JPH0243678A - Cause investigating system - Google Patents

Abstract

PURPOSE:To investigate cell causes to have resulted in an empty set by referring to the final condition of a partial set table when the combination of conditions to be retrieved does not exist and providing a response generating part to prevent the cause why the retrieving result of the combination of the first applied retrieving conditions is the empty set. CONSTITUTION:A partial set table generating part 2 generates the partial sets corresponding to the all combinations of retrieving conditions 1 as a bit column, and a partial set table 3 added with a discriminating symbol to indicate the success of the retrieve is generated on each bit column. Based on the success failure of the retrieve executed by realizing the propagation of the elimination of the two bit columns with the use of a bit operation, a partial set table updating part 7 finds a combination whose retrieving result is discriminated not to be the empty set or to be the empty set, updates data base 6 thereafter. That is, by feeding back a trial result to the partial set table 3 of the retrieving conditions 1, the number of time of trials of the data base retrieve is suppressed to a minimum, and the all causes are detected. Thus, the all causes to have resulted in the empty set can be investigated.

Description

[Detailed Description of the Invention] [Industrial Field of Application] This invention relates to a processing method for determining the cause of an empty set of database search results. In database management systems, when a database search fails, especially when the search result is an empty set, only the fact that the search result is an empty set is communicated to the user, and the cause is discovered by the user through trial and error. This makes it extremely difficult for inexperienced users to use.

In contrast, a method has been proposed in which, when the database search result is an empty set, the database search is repeated with different combinations of search conditions to find out what caused it. For example, Figure 3 is apl from the following document;
an S, J., “Cooperati veResp.
onces From a Portable Nat
ural Language Database Queue
ry System” (In Brody M,a
ndBerwick R, C, (eds,) Comp
utational Models of Discou
rse, The MIT Press, p. 167
-208.1983), the system C○
-Processing method for investigating the cause of empty set adopted in OP (corrective 1direct
In figure 0, 10 is a question from the user as an input, and 11 is a question 10 that is given as an input.
12 is an MQL expression generated by this converter 11, 13 is a subgraph generation unit that cuts out the subgraph from the MQL expression 12, and 14 is a subgraph generation 15 is a search execution unit that performs a database search based on the subgraph 14; 6 is a database to be searched; and 8 is a message to the user based on the search results in the search execution unit 15. 9 represents the response generated by the response generation unit 8.

Next, the operation will be explained. When the MQL converter 11 receives a user's question 10, it analyzes its contents and generates a communication graph called an MQL expression.

In Co-OP, when the set corresponding to the entire graph is empty, it is assumed that some sets corresponding to the subgraph will also be empty, and the subgraph corresponding to the empty set is response as a cue (corrective)
e 1direct responses). Specifically, when the number of nodes included in the MQL expression 12 is n, the subgraph generation unit 13 cuts out subgraphs with the number of nodes from 1 to n-1 from the MQL expression 12 in descending order of the number of nodes, Search execution unit 15
Attempt to search the database 6. If the search result is empty, the response generator 8 generates a response 9 from the subgraph 14 at that point, and this processing method ends. On the other hand, if the search result is not an empty set, the process returns to the subgraph generation unit 13, cuts out the next subgraph, and repeats the database search. If the database search results for all subgraphs are not empty sets, this processing method fails and ends.

[Problem to be solved by the invention] The conventional method is as described above, but the cause is investigated from the subgraph at the point when the search result is an empty set by simply adding the number of nodes one by one. Therefore, there was a problem that if there were multiple causes of failure, it was not possible to investigate all the causes of failure.

This invention was made to solve the above-mentioned problems, and aims to provide a cause investigation method that can investigate all the causes of an empty set.

[Means for Solving the Problems] The cause investigation method according to the present invention obtains all subsets of given search conditions from a combination of search conditions and adds an identification code to each of them to indicate success or failure of the search. a subset table generation unit that generates a subset table based on the above information; a combination extraction unit that extracts a combination of conditions for performing a database search based on the above identification code from this subset table; a command generation/execution unit that generates an access command and executes a search; and a subset table update unit that examines obvious combinations of success or failure in the subset table and updates their identification codes based on the success or failure of the executed search. and a response generation unit that refers to the final state of the subset table when there are no more combinations of conditions to be searched and presents the reason why the search result for the initially given combination of search conditions is an empty set. It is equipped with the following.

[Operation] In the present invention, the subset table generation unit generates subsets corresponding to all combinations of search conditions as bit strings, and also generates subsets with identification codes added to each bit string to indicate success or failure of the search. By generating a table and realizing the propagation of the constraints between the bit strings using bit operations, the subset table update unit can determine that the search result is not an empty set or that there is an empty set based on the success or failure of the executed search. , and update its identification code to prevent unnecessary database searches in the future. That is, by feeding back the trial results to the search condition subset table, the number of database search trials can be minimized and all causes can be detected.

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

In FIG. 1, 1 is a search condition input by the user;
2 is a part in which all subsets of search condition 1 are generated from the combination of given search conditions 1 as bit strings indicating the presence or absence of each element, and identification bits are added to each bit string to indicate the success or failure of the search. A subset table generation unit 4 generates a set table 3; a combination extraction unit 4 extracts a combination of conditions for performing a database search based on the identification bits from the generated subset table 3; and 5 a combination extraction unit 4. Database 6 is created from the combination of extracted search conditions.
A command generation/execution unit 7 generates an access command and executes a search, and a command generation/execution unit 7 performs bit operations on obvious combinations of success/failure in the subset table 3 based on the success/failure of the search executed by the command generation/execution unit 5. A subset table update unit 8 is executed when the combination extraction unit 4 runs out of combinations of search conditions to search the database 6, and updates the identification bits of the subset table 3. a response generation unit that refers to the above and presents to the user the reason why the search results for the initially given combination of search conditions are an empty set;
is the response generated by the response generation section 8.

Next, the operation will be explained.

In the system of this embodiment, a search condition 1, which is a set of elements consisting of pairs of attribute names and conditions for the attribute values, is input to the subset table generation unit 2, and as a subset table 3,
Generate all subsets of the given search conditions. At this time, if the number of elements in the search condition is n, the number of generated subsets is 2'-2. Each of the subsets is represented by n+1 registers as shown in FIG. The first register 20 corresponds to the identification bit, and is set to '1' when it is known that the search for that subset is not the cause of failure.
If you don't know 2, it will be '0', and initially it will be all 101.
is set to . The remaining n registers 2n form bit strings corresponding to combinations of conditions, and are set to '1' if the subset includes each of the n search conditions, and '0' otherwise. The combination extraction unit 4 has a counter, and the counter indicates how far in the subset table 3 the search has been performed until now. The combination extraction unit 4 extracts the combination of search conditions (the bit value of the first register 20 is tOj) for the next database search from the 9-subset table 3 according to the counter and the bit value of the first register 20. do. The command generation/execution unit 5 generates a search command based on the combination of search conditions extracted by the combination extraction unit 4, and actually searches the database 6.

The subset table update unit 7 updates the two-subset table 3 according to the success or failure of the search by the command generation/execution unit 5. If the search by the command generation/execution unit 5 is successful, the bit string (
(excluding the first register 20), the register 20 before X is set to '1', and the register 20 before P1 to column Y where XAY=O is set to '1'. At this time, it is only necessary to check the first register 20 of Y and those below X. Note that if the search by the command generation/execution unit 5 fails, the register 20 in front of X is left as '0'.

The above updating ends when there are no more combinations of search conditions to perform the next database search, that is, when the counter of the combination extraction section 4 reaches 2n-2, and the response generation section 8 is executed. In this case, the fact that the first register 20 of the subset table 3 is '0' causes the database search to fail.

Therefore, the response generation unit 8 generates a response by referring only to the first register 20 of this subset table 3 that is '0'.
Present to the user.

In this way, by feeding back the search results attempted in investigating the causes of database search failures to the search condition subset table, all the causes of failures can be quickly investigated with the minimum number of trials required. Furthermore, it is possible to implement a cause investigation method on an existing database management system when the database search result is an empty set.

In the above embodiment, combinations were extracted from the top to the bottom of the subset table 3 shown in FIG. By examining combinations that are clear that
It is also possible to extract combinations from the bottom to the top, that is, from those with few elements to those with many elements.6 In addition, in the above embodiment, in order to narrow down the combination candidates,
Although it does not use algorithms that depend on the structure of the database or heuristics (rules of thumb) specific to the field to which the contents of the database belong, they can be used in conjunction with the method of this invention to create more efficient failure causes. It is possible to realize an investigation method.

In addition, in the above embodiment, the case of investigating the cause of a database search failure was explained, but it can also be If this method is applied to reduce the number of tests and formulate an efficient experimental plan, the same effects as in the above embodiments will be achieved.

[Effects of the Invention] As described above, according to the present invention, all subsets of a given search condition are obtained from a combination of search conditions, and an identification code is added to each subset to indicate success or failure of the search. A subset table generation unit that generates a set table, a combination extraction unit that extracts from this subset table a combination of conditions for performing a database search based on the above-mentioned identification code, and access to the database from the extracted combination of search conditions.・A command generation/execution unit that generates a command and executes a search, and a subset table update unit that examines obvious combinations of success or failure in the subset table and updates their identification codes based on the success or failure of the executed search. and a response generation unit that refers to the final state of the subset table and presents the reason why the search result for the initially given combination of search conditions is an empty set when there are no more combinations of search conditions to be searched. Therefore, it is possible to investigate all causes quickly with the minimum number of trials required.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing a cause investigation method according to an embodiment of the present invention, FIG. 2 is a diagram showing a subset table of search conditions used in the embodiment, and FIG. 3 is a diagram showing a conventional cause investigation method. be. 1 is a search condition, 2 is a subset table generation unit, 3 is a subset table, 4 is a combination extraction unit, 5 is a command generation/execution unit, 6 is a database, 7 is a subset table update unit, 8 is a response generation unit , 9 is the response, 20 is the first register (identification code), 2n
is n registers (subset). In addition, in the figures, the same reference numerals indicate the same or equivalent parts. Agent Masuo Oiwa (and 2 others) 3rd Ward 20.2n Register

Claims (1)

[Claims] When the database search result is an empty set, the database search is attempted by changing the combination of search conditions and the cause is investigated. A subset table generation unit that obtains a set and generates a subset table in which an identification code is added to each of the sets to indicate the success or failure of the search, and a combination of conditions for performing a database search based on the identification code from this subset table. a combination extraction unit that extracts; a command generation/execution unit that generates a database access command from the extracted combination of search conditions and executes the search;
A subset table update unit that checks the obvious combinations of success or failure in the subset table based on the success or failure of the executed search and updates their identification codes, and a final update unit of the subset table that updates the identification code when there are no more combinations of conditions to be searched. 1. A cause investigation method comprising: a response generation unit that refers to a state and presents the reason why a search result for a combination of initially given search conditions is an empty set.