JP2017062710A - Search method, search program and search device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To shorten the time for extracting a path connecting methods to be an origin of a path search from among a plurality of methods in a calling relation.SOLUTION: The present invention relates to a search method for a computer to execute processing for performing, regarding a path connecting a plurality of methods to be an origin of a path search from among a plurality of methods in a calling relation, the path search in parallel for each method of the origin. The search method includes: if there is a plurality of paths subjected to the path search in an n-th hop from the plurality of methods to be the origin, referring to a storage part in which classes corresponding to the methods are stored, and calculating path priorities corresponding to two methods connected by the path; based on the path priorities, identifying a path to perform the path search thereon from among a plurality of paths subjected to the path search; and based on the identified path, performing the path search of the plurality of methods to be the origin of the path search.SELECTED DRAWING: Figure 8

Description

  The present invention relates to a search method, a search program, and a search device.

  In the analysis of program source code, a method of extracting a call-related path between a plurality of methods included in the program source code is known (for example, see Patent Document 1). According to this, when the source code is analyzed, the method calling relationship can be extracted in the form of a graph structure.

  The extraction of the call relationship path between methods may be used by, for example, a developer who adds a unique function to open source software (OSS) and provides it to a customer. The developer uses a method on the path of the extracted graph structure as an investigation starting point in order to grasp the contents of change and specify the necessary test range when upgrading the OSS. In particular, it is certain that there is a call-related route between a plurality of methods due to OSS version upgrades, etc., but when examining whether the route has changed from before, the above-mentioned call-related route is extracted. The method becomes important.

JP 2012-3752 A

  However, when the size of the source code increases, the search range for the path extraction of the call relation between the two methods to be extracted increases exponentially. There is a problem that extraction is not completed.

  Therefore, in one aspect, an object of the present invention is to shorten the extraction time of a route connecting a method that is a starting point of a route search among a plurality of methods having a calling relationship.

  One proposal is a search method in which a computer executes a process of performing a route search in parallel for each method of the starting point, connecting a plurality of methods that are starting points of route searching among a plurality of methods having a calling relationship. When there are a plurality of routes to be searched for the n-th hop from the plurality of methods that are the starting points, two methods connected by the route are referred to by referring to a storage unit that stores a class corresponding to the method A route priority corresponding to the route search is determined, a route for route search is identified from among the plurality of routes to be route searched based on the route priority, and the origin of the route search is determined based on the specified route. There is provided a search method for performing a route search for a plurality of methods.

  According to one aspect, it is possible to shorten the extraction time of a route connecting a method that is a starting point of a route search among a plurality of methods having a calling relationship.

The figure which shows an example of the graph structure of a source code and a call relationship. The figure which shows an example of the graph structure of a source code and method holding | maintenance relationship. The figure which shows an example in which the search range for extraction of a call related path | route increases exponentially. The figure which shows an example of a function structure of the search device concerning one Embodiment. The figure which shows an example of the call relation graph data storage part concerning one Embodiment. The figure which shows an example of the method holding | maintenance relationship graph data storage part concerning one Embodiment. The figure which shows an example of the graph structure of a calling relationship and a method holding | maintenance relationship. The flowchart which shows an example of the path | route extraction process concerning one Embodiment. The flowchart which shows an example of the edge search instruction | indication process concerning one Embodiment. The figure which shows an example of the route search progress concerning one Embodiment. The flowchart which shows an example of the graph edge search process concerning one Embodiment. The flowchart which shows an example of the path | route output process concerning one Embodiment. The flowchart which shows an example of the waypoint selection process concerning one Embodiment. The flowchart which shows an example of the calculation process of the class similarity concerning one Embodiment. The figure which shows an example of the route search progress concerning one Embodiment. The flowchart which shows an example of the priority search node selection process concerning one Embodiment. The figure which shows an example of the result of the route search concerning one Embodiment. The figure which shows an example of the effect of the route search concerning one Embodiment. The figure which shows the hardware structural example of the search device concerning one Embodiment.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In addition, in this specification and drawing, about the component which has the substantially same function structure, the duplicate description is abbreviate | omitted by attaching | subjecting the same code | symbol.
(Introduction)
First, an example of data conversion of a call-related graph structure will be described with reference to FIG. When the source code shown in Fig. 1 is analyzed, the Tranceiver.transmit () method of the Tranceiver class has three methods: the Transformer.encode () method, the Protocol.makeHeader () method, and the Tranceiver.send () method. Is calling. Among them, the method of Tranceiver.send () is a method of the same class (Tranceiver) as Tranceiver.transmit (). The method of Transformer.encode () and the method of Protocol.makeHeader () are methods of classes different from Tranceiver.transmit ().

  By analyzing the source code in this way, it is possible to create a graph structure indicating the method calling relationship. Here, the “node” of the graph represents a method, and the “edge” of the graph represents a call relationship from the method to the method. According to this, the relationship between which class holds which method can be converted into data in a graph structure.

  FIG. 2 shows an example of the source code and the graph structure of the method holding relationship. The source code shown in FIG. 2 is the same as the source code shown in FIG. By analyzing the source code, it is possible to create a graph structure indicating the method holding relationship of which class holds which method. For example, in Fig. 2, the nodes of the graph represent classes or methods, and the edges of the graph represent relationships where the class holds methods (when annotation is <has>) or implementation relationships (when annotation is <impl>) Represents. That is, in FIG. 2, the annotation <has> indicates that the Tranceiver class has a send method and a transmit () method. <Impl> indicates the relationship between classes.

  FIG. 3 shows a range in which the edge of the method b (node b) is searched for an outside edge centered on a node (hereinafter also referred to as “starting node”) that is the starting point of the route search. Each circle indicates 1 hop (first edge search), 2 hops (second edge search), and 3 hops (third edge search) in order from the inside. The distance from node b ≦ 1 at the time of 1 hop, the distance from node b ≦ 2 at the time of 2 hops, and the distance from node b ≦ 3 at the time of 3 hops.

The first hop has ^one of the methods d, the 2-th hop There are d ² pieces of the method, the 3-th hop there is a d ³ pieces of method. Therefore, the distance the total number of search already node at the time that was searched until r is the approximate d ^r number of methods.

Searching the above has been exploring the edges leaving outwardly method b as the starting point node, the search direction is the reverse direction (i.e., the side to enter the inside from the node) may also approximate 2d ^r pieces of method because there. As the number of hops increases, the number of methods increases exponentially. In other words, as the source code size increases, the search range for extracting call-related paths between two methods (hereinafter also referred to as “origin node s”) that are separated from the path extraction target exponentially increases. Thus, it becomes difficult to complete the extraction of the call relationship path between methods in a practical time.

  On the other hand, in the Dijkstra method, priorities are assigned based on edge weights, and a route between two methods to be extracted is searched. However, since the edge weights are all the same in the call-related graph structure, prioritization by the Dijkstra method does not work. Therefore, in the Dijkstra method, it becomes difficult to complete route extraction in a practical time when the size of the source code increases.

  The A star method searches for a route having a short distance to the goal setting. In a call-related graph structure, a predicted value cannot be obtained unless a new method for calculating a predicted value obtained by the A star method is added. For this reason, prioritization by the A star method does not function in a call-related graph structure. Therefore, even with the A-star method, it becomes difficult to complete the extraction of the path in a practical time when the source code becomes large.

  Furthermore, unlike the general route search on the graph structure, the call-related graph structure generally cannot determine which is the starting point (calling source) of the two starting point nodes to be extracted. Therefore, it is necessary to search for a route from both starting points.

  Therefore, in the present embodiment described below, a search method is proposed that shortens the extraction time of a route connecting a starting method among a plurality of methods having a calling relationship.

[Functional configuration of search device]
First, an example of the configuration of the search device 1 that executes the search method according to the present embodiment will be described with reference to FIG. FIG. 4 shows an example of a functional configuration of the search device 1 according to the embodiment. The search device 1 according to the present embodiment performs a route search in parallel for each node on a route connecting a plurality of origin nodes of the route search among a plurality of nodes having a calling relationship. The search device 1 can be a computer including a recording device and a CPU (processor) or another digital processing device. However, the configuration of the search device 1 is not limited to this. For example, the search device 1 may include a CPU and may be a device that can access the recording device via the Internet, a LAN (Local Area Network), or the like.

  The search device 1 includes a search instruction unit 10, a graph search unit 11, a starting point set (Z) storage unit 12, a storage unit 13, a class similarity calculation unit 14, an intermediate point node selection unit 15, and a priority search node selection unit 16. . The search instructing unit 10 instructs to search for a route connecting a plurality of origin nodes of the route search among a plurality of nodes having a calling relationship.

  The graph search unit 11 includes a graph edge search unit 111, a visited node set (open) storage unit 112, a visited node set (close) storage unit 113, and a preceding and subsequent node (prev · next) storage unit 114.

  The graph search unit 11 searches for a search path for two or more origin nodes in the call-related graph structure. The graph search unit 11 separately performs a search in the forward direction (hereinafter also referred to as “fwd”) from each starting node and a search in the backward direction (hereinafter also referred to as “bwd”) from each starting node. To do. For example, when there are three starting nodes, the graph search unit 11 performs six searches in parallel in the fwd and bwd directions from the three starting nodes, respectively. At this time, the graph search unit 11 synchronizes a plurality of searches performed in parallel (six searches in the above example) one hop at a time using a bulk synchronous parallel method or the like. However, the graph search unit 11 can also perform a plurality of searches asynchronously.

  The visiting node set (open) storage unit 112 stores nodes that need to be visited (searched) from now on. The visited node set (close) storage unit 113 stores visited (searched) nodes. The preceding and succeeding node (prev · next) storage unit 114 stores the relationship information with the preceding method (previous node) in the preceding node dictionary prev, and stores the relationship information with the succeeding method (following node) in the succeeding node dictionary next. . The graph edge search unit searches for a route based on the information stored in the visited node set storage unit 112, the visited node set storage unit 113, and the preceding and subsequent node storage units 114.

  The starting point set storage unit 12 stores starting point nodes. The starting point set Z of the starting point set storage unit 12 stores two starting point nodes to be extracted from the route, and, when an intermediate point node is selected, a specific intermediate point node.

  The storage unit 13 includes a call relationship graph data storage unit 131 and a method holding relationship graph data storage unit 132. As a result of analyzing the source code, the following data is stored in the call relationship graph data storage unit 131 and the method holding relationship graph data storage unit 132.

  The call relationship graph data storage unit 131 stores method call relationship information. FIG. 5 shows an example of the call relation graph data storage unit 131 according to the embodiment. Among the call-related graph structures obtained as a result of the source code analysis, information having a relation type “call” is included. The call relation graph data storage unit 131 includes data items of a source name (calling source method) 131a, a source type 131b, a destination name (calling destination method) 131c, a destination type 131d, and a relation type 131e.

  The method holding relationship graph data storage unit 132 stores the relationship between the class and the method and the relationship information between the class and the class. Of the table storing the call-related graph structure obtained as a result of the source code analysis, the relation type has information of “has” and “impl”. FIG. 6 shows an example of the method holding relationship graph data storage unit 132 according to the embodiment. Of the table storing the call-related graph structure obtained as a result of the source code analysis, the relation type has information of “has” and “impl”. The method holding relationship graph data storage unit 132 includes data items of a source name 132a, a source type 132b, a destination name 132c, a destination type 132d, and a relation type 132e.

  When there are a plurality of routes to be searched for the n-th hop from a plurality of origin nodes, the class similarity calculation unit refers to the storage unit 13 that stores the classes corresponding to the plurality of routes, and the two connected by the route Class similarity corresponding to two or more is calculated.

  The midpoint node selection unit 15 is also referred to as a midpoint method (hereinafter referred to as a midpoint node) serving as a starting point for a road search. ). When a plurality of waypoints are noded, the waypoint node selection unit 15 specifies the priority of the waypoint for performing a route search from among the plurality of waypoints based on the class similarity.

  The priority search node selection unit 16 refers to a storage unit that stores a class corresponding to a method when there are a plurality of routes to be searched for an n-hop route from the origin node, and the two nodes connected by the route The route priority corresponding to (method) is calculated.

  Based on the route priority, the graph search unit 11 specifies a route to be searched from among a plurality of routes to be searched for a route, and performs a route search for a plurality of starting nodes based on the specified route.

  For example, as shown in FIG. 7, a graph structure of the method call relationship held by the call relationship graph data storage unit 131 is created based on the source code group. Also, a method holding relationship graph structure held by the method holding relationship graph data storage unit 132 is created based on the source code group.

  The search device 1 according to the present embodiment is the shortest connecting the two origin nodes (method Q.b and method Vi in FIG. 7) identified in the graph structure of the method call relationship of FIG. Extract the search route.

[Route extraction process]
Below, the route extraction process among the search processes performed by the search device 1 according to the present embodiment will be described with reference to FIG. Thereafter, the edge search instruction process (FIG. 9), the graph edge search process (FIG. 11), the route output process (FIG. 12), the intermediate point selection process (FIG. 13), and the class similarity among the search processes executed by the search device 1 Each of the degree calculation process (FIG. 14) and the priority search node selection process (FIG. 16) will be described.

  When the route extraction process of FIG. 8 is started, the search instruction unit 10 sets 0 to a counter (count). The counter counts the number of search hops. The search instruction unit 10 instructs to search for a route a → b. “A” and “b” indicated by the route a → b include the starting node of the route extraction target. “A” and “b” may be intermediate point nodes.

  Next, the search instruction unit 10 sets so as to return elements of a set of {“fwd”, “bwd”} for each route a → b (step S101). In addition, the search instruction unit 10 sets the detection result dictionary p to return an empty set {} for all routes (step S101). The detection result dictionary p is a dictionary that holds the result of whether or not there is a desired route. In an initial state where the desired route is not found, the detection result dictionary p is an empty set {}.

  The detection result dictionary p is used when a path is found in the forward direction (hereinafter also referred to as “fwd”) when the desired path is found (for example, from the starting node to another starting node or intermediate point). When the node reaches the node, the search instruction unit 10 stores the element “fwd” in the corresponding route (for example, route a → b). In the detection result dictionary p, when a path is found in a backward direction (hereinafter also referred to as “bwd”), an element “bwd” is stored in the corresponding path (for example, path a → b).

  Next, the search instruction unit 10 specifies the starting node s1 and instructs the graph search unit 11 to execute the edge search instruction process of FIG. 9 (step S102). In addition, the search instruction unit 10 specifies the starting node s2 and instructs the graph search unit 11 to execute the edge search instruction process (step S102).

(Edge search process)
The edge search instruction process of FIG. 9 is started, and the graph edge search unit 111 adds the specified starting node s as an element to the starting point set Z (step S201). The starting point set Z stores the node that is the starting point at that time. At this time, the starting point set Z may store starting point nodes s1 and s2. For example, the route extraction target Q.D. b and V. The method i is an example of the starting nodes s1 and s2. When a node at an intermediate point described later (for example, the method of Vi, Uf in FIG. 10) is selected, the starting point set Z stores the nodes at these intermediate points together with the starting point nodes s1 and s2. . The intermediate point stored in the starting point set Z becomes the starting point of the route search as one of the starting point nodes.

  Next, the graph search unit 11 creates an instance process i1, gives graph = G, starting node = s, direction indication = “fwd” as inputs to the instance process i1, and the graph edge of FIG. The asynchronous execution of the search process is instructed (step S202). Also, an instance process i1 of the graph search unit is created, and the instance process i1 is given graph = G, starting node = S, direction instruction = “bwd” to instruct asynchronous execution of the graph edge search (step S203). . This is the end of this process.

  The edge search when the search direction in step S202 is the fwd direction and the edge search when the search direction in step S203 is the bwd direction are performed in parallel for each instance process il of another process.

  When the edge search instruction process in FIG. 9 ends, the process returns to step 103 in FIG. 8, and the search instruction unit 10 increments the counter by one (step S103). In steps S104 to S110, the route search status of each instance process i1 is confirmed. The graph search unit 11 performs a search corresponding to each starting point node and intermediate point node in parallel in the edge search instruction process of FIG. 9 and notifies the search report. The search instruction unit 10 receives a search report from the graph search unit 11.

  Specifically, the search instruction unit 10 determines whether or not the completion reports of both “fwd” and “bwd” have been received from all nodes of the starting point set Z (step S104). The search instruction unit 10 receives a completion report from the graph search unit 11 when there is no direction to advance from the starting node. If it is determined that the completion report has been received, the search instruction unit 10 proceeds to step S118, outputs a “no route” notification, and ends this process (step S119).

  On the other hand, when the search instruction unit 10 determines that there is a node that has not received the completion report among all the nodes of the starting point set Z, the direction instruction “fwd” is obtained from all the elements s (each s∈Z) of the starting point set Z. It is determined whether “waiting for instructions” for both “” and “bwd” has been notified (step S105). For example, the graph search unit 11 notifies “waiting for instruction” when the search of “fwd” and “bwd” of one hop indicated by each circle in FIG. The graph search unit 11 does not proceed to route search for the next hop until it receives a search restart instruction from the search instruction unit 10 and enters a waiting state.

  When the search instruction unit 10 determines that “waiting for instruction” of “fwd” or “bwd” is not notified from at least one of the elements s of the starting point set Z, the search instruction unit 10 reports the detection of the forward route a → b. It is determined whether it has been received (step S106). For example, when a path (side) indicated by an arrow in the graph of FIG. 10 is found from the starting node, the search instruction unit 10 receives a report of the forward path a → b (fwd). In this case, an element of fwd is added to the path a → b of the detection result dictionary p (step S107), and the process proceeds to step S108. As a result, the detection result dictionary p stores the detection result that the side connecting the node a to the node b is in the fwd direction. On the other hand, when the side of the forward path is not found from the starting node, the search instruction unit 10 determines whether a report of detection of the backward path a → b has been received (step S108).

  In step S108, when the search instruction unit 10 receives the report of the backward route a → b (bwd), the search instruction unit 10 adds the element bwd to the route a → b of the detection result dictionary p (step S109), and returns to step S104. As a result, the detection result dictionary p stores the detection result that the side connecting the node a to the node b is in the bwd direction.

  In step S108, when the search instruction unit 10 does not receive the report of the backward route a → b (bwd), the search instruction unit 10 immediately returns to step S104.

  When steps S104 to S109 are repeated and the search instruction unit 10 determines that “waiting for instructions” for both “fwd” and “bwd” is notified from all the elements s of the starting point set Z, the process proceeds to step S110.

  In step S110, the graph search unit 11 determines whether or not the origin nodes s1 and s2 are reachable by combining the detection results stored in the detection result dictionary p. In addition, the graph search unit 11 creates a set T of route groups constituting the arrival route.

  When it is determined that the starting node s1, s2 is not reachable (step S111), the search instruction unit 10 determines whether the counter is equal to or greater than a predetermined value (step S120). If it is determined Yes in step S105 and No in step S111 following step S110, the search instruction unit 10 determines whether the counter is equal to or greater than a predetermined value in step S112 (step S120). If it is determined that the counter is smaller than the predetermined value, the process returns to step S103, 1 is added to the value of the counter, and steps S104 to S109 are repeated.

  When it is determined that the counter is equal to or greater than the predetermined value, the search instruction unit 10 instructs the intermediate point node selection unit 15 to execute the intermediate point selection process shown in FIG. I is designated as the starting node and the asynchronous execution of the edge search instruction process shown in FIG. 9 is instructed. The waypoint selection process shown in FIG. 13 will be described later. For example, in the call-related graph structure of FIG. An example is shown in which the method of g is selected as an intermediate node.

  Returning to FIG. 8, the counter is set to 0 in step S122, 1 is added to the counter in step S103, steps S104 to S109 are repeated, and the process proceeds to step S110. Here, it is assumed that the starting point set Z includes intermediate points. In the example of FIG. b, Method V. i, method V. All of g are starting nodes. Method Q. b, Method V. i, method V. Steps S104 to S109 are performed on the starting node of g. As a result, in the example of FIG. 10, P (Q.b → V.g) = {“fwd”} is stored in the detection result dictionary p. However, at this point, the origin node method Q. b, Method V. i is not reachable. Therefore, it determines with No in step S111, determines with No in step S120, repeats step S104-S109 again, and progresses to step S110. At this time, in the example of FIG. 10, in addition to P (Q.b → V.g) = {“fwd”}, P (V.g → V.i) = {“bwd” is added to the detection result dictionary p. "} Is added and saved. Therefore, the set T includes {Q. b → V. g, V.D. g → V. i} is stored. At this time, as shown in FIG. b, Method V. i can be reached. Therefore, the search instruction | indication part 10 determines with Yes in step S111, and determines whether the set T is an empty set (step S112).

  When it is determined in step S112 that the set T is not an empty set, the search instruction unit 10 extracts one element t of the set T, deletes t from the set T, and defines t as an edge of a → b (step S113). Next, the search instruction unit 10 determines whether there is an element of fwd in the path a → b of the detection result dictionary p (step S114).

  When it is determined that there is an element of fwd in the path a → b of the detection result dictionary p, the search instruction unit 10 sends the arrival node b and the search direction to the instance process responsible for the starting node a created by the graph search unit 11. “Fwd” is input to instruct execution of the route output process of FIG. 12 (step S115).

  When it is determined that there is no fwd element in the path a → b of the detection result dictionary p, the search instruction unit 10 sends the arrival node b and the search direction to the instance process in charge of the starting node a created by the graph search unit 11. “Bwd” is input to instruct execution of the route output process of FIG. 12 (step S116).

(Route output processing)
The route output process executed in steps S115 and S116 will be described with reference to FIG. Assume that the arrival node (node to be a goal) to be route output at the start of this output process is g, and that the search direction d is one element of the directions “fwd” and “bwd”. Here, the best return path from the goal node g is stored in the preceding and succeeding node (prev · next) storage unit 114.

  When this process is started, the graph edge search unit 111 sets the node g of the goal to t and sets an empty list to the edge list P (step S401). The list P is a list of edges from the origin node to the arrival node.

  Next, the graph edge search unit 111 determines whether or not prev (t) of t that is the target of the search at this time is null (step S402).

  When the graph edge search unit 111 determines that prev (t) is not null, the edge of prev (t) → t is added to the list P (step S403), and prev (t) is set to T (step S404). ). For example, in the case of the call-related graph structure of FIG. where g is the previous node. e. In this case, the method P. e → Method V. The edge of g is added as an element, and the method S.G. e is set.

  Returning to step S402 and repeating steps S402 to S404, in the example of FIG. b, Method Q. b can be traced. t is N.N. In the case of b, S. e → V. g, N.I. b → S. The side of e is added as an element. Further, steps S402 to S404 are repeated, and t becomes Q.Q. In the case of b, S. e → V. g, N.I. b → S. e, Q. b → N. The side of b is added as an element.

  In step S402 to be executed next, the graph edge search unit 111 determines that prev (t) is null, reports the list P to the search instruction unit 10 (step S405), and ends this process.

  In the above description, when the direction d is “fwd”, it is determined in step S402 whether prev (t) is null, and the search result path is traced backward. The same processing is performed when the direction d is “bwd”. In step S402, it is determined whether next (t) is null, and the search result is based on the preceding and succeeding node (prev · next) storage unit 114. Follow the path forward.

  Returning to FIG. 8, the processes in steps S112 to S116 are repeated, and the element t is deleted one by one from the set T. Finally, when all the elements are extracted from the set T, in step S112, the search instruction unit 10 It is determined that the set T is an empty set, and the process proceeds to step S117. The search instruction unit 10 connects the output route group and outputs it as a route of a → b (starting node s1 → s2, method Q.b → Vi in FIG. 10) (step S117). Exit.

  In step S104, when the search instruction unit 10 receives completion reports of both “fwd” and “bwd” from all nodes of the starting point set Z, it outputs a “no route” notification (step S118). This process ends.

[Intermediate point selection process / class similarity calculation process]
(Intermediate point selection process)
Next, the waypoint selection process will be described with reference to FIG. In the intermediate point selection process of FIG. 13, the intermediate point node selection unit 15 selects a method that becomes a node of the intermediate point between the starting point nodes included in the call-related graph structure.

  This process is executed in step S121 of FIG. The intermediate point node selection unit 15 refers to the method holding relationship graph data storage unit 132, and refers to the source name whose relation type 132e is “impl”, destination type 132d is “interface”, and source type 132b is “class”. 132a is stored in set C (step S501). As a result, the set C stores a method connected to the interface in the relationship of “impl”. For example, in the example of FIG. 6, classes U, V, and W are stored in the set C.

  Next, the midpoint node selection unit 15 clusters the elements of the set C using the K-Means method (step S502). At this time, the intermediate point node selection unit 15 divides elements having a short distance into the same group. The distance between elements can be calculated in the class similarity calculation process shown in FIG. Therefore, the waypoint node selection unit 15 instructs the class similarity calculation unit 14 to execute the class similarity calculation process, and clusters the elements of the set C based on the class similarity obtained as a result (step S502). ).

  Next, the intermediate point node selection unit 15 selects a cluster L having more elements. The intermediate point node selection unit 15 selects a class D having a large number of methods from the selected cluster L, selects a method M that appears least in other classes in the cluster L from the methods of the class D (step S503), The process ends. However, in step S503, the intermediate point node selection unit 15 may select a class D having a small number of methods from the selected cluster L. Further, the intermediate point node selection unit 15 may select a method M that appears most frequently in other classes in the cluster L among the methods of the class D.

  According to this, the intermediate point node is a method whose destination type corresponds to the class of the interface, and a method as far as possible from the starting point node is determined.

(Class similarity calculation process)
The class similarity calculation process executed in step S502 will be described with reference to FIG. First, the class similarity calculation unit 14 refers to the method holding relationship graph data storage unit 132, the relation type 132e is “has”, the source type 132b is “class”, and the source name 132a is “C”. The destination name 132c of the row is stored in the set M (step S701). When the class V is set to the set C as an input value of this processing, referring to the method holding relationship graph data storage unit 132 shown in FIG. 6, the set M includes methods g, h, i, j, k, and l. Is stored.

  The class similarity calculation unit 14 collects from the method holding relationship graph data storage unit 132 a destination name 132c whose relation type 132e is “has”, source type 132b is “class”, and source name 132a is “D”. (Step S702).

  In FIG. 6, the number of methods of class U is 1, the number of methods of class V is 6, and the number of methods of class W is 5. Therefore, when the class W is set in the set D as an input value of this process, referring to the method holding relationship graph data storage unit 132 shown in FIG. 6, the set N includes methods h, i, j, k, and l. Is stored.

  Next, the class similarity calculation unit 14 outputs the number of names that appear only in the set M and the number of names that appear only in the set N. When the set M = {g, h, i, j, k, l} and the set N = {h, i, j, k, l}, the name that appears only in the set M is “g”, and the number thereof is 1. The number of names that appear only in set N is zero. In this case, the class similarity calculation unit 14 outputs “1” as the class similarity between C (class V) and D (class W). The numerical value indicated by the class similarity indicates how similar the methods of the class are. In other words, the numerical value indicated by the class similarity indicates the distance between classes.

  Note that the name similarity determination method in the class similarity calculation processing is determined to be the same if it is completely matched, or is determined to be the same if it is completely lowercased and completely matched, and is the same if the edit distance between the character strings is within the specified value. Or the like can be used.

  14, it is assumed that the class similarity (U, V) is calculated as 7, the class similarity (U, W) as 6, and the class similarity (V, W) as 1. At this time, when classifying into two clusters in step 502 of FIG. 13, the elements of the set C = {U, V, W} are classified into {U} and {V, W} based on the class similarity. The In step 503 of FIG. 13, a cluster L = {V, W} having more elements is selected. As the class D, the class V having a larger number of methods is selected from the class V and the class W of the cluster L. In the method M, among the methods of the class V selected for the class D, the method V. which appears the least in the other class W in the cluster. g is selected.

  According to this, by selecting one waypoint method for each cluster, a method having a low class similarity is selected as a node of the waypoint. As a result, as shown in FIG. By selecting g, it is possible to select a method that is separated from the method that is already the starting node in the route search as the intermediate node. Thereby, an intermediate point node can be set at a position where the distance from the starting point node is not too close, and the efficiency of route search can be improved.

  In the case of Java (registered trademark) source code, the intermediate point node is selected from a method of a class that implements an interface. In the case of C ++ source code, a method that is connected to a class that implements a pure virtual class by impl Chosen from.

  After the midpoint setting is selected in step S121 of the route extraction process of FIG. 8, as shown in FIG. b and the method of the starting node V. i and waypoint node V. A route search is performed starting from g.

  In the present embodiment, the route search between the starting points is synchronized so that each starting point starts the search for the next second hop after each starting point ends the search for the first hop. Thereby, the timing which elects an intermediate point can be taken appropriately. However, the route search for each starting point may proceed without being synchronized.

[Graph edge search processing / Priority search node selection processing]
(Graph edge search processing)
Next, the graph edge search process will be described with reference to FIG. This process is executed in steps S202 and S203 of the edge search instruction process in FIG. Note that one of the directions “fwd” and “bwd” is stored in the direction indication variable dir in advance. In the following, a set of nodes in the call-related graph structure is referred to as a node set V.

  The graph edge search unit 111 returns a dictionary d that returns the distance from the starting node s (for example, s1, s2, intermediate point nodes, etc.) for each node of the call-related graph structure, and ∞ for all nodes. (Step S301). The distance from the starting node s is the number of sides between the starting node and the node to be searched. In the dictionary d, whether or not it is possible to reach the intermediate node or another starting node s from the starting node s, or the distance (number of sides) indicates some information. The dictionary d is set to return ∞ if it cannot be reached or is unknown.

  Next, the graph edge search unit 111 returns the end point u of the input edge u → v if v is a node on the path with respect to the starting node (including the intermediate node) for each node v of the node set V, and so on. Otherwise return null. The graph edge search unit 111 sets the preceding node dictionary prev to return null for all nodes. The preceding node dictionary prev is a dictionary that holds information on the node immediately before the current search position. At this time, null is set in the preceding node dictionary prev indicating that the node immediately before the current search position has not been found.

  The above description is for the case where the direction indicating variable dir is “fwd”. If the direction indication variable dir is “bwd”, in step S302, the graph edge search unit 111 returns the end point v of the output edge v → u, otherwise returns null. The graph edge search unit 111 sets the subsequent node dictionary next so as to return null for all nodes.

  Next, the graph edge search unit 111 sets the origin node s (for example, the method Q.b and the method Vi, the intermediate point node, etc. of the route extraction target in FIG. 15) to the visited node set open (step S303). ). Hereinafter, of the starting nodes, the two methods for route extraction are also referred to as starting nodes s1 and S2.

  In addition, the graph edge search unit 111 sets 0 (that is, the distance from itself is 0) to the dictionary d (s) that returns the distance (step S303). For example, when the starting nodes are only S1 and S2 (that is, when the starting node does not include an intermediate point), “fwd” and “bwd” of the starting node s1, 4 of “fwd” and “bwd” of the starting node s2 One route search is processed in parallel. For example, when n waypoint nodes are selected, 2 (2 + n) route searches are processed in parallel.

  Next, the graph edge search unit 111 sets the visited node set close to an empty set (step S304). Next, the graph edge search unit 111 sets the next visited node set newopen to an empty set (step S305). In the next visited node set newopen, the next visited node is stored. In addition, the graph edge searching unit 111 transfers the starting point node stored in the starting point set storage unit 12 to ZL (step S305). As an example, if the starting node s is the method Q. b and the direction indicating variable dir is “fwd”, the {N. b} is set, and d (Q.b) = 0 is added to the dictionary d (s) that returns the distance. The visited node set “close” and the next visited node set “newopen” are in an empty set state.

  Next, the graph edge search unit 111 determines whether or not the visited node set open is not an empty set (step S306). Here, the starting node s1 and S2 are stored in the visiting node set open required. At this time, the graph edge search unit 111 determines that the visited node set open is not an empty set (Yes), and sets c having the smallest distance d (c) among the elements c included in the visited node set open required. One is selected, and the node (method) c is visited t (step S307). Further, the graph edge search unit 111 sets the excluded node set X (the set of nodes selected in the node S309 having a high priority) to an empty set. In the excluded node set X, excluded nodes at the time of selecting nodes with high priority are held.

  Next, the graph edge search unit 111 removes the search target t (currently visited method m) from the visited node set open and adds it to the visited node set close (step S308). At this time, there is no element in the visiting node set open required (hereinafter also referred to as “set name = {}”). Also, visited node set close = {Q. b}, next visited node set newopen = {}, excluded node set X = {}, t = Q. b, prev (N.b) = Q. b.

  Next, a search target t = method m, a direction indicating variable dirε {“fwd”, “bwd”}, and an excluded node set X = {} are input to instruct execution of a priority search node selection process (step S309).

(Priority search node selection processing)
The priority search node selection process will be described with reference to FIG. The priority search node selection unit 16 refers to the call relation graph data storage unit 131 shown as an example in FIG. The priority search node selection unit 16 has the relation type 131e = “call”, the source type 131b = “method”, the destination type = “method”, and the source name 131a = m from the call relation graph data storage unit 131. The destination name 131c of the row is stored in the set N (step S601).

  The above shows the case where the direction indicating variable dir is “fwd”. When the direction indication variable dir is “bwd”, the relation type 131e = “call”, the source type 131b = “method”, the destination type = “method”, and the destination name 131c from the call relation graph data storage unit 131. The source name 131a of the row where = m is stored in the set N.

  Next, the priority search node selection unit 16 removes the elements of the excluded node set X from the set N (step S602). Next, the priority search node selection unit 16 refers to the method holding relationship graph data storage unit 132 shown in FIG. 6 as an example, and the source of the row with the relation type 132e = “has” and the destination name 132c = m The name 132a is stored in c (step S603).

  Next, the priority search node selection unit 16 for each element n in the set N, from the method holding relationship graph data storage unit 132, the source of the row with the relation type 132e = “has” and the destination name 132c = n The name 132a is substituted for d (step S604).

  Next, the class similarity calculation unit 14 executes the class similarity calculation process of FIG. 14 to calculate the similarity between d and c. As a result, the priority search node selection unit 16 selects the method m ′ of the class d ′ having the highest similarity (that is, the lowest similarity numerical value) as a search target node (step S604), and ends this processing. To do.

  Thereby, the path priority corresponding to the two methods connected by the path is calculated with reference to the storage unit 13 that stores the class corresponding to the method.

  For example, as an input value of this process, the method m is V.V. i, when the direction indication variable dir is “bwd”, the call relationship graph data storage unit 131 shown as an example in FIG. 5 is referred to in step S601, and the set N = {X. i, Wd. i}. Also in the next step S602, the set N = {X. i, Wd. The state of i} does not change. In the next step S603, c = V (class). In the next step S604, d = X, Wd for each element n (= X.i, Wd.i) in the set N. The similarity between c (= V) and d (= X, Wd) is similarity (X, V) = 5 and similarity (Wd, V) = 1. As a result, the class d ′ having the highest similarity (that is, the lowest similarity value) is “Wd”, and the method m ′ is “Wd.i”. Thereby, for example, in FIG. i is a method X. The route search is prioritized over i. Note that the similarity calculated in the priority search node selection process is an example of a route priority.

  Returning to FIG. 11, the graph edge search unit 111 determines whether a search target node is selected in step S <b> 310. When the graph edge search unit 111 determines that the search target node has not been selected, the graph edge search unit 111 returns to step S306 and repeats steps S306 to S310. When the graph edge search unit 111 determines that the search target node has been selected, the graph edge search unit 111 proceeds to step S311 and adds the node n having the highest priority selected by the priority search node selection unit 16 to X as an element (step S311). .

  Next, the graph edge search unit 111 determines whether the node n is not included in the elements of the visited node set close and satisfies d (n)> d (t) +1 (step S312). If the graph edge searching unit 111 determines that the node n is included in the elements of the visited node set close, the process proceeds to step S314. When the graph edge searching unit 111 determines that d (n) ≦ d (t) +1, that is, when the distance d (n) of the node n is determined to be the shortest distance from the starting point, The process proceeds to step S314. On the other hand, if the graph edge searching unit 111 determines that the node n is not a visited node and the distance d (n) is not the shortest distance from the starting point, the graph edge searching unit 111 sets d (t) +1 to the distance in step S313. The distance d (n) is rewritten by substituting for d (n). As a result, the origin node can be traced with a shorter route.

  Further, the graph edge searching unit 111 sets the value of t to the preceding node prev (n) of the preceding and succeeding node (prev · next) storage unit 114, so that the node t immediately before the node n is the node t. Rewrite. Further, the graph edge search unit 111 adds the node n of the next visit destination to the node set newopen of the next visit destination. At this time, for example, prev (N.b) = Q. b is added, and the node set newopen at the next visit destination is N.P. b is set.

  In step S314, the graph edge search unit 111 determines whether n is ZL (ZL: a starting point or a node at an intermediate point). In S305, the contents of the starting point set storage unit Z are transferred to ZL. Therefore, if the graph search unit 11 determines that n is ZL, the graph search unit 11 reports to the search instruction unit 10 that a forward path of s → n has been detected (step S315), and returns to step S309. Thereby, it is reported to the search instructing unit 10 that a certain starting point or intermediate point has reached another starting point or intermediate point.

  On the other hand, if the graph edge searching unit 111 determines that n is not ZL, the graph edge searching unit 111 immediately returns to step S309. In step S315, when the direction indicating variable dir is “bwd”, the search instructing unit 10 is notified that the n → s forward path has been detected.

  Returning to S309, if the elements are accumulated in the excluded node set X and it is determined in S310 that no node can be selected, the graph edge search unit 111 determines whether there is an element in the visited set “open” (required) Step S306). When it is determined that there is an element in the visited destination set open, the graph edge search unit 111 performs the processing after step S307. When it is determined that there is no element in the visited destination set open, the graph edge search unit 111 proceeds to step S316 and notifies the search instruction unit 10 of “wait for instruction to resume execution”. The graph edge search unit 111 waits for an instruction to resume execution from the search instruction unit 10 (step S317).

  The graph edge search unit 111 determines whether or not an instruction to resume execution has been received from the search instruction unit 10 (step S318), and waits until an instruction to resume execution is received. If the graph edge search unit 111 determines that an instruction to resume execution has been received, the graph edge search unit 111 determines whether there is no element in the node set newopen for the next visited site (step S319). When it is determined that there is no element in the next visited node set newopen, the graph edge searching unit 111 proceeds to step S321 because the next visited node set newopen does not include the next visited node n. Report completion to the search instruction unit 10. When it is determined that there is an element in the next visited node set newopen, the graph edge searching unit 111 sets the value of the next visited node set newopen to “open” (step S320), and the process returns to step S305.

  As a result, the graph edge search unit 111 ends the first hop of the starting node and starts the next second hop search. The visited destination set open contains the nodes found in the previous route search in the second hop search. In the example of FIG. 15, open = {Q. b} In the second hop, open = {N. b, Y. m}.

  In this state, the graph edge search unit 111 executes the processes of steps S305 to S316. At this point, prev (Se) = N. b is added, and the node set newopen at the next visit destination is S.B. e is added.

  In step S317, the graph edge search unit 111 waits for an “execution restart” instruction from the search instruction unit 10 and determines that the instruction is received in step S318. In step S319, the next visited node set newopen is determined. Determine if it is not empty. At this time, S.P is included in the next visited node set newopen. There are elements of e. For this reason, the graph edge search unit 111 substitutes the value of newopen into the visited node set open, and repeats the processing of steps S305 to S316. By this repeated processing, prev (U.f) = S. e is added, and the next visited node set newopen is U.P. f is added. In the next iteration, prev (V.g) = S. e is added, and the next visited node set newopen is U.P. f, V. g is added.

  Here, in the example of the graph of FIG. 15, the method V. Since g is an intermediate node, the forward path Q.g. b → V. The forward path of g is found. Therefore, the graph edge search unit 111 reports that the forward path has been detected to the search instruction unit 10 in step S315. As a result, the forward route and the updated route obtained by the edge search are accumulated in the preceding and succeeding node (prev · next) storage unit 114. Thereby, for example, the method Q. of the starting node which is the path extraction target indicated by the thick line in FIG. b and method V. The path between i is extracted.

As explained above, according to the search method which concerns on one Embodiment, the following 1-4 are performed.
1. In addition to the two methods that are the origin nodes of the route extraction target, a plurality of intermediate node methods (P) are prepared, and the route search is performed in parallel starting from P + 2 methods (corresponding graph nodes). . The route search is performed not only in the forward direction (direction fwd toward the call destination) but also in the reverse direction (direction bwd toward the call source). Search until a node other than the starting point is reached, and finally combine the results of the route search.
In 2.1, the search is started by narrowing down to a small number of P intermediate point nodes, and after the execution of several steps, if the target two methods (nodes) are not connected by the route, the route search is performed by gradually increasing the intermediate point nodes.
In order to prioritize the waypoint nodes starting the search in 3.2, use the low similarity (with respect to the method name list) of the class (holding the method) corresponding to the node.
In 4.1, when there are a plurality of sides for which the route search is advanced, the route priority is calculated as the prioritization for determining the side to be investigated first. The calculation of the route priority uses that the class similarity corresponding to the node is high. By determining the side to be investigated first based on the route priority, it is possible to reduce the time required to extract the call relationship route between two distant methods that are the starting point of the route search among a plurality of methods having a call relationship.

  Further, (a) 1. In the intermediate point method, a method that is likely to be a method on the route is selected as follows. Based on the storage section that has the method retention relationship analyzed from the source code, select the class that is used for the separation and loose coupling of interest in the programming language specification. For example, in Java, a class that implements an interface is selected, and in C ++, a class that implements a pure virtual class is selected. Alternatively, the class specified by dependency injection may be selected.

  (B) 1. It is preferable that the route search performed in parallel is performed by a bulk synchronous parallel method that synchronizes the route search from each starting point for each hop, for example.

  (C) 2. Narrowing is performed as follows. 1. Cluster the class that holds the waypoint method selected in step 1 by similarity. At that time, a K-Means method or the like can be used. Select one class from the resulting cluster, and select the method group held by that class as the midpoint node.

  (D) 3. The prioritization is performed as follows. 2. Each class narrowed down in step 2 is compared with the class similarity to which the route extraction target two methods belong, and the class that is not similar to any of the route extraction targets (that is, the similarity is low) is sequentially selected. Select the method group held by the class as the intermediate node.

  (E) 4. Prioritization of the sides to be investigated is performed as follows. Assume that the currently visited node is v, the side where the route search is advanced is v → w, and the classes holding v and w are c and d, respectively. The sides v → w are selected in descending order of similarity between c and d.

  3. 4. The class similarity in the above (c), (d), and (e) is calculated as follows. For each class, create a sorted list of method names (strings) that the class has. The edit distance (Levenshtein distance) of this list is the class similarity (assuming that the smaller the value, the higher the similarity).

  As described above, according to the search device 1 according to the present embodiment, it is possible to shorten the extraction time of a route connecting a starting method among a plurality of methods having a calling relationship.

  As shown in FIG. 18, it is assumed that the distance between route search target methods is r, and there are P intermediate point methods on the route. The distance f from each starting point method to the adjacent starting point method is r / (1 + 1P).

  The total number n (P) of the callee (or original) methods that visit from each starting method to the adjacent starting method is best calculated by equation (1).

  The total number of methods N (P) visited as a whole is the best and is calculated by equation (2).

For example, the total number _NOM of methods visited by the following well-known methods Dijkstra method [1] and A * method [2] is calculated by equation (3).
Dijkstra method [1] Dijkstra, EW (1959) .A note on two problems in connexion with graphs. In Numerische Mathematik, 1 (1959), S. 269-271.
A * method [2] Peter E. Hart; Nils J. Nilsson; Bertram Raphael (July 1968). "A Formal Basis for the Heuristic Determination of Minimal Cost Paths". IEEE Transactions on Systems Science and Cybernetics 4 (2): 100 -107.doi: 10.1109 / TSSC.1968.300136.ISSN0536-1567.

Therefore, the total number ratio N (P) _{/ N OM} methods visit is calculated by approximation 1 / ^{d r} using equation (4). In other words, according to the search method according to the present embodiment, it is possible to shorten the extraction time of a route connecting a method that is a starting point of route search among a plurality of methods having a calling relationship.

(Hardware configuration example)
Finally, the hardware configuration of the search device 1 according to the present embodiment will be described with reference to FIG. FIG. 19 shows an example of the hardware configuration of the search device 1 according to this embodiment. The search device 1 includes an input device 101, a display device 102, an external I / F 103, a RAM (Random Access Memory) 104, and a ROM (Read Only).
Memory (105), CPU (Central Processing Unit) 106, Communication I / F 107, HDD (Hard Disk Drive) 108, etc. are connected to each other via bus B.

  The input device 101 includes a keyboard and a mouse, and is used to input each operation signal to the search device 1. The display device 102 includes a display and displays various processing results. The communication I / F 107 is an interface that connects the search device 1 to a network. As a result, the search device 1 can perform data communication with other devices via the communication I / F 107.

  The HDD 108 is a non-volatile recording device that stores programs and data. The stored programs and data include basic software and application software that control the entire search device 1. For example, the HDD 108 may store various databases and programs.

The external I / F 103 is an interface with an external device. The external device includes a storage medium 103a. Accordingly, the search device 1 can read and / or write to the storage medium 103a via the external I / F 103. The storage medium 103a includes a CD (Compact Disk), a DVD (Digital Versatile Disk), an SD memory card (SD Memory card), and a USB memory (Universal Serial Bus).
memory).

  The ROM 105 is a nonvolatile semiconductor memory (recording device) that can retain internal data even when the power is turned off. The ROM 105 stores programs and data such as network settings. The RAM 104 is a volatile semiconductor memory (recording device) that temporarily stores programs and data. The CPU 106 is an arithmetic unit that realizes control of the entire apparatus and mounting functions by reading a program and data from the recording device (for example, “HDD 108” and “ROM 105”) onto the RAM 104 and executing processing.

  With this configuration, in the search device 1 according to the present embodiment, the CPU 106 uses the data and programs stored in the ROM 105 and the HDD 108 to perform graph edge search processing, route output processing, intermediate point selection processing, and class similarity calculation. Processing and priority search node selection processing are executed. The information stored in the method holding relationship graph data storage unit can be stored in the RAM 104, the HDD 108, or a server on the cloud connected to the search device 1 via the network.

  The search method, the search program, and the search device have been described in the above embodiment. However, the search method, the search program, and the search device according to the present invention are not limited to the above-described embodiment, and are various within the scope of the present invention. Modifications and improvements are possible. In addition, when there are a plurality of the above-described embodiments and modifications, they can be combined within a consistent range.

Regarding the above description, the following items are further disclosed.
(Appendix 1)
A search method in which a computer executes a process of performing a route search in parallel for each method that is a starting point of a route that connects a plurality of methods that are starting points of a route search among a plurality of methods that are in a calling relationship,
When there are a plurality of routes to be searched for the n-th hop from the plurality of methods that are the starting points, the storage unit that stores a class corresponding to the method is referred to, and two methods connected by the route are supported. Calculate the route priority to
Based on the route priority, specify a route to be searched from among a plurality of routes to be searched,
Based on the identified route, a route search of a plurality of methods that are the starting point of the route search is performed.
Search method.
(Appendix 2)
The method of the intermediate point that is the starting point of the route search is specified, and the route that connects the plurality of methods and the method of the intermediate point is parallel from each of the plurality of methods that are the starting points of the route search including the method of the intermediate point To search for a route,
The search method according to attachment 1.
(Appendix 3)
Based on the class similarity corresponding to the plurality of methods and the method of the intermediate point, increase the method of the intermediate point that is the starting point of the route search,
The search method according to attachment 2.
(Appendix 4)
When a plurality of intermediate point methods are specified, the priority of the plurality of intermediate point methods is specified based on the class similarity,
According to the priority of the method of the plurality of intermediate points that have been identified, the method of the intermediate point that is the starting point of the route search from among the methods of the plurality of intermediate points that have been identified,
A route search is performed in parallel from each of a plurality of methods that are the starting points of the route search including the method of the specified intermediate point.
The search method according to attachment 3.
(Appendix 5)
A search program that causes a computer to execute a process of performing a path search in parallel for each method of the starting point, connecting a plurality of methods that are a starting point of path searching among a plurality of methods having a calling relationship,
When there are a plurality of routes to be searched for the n-th hop from the plurality of methods that are the starting points, the storage unit that stores a class corresponding to the method is referred to, and two methods connected by the route are supported. Calculate the route priority to
Based on the route priority, specify a route to be searched from among a plurality of routes to be searched,
Based on the identified route, a route search of a plurality of methods that are the starting point of the route search is performed.
Search program.
(Appendix 6)
The method of the intermediate point that is the starting point of the route search is specified, and the route that connects the plurality of methods and the method of the intermediate point is parallel from each of the plurality of methods that are the starting points of the route search including the method of the intermediate point To search for a route,
The search program according to attachment 5.
(Appendix 7)
Based on the class similarity corresponding to the plurality of methods and the method of the intermediate point, increase the method of the intermediate point that is the starting point of the route search,
The search program according to attachment 6.
(Appendix 8)
When a plurality of intermediate point methods are specified, the priority of the plurality of intermediate point methods is specified based on the class similarity,
According to the priority of the method of the plurality of intermediate points that have been identified, the method of the intermediate point that is the starting point of the route search from among the methods of the plurality of intermediate points that have been identified,
A route search is performed in parallel from each of a plurality of methods that are the starting points of the route search including the method of the specified intermediate point.
The search program according to attachment 7.
(Appendix 9)
A search device that performs a route search in parallel for each method of the starting point, connecting a plurality of methods that become a starting point of a route search among a plurality of methods having a calling relationship,
When there are a plurality of routes to be searched for the n-th hop from the plurality of methods that are the starting points, the storage unit that stores a class corresponding to the method is referred to, and two methods connected by the route are supported. A priority search node selection unit for calculating a route priority to be
Based on the route priority, a route for performing a route search from a plurality of routes to be searched for the route is specified, and a search for performing a route search for a plurality of methods serving as a starting point of the route search based on the specified route. And
A search device having:
(Appendix 10)
It has an intermediate point node selection part that identifies the method of the intermediate point that is the starting point of the route search,
The search unit performs a route search in parallel with each of a plurality of methods serving as a starting point of the route search including the method of the intermediate point, a route connecting the plurality of methods and the method of the intermediate point.
The searching device according to attachment 9.
(Appendix 11)
The intermediate point node selection unit increases the method of the intermediate point that is the starting point of the route search based on the class similarity corresponding to the plurality of methods and the method of the intermediate point,
The search device according to attachment 10.
(Appendix 12)
The intermediate point node selection unit, when a plurality of intermediate point methods are specified, specifies the priority of the plurality of intermediate point methods based on the class similarity, and According to the priority of the method, identify the method of the intermediate point that is the starting point of the route search from among the methods of the plurality of intermediate points that are identified,
The search unit performs a route search in parallel from each of a plurality of methods that are the starting point of the route search including the method of the specified intermediate point.
The search device according to attachment 11.

1: Search device 10: Search instruction unit 11: Graph search unit 12: Origin set storage unit 13: Storage unit 14: Class similarity calculation unit 15: Intermediate point node selection unit 16: Priority search node selection unit 111: Graph edge search Unit 112: Visited node set storage unit required 113: Visited node set storage unit 114: Predecessor and subsequent node storage unit 131: Call relationship graph data storage unit 132: Method holding relationship graph data storage unit

Claims (6)

A search method in which a computer executes a process of performing a route search in parallel for each method that is a starting point of a route that connects a plurality of methods that are starting points of a route search among a plurality of methods that are in a calling relationship,
When there are a plurality of routes to be searched for the n-th hop from the plurality of methods that are the starting points, the storage unit that stores a class corresponding to the method is referred to, and two methods connected by the route are supported. Calculate the route priority to
Based on the route priority, specify a route to be searched from among a plurality of routes to be searched,
Based on the identified route, a route search of a plurality of methods that are the starting point of the route search is performed.
Search method. The method of the intermediate point that is the starting point of the route search is specified, and the route that connects the plurality of methods and the method of the intermediate point is parallel from each of the plurality of methods that are the starting points of the route search including the method of the intermediate point To search for a route,
The search method according to claim 1. Based on the class similarity corresponding to the plurality of methods and the method of the intermediate point, increase the method of the intermediate point that is the starting point of the route search,
The search method according to claim 2. When a plurality of intermediate point methods are specified, the priority of the plurality of intermediate point methods is specified based on the class similarity,
According to the priority of the method of the plurality of intermediate points that have been identified, the method of the intermediate point that is the starting point of the route search from among the methods of the plurality of intermediate points that have been identified,
A route search is performed in parallel from each of a plurality of methods that are the starting points of the route search including the method of the specified intermediate point.
The search method according to claim 3. A search program that causes a computer to execute a process of performing a path search in parallel for each method of the starting point, connecting a plurality of methods that are a starting point of path searching among a plurality of methods having a calling relationship,
When there are a plurality of routes to be searched for the n-th hop from the plurality of methods that are the starting points, the storage unit that stores a class corresponding to the method is referred to, and two methods connected by the route are supported. Calculate the route priority to
Based on the route priority, specify a route to be searched from among a plurality of routes to be searched,
Based on the identified route, a route search of a plurality of methods that are the starting point of the route search is performed.
Search program. A search device that performs a route search in parallel for each method of the starting point, connecting a plurality of methods that become a starting point of a route search among a plurality of methods having a calling relationship,
When there are a plurality of routes to be searched for the n-th hop from the plurality of methods that are the starting points, the storage unit that stores a class corresponding to the method is referred to, and two methods connected by the route are supported. A priority search node selection unit for calculating a route priority to be
Based on the route priority, a route for performing a route search from a plurality of routes to be searched for the route is specified, and a search for performing a route search for a plurality of methods serving as a starting point of the route search based on the specified route. And
A search device having:

Images