in FIG.

The constraint holding unit 12 stores the presence / absence of tuples in the database holding unit 11 and the constraint expression in which the attribute value in each tuple is a unit variable. Constraint holding means 1
The constraint held in 2 is a rule that describes the semantic relationship that the data in the database holding means 11 must satisfy each other. The rule expression shown in FIG. 4 conforms to the format generally used in the expert system construction support tool, but what kind of description can be used as long as it describes the constraint relationship between data (between attributes of the relational model)? It can be in the form. However, the present invention is premised on that a key attribute (employee code in this embodiment) for checking whether or not a tuple exists in each relation is designated in advance. The key attribute may be different for each relation.

The constraint checking means 13 has a function of verifying the constraint relationship shown in FIG. 4 and a function of deciding and presenting a unit to be corrected or input required when the constraint is not satisfied. ..

The unit selecting means 14 is the constraint checking means 1
If there is more than one unit to be modified presented by 3, select one of them. For example, the unit selection means 14 rewrites the following "employee.tuple" from the constraint inspection means 13 to "none": "single transfer. Tuple" to "present": "employee.transfer" to "accompanied" Rewrite: When three kinds of selection items of are presented, the user is made to select one of them. The unit selection means 14 specifically has a function of presenting to the user which unit should be corrected; acquiring the unit name selected by the user; transmitting the processing of the unit to the sequence storage means; .. If there is only one type of processing to be corrected according to the inspection result of the constraint inspection means 13, basically, the processing to the database is executed without waiting for the operation of the unit selection means 14, and the sequence storage means 15 is executed.
It suffices to store the processing content in the. But,
Even if you only have one option,
It is possible to operate the unit selection means 14 and make the user confirm.

FIG. 8 is a diagram for explaining the operation of the unit selecting means 14. In order to allow the user to select the corrective action, at least in the case of adding or deleting a tuple, the tuple to which the tuple is added or deleted; is displayed to the user. In FIG. 8, the information contents are displayed on the CRT2.
It is displayed above 0. Items to be displayed on the CRT 20 when processing tuple attribute values are-for which tuple; -for which attribute; -if the attribute value is known in advance as a candidate attribute value; Display on top. Since the function of selecting the item displayed on the CRT 20 is required, the user selects this with the cursor 23 and the mouse 24 in FIG. The user selects a candidate 22 for each process according to the message guide in the message area 21 displayed on the CRT 20.

The sequence storage means 15 associates with the trigger, which is an operation of adding / deleting the first tuple that changes the satisfied constraint and makes a transition to the unsatisfied state, and finally the constraint is satisfied again. The unit sequence to be corrected is stored. Further, the sequence storage unit 15 stores not only the unit name to be corrected but also the satisfaction state of the constraint rule after the unit correction.
In the above example, the sequence selected by the user when "Add Employee.tuple" is triggered Step 1: Acquire "Employee.Address" Step 2: Acquire "Employee.Address 1" Step 3: "Single employee transfer. Tuple" Is rewritten to “present” Step 4: Acquisition of “employee.name1” Step 5: Acquisition of “employee.relationship1” is stored. By these steps, the constraint relation is satisfied again. The sequence of the processing described above is stored in the sequence storage means 15 in association with “addition of employee tuple”. Then, when the same process “addition of employee tuple” is performed again, the stored sequence is reproduced.

In this case, the constraint checking means 13 is used only for checking the constraint. The resulting satisfaction state of the constraint rule is compared with the record of the satisfaction state of the constraint rule stored in the sequence storage unit 15 and the comparison unit 16, and as long as they match, the sequence storage unit 1
The process is executed according to the sequence stored in 5.
As described above, it is necessary to store the sequence and the satisfaction status of the constraint rule triggered by various tuple triggers. When a trigger is generated, the stored sequence is generated one after another according to the type of the trigger, so that the database processing is executed.

FIG. 9 is a flow chart showing the processing of the sequence storage means 15 of one embodiment of the present invention. Step 100; When the constraint is unsatisfied by adding / deleting tuples from the condition that the constraint is satisfied, the sequence storage area is first examined. It is determined whether the tuple name and the type of addition / deletion are already registered in the sequence storage area. If they are not registered, the process proceeds to step 106. Step 101; If the name of the tuple and the processing type (addition or deletion) of the tuple are already stored, it is determined whether the sequence exists in the sequence storage area, and if there is no sequence to be read, the process ends. Step 102; If the sequence is in the sequence area, read one processing sequence. Step 103: The processing is executed according to the read sequence to change the unit value or acquire it from the user. Step 104: Whether or not the satisfaction state of the constraint rule stored in the sequence storage unit 15 and the constraint rule satisfaction state are the same is compared by the comparison unit 16, and the process is executed according to the sequence unless the match is broken. Step 105: The satisfaction state of the constraint rule stored in the sequence storage means 15 is compared with the comparison means 16, and if the coincidence is broken, the processing by the sequence stored in the sequence storage means 15 is stopped. Step 106; if the tuple name and the type of addition / deletion are not registered in the sequence storage area, the tuple name and the
Register the types of "add" and "delete". Step 107; subsequently, the operation sequence (satisfaction state of the unit and the constraint rule) instructed by the constraint checking means 13 is stored until the constraint is in the satisfaction state.

FIG. 10 shows a sequence storage area of the sequence storage means 15 according to one embodiment of the present invention. Sequence storage area is tuple name, distinction between addition and deletion, sequence,
The satisfaction state of the constraint rule is stored. In the example of the figure, the tuple name is the “employee” tuple and the trigger type is “addition”. Regarding the sequence, (1) Acquire the attribute of the "address" of the "employee" tuple, and check the satisfaction status of the constraint rule in that case (False, True / False undetermined, False) Commit, true / false undetermined, false), (false, true / false undetermined, false), etc. are stored.

The processing sequence in the case where the employee tuple serving as a trigger is added to the single-employed employee stored in FIG. 10 will be specifically described below. It should be noted that the satisfaction status of the constraint rules only needs to have information that can determine the satisfaction status of each rule, and only the rules satisfying the restrictions are stored in the sequence storage unit 15. In this example,
Since the truth value of the rule has three values (true, false, and true / false), each rule satisfaction state of the constraint table is represented by a list. This will be described with reference to FIGS. 4 and 6. In FIG. 4, the rule number is “1a, 3” on the right side of each constraint table.
b ”and the like.

First, the learning operation of the sequence storage means 15 will be described. The states of the constraint rules corresponding to the unit variable states of FIG. 6 are as follows. [Restrictions on single-person transfer] (1a, 1b, 1c) = (False,
True / False undecided, False) [Reppointment status constraint] (2a, 2b, 2c, 2d) = (True / False undetermined, True / False undetermined, True / False undetermined, False) [Dependent application constraint] (3a, 3b, 3c) = (False, True / False undetermined, False) Next, (1) the "address" attribute of the "employee" tuple is acquired.
There is no change in the satisfaction status of the constraint rule. [Restrictions on single-person transfer] (1a, 1b, 1c) = (False,
True / False undecided, False) [Reppointment status constraint] (2a, 2b, 2c, 2d) = (True / False undetermined, True / False undetermined, True / False undetermined, False) [Dependent application constraint] (3a, 3b, 3c) = (False, True / False undetermined, False) Next, (2) the "address 1" attribute of the "employee" tuple is acquired. As a result, the two rule variables "?
Since the values of “x” and “? y” are fixed, the satisfaction state of the constraint rule is changed and becomes as follows. [Restrictions on single-person transfer] (1a, 1b, 1c) = (False,
False, false) [Relocation assignment constraint] (2a, 2b, 2c, 2d) = (True, False, False, False) [Dependent application constraint] (3a, 3b, 3c) = (False, True / False undetermined, False ) Next, (3) the value of the "single assignment" tuple is rewritten to "present". As a result, the satisfaction state of the constraint rule is as follows. [Restrictions on single-person transfer] (1a, 1b, 1c) = (true,
False, false) [Relocation assignment constraint] (2a, 2b, 2c, 2d) = (True, False, False, False) [Dependent application constraint] (3a, 3b, 3c) = (False, True / False undetermined, False As a result, the "single assignment report constraint" and "assignment status constraint" are satisfied (at least one constraint must be "true" in the multiple rules that make up the constraint table).
Therefore, it is necessary to satisfy the remaining "restriction application restrictions".
Among the rule variables of "Dependent application restrictions", "? Y" and "?
Since the value of “w” is fixed, the “name 1” attribute of the “employee” tuple is acquired. As a result, the following satisfied state is achieved. [Restrictions on single-person transfer] (1a, 1b, 1c) = (true,
False, false) [Restriction on assignment] (2a, 2b, 2c, 2d) = (True, False, False, False) [Dependent application restriction] (3a, 3b, 3c) = (False, False, False) "Dependent In order to satisfy the “application restrictions”, “supporting application”
The tuple value must be "Yes". By this unit value correction, the following sufficiency will be achieved. [Restrictions on single-person transfer] (1a, 1b, 1c) = (true,
False, false) [Restriction on assignment status] (2a, 2b, 2c, 2d) = (true, false, false, false) [Restriction application restriction] (3a, 3b, 3c) = (true / false undetermined, false, false) ) Hereinafter, the "relationship 1" attribute of the "support application" tuple, which is the rule variable of "? Z", is acquired. The state of satisfaction of the constraint rules after acquisition is as follows, and all constraint tables are satisfied. [Restrictions on single-person transfer] (1a, 1b, 1c) = (true,
False, false) [Restriction on assignment status] (2a, 2b, 2c, 2d) = (true, false, false, false) [Restriction application constraint] (3a, 3b, 3c) = (true, false, false) The sequence is stored in the sequence storage means 15.

Next, the operation of utilizing the sequence storage means 15 will be described. It is assumed that an employee tuple is added to another employee as an operation when the same trigger as that of the sequence storage unit 15 that has performed the learning operation occurs.
Since the tuples of the same type are added by the operation of the sequence storage means 15 (FIGS. 9 and 10), the sequence is read from the sequence storage means 15 and used. However, this time, with the "address" attribute of the "employee" tuple,
The "address 1" attribute of the "employee" tuple has the same value (that is,
The wife of this employee lives together).

An employee tuple that becomes a trigger is added,
The satisfied state of the constraint rule corresponding to the unit variable state of FIG. 6 is as follows. This constraint rule satisfaction state is not different from the satisfaction state stored in the sequence storage unit 15. [Restrictions on single-person transfer] (1a, 1b, 1c) = (False,
True / False undecided, False) [Reppointment status constraint] (2a, 2b, 2c, 2d) = (True / False undetermined, True / False undetermined, True / False undetermined, False) [Dependent application constraint] (3a, 3b, 3c) = (False, True / False undetermined, False) Next, according to the sequence stored in the sequence storage unit 15, the "address" attribute of the "employee" tuple is acquired. There is no change in the satisfaction status of the constraint rule, which is the same as the stored satisfaction status of the constraint rule. [Restrictions on single-person transfer] (1a, 1b, 1c) = (False,
True / False undecided, False) [Reppointment status constraint] (2a, 2b, 2c, 2d) = (True / False undetermined, True / False undetermined, True / False undetermined, False) [Dependent application constraint] (3a, 3b, 3c) = (False, True / False undetermined, False) Next, according to the sequence stored in the sequence storage unit 15, the "address" attribute of the "employee" tuple is acquired. As a result, since the values of the two rule variables “? X” and “? Y” of the “assignment status constraint” are determined, the satisfaction state of the constraint rule changes and the following occurs. [Restrictions on single-person transfer] (1a, 1b, 1c) = (False,
False, false) [Restriction on assignment status] (2a, 2b, 2c, 2d) = (False ^* , True ^* , False, False) [Dependent application restriction] (3a, 3b, 3c) = (False, True / False undetermined) , False) The satisfaction state of this constraint rule is the sequence storage means 15
It is different from the one stored in (the different parts are marked with *).

Comparison means 1 in the flow chart of FIG.
This mismatch is detected by 6. As a result, unlike the conventional technique, the use of the sequence in the sequence storage means 15 is abandoned at the stage of mismatch. further,
The constraint checking unit 13 presents a plurality of candidate units to be modified, and the unit selecting unit 14 returns the operation to the learning operation for allowing the user to select one of them. if,
If the processing is continued without using the comparison means 16, even if a single transfer notification is submitted to an employee who is not originally transferred, the inconsistent phenomenon that the address of the single transfer destination is the same as the wife's address. Occurs.

The sequence storage area of the sequence storage means 15 may be secured in a normal RAM area or may be represented by dedicated hardware.

In this embodiment, the sequence storage means 1
Although the information stored in 5 stores only the rules satisfying the constraints, a method of storing the rules not satisfying the constraints may be conversely considered.

[0043]

As described above, according to the present invention, the database processing is realized by using only the declaratively expressed knowledge, so that the program creation is simple and the addition / update of the tuples of the same kind is performed. -When deletion or the like is performed, database processing can be performed in a short time, and semantically erroneous processing can be prevented.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a learning database processing system according to an embodiment of the present invention.

FIG. 2 is a diagram showing an example of a relation.

FIG. 3 is a diagram showing an example of a unit.

FIG. 4 is a diagram showing an example of constraints.

FIG. 5 is a diagram showing a unit value in an initial state.

FIG. 6 is a diagram showing a unit value after acquisition of an employee tuple.

FIG. 7 is a diagram showing a unit value after data acquisition.

FIG. 8 is a diagram for explaining unit selection means.

FIG. 9 is a flowchart showing a process of a sequence storage means according to an embodiment of the present invention.

FIG. 10 is a diagram showing a sequence storage area of a sequence storage means according to an embodiment of the present invention.

[Explanation of symbols]

 1 Learning Type Database Processing System 11 Database Holding Means 12 Constraint Holding Means 13 Constraint Checking Means 14 Unit Selecting Means 15 Sequence Storage Means 16 Comparing Means 20 CRT 21 Message Area 22 Processing Candidate 23 Cursor 24 Mouse

Claims (1)

[Claims] 1. A database holding means for storing a database based on a relational database model, presence / absence of tuples stored in the database holding means, and a constraint expression having an attribute value for each tuple as a unit variable are stored. A constraint check that checks whether the constraint holding unit and the constraint expression with respect to the database are satisfied, and presents one or more units to be modified when the constraint expression has a non-satisfied relationship. Means, a unit selecting means for selecting one unit from the units when there are a plurality of units to be modified designated by the constraint checking means, and a database state satisfying the relation of the constraint expression Is used as a trigger for adding or deleting the first tuple that first transits to unsatisfied Accordingly, a unit to be corrected until the relation of the constraint expressions is finally satisfied, and a sequence storage unit that stores the satisfaction state of each constraint rule in the constraint as a sequence, and the unit stored in the sequence storage unit. A sequence of satisfaction states of the constraint rule and a comparing unit for comparing the satisfaction state of the constraint rules presented by the constraint checking unit are provided, and the unit to be corrected by the constraint checking unit is triggered by the trigger. To be presented and the unit to be modified is 1
If the number of units to be modified is more than one, the unit selected by the unit selection unit is stored in the sequence storage unit together with the satisfaction state of each constraint rule, and the same operation as the trigger is performed again. The learning type database processing system is characterized in that, when the comparison result is matched by the comparing means, the database holding means is modified according to the unit sequence stored in the sequence storing means.