JP4623446B2 - Data management program and data management system - Google Patents

Description

The present invention extracts a part of data stored in a computer in response to a request from a user, or further transmits the extracted data through a network, presents it to a user, a storage medium, etc. In a system that performs storage, etc., by removing, transmitting, presenting, storing, etc. with redundancy in the data removed, it is possible to reduce the cost of retrieval processing, transmission processing, storage area, etc. It relates to a program and system for improving browsing convenience.

In response to a data retrieval request from a user, a data management system that extracts a set of data that is a solution to the retrieval request from the managed data, that is, a data management system that extracts the solution set, removes duplicates in the solution set The following four classification axes can be considered for the system that considers.

(1) If only duplicates where the same data appears multiple times as a solution are considered, or if the target data has a nested structure that can be expressed as a tree structure, a solution is a part of another solution Do you consider duplicates that appear as trees?

(2) Check the retrieved data to find and remove duplicates, or analyze the given data retrieval request and synthesize these retrieval requests so that the data is retrieved without duplicates. , Do you remove duplicates by using these to extract data?

(3) Consider only the duplication in the solution set for one data retrieval request, or if multiple data retrieval requests are received at the same time, consider the duplication between the solutions for these different requests. Or.

(4) Whether or not it has a function of synthesizing a solution set including duplicates as required from the original data as needed.

There are 16 possible cases from the combination of these four classification axes. Among them, with respect to (1) of the classification axis, there are prior arts for some of the eight types classified into those that only consider duplication that the same data becomes a solution multiple times.

First, for a system that considers only one retrieval request by examining the retrieved data and finds duplicates, the retrieval results are obtained according to the retrieval request from the user in Patent Document 1 and Patent Document 2 below. In the system that presents, the contents of the search results are examined before presenting, and when the same data is included multiple times, one of these is deleted and deleted. Similarly, Patent Document 1 mentioned below also points out that multiple inquiries are taken into account by examining the extracted data and finding duplicates.
JP 07-295994 A JP-A-10-143523

In addition, as an example of a technique that considers only one request by modifying the data retrieval request to eliminate duplication, the relational database is “a set of employee numbers of employees over the age of 60”. , SQL search expression representing a request that “take out the sum of a set of employee numbers of employees who have worked for 40 years or more among employees”:
(Select employee number from employee where age> = 60)
union
(Select employee number from employee where years of service> = 40)
Is an SQL search expression that expresses a request to “retrieve a set of employee numbers of employees who are over 60 years old or whose service years are over 40 years”:
select employee number from employee
where age> = 60or years of service> = 40
There is a widely known method that eliminates the possibility of the same employee taking out twice by transforming it into

Non-patent document 1 discloses a technique called “reminder query” as an example of a technique that considers a plurality of requests by modifying a data retrieval request to eliminate duplication. This is because if there was a request to “take out a set of employee numbers of employees over 60 years of age” before, this request content and the resulting data are saved, and then “age 50 If there is a request to “take out the set of employee numbers of employees over the age of”, this is taken out only for the difference from the solution set of the previous request. “Employees who are over 50 and under 60 The data is extracted by transforming it into a request to “take out the set of employee numbers of the employees”, and the result is obtained by taking the set sum of the stored employee numbers of 60 or more employees. This is a method for generating a set of employee numbers of employees. This eliminates the process of repeatedly retrieving information on 60 or more employees that have been previously retrieved. As can be seen from this example, the reminder query has a function for synthesizing data as a solution as originally requested from data extracted by removing redundancy as described in the classification axis in (4). You can see that
S. Dar et al., “Semantic Data Caching and Replacement”, International Conference on Very Large Databases (VLDB) Proceedings, pages 330-341, 1996.

Similarly, in the management of XML data that is data having a tree structure, research using the method of the reminder query is shown in Non-Patent Document 2 below. However, in the method shown in Non-Patent Document 2, even if the data to be handled is tree structure data, only the case where an element in the solution set of the past request becomes the solution of the new request is considered. The case where the solution with a new requirement is a subtree of a solution with a requirement in the past is not considered, and the classification axis (1) is classified in the same manner as the ordinary reminder query method. R
L. Chen, E .; A. RUNDENSTINER "ACE-XQ: A CacheE-aware XQuery answering system" International Works on the Web and Databases (WebDB) Proceedings, pp. 31-36, 2002

This is different from the redundancy when tree structure data appears as a subtree of another tree structure data, but only when a part of the value is common among different elements in the solution set for a request. , A technique for removing the redundancy of the portion is disclosed in Patent Document 3 below.
JP 07-121571 A
In this method, if the retrieved data is tabular data such as Sato, General Affairs Department Tanaka, General Affairs Department Takeda, Human Resources Department Ishii, and Human Resources Department, this is represented by the sign “〃” representing “same value”. Use Sato, General Affairs Department Tanaka, Satoshi
Takeda, Human Resources Department Ishii, Satoshi
Redundancy is removed by transforming into an expression like.

However, with respect to the points listed above as the classification axis (1), there is a kind of duplication that occurs when managing data that can be expressed as a tree structure, where one solution also appears as a subtree that forms part of another solution. No technology has been disclosed so far. Therefore, by taking out, transmitting, presenting, and storing data containing duplicates in this way, the processing cost and required storage capacity are unnecessarily increased, and the user's browsing convenience is reduced. There's a problem.

In addition, even when considering only duplication in which the same data appears multiple times as a solution, the above-described reminder query method extracts the data with the duplication removed and then synthesizes the original solution set. That is not always possible. For example, when there is a request that “a set of employee numbers of employees aged 40 and under 60” can be retrieved, and when the results are saved, a new request is “an employee aged 30 and under 50” Suppose that a request is issued to “take out a set of employee numbers for the current employee”. In this case, all the employee numbers of employees aged 40 and under 50 are already included in the stored data, and it is redundant to retrieve it again. However, it is not possible to determine which employee number is the number of an employee under 50 years old simply by looking at the stored data of “employee number” of an employee under 40 years old and under 60 years old. The only way to retrieve the data is to request the retrieval of a set of employee numbers for employees aged 30 to under 50.

This is because when the request for the condition of “40 years old and under 60 years old” is first given, the condition of “30 years old and under 50 years old” is assumed. It is an unavoidable problem in a situation where the demand is given unpredictably, but there are two requirements: the requirement of “40 to under 60” and the requirement of “30 to under 50”. If these requests are given at the same time, these requests are divided into three requests, “30 to 40 years old”, “40 to 50 years” and “50 to 60 years”. , Obtain data in a form that eliminates duplicates, and then use these data to collect the initial request for an employee number of “40 to under 60” and “30 to under 50” Numbered It is easy to generate a slip. Therefore, when multiple retrieval requests are given “simultaneously” as in this example, simply applying the reminder query method does not retrieve duplicate data, and the original solution set later Although there is a method of combining, there is a problem that the processing cost may be increased by extracting duplicate data unnecessarily.

In view of the above problems, the object of the present invention is in any case when one or a plurality of data retrieval requests are given at the same time, and a solution of another solution having tree structure data. Data that has the function of removing the duplication, including the type of duplication that appears as a partial subtree, and combining the solution set of the original retrieval request as needed from the data from which this duplication was removed By providing a management program and system, it is possible to reduce the processing cost of data retrieval and transmission, reduce the storage capacity of data storage, and improve the browsing convenience when presenting data to the user. .

In the data management program and system according to the present invention, when a set of data fetching requests is given, all the data that are the solutions to any fetching request are included in the answer set for which fetching request. Generates a classification based on whether it is not included in the solution set for the retrieval request. This eliminates the redundancy that the same data appears redundantly in multiple fetch request solution sets, and simultaneously synthesizes the original fetch request solution set from these classifications, as described below. Is possible.

The data management program and system of the present invention generates the above classification after extracting each solution set using a given retrieval request, or uses a retrieval request synthesized based on a given retrieval request. To directly extract each classification. In the latter case, the retrieval of each classification is the product of all the retrieval requests whose classification is included in the solution set and all the sums of the retrieval requests whose classification is not included in the solution set. The retrieval request corresponding to the difference between the two is used.

As described above, the data management program and system of the present invention can synthesize a solution set of the initial retrieval request from each of the above classifications as necessary. This composition is performed by taking the union of all the classifications that were supposed to be included in the solution set of the retrieval request.

Also, the data management program and system of the present invention are given a request to fetch a set of subtrees at various depths in the tree structure data, and one of the subtrees taken out by the request is If a subtree that forms part of another subtree that appears is duplicated, the former subtree is removed from the solution set for the request. Here, an element is a subtree of an element, including the case where two elements are exactly the same tree. As a result, redundancy due to duplication as a subtree specific to the tree structure data existing in the solution set of the retrieval request can be eliminated.

Similarly, according to the data management program and system of the present invention, a plurality of requests for extracting a set of subtrees at various levels in the tree structure data are given, and a request in the plurality of requests is extracted. If one of the subtrees also appears as a subtree that forms part of it in another subtree taken by another request in the multiple requests, this former subtree is assigned to the corresponding request. Remove from the solution set. Again, an element is a subtree of an element, even if two elements are exactly the same tree. As a result, it is possible to eliminate redundancy due to duplication as a subtree peculiar to tree structure data that exists between the solution sets of the multiple fetch requests.

According to the data management program and system of the present invention, the above-described overlapping subtree is removed after a subtree set is extracted using a given fetch request, or is synthesized based on a given fetch request. This is done by directly extracting the subtree set from which duplicates have been removed.

The data management program and system according to the present invention have a function of synthesizing a subtree set as a solution to the initial retrieval request from the subtree set from which duplication is removed as described above.

In addition, the data management program and system of the present invention provide a path expression when a request for extracting a set of subtrees at various levels in the tree structure data is given in the form of a path expression. All the subtrees that do not form part of the other solutions of the path expression are matched to which prefix of the path expression and which prefix of the path expression Generates a classification based on whether does not match. As a result, redundancy due to duplication as a subtree unique to the tree structure data existing in the solution set of the fetch request can be removed, and at the same time, as described later, the initial fetch request can be removed from these classifications. Solution sets can be synthesized. Here, the path expression specifies the node to be extracted from the tree structure data by describing the conditions to be satisfied by each node on the path starting from the root node of the tree structure data and ending at the node to be extracted. Usually, the conditions for each node are described in order from the condition of the node closer to the root node.

Similarly, the data management program and system according to the present invention can be used when a plurality of requests for extracting a set of subtrees at various depths in the tree structure data are given in the form of a path expression. All of the subtrees that are solutions of the path expression that do not become subtrees that form part of the other solutions of any path expression are included in the solution set of which path expression. Classified based on the above two points, which are not included in the solution set of, and which prefix of which path expression matches and which prefix of which path expression does not match Generate. This eliminates redundancy due to duplication as a subtree specific to the tree structure data that exists between the solution sets of the multiple fetch requests, and at the same time, as described later, the initial fetch from these classifications. You can compose a solution set of requirements.

The data management program and system according to the present invention generates the above classification after extracting all data using a given path expression, or uses a path expression synthesized based on the given path expression. To extract each classification. In the latter case, if a single path expression is given, the composition of the path expression that retrieves each class is combined with all the products of the given path expression and the prefix that was supposed to match that class, This is done using the path expression corresponding to the difference between these two sums of all the prefixes that were not considered to match the classification. In the latter case, if multiple path expressions are given, the composition of the path expression that extracts each classification is considered to match the retrieval request and the classification that were included in the solution set. A path expression corresponding to the difference between these two products, the product of all the prefixes, the retrieval request whose classification was not included in the solution set, and the sum of all the prefixes that were not considered to match that classification. This is done using.

In addition, as described above, the data management method and system of the present invention can synthesize a solution set of the original path formula from the above-mentioned classifications as necessary. This synthesis is performed in the path expression from all the classifications that were supposed to be included in the solution set of the path expression and the subtree in each classification that was supposed to match each prefix of the path expression. This is done by taking the union of all of these with the set of subtrees extracted using the suffix paired with the prefix.

If multiple search formulas with overlapping parts of the solution are given at the same time, the data is extracted by the method described above, and even if it cannot be handled by the conventional reminder query method, duplicates are included. The data can be retrieved in a form that is less than the conventional methods and systems, the processing cost of data retrieval and transmission is reduced, the storage capacity for data storage is reduced, and the user who is presented with the data is viewed Convenience improvement can be realized.

In addition, with regard to the type of duplication that occurs when managing tree-structured data such that a solution appears as a subtree that forms part of another solution, data is extracted, transmitted, and presented in a state where this is removed. Compared with conventional methods and systems, the processing cost of data retrieval and transmission can be reduced compared to conventional methods and systems, storage capacity for data storage can be reduced, and data can be presented It is possible to improve the viewing convenience of the users who have been sent.

In addition, since the data management method or system of the present invention has a function of synthesizing data as originally requested from data in a state where duplication has been removed, for example, data can be sent via a network or the like in a state where duplication has been removed. By transmitting and receiving this data, the data requested in the original retrieval request is synthesized, so that the content presented to the user is maintained as the original request, and the transmission cost of the network is reduced. be able to. Similarly, save the data with duplicates removed, and synthesize the data requested in the original retrieval request when it is needed later, so that the content presented to the user is the one as originally requested. It is possible to reduce the storage capacity for storing data while maintaining the above.

(First embodiment)
Next, the details of the first embodiment of the present invention will be described in detail. In this description, an embodiment will be described in which the present invention is applied to a relational database that makes a data retrieval request using an SQL search formula that is a JIS standard. In the present embodiment, the solution sets of a given plurality of search expressions can include elements that overlap each other. Therefore, the system will include all the data that is included in the solution set of any one of the given search expressions, in which search expression solution the data is included in which search expression solution. Create a classification based on whether it is not included. This eliminates the redundancy that the same data appears redundantly in the solution set of multiple retrieval requests. Furthermore, as described later, the initial search formula can be changed from these classifications as necessary. You can synthesize the solution. On the other hand, since the data to be handled is not tree-structured data, the type of duplication in which one element in the solution set appears redundantly as a subtree of another element in the same or another solution set is the present embodiment. There is no need to consider in.

Figure 2 shows the overall configuration of the system in this embodiment. The system according to the present embodiment receives a storage device 4 that stores relational data, an input device 1 that receives a search expression described in SQL from a user, and one or a plurality of search expressions received by the input device 1. As for the retrieval formula, from the relational data in the storage device 4, the retrieval device 3 that extracts a set of data that is the solution of the retrieval formula, and the data retrieved by the retrieval device 3 are received, and the above-mentioned classifications are obtained from these data. And an output device 6 that receives and outputs data generated by the deduplication device.

The processing procedure when the input device receives a set of search expressions in this embodiment will be described below. As an example, consider the case where three search expressions are given to the input device. In this case, the solution set of these search expressions may contain elements that overlap each other. In general, when three search expressions are given and their solution sets are A1, A2, and A3, the relationship between these three sets is as shown in the Venn diagram shown in FIG. Therefore, in this system, when the deduplication apparatus receives A1, A2, and A3, the data included therein are classified into seven groups corresponding to the seven regions in the Venn diagram shown in FIG. Generate seven sets defined by the following equations.
A1∩A2∩A3
(A1∩A2)-A3
(A1∩A3)-A2
(A2∩A3)-A1
A1-(A2∪A3)
A2-(A1∪A3)
A3-(A1∪A2)
Here, ∩ represents a set product operation, ∪ represents a set sum operation, and − represents a set difference operation. It is clear from the Venn diagram shown in Fig. 3 that these seven groups do not contain overlapping elements.

で定義される集合を求めていけばよい．これは，別の言い方で言えば，Ａ１，…，Ａｎのいずれかに含まれる全データを，それがＡ１，…，Ａｎのうちのどれには含まれ，どれには含まれないかに基づいて分類していることに相当する．More generally, when n search expressions are received and their solution sets are A1,..., An, for any non-empty subset T of S = {A1,.

Find the set defined by. In other words, this is based on whether all data included in one of A1,..., An is included in which of A1,. Corresponds to classification.

If the user only needs to be able to browse all the data contained in any solution set of the original search expression, the user can output the original three categories by outputting the classification with duplicates removed in this way. Compared to browsing the solution of the search formula itself, the same data need not be viewed multiple times, and the convenience of browsing is improved.

It is also possible to compose the solution set of the original search formula from the classification generated in this way, if necessary. For example, in the example of the solution sets A1, A2, and A3 of the three search formulas described above, the solution set A1 of the first search formula is the following four solution sets A1∩A2∩A3.
(A1∩A2) -A3
(A1∩A3) -A2
A1- (A2UA3)
Find the set sum of. The same applies to A2 and A3. This point can also be easily confirmed from Fig. 3.

これは別の言い方で言えば，Ａｉに含まれるとされていた分類の全ての和集合を取ることに相当する．上式が成り立つことについては，複数の集合をベン図で表わした際の，ベン図中の各領域と各集合の間の関係に関する一般的に知られている数学的性質より明らかである．More generally, in order to synthesize a certain solution set Ai, the following equation may be used.

In other words, this is equivalent to taking all unions of classifications that were supposed to be included in Ai. The fact that the above equation holds is clear from the generally known mathematical properties of the relationship between each region in the Venn diagram and each set when multiple sets are represented by a Venn diagram.

Therefore, each classification that does not contain duplicates generated by the above method is saved to disk or sent to other computers via a network, etc., and then the solution set of the original search formula is synthesized. By presenting it to the user, it is possible to reduce the storage area for storing data and the communication cost for transmitting the data without changing the content presented to the user from the solution set of the original search expression.

(Second embodiment)
Next, the details of the second embodiment of the present invention will be described in detail. In the present embodiment, as in the first embodiment, an embodiment in which the present invention is applied to a relational database in which an extraction request is made by SQL will be described. The difference from the first embodiment is that the above classification is not directly generated by using a search expression synthesized based on a given search expression, instead of generating the above classification after extracting data. This is the point where the data corresponding to is extracted.

Figure 4 shows the overall configuration of the system in this embodiment. The system according to the present embodiment includes a storage device 4 that stores relational data, an input device 1 that receives input of a search expression described in SQL from a user, and one or more search expressions received by the input device 1. And a search expression synthesizer 2 for synthesizing a new search expression, and one or a plurality of search expressions generated by the search expression synthesizer 2 are received from the relational data in the storage device 4 for each search expression. Generated by a search device 3 that extracts a set of data that is a solution of the search formula, a search solution synthesis device 5 that receives the data extracted by the search device 3 and synthesizes the solution of the original search formula, and a search solution synthesis device 5 Output device 6 for receiving and outputting the received data.

These components may all be on the same computer, or, as shown in FIG. 4, only the storage device 4 and the search device 3 are on the computer 7 serving as a server, and the input device 1, The search expression synthesizer 2, the search solution synthesizer 5, and the output device 6 may be arranged on a client computer 9 connected to the server 7 via the network 8.

The processing procedure when the input device receives a set of search expressions in this embodiment will be described below. As an example, consider the case where the following three search expressions are given.
SELECT Name FROM Employee WHERE Age> 50
SELECT name FROM employee WHERE age> 40 AND affiliation = account SELECT name FROM employee WHERE age> 30 AND affiliation = sales The first search formula is a request to retrieve the names of all employees over the age of 50 from the list of employee data Represents. The second search expression represents a request to retrieve the names of all employees whose ages are over 40 and whose affiliation is accounting from the list of employee data. Similarly, the third search expression represents a request to retrieve the names of all employees whose ages are over 30 and whose affiliation is sales from the list of employee data.

In this case, the solution set of these search expressions may contain elements that overlap each other. For example, the names of accounting employees over the age of 50 appear as elements in both the solution set of the first search expression and the solution set of the second search expression. Similarly, the names of employees who belong to the sales department who are over 50 years of age appear as elements in both the solution set of the first search expression and the solution set of the third search expression.

Therefore, in this system, the search expression synthesizer synthesizes a search expression that extracts a solution set represented by the following expression from these search expressions, sends it to the storage device, and uses these to eliminate duplication. Extract data in the form.
A1∩A2∩A3
(A1∩A2)-A3
(A1∩A3)-A2
(A2∩A3)-A1
A1-(A2∪A3)
A2-(A1∪A3)
A3-(A1∪A2)
Here, A1, A2, and A3 represent solution sets of the above three search expressions, respectively, ∩ represents a set product operation, ∪ represents a set sum operation, and − represents a set difference operation. . An SQL search expression that extracts such a solution set can be created using the SQL INTERSECT syntax, UNION syntax, and DIFFERENCE syntax. For example, the search expression for extracting the set (A1∩A2) -A3 is as follows. .
(SELECT name FROM employee WHERE age> 50
INTERSECT
SELECT Name FROM Employee WHERE Age> 40 AND Affiliation = Accounting) DIFFENCE
SELECT Name FROM Employee WHERE Age> 30 AND Affiliation = Sales

Similarly, the retrieval formula for retrieving the set A1- (A2∪A3) is as follows.
SELECT Name FROM Employee WHERE Age> 50
DIFFERENCE
(SELECT name FROM employee WHERE age> 40 AND affiliation = accounting UNION
SELECT Name FROM Employee WHERE Age> 30 AND Affiliation = Sales)
The same is true for the other six.

In fact, the search expression corresponding to (A1∩A2) -A3 above can be equivalently transformed into the following simpler SQL.
SELECT name FROM employee WHERE age> 50 AND affiliation = accounting Also, considering that the value of the item belonging to the same employee data cannot be both accounting and sales, it can be A1∩A2∩A3 or (A2∩A3 ) It can be seen that the search formula corresponding to -A1 is not necessary because the solution set is always an empty set. However, the technique for determining a search expression in which simplification of a given search expression or a solution is always an empty set is a technique independent of the present invention in which prior art exists, and is described in detail here. Is not mentioned.

The fact that the solution set of the seven retrieval formulas created as described above does not include elements that overlap each other, as described in the description of the first embodiment, intuitively, these seven retrieval formulas. It can be easily understood from the fact that the expression is a retrieval expression with the set corresponding to the seven regions in the Venn diagram representing the relationship between A1, A2 and A3 shown in FIG.

The search solution synthesizer that receives the solutions of the above seven search formulas uses these to synthesize the solution sets A1, A2, and A3 of the initial three search formulas. For example, the solution set A1 of the first retrieval formula is represented by the following four solution sets A1∩A2∩A3.
(A1∩A2) -A3
(A1∩A3) -A2
A1- (A2∪A3)
Find the set sum of. The same applies to A2 and A3. This point can also be easily confirmed from Fig. 3.

By retrieving data using the retrieval formula synthesized in this way, the retrieval device can avoid retrieving the same data from the storage device multiple times, and the processing efficiency of the retrieval device portion can be improved. Is possible. For example, if you simply execute the three given search formulas in order, the data of the names of employees whose ages are over 50 and whose affiliation is accounting are used when the first search formula is executed. When the second retrieval formula is executed, it is fetched twice. When data is fetched using the above seven retrieval formulas, any data is retrieved twice or more. Is not taken out by.

In addition, if the search synthesizer or the search solution synthesizer is on a client connected via a network or the like, the data from which the duplicate is removed is transmitted from the server to the client, and the client initially includes the duplicate. The network communication cost can be reduced by synthesizing the solution set of search expressions.

If the user can browse all the data included in the solution set of the original search formula, the function of the search solution synthesizer is not necessary if it is not necessarily synthesized into the form of the solution set of the original search formula. You can turn off and send the solution of the combined query to the output device as it is. As a result, the user is not required to view the same data multiple times compared to the case of browsing the solutions of the original three search expressions, thereby improving the convenience of browsing.

で定義される検索式Ｑ（Ｔ）を求める．ここでは，∩，∪，−はそれぞれ，検索式の積演算，和演算，差演算を表わし，ＳＱＬでは，前述のように，各々，ＩＮＴＥＲＳＥＣＴ構文，ＵＮＩＯＮ構文，ＤＩＦＦＥＲＥＮＣＥ構文で表現できる．ここでは，取出し要求の記述にＳＱＬ検索式を用いる場合を例にとっているが，他の多くの取出し要求記述言語においても，同様の積，和，差にあたる取出し要求の合成が容易に可能である．そして，全ての空集合でない部分集合Ｔに対するＱ（Ｔ）を使って，データを取り出す．上の式は，複数の集合をベン図で表わした際の，ベン図中の各領域に対応する集合を表わす式を，そのまま検索式の式に置き換えたものであり，よって，この式で定義される各検索式の解集合が重複を含まないことは明らかである．Up to this point, we have explained using an example with three search expressions, but in general, search expressions are synthesized in the following procedure. First, suppose that a set of search expressions S = {Q1,..., Qn} is given. In that case, for any non-empty subset T of S, the product of all the search expressions included in T (that is, the search expression that retrieves the data satisfying all the search expressions) and not included in T The difference between these two search expressions and the sum of all the search expressions (that is, a search expression that retrieves data that satisfies the conditions of either search expression) (that is, the former search expression satisfies the conditions, but the latter search) Create a search expression that retrieves data that does not satisfy the expression conditions. That is, for any non-empty subset T of S,

The search expression Q (T) defined by is obtained. Here, ∩, ∪, and-represent the product operation, sum operation, and difference operation of the search expressions, respectively, and can be expressed in the INTERSECT syntax, the UNION syntax, and the DIFFERENCE syntax, respectively, in SQL. In this example, SQL retrieval formulas are used to describe fetch requests. However, in many other fetch request description languages, fetch requests corresponding to similar products, sums, and differences can be easily combined. Then, data is extracted using Q (T) for all non-empty subsets T. The above expression replaces the expression that represents the set corresponding to each area in the Venn diagram when multiple sets are represented by the Venn diagram, and is therefore defined by this formula. It is clear that the solution set of each search expression does not contain duplicates.

である．上記の式が成り立つことについても，複数の集合をベン図で表わした際の，各領域と，各集合の間の関係に関する，一般的に知られている数学的性質より明らかである．Then, in order to synthesize the solution of the original search formula Qi from these solutions, out of all the search formulas used for extracting data, all solution sets of the search formula created using T including Qi are used. Take the union. That is, if the solution set of Qi is written as Ai and the solution set of Q (T) is written as A (T),

It is. The fact that the above equation holds is also clear from the generally known mathematical properties of each region and the relationship between each set when multiple sets are represented by a Venn diagram.

Here, when simplification of the search expression or exclusion of unnecessary search expressions as described above, the search solution synthesis apparatus is created using T including Qi for the synthesis of the Qi solution. This is done by taking the union of the solution sets for all of the search expressions that have not been excluded.

(Third embodiment)
Next, details of the third embodiment of the present invention will be described in detail. In this description, the present invention stores XML data, which is a kind of data that can be expressed in a tree structure, and is a kind of path expression called XPath in the description of a data retrieval request for this (more specifically, the following non-patent document 3), an embodiment when applied to an XML database system will be described.
J. et al. Clark, S.M. By DeRose, “XML Path Language XPath Version 2.0—W3C Working Draft” http: // www. w3. org / TR / xpath20

XPath is a language that describes a request for retrieving a set of subtrees rooted at nodes of arbitrary depth in an XML tree structure. Therefore, in this embodiment, in the first and second embodiments, In addition to the kind of duplication in which the same data appears in multiple search expression solution sets as it occurs, a subtree from which a fetch request is taken is either the fetch request itself or another fetch request given at the same time. There can also be a type of duplication where another subtree is taken out and appears as a subtree that forms part of it. Therefore, in the present embodiment, processing for removing such duplication is performed. In this embodiment, these duplications are found by examining the extracted data.

The overall configuration diagram of the XML database system in this embodiment is as shown in FIG. 2 as in the case of the first embodiment. The XML database system in the present embodiment includes a storage device 4 that stores a large XML data tree structure, an input device 1 that receives input of a search expression described in XPath from a user, and one or more received by the input device 1 A plurality of search expressions are received, and for each search expression, a search apparatus 3 that extracts a set of subtrees as a solution of the search expression from the XML data in the storage device 4, and the data extracted by the search apparatus 3 are received, It consists of a deduplication device 10 that removes duplicates found by examining this data, and an output device 6 that receives and outputs data generated by the deduplication device.

The processing procedure when the input device receives a search expression or a set of search expressions in this embodiment will be described below. For example, assume that the input device receives a set of the following three XPath search expressions.
/ A / b
/ A / c / *
/ A / * / d
The first / a / b starts from the root of the tree structure of the stored XML data, and has a node with a tag name of the child a and a node with the tag name of the child b. Represents a request to retrieve. On the other hand, * in the second search expression represents "arbitrary tag name", and / a / c / * starts from the root and has a node with the tag name of the child a. This represents a request to extract a node having a tag name of child c and an arbitrary tag name of the child. Similarly, the third / a / * / d starts from the root and has a child with a tag name of the child a, a node with an arbitrary tag name of the child, and the child. Represents a request to retrieve a node with a tag name of d.

Therefore, the solution set of these three search expressions can contain elements that appear overlapping each other. Specifically, first, all nodes that match the search expression / a / c / d appear redundantly as elements of the solution sets of both / a / c / * and / a / * / d. . Of the elements in the solution set of / a / * / d, a node that matches the search expression / a / b / d is also a subtree of an element in the solution set of / a / b. It appears in duplicate.

For example, when the above three search expressions are evaluated for XML data represented by the tree structure shown in FIG. 5, the solution set for each search expression is as shown in FIG. Here, of the two subtrees in the solution set of / a / c / *, the left one appears also in the solution set of / a / * / d. Of the three subtrees in the solution set of / a / * / d, the leftmost one is also duplicated as a subtree that forms part of the subtree in the solution set of / a / b And appear.

Therefore, in this case, after the input device sends the above three retrieval formulas to the retrieval device, the retrieval device extracts the solution set for these, and sends them to the deduplication device, the deduplication device then sends the solution The set is examined and either one of these duplicates is removed. For example, the result shown in FIG. 7 is sent to the output device, which outputs it.

In this way, duplication such that the same subtree appears as an element in multiple solution sets, or duplication such that one element in one solution set appears as a subtree of another element in another or the same solution set. By removing and outputting, the user does not see the same data multiple times, and the convenience of the user's browsing can be increased.

In the above example, multiple search expressions are received and the solution of one search expression appears redundantly as a solution of another search expression or a subtree that forms part of it. When a search expression including the syntax "|" or "//" is given, a subtree in the solution set of one search expression is replaced with a part of another subtree in the same solution set. In some cases, they appear as subtrees formed. For example, given the following search expression:
/ A / b | / a / * / d
“|” Represents the “sum” of the two search expressions. In the above search expression, the union of the solution sets of the two search expressions / a / b and / a / * / d is the solution set. Become. That is, the union of the leftmost set and the rightmost set in Fig. 6 is the solution set. Therefore, the leftmost subtree of the rightmost set appears redundantly as a subtree in the leftmost set, which is another solution. Therefore, the duplicate elimination device also removes duplicates in one solution set.

Similarly, consider the following search expression as an example of a search expression including "//".
/ A / b // *
Intuitively, “//” means “descendant of any depth below it”, and the above search formula starts from the root and has a child node with the tag name “a”. , Represents a request to extract a node having an arbitrary tag name of a descendant of an arbitrary depth below the node having the tag name b of the child. Therefore, when this search expression is evaluated for the data shown in FIG. 5, the subtree set shown in FIG. 8 becomes the solution set. The solution shown on the right in this solution set also appears as a subtree of the solution shown on the left. Therefore, the duplicate elimination device removes such duplicates.

(Fourth embodiment)
Next, details of the fourth embodiment of the present invention will be described in detail. In this description, similarly to the third embodiment, the present invention stores XML data, which is a kind of data that can be expressed in a tree structure, and is a kind of path expression called XPath in the description of a data retrieval request for this data. (For more details, see Non-Patent Document 3 above.) An embodiment when applied to an XML database system will be described.

As in the case of the third embodiment, in this embodiment, the same type of duplication that the same data appears in the solution sets of a plurality of search formulas as occurred in the first and second embodiments. In addition, a subtree that a fetch request fetches appears as a subtree that forms part of the fetch request itself or another subtree that another fetch request given at the same time fetches. Duplication can also occur. Therefore, in this embodiment, processing for removing both types of overlap is performed. In the present embodiment, data is extracted in a form in which these duplicates are removed by extracting data using a search expression synthesized based on a given search expression.

Fig. 9 shows the overall configuration of the XML database system in this embodiment. The XML database system in the present embodiment includes a storage device 4 that stores a large XML data tree structure, an input device 1 that receives input of a search expression described in XPath from a user, and one or more received by the input device 1 A search expression synthesizer 2 that synthesizes a new search expression based on a plurality of search expressions, and one or more search expressions generated by the search expression synthesizer 2 are received, and each search expression is stored in the storage device 4. The search device 3 extracts a set of subtrees as a solution of the search formula from the XML data of the data, and the output device 6 receives and outputs the data extracted by the search device 3.

The processing procedure when the input device receives a search expression or a set of search expressions in this embodiment will be described below. As an example, assume that the input device receives the following three sets of XPath search expressions, as in the case of the third embodiment.
/ A / b
/ A / c / *
/ A / * / d

In this case, the retrieval formula synthesizer first modifies the above three retrieval formulas as follows in order to eliminate duplication of data between solution sets of different retrieval formulas.
/ A / b
/ A / c / *-/ a / b
/ A / * / d− / a / b− / a / c / *
More generally, an appropriate order is determined for all search expressions, and all the search expressions that are younger than the search expression are connected at the end of each search expression with "-". The search expression generated in this way is further transformed as follows to eliminate duplicates that cause a solution to appear as a subtree of another solution.
/ A / b- / a / b // *-/ a / c / * // *-/ a / * / d // *
/ A / c / *-/ a / b
-/ A / b // *-/ a / c / * // *-/ a / * / d // *
/ A / * / d− / a / b− / a / c / *
-/ A / b // *-/ a / c / * // *-/ a / * / d // *
More generally, after each search expression, add / * to the end of all the search expressions and put them together with-. By extracting data using the modified search expression, data can be extracted with duplicates removed, and the cost of data extraction processing can be reduced and the convenience of user browsing can be improved. .

Here, for the modified query shown above, it is possible to simplify or remove the query that makes the solution set always empty. However, as described above, these are technologies independent of the present invention, and will not be described in detail here. In addition, here, the method of synthesizing search expressions has been described by taking the case of using XPath to describe the extraction request of a subtree set from tree structure data, but the same applies to many other extraction request description languages. Synthesis is possible.

(Fifth embodiment)
Next, details of the fifth embodiment of the present invention will be described in detail. In this description, as in the third and fourth embodiments, the present invention stores XML data, which is a kind of data that can be expressed in a tree structure, and a path called XPath in the description of the data retrieval request for this data. An embodiment when applied to an XML database system using a kind of formula (more specifically, see Non-Patent Document 3 above) will be described.

Also in this embodiment, in addition to the kind of duplication that the same data appears in the solution sets of a plurality of retrieval formulas as in the first and second embodiments, the third and fourth implementations A subtree taken out by a fetch request, such as occurs in the form of, appears as a subtree that forms part of the fetch request itself or another subtree taken by another fetch request given at the same time The kind of duplication that occurs is possible. Therefore, in this embodiment, processing for removing both types of overlap is performed. In the present embodiment, data is extracted in a form in which these duplicates are removed by extracting data using a search expression synthesized based on a given search expression. Furthermore, in the present embodiment, it is possible to synthesize a solution set of the initial search formula from the data thus removed with duplication as necessary.

Fig. 1 shows the overall configuration of the XML database system in this embodiment. The XML database system in the present embodiment includes a storage device 4 that stores a large XML data tree structure, an input device 1 that receives input of a search expression described in XPath from a user, and one or more received by the input device 1 A search expression synthesizer 2 that synthesizes a new search expression based on a plurality of search expressions and simultaneously generates a description indicating how to extract the solution of the original search expression from the solution of the combined search expression. And a search device that receives one or a plurality of search formulas generated by the search formula synthesizer 2 and extracts a set of subtrees as a solution of the search formula from the XML data in the storage device 4 for each search formula. 3 and the data retrieved by the retrieval device 3 and the description generated by the retrieval formula synthesis device 2 that indicates how to retrieve the solution of the original retrieval formula are received, and the solution of the original retrieval formula is synthesized. That a search solution synthesizer 5, and an output device 6 for outputting receiving data search solutions synthesizer 5 to produce.

These components may all be on the same computer, or only the storage device 4 and the search device 3 are on the computer 7 serving as a server, and the input device 1, the search formula synthesis device 2, and the search The disassembly / synthesis device 5 and the output device 6 may be arranged on the client computer 9 connected to the server 7 via the network 8.

In the following, in this embodiment, the processing procedure when the input device receives a search expression or a set of search expressions will be described. Here, this will be described in three cases for ease of understanding. . First, processing when the received XPath search expression does not include the above-mentioned syntax “//” and all the search expressions are requests for retrieving nodes of the same depth in the tree structure of the XML data. The procedure is explained. The processing procedure in this case does not consider the case where a solution overlaps as a subtree of another solution, such as when the present invention is applied to a data management system other than tree structure data as in the first embodiment. The procedure is the same as in the case. Second, the processing procedure when the search formulas do not contain “//” but the depths of the nodes extracted by each search formula are not all the same will be described. Finally, the processing procedure when the search expression includes “//” is explained. In the above, for ease of understanding, the explanation will be divided into three cases. Actually, however, the last explained procedure is the most general procedure for all cases. If the procedure is applied to special cases such as the first and second, the final result equivalent to the first and second processing procedures can be obtained.

The processing procedure when the received XPath retrieval formula does not include “//” and all retrieval formulas retrieve data having the same depth is as follows. For example, assume that the input device receives a set of the following three XPath search expressions.
/ A / b
/ A / *
/ A / * [c / d]
The first / a / b starts from the root of the tree structure of the stored XML data, and has a node with a tag name of the child a and a node with the tag name of the child b. Represents a request to retrieve. On the other hand, the second / a / * indicates a request to start from the root and extract a node having a tag name of the child a and a node having an arbitrary tag name of the child. In addition, [c / d] in the third retrieval formula represents the condition “having a path c / d below”, and the retrieval formula / a / * [c / d] is , Starting from the root, a node having a tag name of the child a, and a node having an arbitrary tag name of the child, and having a child node c below the node, and the c Represents a request that a child node called “d” has a child node called “d”.

Therefore, the solution set of these three search expressions can contain elements that appear overlapping each other. Specifically, all nodes that are elements of the solution set of / a / b or / a / * [c / d] also appear as elements of the solution set of / a / *. Also, the solution sets of / a / b and / a / * [c / d] can have the same elements redundantly. On the other hand, in this example, since all three search expressions extract a node with a depth of 2 in the tree structure, an element in one solution set is duplicated as a subtree of elements in the same or another solution set. It cannot appear.

Therefore, in this system, in this case, an arbitrary non-empty subset of the given search expression set, that is, {/ a / b, / a / *, / a / * [c / d]} in this case
{/ A / b, / a / *}
{/ A / b, / a / * "c / d]}
{/ A / *, / a / * [c / d]}
{/ A / b}
{/ A / *}
{/ A / * [c / d]}
Is a product of all search expressions in the subset and a set of search expressions not included in the subset {}
{/ A / * [c / d]}
{/ A / *}
{/ A / b}
{/ A / *, / a / * "c / d]}
{/ A / b, / a / * [c / d]}
{/ A / b, / a / *}
Search formulas corresponding to the difference between them and the sum of all the search formulas included in the / a / bｂ / a / * ∩ / a / * "c / d" (search formula 1)
(/ A / b∩ / a / *) − / a / * [c / d] (Search formula 2)
(/ A / b∩ / a / * "c / d])-/ a / * (search formula 3)
(/ A / * ∩ / a / * [c / d]) − / a / b (search formula 4)
/ A / b-(/ a / * ∪ / a / * [c / d]) (Search formula 5)
/ A / *-(/ a / b∪ / a / * [c / d]) (Search formula 6)
/ A / * [c / d] − (/ a / b∪ / a / *) (search formula 7)
The search formula synthesizer searches for these seven search formulas and sends them to the search unit. This is the same procedure as in the second embodiment described above.

Here, ∩ represents the product of two search expressions, that is, a request to extract a node that satisfies both conditions of the two search expressions. Similarly, ∪ represents the sum of two search expressions, that is, a request to retrieve a node that satisfies either of the conditions of both search expressions, and − represents the difference between the two search expressions, that is, the former search expression. Represents a request to retrieve a node that satisfies the condition but does not satisfy the latter search condition. These notations are different from the notations in the actual XPath (for example, ∪ is described as “|” in the actual XPath as described above). Will be used. It is almost the same as in the case of the second embodiment that the solution sets for the above equations 1 to 7 do not contain overlapping elements. By retrieving the data using the search expression synthesized in this way, it is possible to avoid retrieval of duplicate data or transmission of duplicate data via a network.

The search device receives the combined search expression as above, searches for the solution, and sends seven solution sets corresponding to the search solution combiner. The search solution synthesizer takes the set sum of the solution sets for the search expressions 1, 2, 3, and 5 created using the subset including / a / b among these solution sets. Generate a solution set for the search expression / a / b. The solution of / a / * is the set sum of the solution sets for search formulas 1, 2, 4, and 6, and the solution of / a / * [c / d] is search formulas 1, 3, 4, and 7 Generate by taking the union of the solution sets for.

In this example, Equation 1 is equivalent to a simpler retrieval formula / a / b [c / d], and Equation 3 is a retrieval formula in which the solution set is always an empty set. Other equations can be simplified in the same way, or have solutions that are always empty. Therefore, the query formula synthesizer may simplify such query formulas or exclude search formulas that have empty solutions. Non-patent documents 4 and 5 below describe techniques for simplifying a search expression described in the form of a path expression and determining a search expression in which a solution set is always an empty set. Since there are various prior arts such as those that are independent of the present invention, they are not described in detail here.
G. Miklau, D.M. Suciu, “Containment and Equivalence for an XPath Fragment”, ACM Symposium on Principles of Database Systems (PODS) Proceedings, pages 65-76, 2002 M.M. Benedikt et al., “Structural Properties of XPath Fragments”, International Conference on Database Theory (ICDT) Proceedings, pages 79-95, 2003.

When simplification or exclusion of the search expression as described above is performed, the search solution synthesis apparatus is obtained and simplified by simplifying the search expression generated based on the subset including / a / b. The solution of / a / b is generated by finding the union of all solution sets of those that did not exist. The same applies to the solutions of other search expressions. This is also the same as in the case of the second embodiment described above.

Up to this point, the procedure has been described using a specific example, but more generally, as with the second embodiment, for all subsets of a given search expression set that are not empty. , The subset is defined as T, and a search expression defined by Equation 3 is generated, and data is extracted by all these search expressions. In addition, the synthesis of the solution of the initial search formula from the extracted data is also performed based on Formula 4 as in the case of the second embodiment.

にあたるパス式を作成する．また，このようにして生成された各パス式に対して，順に，Ｖ１，Ｖ２…という識別子を割り当てていく．
（６）また，それと同時に，Ｔに含まれる検索式Ｑｘ，Ｑｙ，…の各々について，今，作成したＶａの解となる部分木が，そのまま，Ｑｘ，Ｑｙ，…の解として取り出されるべきであることを表わす，
Ｑｘ←（Ｖａ，＊）
Ｑｙ←（Ｖａ，＊）
：
：
という項を生成する．同様に，Ｒに含まれる接頭辞ｐの各々について，その接頭辞を取り出した検索式をＱｘ，また，Ｑｘからその接頭辞ｐを取り除いた残りの部分にあたる接尾辞をｐｐとして，Ｖａの解となる部分木の根からパス＊／ｐｐにしたがってたどって到達するノードが，Ｑｘの解として取り出されるべきであることを表わす，
Ｑｘ←（Ｖａ，＊／ｐｐ）
：
：
という項を生成する．Next, the processing procedure when the retrieval formula does not include “//” but all retrieval formulas are not retrieval requests of nodes having the same depth will be described. In this case, the retrieval formula synthesizer synthesizes the retrieval formula in the following procedure.
(1) First, identifiers Q1, Q2,... Are assigned to all the given search expressions, and all the given search expressions are classified into groups according to the path length.
(2) The following processing is performed for each group obtained in (1) above.
(3) Let S be the set of all search expressions included in the group, and T be all non-empty subsets of S.
(4) From all the search expressions longer than the group, the prefix of the length of the group currently considered is taken out, a set of these prefixes is set as P, and each of all subsets of P is set as R. Perform processing.
(5) For all search formulas shorter than the current group, create a search formula with /*/.../* added to the end to make it the same length as the current group. Using the above formula,

Create a path expression corresponding to. In addition, identifiers V1, V2,... Are sequentially assigned to the path expressions generated in this way.
(6) At the same time, for each of the search expressions Qx, Qy,... Included in T, the subtree that is the solution of Va that has just been created should be taken out as the solution of Qx, Qy,. Represent something,
Qx ← (Va, *)
Qy ← (Va, *)
:
:
Generates the term Similarly, for each prefix p included in R, the search expression for extracting the prefix is Qx, and the suffix corresponding to the remaining part obtained by removing the prefix p from Qx is pp. Indicates that the node to be reached from the root of the subtree following the path * / pp should be taken as the solution of Qx.
Qx ← (Va, * / pp)
:
:
Generates the term

As an example, let us explain the case where the following three search expressions are given.
/ A / b / c
/ A / * / c / d
/ A / b / c [d] / *
The first search formula starts from the root and takes out a child node having a tag name of a, a node of that child having a tag name of b, and a child of the child having a tag name of c. The second search formula starts from the root and has a tag name of a child node with a tag name of a, a child node of that child node with a tag name of c, and a child with a tag name of d Get the node with. The third retrieval formula starts from the root, has a tag name of c of the child node of the child node having the tag name of a, the child of the node having the tag name of b, and Extract any child nodes of a node with children with tag name d.

The solution set of these three search expressions can include the following duplication of data. First, the same element can appear twice in the solution set of the second search expression and the solution set of the third search expression. In addition, some elements in the solution set of the second and third search expressions can appear as a subtree of a solution in the solution set of the first search expression. For example, if there is a node that matches the search expression / a / b / c / d in the XML data currently stored in the system, this node is the solution of the second search expression. On the other hand, since the parent node of that node becomes the solution of the first search expression, this node appears as an element in the solution set of the second search expression, and at the same time, the solution of the first search expression It will also appear as a subtree of an element in the set.

Therefore, the retrieval formula synthesizer synthesizes retrieval formulas according to the following procedure. First, identifiers Q1, Q2, and Q3 are assigned to these search expressions. Next, these search expressions are classified into a group {Q1} of length 3 and a group {Q2, Q3} of length 4. First, considering a group of length 3, the only non-empty subset is {Q1}. In the next (4), the set P obtained by extracting the prefix of length 3 from all the search expressions longer than this group is {/ a / * / c, / a / b / c [d]}. A subset of this {/ a / * / c, / a / b / c [d]}
{/ A / * / c}
{/ A / b / c [d]}
{}
Repeat (5) for the four cases where each of R is R. In this case, there is no search expression shorter than the current group, and there is no search expression corresponding to E. Therefore, when {/ a / * / c, / a / b / c [d]} is selected as R, the following path expression is created.
/ A / b / c∩ / a / * / c∩ / a / b / c [d]
At the same time, an identifier V1 is assigned to this search expression. Further, since this search formula uses {Q1} as T used to generate the search formula, it is a term indicating that the solution of V1 should be taken out as the solution of Q1 as it is Q1 ← (V1, *)
Is generated. Further, in this search expression, R used for generating the search expression includes / a / * / c which is a prefix extracted from Q2 and / a / b / c [d] which is a prefix extracted from Q3. Since the set was used, this means that the element that arrives by following the path d that is the suffix corresponding to / a / * / c from the root of the subtree that is the solution of V1 should also be taken out as the solution of Q2 Q2 ← (V1, * / d)
And an element that arrives from the root of the subtree that is the solution of V1 by following the path * that is the suffix corresponding to / a / b / c [d] should be taken out as the solution of Q3 Q3 ← (V1, * / *)
Is generated.

Here, since V1 can be simplified to / a / b / c [d], the retrieval formula synthesis apparatus may perform such simplification, but as described above, the simplification is simplified. The specific method of removing unnecessary search expressions is a technology independent of the present invention, and will not be described in detail here.

Similarly, in the procedure (4), the remaining three ways of selecting R, ie,
R = {/ a / * / c}
R = {/ a / b / c [d]}
R = {}
When (5) is executed as
V2: / a / b / c∩ / a / * / c− / a / b / c [d]
Q1 ← (V2, *)
Q2 ← (V2, * / d)
V3: / a / b / c∩ / a / b / c [d] − / a / * / c
Q1 ← (V3, *)
Q3 ← (V3, * / *)
V4: / a / b / c− / a / * / c− / a / b / c [d]
Q1 ← (V4, *)
And a term that indicates how to extract the solution of the original search formula therefrom.

Next, consider a search expression group {/ a / * / c / d, / a / b / c [d] / *} of length 4. The non-empty subset of this group is
{/ A / * / c / d, / a / b / c [d] / *}
{/ A / * / c / d}
{/ A / b / c [d] / *}
It is. Therefore, (4) is performed with each of these as T. Since there is no search expression longer than this group, the prefix set P is an empty set, and its subset is only the empty set. Therefore, in any case where T is set, (5) is performed with R as an empty set. For E, the search expression shorter than the current group is only / a / b / c with a length of 3. Therefore, in order to set the length to 4, / a / b is added to the end. / C / *. As a result, when {/ a / * / c / d, / a / b / c [d] / *} in the above three sets is T,
V5: / a / * / c / d∩ / a / b / c [d] / *-/ a / b / c / *
The search expression V5 and
Q2 ← (V5, *)
Q3 ← (V5, *)
A term indicating how to extract the solution of the original search formula from V5 is generated. Similarly, when {/ a / * / c / d} and {/ a / b / c [d] / *} are T,
V6: / a / * / c / d− / a / b / c [d] / * − / a / b / c / *
Q2 ← (V6, *)
V7: / a / b / c [d] / *-/ a / * / c / d- / a / b / c / *
Q3 ← (V7, *)
Search terms V6 and V7 and a term indicating how to extract the solution of the original search formula therefrom are generated.

From the above, for the three search expressions given initially, seven search expressions and how to extract the solution of the original search expression from the solutions are generated. The solution sets of these seven search formulas do not contain elements that overlap each other, and no element in one solution set appears twice as a subtree of an element in the same or another solution set . On the other hand, the solution set of these seven search formulas contains all the data necessary for the synthesis of the solution set of the original three search formulas. From the solution set of these seven search formulas, it is shown above. According to the retrieval method, the solution set of the initial three search expressions can be synthesized.

の部分は，長さが同じ検索式のグループについて，前述の第二の実施形態の場合や，与えられた検索式が全て同じ長さだった場合において行ったのと同様の，ベン図の各領域に対応する解集合への分類を行っていることになる．よって，数式５によって作成された各検索式は，その解集合中に重複する要素を含まない．The reason why the search formula created by the above procedure satisfies the above properties for an arbitrary set of search formulas can be explained as follows. First, the point that the solutions of these search expressions do not contain overlapping elements. Only search expressions with the same length can contain duplicate solutions in the solution set. On the other hand, in Formula 5 used for generating the search formula

The part of Venn diagram is the same as that performed in the case of the second embodiment described above or when all of the given search formulas have the same length for a group of search formulas having the same length. Classification into solution sets corresponding to. Therefore, each search formula created by Equation 5 does not contain duplicate elements in the solution set.

Next, we explain that an element in a solution set of a search expression does not appear as a subtree of another element in the solution set of the same search expression or another search expression. A solution of a query can appear as a subtree of a solution of a query when the former query is longer than the latter. However, in Formula 5, for all search formulas that are shorter than the initial search formula included in T, the sum of the same length by adding * /... -E is removed, so that what appears as a subtree of a shorter query is always removed.

Next, we explain that the node that is the solution of the initial search expression always appears in the solution set of one of the composite search expressions. This can be divided into two cases: some of the ancestor nodes of that node will be the solution of any of the initial search expressions, and no such ancestor node. If there is such an ancestor node, the node closest to the root among those ancestor nodes is included in the solution set of a certain composite search expression. Therefore, the node currently in question appears as a subtree that forms part of the solution. Note that the combined search expression in which the ancestor node appears as a solution is T, which is the set of all search expressions that match the ancestor node, and matches the ancestor node among the search expression prefixes longer than the search expression in T. This is a search expression generated with R as the set of all things to do. On the other hand, if there is no such ancestor node, the current node itself appears in the solution set of a single synthetic search expression. The composite search expression generates a set of all search expressions that match the subtree as T, and generates a set of all the search expression prefixes in T that match the node as R. This is the retrieved search expression.

の部分は，これらのノードを，そのノードを根とする部分木からは，他の，より長い検索式のうちのどれとどれの解を，どんなパス式に基づいて取り出せばよいかという観点から分類していることに相当する．よって，数式５は，必要な全ノードを，それをどの検索式の解とすればよいか，そこからどのような部分木を他の解として抜き出せば良いかに基づいて，分類していることに相当し，よって，正しく当初検索式の解を取り出すことができる．Next, the point that the solution of the initial search formula can be correctly synthesized by the above-mentioned method from the solution of each search formula generated by the above method is explained. The portion of Equation 6 of Equation 5 used for synthesis includes all nodes that are solutions of any one of the initial search equations, and which solution is included in any solution set of the original search equations. This is equivalent to classification based on whether it is not included in the set. Similarly, in Equation 5

The part of is from the point of view of what path expression should be taken from these nodes, and from the subtree rooted at that node, which of other longer search expressions and which solutions. Corresponds to classification. Therefore, Equation 5 classifies all the necessary nodes based on which search expression should be used as a solution and what subtrees should be extracted as other solutions. Therefore, the solution of the initial search formula can be extracted correctly.

Next, the processing procedure when the search expression includes the syntax "//" is explained. The syntax of “//” in XPath represents the condition of “any depth below”, for example, / a // b starts from the root of the data and is a child node with a tag name of a Represents a request to extract a node with a tag name of b that is a descendant of any depth below. // a / b represents a request to retrieve a node having a tag name of a at an arbitrary depth in the data and a child node having a tag name of b.

In order to describe the processing procedure when a search expression including the syntax “//” is given, first, the concept of i-prefix of the search expression is defined as follows. Now, the given search formula is: / 1 p1 / 2 p2 ... / mpm
Suppose that Here, / 1,..., / M are either / or //, respectively, and p1,..., Pm are respectively sandwiched between them. This part does not contain //. At this time, for i where 0 ≦ i ≦ m−1, the i-prefix of this search expression is defined as follows.
(1) When i ≠ 0 and / i + 1 is /:
/ 1 p1 ... / ipi
(2) When i ≠ 0 and / i + 1 is //:
/ 1 p1 ... / ipi∪ / 1 p1 ... / i pi // *
(3) When i = 0 and / 1 is /:
φ (special path expression that does not match any node)
(4) When i = 0 and / 1 is //:
// *
For example, for a search expression of length 1 of // a, an i-prefix is defined only for i where 0 ≦ i ≦ 1-1, i.e., i = 0, and the 0-prefix of this search expression. The acronym is // *. For a search expression having a length of / b // c of 2, an i-prefix is defined for i where 0 ≦ i ≦ 2-1, i.e., i = 0 and i = 1. In this search expression, the 0-prefix is φ and the 1-prefix is / b∪ / b // *.

The definition of the i-prefix above expresses the infinite number of cases where the syntax “//” is “/, or / * /, or / * / * /, or…”. Considering this, it is a natural extension of the usual path prefix concept.

にあたるパス式を作成する．また，このようにして作成された各パス式に対して，順に，Ｖ１，Ｖ２，…という識別子を割り当てていく．
（５）また，それと同時に，Ｔに含まれる検索式Ｑｘ，Ｑｙ，…の各々について，今，作成したＶａの解となる部分木が，そのまま，Ｑｘ，Ｑｙ，…の解として取り出されるべきであることを表わす，
Ｑｘ←（Ｖａ，＊）
Ｑｙ←（Ｖａ，＊）
：
：
という項を生成する．同様に，Ｒに含まれるｉ−接頭辞ｐの各々について，その接頭辞を取り出した検索式をＱｘ，また，Ｑｘからその接頭辞ｐを取り除いた残りの部分にあたる接尾辞をｐｐとして，Ｖａの解となる部分木の根からパス＊ｐｐにしたがってたどって到達するノードが，Ｑｘの解として取り出されるべきであることを表わす，
Ｑｘ←（Ｖａ，＊ｐｐ）
：
：
という項を生成する．If the i-prefix is defined as above, the processing procedure for the search expression including the syntax “//” is as follows.
(1) First, identifiers Q1, Q2,... Are assigned to all given search expressions.
(2) The following processing is performed with S being a set of all given search expressions and T being a subset of all non-empty subsets of S.
(3) The following processing is performed with a set of all i-prefixes defined for each given search expression as P and each subset of P as R.
(4) Create a search expression with // * added to the end of all the search expressions.

Create a path expression corresponding to. In addition, identifiers V1, V2,... Are sequentially assigned to the path expressions created in this way.
(5) At the same time, for each of the search expressions Qx, Qy,... Included in T, the subtree that is the solution of Va that has been created should be taken out as a solution of Qx, Qy,. Represent something,
Qx ← (Va, *)
Qy ← (Va, *)
:
:
Generates the term Similarly, for each of the i-prefix p included in R, the search expression from which the prefix is extracted is Qx, and the suffix corresponding to the remaining part obtained by removing the prefix p from Qx is pp. Indicates that the node that is reached from the root of the subtree to be solved along the path * pp should be taken out as a solution of Qx.
Qx ← (Va, * pp)
:
:
Generates the term

The above procedure takes into account the fact that a search expression including the syntax “//” has a variable length, that is, a depth at which the subtree to be extracted appears. It is a natural extension of the procedure. In other words, the above procedure, in other words, considers all the subtrees that are solutions of any given path expression that are not part of other solutions. Is included in which path expression's solution set and not in which path expression's solution set, and it matches which prefix of which path expression matches which prefix of which path expression This is equivalent to generating a path expression that classifies each class based on the above two points.

As an example, the case where the following two search expressions are given is explained.
// a
/ B // *
The first search expression extracts a subtree rooted at a node at an arbitrary depth and having a tag name a. The second search formula starts from the root of the data and extracts a subtree rooted at a node with an arbitrary tag name at an arbitrary depth below a child node with a tag name of b . These two search expressions can include duplication of data as follows. First, if there is a node that matches the search expression / b // a, that node is included as an element in the solution set of both of the above two search expressions. If there is a node that matches the search expression / b // * // a, the node appears as an element in the solution set of the first search expression and at the same time, the second search expression. It also appears as a subtree of an element in the solution set of. Similarly, if there is a node that matches the search expression / b // a // *, that node appears as an element in the solution set of the second search expression and at the same time the first search It will also appear as a subtree of an element in the solution set of the expression.

Therefore, the retrieval formula synthesizer synthesizes retrieval formulas according to the following procedure. First, identifiers Q1 and Q2 are assigned to these search expressions. Next, a non-empty subset {// a, / b // *} of the set of search expressions {// a, / b // *}
{// a}
{/ B // *}
(3) The following processing is performed for each of these. The set of all i-prefixes defined for a given search expression is {// *, φ, / b∪ / b // *}
Therefore, the following processing is performed on all the subsets, each of which is R. For example, when {// a, / b // *} is T and {// *, φ, / b∪ / b // *} is R,
V: // a ∩ / b // * ／ // * ∩ φ ∩ (/ b ∪ / b // *)
-(// a // * ∪ / b // * // *)
And a term representing the retrieval method of the original retrieval formula from V.
Q1 ← (V, *)
Q2 ← (V, *)
Q1 ← (V, * // a)
Q2 ← (V, * / b // *)
Q2 ← (V, * // *)
Is generated. Where // a is the suffix for the 0-prefix of Q1, / b // * is the suffix for the 0-prefix of Q2, and // * is the suffix for the 1-prefix of Q2.

However, the above V is a search expression in which the solution set is always an empty set because φ that does not match all nodes is included as a product component. Therefore, those that contain φ as a product component and can easily determine that the solution set is an empty set may be removed by the search expression synthesizer. Similarly, a retrieval formula including // * as a negative component can be easily determined that the solution set is always an empty set, and may be removed by the retrieval formula synthesis device. Furthermore, since it is easy to see that the meaning of the entire search expression does not change even if "∩ // *" and "-φ" are omitted, the search expression synthesizer can simplify the search expression by omitting them. May decide to do. Here, it is assumed that the retrieval formula synthesizer performs removal and simplification of these particularly easy retrieval formulas. In that case, it is only necessary to consider the cases of {// *, / b∪ / b // *} and {// *} as R. As a result, the following six search formulas are generated by the combination of three kinds of T and two kinds of R.
V11: // a∩ / b // * ∩ (/ b∪ / b // *)
-(// a // * ∪ / b // * // *)
V12: // a∩ / b // *-(/ b∪ / b // *)
-(// a // * ∪ / b // * // *)
V13: // a ∩ (/ b ∪ / b // *)-/ b // *
-(// a // * ∪ / b // * // *)
V14: // a− / b // * − (/ b∪ / b // *)
-(// a // * ∪ / b // * // *)
V15: / b // * ∩ (/ b∪ / b // *) − // a
-(// a // * ∪ / b // * // *)
V16: / b // *-// a- (/ b∪ / b // *)
-(// a // * ∪ / b // * // *)
Furthermore, from these, as a term that represents the method of extracting the original solution of Q1 and Q2,
Q1 ← (Vi, *) where i = 11, 12, 13, 14
Q2 ← (Vi, *) where i = 11, 12, 15, 16
Q1 ← (Vi, * // a) where i = 11, 12, 13, 14, 15, 16
Q2 ← (Vi, * // *) where i = 11, 13, 15
Is generated. Here, // a is the suffix for // 1 which is the 0-prefix of Q1, and // * is the suffix for / b∪ / b // * which is the 1-prefix of Q2.

Actually, the above V11 to V16 can be further simplified as follows.
V11: / b / a
V12: φ
V13: φ
V14: // a- / b // *-// a // *
V15: / b / *-// a-// a // *
V16: φ
Therefore, the retrieval formula synthesizer may perform such further simplification. However, as mentioned above, the specific technology for such simplification is not described in detail here.

By retrieving data using the combined search formula above, it is possible to eliminate the retrieval of duplicate data. This point is shown below using specific data examples. The solution when // a and / b // * are executed on the data shown in FIG. 10 is as shown in FIG. As can be seen from FIG. 11, these solutions contain a lot of data duplication and are very verbose. On the other hand, when V11 to V16 are executed on the data shown in FIG. 10, the solutions of V12, V13, V14, and V16 are empty sets, and the solutions of V11 and V15 are as shown in FIG. From these solutions, it can be seen that the same solution as in FIG.

The same duplication problem occurs when only one search expression is given if the search expression contains "//". As an example, for the data shown in FIG.
// a / * / *
Is given. The solution in this case is as shown in Fig. 14. However, as can be seen from this figure, in this solution, the subtree rooted at f appears redundantly, once as a solution and once as part of another solution.

Therefore, consider synthesizing search expressions by the above procedure. First, let the above search formula be Q1. Since there is only one search expression, the only way to select T is T = {Q1}. The set of prefixes is {// *, // a, // a / *}, and there are 8 ways to select this subset. However, for // *, as in the previous example, the search expression generated when this is not included in R always has an empty set. So for R,
R = {// *, // a, // a / *}
R = {// *, // a}
R = {// *, // a / *}
R = {// *}
The following four search expressions are generated.
V21: // a / * / * ∩ // * ∩ // a∩ // a / *-// a / * / * // *
V22: // a / * / * ∩ / * ∩ // a-// a / *-// a / * / * // *
V23: // a / * / * ∩ / * ∩ // a / *-// a-// a / * / * // *
V24: // a / * / * ∩ // *-// a-// a / *-// a / * / * // *
Furthermore, as a term indicating how to extract the original Q1 solution from these,
Q1 ← (Vi, *) where i = 21, 22, 23, 24
Q1 ← (Vi, * // a / * / *) where i = 21, 22, 23, 24
Q1 ← (Vi, * / * / *) where i = 21, 22
Q1 ← (Vi, * / *) where i = 21, 23
Is generated. // a / * / * in the second term is the suffix corresponding to 0-prefix // *, and similarly, / * / * and / * in the third and fourth terms , 1-prefix and 2-prefix suffix.

Actually, the retrieval formulas for V21 to V24 can be further simplified as follows. The retrieval formula synthesis apparatus may perform such a simplification, but as described above, it will not be described in detail here.
V21: // a / a / a-// a / * / * // *
V22: // a / * / a-// a / *-// a / * / * // *
V23: // a / a / *-// a-// a / * / * // *
V24: // a / * / *-// a-// a / *-// a / * / * // *
When these search expressions are executed on the data shown in FIG. 13, the solutions of V21 and V22 are empty sets, and the solutions of V23 and V24 are as shown in FIG. As can be seen from FIG. 15, these solution sets do not include any data duplication. And if each term showing the above-mentioned extraction method is executed for these results, the solution of Q1 can be correctly extracted.

As described above, four search expressions and how to extract the solution of the original search expression from the solution are generated for one initially given search expression, and the solution set of these four search expressions is It does not contain elements that overlap each other, and an element in one solution set does not appear as a subtree of an element in the same or another solution set. The solution set contains all the data necessary to synthesize the solution set of the original search expression. From the solution set of these four search expressions, the solution of the original search expression is determined according to the retrieval method shown above. You can synthesize sets.

The reason why the search formula created by the above procedure satisfies the above properties for an arbitrary set of search formulas is almost the same as the description for the procedure when “//” is not included. Explain.

In the embodiment shown in this specification, the processing when applying the present invention to a system for extracting a set of subtrees from tree structure data has been described by taking as an example a case where a search expression is given in the form of a path expression. The present invention can be applied based on the same concept even when the retrieval request is given in various other forms.

(Sixth embodiment)
Next, a sixth embodiment of the present invention will be described. In this description, an embodiment in which the present invention is applied to a proxy server for an XML data continuous inquiry system published on the Internet will be described.

FIG. 16 shows an overall configuration diagram of the XML data continuous inquiry system and the proxy server in this embodiment. In the present embodiment, it is assumed that the XML data continuous inquiry system 11 is disclosed on the Internet 12. The XML data continuous inquiry system stores XML data, receives a search expression registration request from a user, registers the search expression permanently, and periodically searches these search expressions registered by the user (for example, This is a system that evaluates (at regular intervals) and sends the evaluation result to the user who registered the query every time the search expression is evaluated, and is used in news information notification services, stock price information notification systems, etc. It is assumed that there are a plurality of users 14 who use this XML data continuous inquiry system 11 on a local network 13 such as an in-house network connected to the Internet 12.

In such an environment, in this embodiment, the proxy server 15 is placed on the local network 13, and each user 14 on the local network 13 directly registers his / her search formula in the XML data continuous inquiry system 11. Instead, a search expression is registered for this proxy server 15. Then, the proxy server 15 synthesizes a new search expression based on a plurality of search expressions registered from a plurality of users 14 by the same method as in the fifth embodiment, and the combined search expression is converted into XML data. Register in the continuous inquiry system 11. Each time a result is periodically received from the XML data continuous inquiry system, the result is used to synthesize a solution of the search formula registered by each user 14 and send it to each user 14.

In this way, when a plurality of users register a large number of search formulas with overlapping solutions, these search formulas are replaced by the proxy server 15 with queries that do not contain duplicates. The result is registered in the continuous inquiry system 11, and the result is transmitted to the proxy server 15 via the Internet 12. The proxy server 15 synthesizes the solution of the original search formula including a partial overlap, and this is synthesized via the local network 13. Since it is transmitted to each user 14, compared to the case where each user 14 directly registers their search expressions in the continuous inquiry system 11, the transmission of duplicate data on the Internet 12 is eliminated, and communication on the Internet 12 is performed. Cost can be reduced.

Overall configuration diagram of system in fifth embodiment Overall configuration diagram of the system in the first and third embodiments Venn diagram representing the relationship between the three solution sets A1, A2, A3 Overall configuration diagram of the system in the second embodiment Example of XML data (part 1) Solution set for / a / b, / a / c / *, / a / * / d Data with duplicates removed from the solution set in FIG. Solution set for / a / b // * Overall configuration diagram of a system according to the fourth embodiment Example of XML data (part 2) Solution set for // a, / b // * Solution set for V11 and V15 Example of XML data (part 3) Solution set for // a / * / * Solution set for V23 and V24 Configuration diagram of the entire system environment in the sixth embodiment

Explanation of sign

1 input device, 2 search type composition device, 3 search device, 4 storage device,
5 search solution synthesizer, 6 output device, 7 server, 8 network, 9 client, 10 deduplication device,
11 XML data continuous inquiry system, 12 Internet,
13 local network, 14 users, 15 proxy server

Claims (20)

A data management program for operating a computer connected to a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, and an output device for outputting data,
When a plurality of search expressions for retrieving a set of subtrees at various depths in the tree structure data stored in the storage device are given to the input device in the form of a path expression,
For all combinations of any non-empty subset of the set of all of the given multipath expressions and any subset of the set of all the prefixes of the given multipath expressions, The product of all of the multipath expressions contained in the former subset and all of the prefixes contained in the latter subset;
The sum of all of the multipath expressions not included in the former subset and all of the prefixes not included in the latter subset,
From the difference between these products and sums,
Further, a path expression corresponding to the difference obtained by removing all descendants (not including itself) of data matching all the sums of the given multiple path expressions is synthesized,
For each of these composite path expressions, the subtree set that is the solution set is extracted from the storage unit,
Of the sub-trees that are solutions of one or more path expressions of the given multi-path expressions, they are not sub-trees that form part of other solutions of any of the multi-path expressions. All of what is included in which path expression's solution set, not in which path expression's solution set, and which prefix of which path expression matches which prefix of which path expression Generate a set of groups that correspond to the words that do not match
A data management program for operating a computer so as to output these set groups to the output device. A data management program for operating a computer connected to a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, and an output device for outputting data,
When a search expression for retrieving a set of subtrees at various depths in the tree structure data stored in the storage device is given in the form of a path expression to the input device,
For all subsets of the set of all prefixes of the path expression,
The product of the given path expression and all of the prefixes contained in the subset;
The sum of all the prefixes not included in the subset,
From the difference between these products and sums,
Further, a path expression corresponding to the difference obtained by removing all descendants (not including itself) of data matching the given path expression is synthesized,
For each of these composite path expressions, the subtree set that is the solution set is extracted from the storage unit,
All subtrees that are solutions of the given path expression that do not form subtrees that form part of the other solutions of the path expression match any prefix of the path expression that matches the path Create a set of groups that are classified based on which prefixes in the expression do not match,
A data management program for operating a computer so as to output these set groups to the output device. A data management program for operating a computer connected to a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, and an output device for outputting data,
When a plurality of search expressions for retrieving a set of subtrees at various depths in the tree structure data stored in the storage device are given to the input device in the form of a path expression,
After extracting from the storage device a subtree set that is a solution set for each of the multipath expressions using the multipath expression,
Among subtrees that are solutions of one or more path expressions in the given multipath expression, a subtree that forms a part of another solution of any path expression of the multipath expressions Remove what appears, and for all remaining subtrees, it is included in which path expression solution set and not in which path expression solution set, and which path expression it is in Find out which prefix in the path expression matches which prefix, and classify based on the result of the above two points, and generate a set group corresponding to each of these classifications,
A data management program for operating a computer so as to output these set groups to the output device. A data management program for operating a computer connected to a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, and an output device for outputting data,
When a search expression for retrieving a set of subtrees at various depths in the tree structure data stored in the storage device is given in the form of a path expression to the input device,
After the subtree set that is the solution set is extracted from the storage device using the path expression,
Among the subtrees that are solutions of the path expression, those that also appear as subtrees that form part of other solutions of the path expression are removed, and for all remaining subtrees, the path tree Find out which prefixes match and which prefixes in the path expression do not match, and classify based on the results, generating a set of groups for each of these classifications,
A data management program for operating a computer so as to output these set groups to the output device. A data management program for operating a computer connected to an input device for receiving a set of data,
A subtree set group representing each classification output by the data management program according to claim 1 or 3 is received by the input device, and a solution set for each of the multiple path expressions initially given to the data management program is obtained. , From all the classifications that were supposed to be included in the solution set of the path expression, and from the subtree in each classification that was supposed to match each prefix of the path expression, the prefix and A data management program that operates a computer to compose a set of subtrees extracted using a pair of suffixes by taking the union of all of them. A data management program for operating a computer connected to an input device for receiving a set of data,
5. The subtree set group representing each classification output by the data management program according to claim 2 or 4 is received by the input device, and the path expression solution set initially given to the data management program A set of subtrees extracted from all subclasses and subtrees in each class that were supposed to match each prefix of the path expression using a suffix paired with the prefix in the path expression A data management program characterized by operating a computer to synthesize by taking all these unions. A data management program for operating a computer connected to a storage device that stores a plurality of data, an input device that receives a data retrieval formula, and an output device that outputs data, the program being stored in the storage device When a plurality of retrieval expressions for retrieving a subset of existing data are given to the input device,
For all non-empty subsets of the set of all the multiple search expressions,
A search expression corresponding to the difference between these products and the sum of all the products included in the subset of the plurality of search expressions and the sum of all of the plurality of search expressions not included in the subset. Synthesized,
For each of these composite search expressions, by extracting from the storage device the data set that is the solution set,
Based on which search expression solution set contains all the data that is the solution of one or more of the given search expressions, and which search expression solution set does not. A set group corresponding to the classification
A data management program for operating a computer so as to output these set groups to the output device. A data management program for operating a computer connected to an input device for receiving a set of data,
A set group representing each classification in the previous period output by the data management program according to claim 7 is received by the input device, and a solution set for each of the plurality of search formulas initially given to the data management program is obtained as the search formula. A data management program that operates a computer to synthesize by taking the union of all the classifications that were supposed to be included in the solution set. A data management program for operating a computer connected to a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, and an output device for outputting data,
When a plurality of retrieval formulas for retrieving a set of subtrees at various levels in the tree structure data stored in the storage device are given to the input device,
After extracting from the storage device a subtree set that is a solution set for each of the plurality of search expressions using the plurality of search expressions,
If one of the sub-expressions of the plurality of search expressions is a sub-tree that forms part of another solution of any one of the search expressions A data management program for operating a computer so as to output each solution set to the output device after removing the subtree that appears as a part of another solution from the solution set. A data management program for operating a computer connected to a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, and an output device for outputting data,
When a plurality of retrieval formulas for retrieving a set of subtrees at various levels in the tree structure data stored in the storage device are given to the input device,
Determine an appropriate total order for all of the given multiple search expressions,
For each of the multiple search expressions,
From the difference obtained by removing data that matches all search formulas having a younger order than the search formula from the search formula,
Further, a search expression corresponding to a difference obtained by removing all descendants (not including itself) of data matching any of the given multiple search expressions is synthesized,
For each of these composite search expressions, a subtree set as a solution set is extracted from the storage device,
Of the sub-trees that are solutions of one or more of the given search expressions, the sub-tree is not a sub-tree that forms part of another solution of any of the search expressions. A data management program characterized in that a computer is operated to take out all of them without duplication and output them to the output device. A data management system comprising a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, an output device for outputting data, and a computer connected thereto,
When a plurality of search expressions for retrieving a set of subtrees at various depths in the tree structure data stored in the storage device are given to the input device in the form of a path expression,
For all combinations of any non-empty subset of the set of all of the given multipath expressions and any subset of the set of all the prefixes of the given multipath expressions, The product of all of the multipath expressions contained in the former subset and all of the prefixes contained in the latter subset;
The sum of all of the multipath expressions not included in the former subset and all of the prefixes not included in the latter subset,
From the difference between these products and sums,
Further, a path expression corresponding to the difference obtained by removing all descendants (not including itself) of data matching all the sums of the given multiple path expressions is synthesized,
For each of these composite path expressions, the subtree set that is the solution set is extracted from the storage unit,
Of the sub-trees that are solutions of one or more path expressions of the given multi-path expressions, they are not sub-trees that form part of other solutions of any of the multi-path expressions. All of what is included in which path expression's solution set, not in which path expression's solution set, and which prefix of which path expression matches which prefix of which path expression Generate a set of groups that correspond to the words that do not match
A data management system, wherein a computer operates so as to output these set groups to the output device. A data management system comprising a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, an output device for outputting data, and a computer connected thereto,
When a search expression for retrieving a set of subtrees at various depths in the tree structure data stored in the storage device is given in the form of a path expression to the input device,
For all subsets of the set of all prefixes of the path expression,
The product of the given path expression and all of the prefixes contained in the subset;
The sum of all the prefixes not included in the subset,
From the difference between these products and sums,
Further, a path expression corresponding to the difference obtained by removing all descendants (not including itself) of data matching the given path expression is synthesized,
For each of these composite path expressions, the subtree set that is the solution set is extracted from the storage unit,
All subtrees that are solutions of the given path expression that do not form subtrees that form part of the other solutions of the path expression match any prefix of the path expression that matches the path Create a set of groups that are classified based on which prefixes in the expression do not match,
A data management system, wherein a computer operates so as to output these set groups to the output device. A data management system comprising a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, an output device for outputting data, and a computer connected thereto,
When a plurality of search expressions for retrieving a set of subtrees at various depths in the tree structure data stored in the storage device are given to the input device in the form of a path expression,
After extracting from the storage device a subtree set that is a solution set for each of the multipath expressions using the multipath expression,
Among subtrees that are solutions of one or more path expressions in the given multipath expression, a subtree that forms a part of another solution of any path expression of the multipath expressions Remove what appears, and for all remaining subtrees, it is included in which path expression solution set and not in which path expression solution set, and which path expression it is in Find out which prefix in the path expression matches which prefix, and classify based on the result of the above two points, and generate a set group corresponding to each of these classifications,
A data management system, wherein a computer operates so as to output these set groups to the output device. A data management system comprising a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, an output device for outputting data, and a computer connected thereto,
When a search expression for retrieving a set of subtrees at various depths in the tree structure data stored in the storage device is given in the form of a path expression to the input device,
After the subtree set that is the solution set is extracted from the storage device using the path expression,
Among the subtrees that are solutions of the path expression, those that also appear as subtrees that form part of other solutions of the path expression are removed, and for all remaining subtrees, the path tree Find out which prefixes match and which prefixes in the path expression do not match, and classify based on the results, generating a set of groups for each of these classifications,
A data management system, wherein a computer operates so as to output these set groups to the output device. A data management system comprising an input device for receiving a set of data and a computer connected to the input device,
14. A subtree set group representing each classification output by the data management system according to claim 11 or 13 is received by the input device, and a solution set for each of the plurality of path expressions initially given to the data management system is obtained. , From all the classifications that were supposed to be included in the solution set of the path expression, and from the subtree in each classification that was supposed to match each prefix of the path expression, the prefix and A data management system in which a computer operates to synthesize a set of subtrees extracted using a pair of suffixes by taking the union of all of them. A data management system comprising an input device for receiving a set of data and a computer connected to the input device,
15. The subtree set group representing each classification output by the data management system according to claim 12 or 14 is received by the input device, and the path-type solution set initially given to the data management system A set of subtrees extracted from all subclasses and subtrees in each class that were supposed to match each prefix of the path expression using a suffix paired with the prefix in the path expression A data management system characterized in that the computer operates to synthesize by taking the union of all of these. A data management system comprising a storage device for storing a plurality of data, an input device for receiving a data retrieval formula, an output device for outputting data, and a computer connected thereto,
When a plurality of retrieval formulas for retrieving a subset of data stored in the storage device are given to the input device,
For all non-empty subsets of the set of all the multiple search expressions,
A search expression corresponding to the difference between these products and the sum of all the products included in the subset of the plurality of search expressions and the sum of all of the plurality of search expressions not included in the subset. Synthesized,
For each of these composite search expressions, by extracting from the storage device the data set that is the solution set,
Based on which search expression solution set contains all the data that is the solution of one or more of the given search expressions, and which search expression solution set does not. A set group corresponding to the classification
A data management system, wherein a computer operates so as to output these set groups to the output device. A data management system comprising an input device for receiving a set of data and a computer connected to the input device,
18. A set group representing each classification in the previous period output by the data management system according to claim 17 is received by the input device, and a solution set for each of the plurality of search expressions initially given to the data management system is obtained as the search expression. A data management system in which a computer operates to synthesize by taking the union of all the classifications included in the solution set of. A data management system comprising a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, an output device for outputting data, and a computer connected thereto,
When a plurality of retrieval formulas for retrieving a set of subtrees at various levels in the tree structure data stored in the storage device are given to the input device,
After extracting from the storage device a subtree set that is a solution set for each of the plurality of search expressions using the plurality of search expressions,
If one of the sub-expressions of the plurality of search expressions is a sub-tree that forms part of another solution of any one of the search expressions A data management system, wherein a computer operates to output each solution set to the output device after removing the subtree that appears as a part of another solution from the solution set. A data management system comprising a storage device for storing one or more tree-structured data, an input device for receiving a data retrieval formula, an output device for outputting data, and a computer connected thereto,
When a plurality of retrieval formulas for retrieving a set of subtrees at various levels in the tree structure data stored in the storage device are given to the input device,
Determine an appropriate total order for all of the given multiple search expressions,
For each of the multiple search expressions,
From the difference obtained by removing data that matches all search formulas having a younger order than the search formula from the search formula,
Further, a search expression corresponding to a difference obtained by removing all descendants (not including itself) of data matching any of the given multiple search expressions is synthesized,
For each of these composite search expressions, a subtree set as a solution set is extracted from the storage device,
Of the sub-trees that are solutions of one or more of the given search expressions, the sub-tree is not a sub-tree that forms part of another solution of any of the search expressions. A data management system in which a computer operates so as to take out everything without duplication and output them to the output device.

Images