JP3346014B2 - Directed graph editing processor - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a directed graph editing and processing apparatus, and more particularly to a method for efficiently using a memory space and reducing a small amount of data when a tree structure data structure is represented by a directed graph and used for various data processing. The present invention relates to a directed graph editing processing device that expresses a directed graph by a quantity and performs editing processing on the directed graph.

[0002]

2. Description of the Related Art Directed graphs can be used as elements to express a tree structure and a network structure as they are, and are therefore often used as data structures representing data in recent data processing. For example, when configuring a database, a tree-structured data structure is created and used for hierarchical search processing. In addition, the network expression enables a search for related data.

Conventionally, since a directed graph is represented and handled on a computer, in a generally used method of displaying a directed graph, first, each clause is specified, a clause indicated by an edge is represented by a pointer, and a certain clause is represented. In each section, each of the edges to be emitted is expressed as a set of pointers indicating the node to which the specified side is directed. According to such an expression method, when an edge is expressed by a data expression of a network composed of a directed graph, first, a position of each node is specified, an edge originating from one node is recorded, and then, The sibling sides immediately after starting from the same node as this side are sequentially recorded. In such a data representation form, the data of each side explicitly expresses the pointer of the referenced node, and in this case, the number of pointers is equal to the number of sides.

FIGS. 14 and 15 are diagrams for explaining a conventional method of expressing a network structure using a directed graph. FIG. 14 schematically shows the relationship between nodes and edges of a network using a directed graph, and FIG. 15 shows a table of a set of data elements expressing the directed graph. In FIG. 14, circles to which numbers from 1 to 10 are added indicate respective nodes, and arrows connecting the nodes with each other are indicated by arrows.
The edges of each directed graph are shown. Clauses have types of start clause, middle clause, and end clause based on the attributes of the directed graph. In FIG. 14, the start clause 81 is indicated by a broken-line circle, and the end clauses 82 and 83 are indicated by a double-line circle.
Each side indicated by an arrow has, as a pointer, the number of the designated node and a label. Here, the label of each side is described near the line segment of the arrow indicating each side.

As a data structure for expressing each of these edge data, for example, a data structure represented by a set of record data (entries) in a table format as shown in FIG. 15 is used. A data table 90 storing record data representing each side has a section number field 91 storing a section number, and an end section flag field 92 storing an end section flag indicating whether the section is an end section. , The number-of-sides field 93 for storing the number of sides originating from the node, and the label field 9 for storing the label of each side
4, and a pointer field 95 for storing a pointer indicating a node to which the side is pointed. In the table shown in FIG.
Is equal to the number of edges of the directed graph represented here.

[0006] In the directed graph shown in FIG. 14, in each of the edges originating from each node, the order in which the recording order is implicitly determined is enclosed in parentheses. It is as follows when attached.
That is, here, there are 10 nodes, and 1
Take a directed graph with five edges as an example. In that notation,
Since nodes that do not emit edges are excluded, they are represented by nine nodes and fifteen edges emanating from this node. Expressing each with the number of the section to which it points, the section: ｛→ 2 → 3 →
4｝, verse: ｛→ 5｝, verse: ｛→ 5 →→ 6｝, verse:
｛→ 7｝, clause: ｛→ 8, → 6｝, clause: ｛→ 10｝, clause
: ｛→ 6 →→ 9｝, Section: ｛→ 10｝, Section: ｛→ 6
→ 10｝. In this notation, it is assumed that the sides originating from each node have an order, and in FIG. 14, the upper side comes first.

By the way, when traversing the directed graph in order from the starting clause, the edge pointing first to a certain node will be simply referred to as “first reference edge”, and edges other than the first reference edge will be referred to as “rereference edges”. Then, all sides are either the first reference side or the re-reference side. Also, the first reference edge is 1
Corresponding to one, such clauses are all clauses except the start clause in the entire directed graph. In the directed graph shown in FIG. 14, starting from the starting clause 81, the first reference edge and the re-reference edge in each clause are ordered in the order of the rule as described above. 86, and if the data of each side is sequentially grouped in this order into a data structure in a table format, a table as shown in FIG. 15 can be constructed.

[0008]

By the way, when the network of the directed graph (FIG. 14) is represented by the table of the data representation format (FIG. 15) as described above, it is apparent that only the number of edges equal to the number of each side is required. A pointer is required, and an area for storing this number of pointer data is required. Therefore, when expressing a large-scale network of a directed graph, the number of pointers that need to be stored increases according to the number of edges. Also, when the number of pointers increases, the address width representing the individual pointers must be further increased, and the amount of data for expressing these data becomes very large.

Therefore, when expressing a large-scale directed graph network, the data area for storing a large number of pointers becomes very large according to the number of edges, and in data processing that handles such a directed graph, For the data processing involving many pointers, a large load is imposed on the data storage area and the data processing speed due to the large amount of data.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a memory for storing a tree-structured data structure by using a directed graph for use in various data processing. It is an object of the present invention to provide a directed graph editing apparatus capable of efficiently using a space, expressing a directed graph with a small amount of data, and performing an editing process on the directed graph.

[0011]

In order to achieve the above object, a directed graph editing apparatus according to the present invention has, as a first feature, a first reference edge, an object which is an edge that first designates a certain node, and an object. A directed graph having a structure consisting of the child edge that is the first edge emanating from the node specified by the target edge and the sibling edge that is an adjacent edge emanating from the same node as the target edge is defined as A directed graph editing processing device that records and forms one-dimensionally in a table according to an order and performs an editing process on each side data of a directed graph formed in the table, wherein a first reference side is recorded, and the first reference side is recorded. Next, the child side is recorded, and if there is a sibling side, after the recording of the child side or less is completed, record data in the side serial format for recording the sibling side is formed, and the first reference side, the child side, and the sibling side are formed. Each side data of Next, storage means (30) for storing in a table as record data, and when a directed graph to be edited is inputted, each of edge data of the inputted directed graph to be edited and record data stored in said storage means Edge data is sequentially searched and compared, edge data up to the matching edge and remaining edge data are separated from the edge data of the directed graph to be edited, and editing operation is performed on the record data of the edge data. After performing the record data editing operation on the table with the first editing processing means (37), the record data of each side data up to the position of the record data on which the edit operation is performed and on and after the position are respectively And a second edit processing means (38) for performing an edit process for updating the reference value of the reference.

Further, as a second feature, the directed graph editing processing apparatus of the present invention has a first reference flag for designating whether or not each edge data in the edge serial format record data is a first reference edge; Label data, an end flag indicating whether or not the node issuing this edge is an end node, a child flag indicating whether a child edge exists, and a sibling flag indicating whether a sibling edge exists. Features.

The directed graph editing apparatus according to the present invention has a third
As a feature of the first embodiment, when a directed graph of a tree expression representing a word group is stored as edge serial record data and an edit process for registering a new word is performed, the first edit processing unit attempts to register. when a word and a ₁ a ₂ ... a _n, this elements of a word a _1, a _2, ..., sequentially searched results from each side data of the record data to a _n, from a ₁ to a _i is equal label edge data exists, when the comparison fails for a i _{+ 1} of the label, and edit to insert records data edge data from a _{i + 1} in that position until a _n,
The second edit processing means performs an edit process for updating the value of the reference to the younger brother of the side data having the side data from a ₁ to a _i and the side data after a _{i + 1} .

Further, as a fourth feature, the directed graph editing and processing apparatus of the present invention is characterized in that a directed graph of a tree expression representing a word group is stored in edge-series record data, and the already registered word when performing editing processing deleting, the first editing processing means, when the word to be deleted and d ₁ d ₂ ... d _n, each element d ₁ of the word,
If d ₂ ,..., d _n are sequentially searched from each side data of the record data and succeeded, the end flag of d _n side data is deleted from each side data obtained by the search, and d 2 _n
If there is edge data to the child flag, the editing without any Thereafter the ends, if no child flag side data d _n, j
The varied to from 1 (n-1), a maximum of j with end flag in the side data d _j and jmax, further, by changing the k from jmax to n, siblings flag side data d _k Let the maximum k be kmax, and from kmax be 1
When the value obtained by subtracting is taken as i, the editing is completed by deleting the edge data of d _{i + 1 to} the edge data of d _n , and the second editing processing means executes the edge data from d ₁ to d _i And an editing process for updating the value of the younger brother reference of the side data with the side data after d _{i + 1} as the younger brother.

[0015]

In the directed graph editing processing apparatus of the present invention,
The first reference edge, which is the edge that first points to a certain node, the child edge, which is the first edge originating from the node specified by the target edge, and the sibling, which is the adjacent edge originating from the same node as the target edge A directed graph having a structure composed of edges is formed by recording each edge data one-dimensionally in a table in a predetermined order. Then, an editing process is performed on each edge data of the directed graph formed in the table.

For this reason, the directed graph editing processor includes:
A storage means (30), a first editing processing means (37), and a second editing processing means (38) are provided.

The storage means (30) records a first reference edge for the directed graph formed with the above structure, records a child edge next to the first reference edge, and stores a sibling edge when there is a sibling edge. After the recording of the child side or less is completed, record data in the side serial format for recording the sibling side is formed, and the first reference side, child side,
In addition, each side data of sibling sides is sequentially stored as record data in a table, and the directed graph is stored.

To perform editing processing on each edge data of the directed graph, the first editing processing means (37), when a directed graph to be edited is input, inputs the edge data of the input directed graph to be edited and the edge data of the directed graph. The edge data of the record data stored in the storage means is sequentially searched and compared, and among the edge data of the directed graph to be edited, the edge data up to the matching edge and the remaining edge data are separated, and Edit the record data of the edge data. Accordingly, the second editing processing means (38) edits the record data for the table, and then adds the record data of each side data up to the position of the record data where the edit operation is performed and thereafter. On the other hand, an editing process for updating the value of each reference is performed.

Each edge data in the edge series record data forming the directed graph here includes an initial reference flag indicating whether or not the edge is an initial reference edge, label data of the edge, and When a directed graph is searched, which is composed of an end flag indicating whether the clause to be emitted is an end clause, a child flag indicating whether a child edge exists, and a sibling flag indicating whether a sibling edge exists. By referring to these flags, edges and nodes of the directed graph can be traced at high speed, and editing processing of the directed graph can be performed.

For this reason, a directed graph of a tree expression representing a group of words is stored as edge-series record data, and when editing processing for registering a new word, the first graph is used.
Edit processing means, the word to be registered a ₁ a ₂ ... a _n
, A _n , as a result of sequentially searching each element a ₁ , a ₂ ,..., An of this word from each side data of the record data, there exists side data of the same label from a ₁ to a _i , When the comparison of the label of a _{i + 1} fails, a
from _{i + 1} edits to insert records data side data until a _n. Along with this, the second editing processing means performs an editing process for updating the value of the younger brother reference of the side data with the side data from a ₁ to a _i and the side data after a _{i + 1} .

When a directed graph of a tree expression representing a word group is stored in edge-series record data and an edit process for deleting a registered word is performed, the first edit processing means includes: The words to be tried are d ₁ d ₂ … d
When the _n, each element of the word d _1, d _2, ..., if the d _n was successful sequentially searched from the edge data of the record data, among each of the edge data obtained by the search, d _n
The end flag of the side data is deleted, and if there is a child flag in the side data of d _n , the editing is terminated without doing anything thereafter.
Without child flag side data d _n, j from 1 (n
-1), the maximum j having an end flag in the edge data of d _j is set to jmax, and k is changed from jmax to n, and the maximum k having a sibling flag in the edge data of d _k is determined. Is defined as kmax, and the value obtained by subtracting 1 from kmax is defined as i, the editing is completed by deleting the data from the edge data of d _{i + 1 to} the edge data of d _n . Then, along with this, the second editing processing means executes the side data from d ₁ to d _i and d
An editing process is performed to update the value of the reference to the younger brother of the edge data with the edge data subsequent to _{i + 1} .

According to this, this directed graph is formed,
In data processing performed using the characteristics of the directed graph, data processing relating to the directed graph can be executed by reading flag data from a table containing only flag data and determining a flag, thereby enabling high-speed processing.
Also, when editing the directed graph, the editing process may be performed on the record data relating to the edge data of the partial directed graph to be edited, so that the processing can be speeded up.

[0023]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be specifically described below with reference to the drawings. 1 and 2 are diagrams illustrating a tree-structured directed graph represented by a data structure of a directed graph according to the present invention. FIG. 1 schematically shows the relationship between nodes and edges of a tree structure using a directed graph, and FIG. 2 shows a directed graph table having a data structure of an edge serial expression expressing the directed graph. In FIG. 1, circles indicate respective nodes, and arrows connecting the nodes indicate edges of the directed graph. Clauses have types of start clause, middle clause, and end clause based on the attributes of the directed graph. In FIG. 1, the start clause 11 is represented by a broken-line circle, and the end clause 12 is represented by 2
This is indicated by a double circle. Since each side indicated by an arrow always points to the next node, the number of the node indicated by each side is used as the address of each side. The number of each section is shown in a circle. Also, the labels of each side are described near the line segment of the arrow indicating each side.

As shown in FIG. 2, in the data structure representing these edge data, in the directed graph table 20, each edge data is represented as record data (entry data) as table format data by edge serial representation. Is done. A directed graph table 20 storing record data representing individual edges includes an address field 21 storing an edge number, a label field 22 storing a label, and a firstp flag storing a firstp flag indicating whether or not the edge is a first reference edge. Field 23, somp flag field 2 for storing a somp flag indicating whether or not a child side is provided
4, a brotherp flag field 25 for storing a brotherp flag indicating whether or not the side has a sibling edge; a finalp flag field 26 for storing a finalp flag indicating whether the node pointed to by the side is an end node; A sibling address field 27 stores the side number address of the sibling side of the side together with its relative address.

The meaning of each flag used in the directed graph table 20 will be described. The firstp flag is a flag indicating whether or not a first reference edge. The sonp flag is a flag indicating the presence or absence of a child side. The brotherp flag is a flag indicating the presence or absence of a brother side. The finalp flag is the first reference edge (firstp = "1")
And a flag indicating whether or not the instruction destination is an end clause. Note that a sibling address is provided together with these flags, but when the side has a sibling side (br
It exists when otherp = "1"), and indicates the relative address of the side data of the sibling along with the address of the side number of the sibling.

In the directed graph table 20, by reducing the number of pointers included in the directed graph, the directed graph is efficiently represented, and the load when the directed graph expression is used in data processing is reduced. Therefore, in the directed graph table 20, each edge data is recorded as tabular data in edge serial representation.
That is, if data of a certain side is recorded as record data (entry), the data position of the next entry is implicitly set to the data of the child side of the first reference side (the first side originating from the node indicated by the target side). Record. If the side to be recorded is a re-reference side and its child side has already been recorded, a pointer to the child side is recorded. Such a case does not occur in a tree-structured directed graph. For this reason, the directed graph table 20 does not have a pointer field.

As described above, the child side is recorded implicitly in order. If there is a brother side, a flag for explicitly recording the existence of the brother side is required. Depending on the presence / absence of this flag, after recording a certain side as record data (entry), if a pointer from that side to a child side is required (that is, if the child side has already been recorded), Immediately after that, there is a degree of freedom to record anything, so in the directed graph table 20, a sibling edge (if any) is recorded at this entry position.

Further, the data structure of the directed graph table 20 according to the embodiment of the present invention will be described in detail. In this edge serial expression table, the following flags are prepared and the edges and child edges are recorded in units in order to efficiently express the directed graph and to correspond to any form of the directed graph. . That is, the entry of each side data includes a flag (firstp flag) indicating whether it is the first reference side or a re-reference side, and a flag (sons flag) indicating presence / absence of a child side in the first reference side. Prepare Furthermore,
Prepare a flag (brotherp flag) indicating the presence or absence of a sibling edge.

In the directed graph table 20, the edges and child edges are recorded in units. In a tree structure composed of a directed graph, all nodes except for the starting node (one point) have one-to-one reference edges that point to them (hereinafter, the pair of the first reference edge and the node here is appropriately set). However, nodes that do not emit edges (nodes without child edges) exist irregularly. Therefore, if the information of the child side is combined with the first reference side, the information of the node becomes unnecessary. The only information that a node that does not emit an edge must have is information indicating whether or not it is an end node.

Next, an algorithm for creating a directed graph having a structure as shown in FIG. 1 from the directed graph table 20 configured as shown in FIG. 2 will be described. The algorithm in this case makes it possible to generate a corresponding directed graph from the representation by the directed graph table 20 of the edge serial record data by repeatedly performing the following steps (A) to (E).

In this directed graph generation process, the stack brothers = {} is set as initial setting data, the (virtual) edge indicating the start node is set to edge 0, and edge 0 is the first reference edge (fir
stp = "1"). (A) Create a start clause. (B) i = 1. (C) If there is no side, end. (C-1) The side (i-1) is the first reference side (firstp = "1"),
If there is a child side (sonp = "1"), the side i
Generate If brotherp = "1", i stack brot
Push to hers. (C-2) The side (i-1) is the first reference side (firstp = "1"),
If there is no child side (sonp = "0"), the stack brothers
Is obtained, and an edge i is created as a sibling edge of the edge j. If brotherp = "1", push i to the stack brothers. (C-3) When the side (i-1) is a re-reference side (firstp = "0"), a value j obtained by popping the stack brothers is obtained, and the side i is created as a brother side of the side j. And brotherp =
If "1", push i to stack brothers. (D) Let i = i + 1. (E) Return to (C).

FIG. 3 is a block diagram showing a configuration of a main part of the directed graph editing processing apparatus according to one embodiment of the present invention. In FIG. 3, 30 is a directed graph table, 31 is an input device, 32 is a data processing device, 33 is a storage device,
4 is an output device, 35 is a directed graph generation processing unit, 36 is a character string analysis processing unit, 37 is a first editing processing unit, and 38 is a second editing processing unit. In the storage device 33, a directed graph table 30 that represents a directed graph in which label data and the like are set according to an analysis target is prepared and provided in advance. The directed graph generation processing unit 35 of the data processing device 32 creates a directed graph using the data of the directed graph table 30, and executes the character string analysis processing unit 3.
6 performs a character string analysis process using the data of the directed graph table 30. The character string to be analyzed is input from the input device 31, and the output of the analysis result is output from the output device 34.

The first edit processing unit 37 performs an edit process on each edge data of the directed graph of the directed graph table 30, so that when a directed graph to be edited is input,
The inputted edge data of the directed graph to be edited and the respective edge data of the record data stored in the directed graph table 30 are sequentially searched and compared, and the edges up to the matching edge among the edge data of the directed graph to be edited. The data and the remaining side data are separated, and an editing operation is performed on the record data of the side data. Accordingly, the second edit processing unit 38 performs an edit operation on the record data for the table, and then, for the record data of each side data up to the position of the record data on which the edit operation is performed,
An editing process for updating the value of each reference is performed.

FIG. 4 is a flowchart showing a processing flow of a directed graph creation process performed by the directed graph creation processing unit 35. With reference to FIG. 4, the directed graph creation processing here will be described. When the process is started, first, in step 40, the stack brothers is set to an empty list. Next, in step 41, a start clause is created, and this is set as a side clause 0. Next, at step 42, the control variable i is set to 1. Then, the process proceeds to a step 43, wherein a side node i is created as a child side of the side node (i-1). Next, in step 44, by referring to the flag data from the directed graph table, br
If otherp = "1", to create a sibling edge later,
Push the value of control variable i onto the stack brothers. Next, in step 45, the control variable i is incremented, and in the next step 46, it is determined whether or not there is edge data of the edge i. If there is no edge data, the process is terminated.

If it is determined in step 46 that the side data of the side i exists, the process proceeds to step 47. In step 47, it is determined from the side data whether the side node (i-1) is the first reference side (firstp = "1"). If it is the first reference side, the process proceeds to step 48, where it is further determined whether or not there is a child side (sonp = "1") in the side node (i-1) from the side data. In this case, the process proceeds to step 49.

In the determination processing of step 48, the side nodes (i
If it can be determined that (-1) has a child side (sonp = "1"),
Returning to step 43, the next processing is continued from the child side creation processing in step 43. When it is determined in step 47 that the side node (i-1) is not the first reference side (firstp = "1"), and in step 48, the child node is added to the side node (i-1).
If it is not determined that (sonp = “1”) is present, the process proceeds to step 49 to perform processing for creating a sibling edge. In this process, in step 49, a value j obtained by popping the stack brothers is obtained, and in the next step 50, an edge node i is created as a sibling edge of the edge node j. And the next step 4
4, the process is repeated from step 44.

As described above, in this processing, child edges are sequentially created when child edges exist from the edge data of the record data (entries) of the directed graph table, and during that time, sibling edges are generated. Then, the position where the sibling edge exists is pushed onto the stack brothers. When there is no child side, a sibling edge is created at the position where the sibling side pushed to the next stack brothers exists. This process is repeated to create a directed graph.

By this processing, for example, the directed graph 10 as shown in FIG. 1 is created from the flag data of the directed graph table 20 shown in FIG. In the directed graph 10,
As shown in FIG. 1, starting from the start clause of section number 0, tracing the side of the first reference side above it, the child side is sequentially created, and when there are no child sides, it returns to the position where the sibling side exists. , Create a sibling edge. The creation order corresponds to the recording order of the side data in the directed flag table 20.
The address indicating the side number is the node number as it is.

The directed graph 10 created in this way is
For example, in accordance with the same notation as in the description of the directed graph in FIG. 14 described above, in each clause and each edge emanating from the clause, a record whose order is determined implicitly is enclosed in parentheses, and a pointer is explicitly set in each element. If a necessary part is indicated by attaching a mark →, it is expressed as follows. In particular, there is no part requiring a pointer in a tree structure. That is, directed graph 1
In the example of 0, since the recording order is implicitly determined for all side data, it can be enclosed in one parenthesis. If the part that records that there is a brother side is underlined,
{1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11,
12, 13, 14}.

FIG. 5 is a flowchart showing a processing flow of the input symbol string analysis processing performed by the character string analysis processing unit 36. The input symbol string analysis processing here will be described with reference to FIG. In this input symbol string analysis process, for example, referring to a tree structure formed by a directed graph as shown in FIG. 1, tracing the tree structure, and determining whether to accept or reject the character string of the input symbol input. Here is the process to be performed,
Instead of using the form of the directed graph as shown in FIG. 1, the input symbol string analysis processing is performed using the data of the directed graph table 20 as shown in FIG.

When the process is started, first, in step 51, the first character of the input character string is set as the current input symbol. Next, in step 52, the side 1 is set as the current side. Then, the process proceeds to a step 53, wherein it is determined whether or not the current input symbol matches the label of the current side. If they match, in step 54, the child side is set as the next side. Next, in step 55, it is determined whether or not there is a next input symbol. If there is a next input symbol, the process proceeds to the next step 56, and the next input symbol is set as the current input symbol. Next, in step 57, the next side is set as the current side, the process returns to step 53, and the processing is repeated from step 53.

Further, in the determination processing of step 53,
If the current input symbol and the label of the current side do not match, the process proceeds to step 58 to determine whether or not there is a similar side for the sibling side, and determines whether there is a sibling side. If there is a sibling edge, the process proceeds to step 59, where the sibling edge is set as the current edge, the process returns to step 53, and the process is repeated from step 53. If there is no sibling edge in step 58, the directed graph cannot be traced, and the process ends in failure.

In this way, the processing from step 53 to step 57 is repeated, and the processing from step 53 to step 57 including the processing from step 58 and step 59 is repeated. Then, if it is determined in the determination processing of step 55 that the next input symbol has disappeared, the process proceeds to step 60, where the final determination processing of the input symbol string analysis processing is performed. That is, if the node pointed to by the current side is an end node, the input symbol can be accepted.
It is determined whether or not finalp = "1". As a result, if the current side is finalp = “1”, the process ends with success, otherwise, ends with failure.

The input symbol string determination process is a process of determining whether or not the label of each edge of the directed graph matches the input symbol of the input character string, and tracing each edge of the tree structure of the directed graph. . In this process, when tracing each edge of the tree structure of the directed graph, the edge data of the directed graph table as shown in FIG. 2 is sequentially read, each flag in the edge data is determined, and the character string of the input symbol is determined. Do. In this determination processing, the child side is traced, and if the labels do not match, the sibling side is traced. However, the tracing to the sibling side in the directed graph table is performed by the record data in the edge serial format corresponding to the directed graph. It is clear from the equivalence of the directed graph table based on the data structure described above. If the order of edges in the directed graph is not deterministic,
In other words, if there are multiple sides with the same label from the same clause, success and failure are distinguished in the way of tracing from the child side. Similarly, it follows the siblings.

Next, regarding an access method using each edge data of such a directed graph table, an algorithm that can cope with the current state of the clause and the case where there is no input symbol is summarized. A directed graph can be traced by the following algorithm. However, this process is a process in the case where the start clause is not the end clause, and the process is performed by taking in the character string of the input symbol. (1) Initial setting; i = 0 (2) Set edge 1 as the current edge. (3) Compare the current input symbol with the label of the current side. (3-1) If they match, the child side is set as the next side. (3-2) When they do not match, (3-2-1) If there is a brother side, the brother side is set as the next side. (3-2-2) If there are no siblings, the process ends abnormally. (4) There is no next input symbol. If finalp = "1", the process ends successfully. If finalp = "0", the process ends unsuccessfully. (5) Let the next input symbol be the current input symbol. The next side is the current side. (6) Return to (3). With such an algorithm, it is possible to trace each edge of the network of the directed graph corresponding to each character string of the input symbol.

By the way, in the case where the processing of tracing the directed graph is performed by using the edge data of the directed graph table as described above, brot indicating that there is a sibling edge in the first reference edge.
If the herp flag is set, the child side is followed by the sibling side, but in order to improve the reference speed, here, the sibling address indicating where the entity is is specified. The record data of the side data is held (sibling address field 27). By referring to this sibling address, a directed graph can be searched at high speed.

Next, a description will be given of a processing example in the case where editing processing is performed on the directed graph table 30 by partially using the input symbol analysis processing performed by the character string analysis processing unit 36. The editing process includes a first-stage editing process performed by the first editing unit 37 and a second editing process performed by the second editing unit 38.
The editing process at the stage is executed.

To perform an editing process on each edge data of the directed graph in the directed graph table 30, when a word to be edited (input character string) such as a newly registered word or a word to be deleted is given, a first editing process is performed. Part 37
Controls the character string analysis processing unit 36 to input the edge data of the input directed graph to be edited and the record data stored in the directed graph table 30 for the directed graph to be edited (character string to be edited). Are sequentially searched and compared with each side data. Then, of the edge data of the directed graph to be edited, edge data up to the matching edge and the remaining edge data are separated, and an editing operation is performed on the record data of the edge data.

In addition, the second editing processing unit 38
Performs an editing process of updating each reference value to the record data of each side data up to the position of the record data where the editing operation is performed after performing the editing operation of the record data for the table.

Next, with regard to the editing processing of the directed graph table, the editing processing of registration processing and deletion processing will be described more specifically. FIG. 6 is a flowchart showing a processing flow of inserting an edge node into the directed graph table by word registration, and FIGS. 7 and 8 are schematic diagrams of a directed graph for explaining an operation relating to the word registration processing. Processing on these directed graphs is actually performed by data processing operations on edge serial record data in the directed graph table. FIG. 9 is a flowchart showing a processing flow of deleting a clause by deleting words from the directed graph table.
FIG. 4 is a schematic diagram of a directed graph illustrating an operation related to a word deletion process.

First, with reference to FIGS. 6, 7 and 8, a description will be given of a process of inserting an edge clause for registering a word in the directed graph table. The directed graphs of FIGS. 7 and 8 show a part of the directed graph expressing a tree structure, and particularly show only a part related to a character string of a word to be registered. 7 and 8, arrows indicate edges of the directed graph, circles indicate nodes, and dotted lines above and below indicate that edges and nodes exist in addition to the illustrated portions. Each character element of the character string forming the word is assigned as a label of the side. As described above, these directed graphs are formed as edge serial record data, and operations on the directed graph are executed as operations on edge serial record data.

As described above, in the edge serial format record data, a set of an edge and a node indicated by the edge is called an edge node, and the edge node is recorded as edge data in table format record data. . Each edge (edge data) of the directed graph forming the tree structure holds a child flag (sonp) indicating whether there is a child edge, a sibling flag (brotherp) indicating whether there is a younger edge, and a sibling edge If there is, an address (brother address) indicating the side number address, an end flag (finalp) indicating whether or not the end clause, a label (label) of the side, and the like. In the record data of the side serial format, if there is a child side for each side data, it is stored immediately after the child side (see the description of FIGS. 1 and 2).

According to the flowchart shown in FIG.
With reference to FIG. 8 and FIG. 8, the process of inserting a margin by word registration here will be described. In registration processing of a word with respect to a directed graph table, strings of words a ₁ a ₂ ... a _n to be registered is input, the processing of the word registration is started, first, in step 61, with respect to the directed graph table, and registers word a ₁ a ₂ ... each character elements a ₁ of a _n, a _2, ..., sequentially to search for a _n. Results of the search of the directed graph table, then, in step 62, the string a ₁ a ₂ ... a _i
Is succeeded, and the label comparison of the next character element a _{i + 1} fails, the position of the side data (address of the side number) at that time is held as p. Next, in step 63, the edge data from the edge node (a _{i + 1} ) to the edge node (a _n ) is inserted at the position p of the record data in the edge serial format.
That is, the remaining character element strings a _{i + 1} a excluding the character element strings a ₁ a ₂ ... A _i in which the character strings of the words to be registered match.
_{i + 2} ... to create each of the sides nodes (edge data) corresponding to a _n, coupling (insertion) in the edge nodes of the failed position in the label comparison. Then, in step 64, corresponding to the edge data newly inserted into the directed graph table, consistency of each flag and reference sibling address for the edge serial format record data of the existing edge data is obtained. Is performed. In other words, in this process, the values of the side data from the side nodes (a ₁ ) to (a _i ) and the side data after the side node (a _{i + 1} ) are increased by inserting the younger brother reference values. This is performed by a process that increases by the amount.

[0054] That is, the processing of the word registration, as shown in FIG. 7, the label comparator edge of the directed graph, beginning a label comparison with the first character element a ₁ of registered word, a successful comparison, then a character performs label comparison element a _2, a successful comparison, further and so do the following label comparison of character elements a _3, carries out the search to continue these labels comparison sequentially, a label comparison of character elements a _i Done successfully,
Next, if the label comparison of the character element a _{i + 1} fails, the storage position (side number address) of the side data of the record data in the side serial format at this time is held as p. In the digraph 65 shown in FIG. 7, the character elements a ₁ ,
The state immediately after the search is sequentially performed up to the side nodes of a ₂ ,..., a _i , and the position of the side node indicated by ★ fails.

Next, in the directed graph table, the storage position p of the record data of the edge serial format storing the failed edge node (edge data) is obtained, and from this position,
In the character string of the word to be registered, each side segment (side data) of the character element of the part where the character string does not match is inserted. That is, as shown in FIG. 8, in the data structure of the directed graph, the existing directed graph 66 is connected to the sub-directed graph 67 of the edge clause to which the newly registered word is connected, and in this case, connect the side section of the character elements a _{i + 1} as a sibling side edge section of the connection position, as it rows of subsequent character element is the child sides, the sides nodes from character elements a _{i + 2} until a character element a _n Connect (insert).

In this case, the connection of the side nodes is performed in the data structure of the record data in the side-by-side format, from the side node (a _{i + 1} ) to the side node (a _i ) at the position next to the side data of the side node (a _i ). edge data up to a _n ) will be inserted. For this reason, as shown in FIG. 8, the insertion position in the directed graph 66 is defined as a boundary between the edges of the search path in which each edge of the directed graph 66 has been searched in correspondence with the edge serial format record data. Is separated, and the edge of the directed graph is divided into a first half (66) and a second half 68. Then, the sub-directed graph 67 is inserted during the division. The sub-directed graph 67 is a character element string a of the registered word.
Of the sides clause ₁ a ₂ ... a _n, a directed graph consisting of edges clause character element sequence a _{i + 1} ... a _n which is not registered. In the insertion process of an edge clause for new word registration, the first half (66), the inserted portion (67), and the second half (68) are stored in the memory in this order to form an updated directed graph table. . In addition, with the insertion of the record data of the side section, the value (address of the side number) of the younger reference of the side section (a ₁ ) to the side section (a _i ) is converted from the side section (a _{i + 1} ). It is increased by the amount of the side clause (a _n).

Next, with reference to FIG. 9, FIG. 10, and FIG. 11, a description will be given of a process of deleting a clause relating to word deletion from the directed graph table. The directed graphs of FIGS. 10 and 11 show a part of the directed graph expressing the tree structure, as in the case of the above-described word registration processing, and here also show only the part related to the character string of the word to be deleted. ing. In the process of deleting a registered word from the directed graph table in which a word is already registered, a word to be deleted is searched from the directed graph table, and if the search is successful, the edge of the directed graph on the registered word to be deleted is searched. The deletion process is performed according to the state of the node.

According to the flowchart shown in FIG.
With reference to FIG. 0 and FIG. 11, the word deletion processing here will be specifically described. In the deletion process of a word with respect to the directed graph table, the input string of words d ₁ d ₂ ... d _n to be deleted, the processing of the word deletion is started, first, in step 71, with respect to the directed graph table, deletes word d ₁ d ₂ ... each character element d ₁ of d _n, d _2, ..., sequentially searches for d _n, a successful this search process, performs a process of deleting from the next. Here, if the search for the word to be deleted is not successful, the deletion process cannot be performed because the word to be deleted has not been registered in the directed graph table.

If the directed graph table is successfully searched for the word to be deleted, then, in step 72, the end flag (final) in the edge data of the edge node (d _n )
p) is set to “0”. Next, the process proceeds to a step 73, wherein it is determined whether or not the side node (d _n ) has a child side. This process
Carried out by judging (which is "1") or not to the side data of the side section (d _n) is sonp flag. Sides clause (d _n)
If there is a child side, the deletion process is terminated. That is, in this case, the character element string d _{1 of} the word to be deleted here
side section of the d ₂ ... d _n are all used as part of another registered word, it has become a Henbushi necessary to a state of being registered in the other registered word. For this reason, the side section of the successful search character element sequence d ₁ d ₂ ... d _n does not perform or delete any side clause. The process of step 72, the end flag of the side section of the last character string elements of a word to be deleted (d _n) (final
Since p) is "0", words that end with the letter element d _n is, it will have been deleted from the registration state.

If it is determined in step 73 that there is no child side in the side node (d _n ), the character element string d of the word to be deleted is determined by the processing from step 74 onward.
₁ d ₂ … d _n , each part used as a part of other registered words, and a part that is necessary to register other registered words in the registered state Except for the side node of, the processing of deleting the other side nodes is performed.

Therefore, in the next step 74,
By changing j from 1 to (n-1), each edge node (d
_{In j} ), the maximum j in which the end flag of the side data is “1” is jmax. Next, in step 75,
k is changed from jmax to n, and the maximum k having a sibling flag in the d _k side data in the edge node (d _k ) is k ma
Let x be a value obtained by subtracting 1 from kmax. next,
In step 76, the edge data from the edge node (d _{i + 1} ) to the edge node (d _n ) is deleted from the edge serial record data.

That is, the character element string d _{1 of} the word to be deleted
The last character element d of each of the edge segments d ₂ ... dn _{-1 up} to the edge segment having the immediately preceding end flag and among the edge segments having sibling edges therebetween. The position i of the side node close to _n-1 is obtained, and the side nodes after that side node are deleted. That is,
In the search path of the character element strings d ₁ d ₂ ... Dn _-1 of the word to be deleted, the last branching part up to the edge with the previous end flag is determined. A process for deleting only the side nodes (edge data) is performed. If there is no branch portion, the portion after the edge node having the immediately preceding end flag becomes unnecessary, so that these are deleted. Then, next, step 77
In step (2), processing is performed to match the existing edge data as serial record data in correspondence with the edge data deleted from the directed graph table. This process
The processing is performed by reducing the edge data from the side node (d ₁ ) to the edge node (d _i ) and the value of the younger brother reference that uses the edge data after the edge node (d _{i + 1} ) as the younger brother by the amount reduced by deletion. .

In other words, in the processing of edge section deletion for the word deletion, as shown in FIG. 10, the label comparison of the edges of the directed graph starts from the first character element d ₁ of the word to be deleted. Succeeds, then the character element d
For ₂ label comparison, a successful comparison, further and so do the following label comparison of character elements d _3, carries out the search to continue these labels comparison sequentially, the label comparator of the final character elements d _n If the search is completed successfully, the deletion process is performed here.

Of the edge nodes obtained by such a search, the end flag of the last edge node (d _n ) is set to “0” to cancel the end node, and the child edge is added to the edge node (d _n ). If there is an edge clause that satisfies, then nothing is done and the process ends. If there is no child side edge, from the last side node (d _n ) to the immediately preceding end node, trace the edge node in the opposite direction to the side node with the nearest sibling side, and determine this Then, each subsequent edge segment is deleted. For this reason, in the processing here, j is changed from 1 to (n-1), and the largest j whose end flag is "1" in the side nodes (d _j ) therebetween is jmax. Further, k is changed from jmax to n, the maximum k having a sibling edge in the side node (d _k ) is set to kmax, and a value obtained by subtracting 1 from this kmax is i (i
= Kmax-1). Directed graph 10 shown in FIG.
At 0, the position i of the edge node thus obtained is indicated by ★. Edges after the position i of the edge indicated by ★ are deleted.

[0065] Removing the side section in this case, the data structure of the record data side serial form, the sides clause (d _i) sides clause of the following positions of the edge data of the (d i _{+ 1)} from the side section (d _Since the edge data up to _n ) is to be deleted, as shown in FIG. 11, at the position i to be deleted from the directed graph 101, the sub-directed graph 102 of the edge node to be deleted is separated. Also in this processing, the sub-directed graph 102 of the edge node to be deleted is removed from the edge searched in the directed graph 101 in correspondence with the record data in the edge serial format. As a boundary, the latter half portion 103 of the subsequent side segment is separated, and is divided into the first half portion (101) and the second half portion 103. Sub directed graph 102 Here, of the side section of the character element sequence d ₁ d ₂ ... d _n words to be deleted is the directed graph consisting of edges clause character element sequence d _{i + 1} ... d _n to be removed . In the word deletion, the first half (101) and the second half (103) here are stored in the memory in this order and reconfigured to obtain an updated directed graph table. With the deletion of the record data of this edge clause,
The value (address of the side number) of the younger reference of the side node (d ₁ ) to the side node (d _i ) is reduced by the side node (d _n ) from the side node (d _{i + 1} ).

Next, as a specific example of the editing process of the directed graph, a case will be described in which the above-described directed graph of FIGS. 1 and 2 is deleted. The directed graphs shown in FIGS. 12 and 13 show changes in the state when edge nodes are deleted from the directed graphs shown in FIGS. 1 and 2 described above. As shown in FIG. 12, when the sub-directed graph 122 composed of the edge node of the label A and the edge node of the label B is deleted from the directed graph 121, the edge series record data of the directed graph table corresponds to the deletion. As shown in FIG. 13, the edge data 132 is deleted from the directed graph table 131. Along with this, the side number address of each side data after the position of the deleted side data 132 is incremented, and the reference value of the sibling address for reference in the sibling address field is changed to
Only that amount goes up.

That is, as for the edge number address of the directed graph table 131 shown in FIG. 13, two record data are deleted by the deletion processing of the edge data 132, and the edge data after the edge data 132 is shown in FIG. As is clear from the comparison with the directed graph table 20, the edge number address is reduced, and the value of the reference of the sibling address for reference in the sibling address field is further reduced accordingly.

Also, when looking at the directed graph 10 and the directed graph table 20 shown in FIGS. 1 and 2, respectively, from the directed graph 121 and the directed graph table 131 shown in FIGS. 12 and 13, respectively, Is a diagram showing a state change in which a process of inserting a side segment of the directed graph according to the registration of is performed. That is, for the word to be registered, the insertion position of the side node of the character element string (the position of the side data 132) is determined, and the edge data corresponding to the side node to be inserted is stored at the corresponding position of the edge serial format record data. Is inserted. In this case, with the insertion of the record data, the side number address is increased, and the reference value of the sibling address for reference in the sibling address field is increased by that amount.

[0069]

As described above, according to the present invention, when a tree-structured data structure is represented by a directed graph and used for various data processing, a small amount of data can be efficiently used in the memory space. Can provide a directed graph, and can further provide a directed graph editing apparatus capable of performing editing processing on the directed graph. Also,
In data processing performed by forming a directed graph and utilizing characteristics of the directed graph, data processing relating to the directed graph can be executed by reading flag data from a table including only flag data and determining a flag. It becomes possible. Also, when editing the directed graph, the editing process may be performed on the record data relating to the edge data of the partial directed graph to be edited, so that the processing can be speeded up.

[Brief description of the drawings]

FIG. 1 is a diagram showing the relationship between nodes and edges of a tree structure using a directed graph,

FIG. 2 is a diagram showing a directed graph table having a data structure based on an edge serial expression expressing the directed graph of FIG. 1;

FIG. 3 is a block diagram showing a configuration of a main part of a directed graph editing processing device according to an embodiment of the present invention;

FIG. 4 is a flowchart illustrating a processing flow of a directed graph creation process performed by a directed graph generation processing unit;

FIG. 5 is a flowchart showing a processing flow of an input symbol string analysis process performed by a character string analysis processing unit;

FIG. 6 is a flowchart showing a processing flow of inserting an edge by registering a word in a directed graph table;

FIG. 7 is a first schematic diagram of a directed graph illustrating an operation related to a word registration process,

FIG. 8 is a second schematic diagram of a directed graph illustrating an operation related to a word registration process;

FIG. 9 is a flowchart showing a processing flow of deleting a clause by deleting words in the directed graph table;

FIG. 10 is a third schematic diagram of a directed graph illustrating an operation related to a word deletion process,

FIG. 11 is a fourth schematic diagram of a directed graph illustrating an operation related to a word deletion process;

FIG. 12 is an explanatory diagram of a directed graph showing a change in state when edge processing is performed on the directed graph of FIG. 1;

FIG. 13 is an explanatory diagram of a directed graph table showing a change in state when edge node deletion processing is performed on the directed graph corresponding to FIG. 12;

FIG. 14 is a diagram schematically showing the relationship between nodes and edges of a network using a directed graph according to a conventional expression method;

FIG. 15 is a diagram showing a table of a set of data elements expressing a network using a directed graph according to a conventional expression method.

[Explanation of symbols]

10: directed graph, 11: start clause, 12: end clause, 20
... Directed graph table, 21 ... Edge number field, 22
... label field, 23 ... firstp flag field,
24 ... sonp flag field, 25 ... brotherp flag field, 26 ... finalp flag field, 27 ... brother address field, 30 ... directed graph table, 31
... input device, 32 ... data processing device, 33 ... storage device,
34 output device, 35 directed graph generation processing unit, 36
Character string analysis processing unit, 37: first editing processing unit, 38: second
Edit processing unit, 65: directed graph, 66: directed graph, 6
7 ... sub-directed graph, 68 ... second half, 80 ... network representation of directed graph, 81 ... start clause, 82 ... end clause,
83 ... end clause, 84, 85, 86 ... side, 90 ... data table, 91 ... section number field, 92 ... end clause flag field, 93 ... side number field, 94 ... label field, 95 ... pointer field, 100 ... directed graph , 101: directed graph, 102: sub-directed graph, 1
03: second half, 121: directed graph, 122: sub-directed graph, 131: directed graph table, 132: edge data.

Continuation of the front page (56) References JP-A-6-162088 (JP, A) JP-A-7-73202 (JP, A) JP-A-2-148173 (JP, A) JP-A-2-202620 (JP) JP-A-5-307476 (JP, A) JP-A-7-152919 (JP, A) JP-A-5-20124 (JP, A) JP-A-4-107789 (JP, A) 2-272622 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G06F 17/30 JICST file (JOIS)

Claims (4)

(57) [Claims] 1. A first reference side which is a side which first designates a certain node, a child side which is a first side emanating from the node which the target side designates, and an adjacent side which emanates from the same node as the target side A directed graph having a structure composed of sibling edges that are edges to be formed is formed by recording each edge data one-dimensionally in a table in a predetermined order, and performing an editing process on each edge data of the directed graph formed in the table. A directed graph editing processor for performing, recording a first reference edge, recording a child edge next to the first reference edge,
If there is a sibling side, after the recording of the sibling side or less is completed, record data in the side serial format for recording the sibling side is formed, and each side data of the first reference side, the child side, and the sibling side are sequentially recorded. When a directed graph to be edited is input, the edge data of the input directed graph to be edited and each edge data of the record data stored in the storage unit are sequentially stored. First editing processing means for searching and comparing, separating edge data up to a matching edge and remaining edge data from the edge data of the directed graph to be edited, and performing an editing operation on record data of the edge data; After editing the record data for the table, the record of each side data up to and after the position of the record data where the edit operation was performed For the data, the editing process of updating the value of each of the reference 2
A directed graph edit processing device, comprising: an edit processing unit. 2. Each edge data in the edge serial format record data is an initial reference flag indicating whether or not the edge is an initial reference edge, label data of the edge, and whether or not a node issuing this edge is an end node. 2. The directed graph editing processing apparatus according to claim 1, wherein the directed graph editing processing device is constituted by an end flag for designating whether a child edge exists, a child flag for designating the presence or absence of a child edge, and a sibling flag for designating the presence or absence of a sibling edge. 3. The directed graph editing processing device according to claim 2, wherein the directed graph of a tree expression representing a word group is stored as edge serial record data, and a new word is registered and edited. The first editing processing means stores a word to be registered in a ₁
a ₂ ... a _n , each element a ₁ , a ₂ ,.
result of sequentially searching for a a _n from each side data of the record data, from a ₁ to a _i are present equal label edge data, when the comparison fails for a i _{+ 1} of the label, a in position from _{i + 1} edits to insert records data side data up a _n, the second editing means, and brother edge data and a _{i + 1} subsequent side data from a ₁ to a _i A directed graph edit processing device, which performs an edit process for updating a value of a brother reference of edge data. 4. The directed graph editing processing device according to claim 2, wherein the directed graph of the tree expression representing the word group is stored in edge serial format record data, and the editing process of deleting a registered word is performed. Is performed, the first editing processing means sets the word to be deleted to d ₁
When the d ₂ ... d _n, each element of the word d _1, d _2, ...,
If the d _n was successful sequentially searched from the edge data of the record data, among each of the edge data obtained by the search, and delete the end flag side data d _n, the edge data of the d _n if the child flag, and ends the editing without any Thereafter, if there is no child flag side data d _n, by changing the j from 1 to (n-1), terminated by an edge data d _j The maximum j with a flag is jmax, and k
The varied from jmax to n, and kmax maximum k of the edge data has siblings flags d _k, when the value obtained by subtracting 1 from kmax and i, the d _{i + 1} side data d _n The editing is completed by deleting up to the side data, and the second edit processing means sets a value of a brother reference to the side data from d ₁ to d _i and the side data from d _{i + 1} and on. A directed graph editing processing device, which performs editing processing for updating.