JP3062119B2 - Character string search table, method for creating the same, and character string search method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character string search method, and more particularly to a character string search method capable of performing a high-speed character string search by using a character string search table. In addition, the present invention relates to the character string search table and a method of creating the same.

[0002]

2. Description of the Related Art As a conventional general character string search method,
There are a sequential search method, a binary tree search method, and a hash search method, and the average number of searches in each search method is the number of reserved words for searching whether a given character string matches any of them. Let N be the sequential search method: (N +
1) / 2 times binary tree search method: Log ₂ N times, hash search method: once (however, depending on conditions), except for the hash search method, the number of searches increases as the number of reserved words increases, so that it is difficult to search for reserved words at high speed. Therefore, in the related art, the hash search method is the fastest search method.

Hereinafter, each method will be described in detail.

[Prior Art 1] Sequential Search Method A sequential search method is a method in which N (N> 0) reserved words are stored in an array of character strings in advance, and this array is stored in an array of character strings. While increasing the array elements from 1 to N, the reserved words sequentially stored in the array elements are referred to, and the input character string and the character string between the referenced reserved words are compared to obtain the reserved words. This is a method for searching. Because the search direction is linear, it is also called a linear search method.

[Related Art 2] Binary Tree Search Method A binary tree search method is a method in which N (N> 0) reserved words are
Data and two left and right pointers ref at each node (node)
Descendants that have t and right, and the descendant data connected by the left pointer left are smaller than their own, and the right pointer rig
ht is stored in a binary tree structure that is larger than itself, and a search is started for the input character string from the root node of the binary tree structure. The character strings are compared in magnitude. If the comparison result is the same, the search ends. If the comparison result is small, l
move to the descendant node connected to the left pointer, and if the result of the comparison is large, move to the descendant node connected to the right pointer, and again save the input character string and the destination node This is a method of comparing a word and searching for a character string.

[Prior Art 3] Hash search method A function h (x) for converting data x into integers distributed as uniformly as possible within a range of 0 ≦ h (x) <n is called a hash function. A method of searching for a reserved word using is called a hash search method. As a simple hash function, the data x is converted to an integer, divided by the value M, and the remainder is used as a function value. Usually, the value M is a prime number.

A typical method of linking a hash function value to a reserved word registration location is to prepare n binary search trees and register data x in a tree of number h (x), An array for storing reserved words is prepared, and the value h (x)
Is a method of registering a reserved word directly in an element as an element of this array. Since a method using a binary tree requires a longer search time than a method of preparing an array storing n reserved words, a method using an array storing n reserved words will be described below.

[0008] This is done by using this hash function in advance.
The reserved word is converted to an integer value, and the reserved word is stored as an element of an array for storing the reserved word. The input character string is subjected to a hash function to obtain an integer value. Is a method of searching for a character string as an element of an array in which is stored.

As an example, FIG. 9 shows a flow of creating a reserved word table in the hash search method, FIG. 10 shows a process flow of the hash function, and FIG. 11 shows a process flow of the hash search method.

[0010] As preprocessing of the hash search method, 901
Prepare an array (hash table) for storing reserved words.
Next, a hash function is created in 902 (described later). Next, in 903, the reserved word is passed to the hash function of 902, a hash value is obtained, and this hash value is stored as an element of the array of 901 in the element. 90 reserved words only
Steps 2 to 903 are repeated.

Now, the internal processing of the hash function 902 will be described. FIG. 10 shows an example of internal processing of a simple hash function 902 in which an input character string is converted to an integer, divided by a prime number of a value M, and the remainder is used as a function value.
First, in step 1001, the value of a hash value h, which is one of the internal variables, is initialized to zero. Similarly, a character pointer s, which is one of the internal variables, is initialized to indicate the first character of the input character string. At 1002, the character indicated by the pointer s is obtained, and at 1003, it is determined whether or not this character is the end of the character string. If it is the end, go to 1008; otherwise, go to 1004. At 1004, the value of the variable h is shifted left by 8 bits (256 times), and
In 005, the character code value of the character indicated by the pointer s is added to the variable h. Next, in 1006, the value of the variable h is
The value is divided by the number of elements in the hash table prepared in step 1, and the remainder is substituted into h again. Next, at 1007, the character pointer s is advanced by one, and the processing from 1002 is repeated again.
When all the input character strings have been read in this repetition processing, the hash value h determined in 1008 is returned.

Finally, the processing flow of the hash search method will be described with reference to FIG. In step 1101, the input character string is obtained, and in step 1102, it is determined whether or not the input of the character string is terminated by checking whether the input character string length is 0. If not, the process proceeds to step 1103. The character string input in 1103 is passed to the hash function of 902,
The return value is used as the element of the array 901. Checks whether a reserved word is stored in this element. 11
If there is a reserved word in 04, the processing in the case where the reserved word of 1105 is found is executed, and the process returns to the input of the character string of 1101 again. If no reserved word is found in 1103, the process returns to 1101 in the same manner. When the input character string ends, the end of the character string input is recognized in 1102, and the processing flow ends.

[0013]

The first problem here is that the hash search method previously converts each reserved word into an integer value that is distributed as uniformly as possible by using a hash function, and the integer value is converted to the integer value. Each reserved word is stored as a numerical value as an element number of an array for storing the reserved word. However, hash values may overlap depending on the number of reserved words to be registered and the number of characters constituting one reserved word.

The reason is that, in general, when one character is represented by n bits in the internal representation of a computer, if the number of characters constituting one character string is m, it is necessary to express an integer value unique to the character string. Although n × m bits are required, if the hash value cannot be represented by n × m bits, the integer value cannot be mapped one-to-one. For example, if an attempt is made to map a hash value that can be handled by a 32-bit CPU register length (32 bits) from an original character string, even if the character is an ASCII code (1 character = 8 bits), a 4-character length is used. Only character strings can be mapped one-to-one. Therefore, when obtaining the hash value of a character string exceeding four characters, the hash values may overlap.

The second problem is that even if the hash function is made plural (multi-stage) so that a unique value can always be generated, the time required for the operation in the hash function increases, and the unique value It takes time to generate.

The reason is that, among the current CPUs, arithmetic instructions, especially multiplication / division instructions, take more time than other instructions, for example, branch instructions, but are described in the first problem. As described above, in the hash function, when the hash function value exceeds the number of data bits that can be handled by the CPU, division is always required. If a plurality of hash functions are used, a multiplication / division instruction must be used many times, which increases the operation time.

As an example, each operation time in the μPD 70208 is as follows.

Addition: 14 clocks Subtraction: 14 clocks Multiplication: 39 clocks Division: 34 clocks Comparison: 16 clocks 8 bits left shift: 15 clocks Branch: 15 clocks The above two problems hinder the speed-up of character string search. .

Therefore, an object of the present invention is to provide a high-speed character string search method.

[0020]

[0021]

According to the present invention, there is provided a method for creating a character string search table according to the present invention, in which a two-dimensional array structure is provided in which all types of characters that can be input are set as rows, and state values indicating the state of the character string search processing are set as columns. A branch destination address for branching to a processing procedure for changing a state value determined in advance according to a search processing state of all reserved words in each array element, or processing when the reserved word is searched for In the character string search table creating method for creating a character string search table in which a branch destination address for branching to a procedure is stored, all types of characters that can be input are set to the size of the character string search table in the row direction. A first procedure, a second procedure of preparing a state value “0” indicating an initial value of the character string search situation, and a processing unit “processing 0” for changing the state value to “0”; table A third procedure for sorting all reserved words to be registered, all reserved words which are the sorted first
Of reserved word "reserved word 1", and the first to nth characters constituting "reserved word 1" (where n is the character string length of reserved word 1)
And a fourth procedure for preparing state values “1” to “n” from the state values “0” to “n” in the column direction of the character string search table. 5
And the processing unit "processing 1" for changing the state value to "1"
, A sixth procedure for preparing n processing units up to a processing unit “processing n” for changing the state value to “n”, and a processing “processing reserved word” when the reserved word “reserved word 1” is searched for In the seventh procedure for preparing "1" and the character string search table,
Each of the characters from the first character to the n-th character of the “reserved word 1” is a line, and each of the elements having a state value “0” to a state value “n−1” is a column. Eighth procedure for storing the branch address to "n", and processing in the element having the delimiter character as a row and the state value "n" as a column in the character string search table The address to "processing reserved word 1" A ninth procedure of storing the reserved word “reserved word m” after the sorting, a tenth procedure of obtaining the number of characters A of the “reserved word m”, and a reserved word “reserved word m” The word "reserved word m
-1 ", and the eleventh procedure for calculating the number B of the same characters consecutive from the first character to the n-th character of both reserved words, and AB
From the state value “n + 1” to the state value “n + A−
B), a thirteenth procedure in which the state value “0” to the state value “n + A−B” are again set in the column direction of the character string search table, and a state value From the processing unit “processing n + 1” that changes to “n + 1”, the processing unit “processing n” that changes the state value to “n + AB”
+ A-B "to prepare AB processing units, respectively.
And a fifteenth procedure for preparing a process “processing reserved word m” when the reserved word “reserved word m” is searched. In the character string search table, B + 1 of the “reserved word m”
A sixteenth procedure of storing a branch address to the processing unit “processing n + 1” in an element having a character as a row and a status value “B” as a column, and in the character string search table, the “reserved word m” Each character from B + 2 character to A character is lined,
From the state value “n + _100; 50 B1 —100; 100 B” to “n”
+ A-B-1 ”is included in each of the elements having“ process n +
A seventeenth procedure for storing the branch address from "2" to the processing "n + AB", and processing in an element having a delimiter character as a row and a status value "n + AB" as a column in the character string search table An eighteenth procedure for storing the address to the “process reserved word m”, and the state value “n + A−
B) and the tenth to nineteenth steps are repeated for all the remaining reserved words that have been sorted, and after all the reserved words have been registered, the character string search A twentieth procedure of storing a branch address to the processing unit “processing 0” in an element that does not store anything among the elements of the table.

A recording medium according to the present invention is characterized by recording a program for causing a computer to execute the above character string search method.

Further, a recording medium according to the present invention is characterized by recording a program for causing a computer to execute the above-described method for creating a character string search table.

[0024]

[0025]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The description will proceed in the following order.

A. Method of creating character string search table B. String search method using a string search table

In this embodiment, the characters to be handled are ASCII code type characters, and the characters that can be input are from 0x00 in hexadecimal notation.
128 types of characters up to 0x7F. Since there are 128 types of characters that can be input, the number of character elements in the character string search table is 1
It becomes 28 pieces. In addition, the character that separates a character string from the character string (this is defined as a delimiter) is replaced by a blank character (character code 0x2
0). Here, reserved words are defined as follows.

NEC XYZ NIMS NET First, a state value “0” of the character string search processing state is prepared.
The processing unit “process 0” that changes the status value to “0” is defined as follows.

[0029]

[Table 1] Next, the reserved words are sorted. This can be in ascending or descending order. In this example, sorting is performed in ascending order. After sorting, the sequence is as follows.

NEC NET NIMS XYZ Since the first reserved word NEC is composed of three characters, three new state values “1”, “2”, and “3” are prepared. Therefore, a character string search table for recognizing the reserved word NEC is prepared as shown in FIG.

Next, three processing units for updating state values corresponding to these three state values are prepared. Here, the three processing units are defined as “processing 1”, “processing 2”, and “processing 3”, and are labeled so that addresses to the respective processing units can be obtained.
In addition, a processing unit “processing NEC” when the reserved word NEC is searched is also created.

Each of these processing units is described below.

[0033]

[Table 2] The “read processing unit” in each processing content will be described later in the flow description. Further, the processing when the reserved word NEC in the “processing NEC” is searched is an arbitrary processing.

Next, referring to the character string search table of FIG. 4 from the first character "N" of the reserved word NEC and the state value "0", an element having the state value "0" and the character element "N" Stores the branch address to "Process 1". Similarly, referring to the character string search table of FIG. 4 based on the second character “E” and the state value “1”, “processing” is included in the element having the state value “1” and the character element “E”.
Store the branch address to "2". Similarly, referring to the character string search table of FIG. 4 from the third character “C” and the state value “2”, “processing 3” is included in the element having the state value “2” and the character element “C”. Stores the branch address to Finally, referring to the character string search table of FIG. 4 based on the delimiter "blank character" and the state value "3", the state value is "3", and the character element is branched to "processing NEC" in the delimiter element. Store the address. This completes the character string search table for searching for the reserved word NEC. FIG. 5 shows a character string search table obtained as a result.

Subsequently, the second and subsequent reserved words are registered. When registering the second and subsequent reserved words, it is necessary to determine whether it is necessary to prepare a new state value in accordance with the following formula.

When the number of characters constituting a reserved word to be registered from now on is assumed to be A, and the number of the same character continuing from the first letter to an arbitrary character in the reserved word registered immediately before and the reserved word to be registered is assumed to be B, AB is the newly required number of states. When a new state value is required, the character string search table must be extended in the direction of the state value element.

In this example, the second reserved word is NET
Since the reserved word NET has a total of three characters, A = 3. Next, in the reserved word NEC registered immediately before and the reserved word NET registered this time, the same number of characters is checked from the initial letter to the arbitrary character. In the reserved word NEC and the reserved word NET, “N” and “E” are the same, and B = 2 because two characters are the same. Therefore, AB = 1, and it is necessary to prepare one new state value. Therefore, a new state value “4” is prepared, and the elements in the state value element direction (column direction) of the character string search table are changed from the state value “0” to the state value “4”. Further, a process "process 4" for newly updating the state value and a process "process NET" for searching for a reserved word NET are provided. The "process 4" and "process NET" are defined as follows.

[0038]

[Table 3] Next, the character string search table is referred from the "T" of the third character (B + 1 character) and the state value "2" (state value "B") in the reserved word NET, and the character having the state value "2" The branch address to the processing unit “processing 4” is stored in the element whose element is “T”. Next, the delimiter character and the status value “4” (status value “n + A
−B ”, where n is the highest state value at the end of the registration after the previous reservation) and refers to the character string search table, and the state value is“ 4 ”and the character element is within the delimiter element Stores the branch address to the processing "reserved word NET". FIG. 6 shows the character string search table obtained as a result.

Similarly, consider the next reserved word NIMS. The reserved word registered immediately before is NET. Since the number of characters A constituting the reserved word to be registered from now on is A = 4 (characters), and the number of the same character continuous from the initial character to the arbitrary character in the reserved word registered immediately before and the reserved word to be registered is B = 1 (character). , AB = 3, and three new state values “5”,
“6” and “7” are provided, and the size in the state value element direction (column direction) of the character string search table is set to a size from the state value “0” to the state value “7”. Further, processing units “processing 5”, “processing 6”, and “processing 7” for changing the state values to “5”, “6”, and “7” are provided. Also, a processing unit “processing NIMS” for searching for a reserved word NIMS is provided. These processing units are described below.

[0040]

[Table 4] Next, refer to the character string search table from the second character "I" of reserved word NIMS and the state value "1" (since the state value is "1" if the first character is "N"), and the state value The branch address to the processing unit “processing 5” is stored in the element having “1” and the character element “I”. Similarly, refer to the character string search table from the third to fourth characters of the reserved word NIMS and the state values "5" and "6", and find an element whose state value is "5" and whose character element is "M". The branch address to “processing 7” is stored in an element whose processing value is “processing 6”, the state value is “6”, and the character element is “S”. Finally, the character string search table is referred to based on the delimiter character and the state value “7”, and the branch address to the processing unit “processing NIMS” is stored in the delimiter element with the state value “7”. FIG. 7 shows a character string search table obtained as a result.

Similarly, consider the last reserved word XYZ.
The reserved word registered just before is NIMS. A, as before
When B is obtained, since A = 3 characters and B = 0 characters, AB
= 3, three new state values "8", "9", "10"
Is provided, and the size from the state value “0” to the state value “10” is set as the size in the column direction of the character string search table. In addition, processing units “processing 8”, “processing 9”, and “processing 10” for changing the state values to “8”, “9”, and “10” are provided. Also reserved word XY
A process “process XYZ” when Z is searched is also provided. These processing units are described below.

[0042]

[Table 5] Next, the character string search table is referred from the first character “X” of the reserved word XYZ and the state value “0”, and the processing unit “processing 8” is included in the element whose state value is “0” and whose character element is “X”. Is stored. Similarly, referring to the character string search table from the second character "Y" to the third character "X" of the reserved word XYZ and the state values "8" and "9", the character with the state value "8" Element is "Y"
The branch address to “processing 10” is stored in the element having the processing unit “processing 9” in the element and the element having the status value “9” and the character element “Z”. Next, the character string search table is referenced from the character of the delimiter and the state value “10”, and the branch address to “processing XYZ” is stored in the element. FIG. 8 shows a character string search table obtained as a result.

Finally, in the character string search table of FIG. 8, the branch address to "processing 0" is stored in all the other elements not used when registering the reserved words NEC and NET, NIMS and XYZ. I do. This is a reserved word NEC, NET, NI
The setting of the character string search table for MS and XYZ is completed.

The procedure for creating a generalized character string search table will be described with reference to FIG.

In the first procedure (201), all types of characters that can be input are set to the size in the row direction of the character string search table.

In the second procedure (202), an initial state value "0" is prepared, and an initialization processing unit "process 0" for setting the state value to the initial state value is prepared.

In the third step (203), all the reserved words to be registered are sorted in ascending or descending order according to the sorting criterion.

In the fourth procedure (204), a state value corresponding to the first reserved word is prepared. The number “n” of this state value is the character string length of the first reserved word. Normally, state values are prepared in order from 1.

In the fifth procedure (205), the size of the character string search table in the column direction is determined. Column is "0"
To “n”, and the size in the column direction is the character string length of the first reserved word + 1.

In the sixth procedure (206), a state value change (update) processing unit is prepared. In the above example, the processing unit “processing 1” that changes (updates) the state value to “1” to the processing unit “processing 3” that changes the state value to “3 (= n)” is prepared.

In the seventh procedure (207), a processing unit for the case where the first reserved word is searched is prepared. In the above example, “processing NEC” is prepared.

In the eighth procedure (208), the branch address to each state change processing unit is stored in the character string search table. In the above example, for example, the branch address to “processing 1” is stored in an element whose state value is “0” and whose character element is “N”.

In the ninth procedure (209), a branch is made to the processing section when the first reserved word is searched for in the element having the row of delimiter characters in the character string search table and the state value "n" as a column. Store the address. In the above example, the state value is “3”, and the character element stores the branch address to “processing NEC” in the delimiter element.

In the tenth procedure (211) to proceed when there is a next reserved word, the m-th
The number A of characters constituting the reserved word is obtained from the reserved word (m is an integer of 2 or more).

In an eleventh step (212), the number B of consecutively matching characters is obtained from the first letter of the m-th reserved word and the (m-1) -th reserved word.

In the twelfth procedure (213), AB
A new state is prepared in which the state values “n + 1” to “n + AB” are prepared.

In the thirteenth procedure (214), the size of the character string search table in the column direction is enlarged by AB, and the range is updated from the state value "0" to the state value "n + AB". I do.

In the fourteenth procedure (215), the processing unit “processing n + 1” for changing the state value to “n + AB” is changed from the processing unit “processing n + 1” for changing the state value to “n + 1”. In the above example of preparing AB processing units up to "AB", for example, "NIMS"
When processing is performed, a processing unit “processing 5” that updates the state value to “5” is prepared.

In the fifteenth procedure (216), a processing unit is prepared for searching for the m-th reserved word. In the above example, for example, “processing NIMS” is prepared.

In the sixteenth procedure (217), in the character string search table, the processing unit “processing n +” is included in an element having the B + 1st character of “reserved word m” as a row and the state value “B” as a column.
Store the branch address to "1". In the above example, for example, when handling “NIMS”, “I” which is the second character of “NIMS” is placed in a row, and “Process”
Store the branch address to "5".

In the seventeenth procedure (218), in the character string search table, the characters from the B + 2nd character to the Ath character of the “reserved word m” are lined, and the state values “n + 1” to “n” are written. The branch address from "process n + 2" to process "n + AB" is stored in each element having "+ AB-1" as a column. In the above example, for example, when dealing with "NIMS", "NIMS"
Lines each character from the third character to the fourth character of
(= 4 + 1) ”and a branch address from“ process 6 ”to“ process 7 ”is stored in each element having the state value“ 6 (= 4 + 4-1-1) ”as a column.

In the eighteenth procedure (219), in the character string search table, the address of the processing “processing reserved word m” is stored in the element having the delimiter character as a row and the state value “n + AB” as a column. I do. In the above example, for example, when handling "NIMS", the delimiter character is lined, and the status value "7 (= 4 + 4
-1) ”, the address to the“ processing NIMS ”is stored in the element whose column is“ ”.

In the nineteenth procedure (220), the state value “n + AB” is substituted for the state value “n”. In the above example, for example, when “NET” is handled, the status value “4 (= 3 + 3−2)” is substituted for the status value “3”.

In the twentieth procedure (221), which proceeds when there is no next reserved word, among the elements of the character string search table, the processing unit "processing 0" is placed in an element storing nothing.
Stores the branch address to

Next, the processing flow of the character string search of this embodiment will be described. FIG. 3 is a processing flow of the character string search method according to the present embodiment. First, at 300, the state value is initialized to "0". Here, as an example, consider the case where a character string NEC and then a delimiter blank character are input. First, the reading processing unit 301 reads the character “N”.

Next, at 302, the address of the "process 1" stored at the position where the character element is "N" and the state value element is "0" in the character string search table shown in FIG. 303
Jumps to one of the processes “process 1”, updates the state value to “1” in “process 1”, and returns to the read processing unit 301 again. Similarly, at 302, the address of "processing 2" in which the character element in the character string search table shown in FIG. 6 is the next input character "E" and the state value element is stored at the position "1" is obtained. Then, the process jumps to “Process 2” of 302, updates the state value to “2” in “Process 2”, and returns to the read processing unit of 301 again.

Further, at 302, the address of "process 3" in which the character element in the character string search table shown in FIG. 6 is stored at the position of the next input character "C" and the state value element of "2" Then, the process jumps to “Process 3” of 303, updates the state value to “3” in “Process 3”, and returns to the read processing unit 301 again.

Further, at 302, the character element in the character string search table shown in FIG. 6 is the next input character "blank character" and the state value element is stored at the position of "3". The address is obtained, and the process jumps to “processing NEC” 303. When jumping to “processing NEC”, the reserved word NEC has been searched.

Next, the processing for searching for the reserved word NEC in the "processing NEC" is executed, the state value is updated to "0", and the process returns to the character reading unit 301 again to return to the next character string. Prepare for the search.

As another example, consider a case where a character string "ABC" not included in a reserved word is input. In the initial state of the program, the state value is “0”. Introduction 30
The character “A” is read by the read processing unit 1. Next, at 302, the address "processing 0" stored at the position where the character element is "A" and the state value element is "0" in the character string search table shown in FIG. Jump to "0", update the status value to "0" in "Process 0",
The process returns to the read processing unit again. That is, the state value does not change from “0”. The same applies to the subsequent input characters "B" and "C", so that the input character string ABC is ignored.

FIG. 1 shows a generalized processing flow of the character string search. In FIG. 1, processing blocks denoted by reference numerals 100 to 102 correspond to reference numerals 300 to 3 respectively.
It is the same as the processing block indicated by 02. "Process M + 1"
(Reference numeral 105) indicates “processing NEC”, “processing NET”, “processing
NIMS "and" Process XYZ ". "Process N"
10 ".

In this embodiment, the characters to be handled are
The ASCII code system is used, but the characters are not limited to this system, and the present invention can be applied to characters in any system such as JIS kanji and UNICODE.

The character string search method and the character string search table creation method in this embodiment are realized by a computer-executable program, which is recorded on a computer-readable recording medium.

The character string search table in this embodiment is recorded on a computer-readable recording medium. This character string search table is created by the computer using the program of the creation method according to the present embodiment, recorded on a recording medium, and read from the recording medium and used when the computer executes the program of the search method according to the present embodiment. .

[0075]

As described above, according to the present invention, a character string can be searched at a higher speed than in a hash search method.

The reason is that the character string search method according to the present invention uses the input character and the state value at that time as the position information of the elements of the array of the character string search table, at the position specified by the state element value and the character element value. Operation instructions such as multiplication and division, which take a long time in CPU operation, are performed by recognizing a character string by branching by referring to the address of the branch destination process stored in a certain element and repeating the branch destination process. Is not used. Further, the input character string can be determined as a reserved word by a single search, so that the character string can be searched at high speed.

Here, as a guide, a comparison is made between the simple hash search method shown in FIG. 10 and the character string search method according to the present invention, in which the input character string is converted to an integer, divided by the prime number of the value M, and the remainder is used as a function value. Then, how fast the character string search method according to the present invention is will be described.

Taking the μPD70208 CPU of NEC as an example, the number of clocks required for each operation is as follows.

Addition: 14 clocks Subtraction: 14 clocks Multiplication: 39 clocks Division: 34 clocks Comparison: 16 clocks 8 bits left shift: 15 clocks Branch: 15 clocks From these, the CPU time required for the hash search method and the search for this method The comparison of the CPU time required for the method is as follows.

CPU time in the hash search method = 77 clocks × number of characters (left 8 bit shift operation 15 clocks × 1 time = 15 clocks) (addition 14 clocks × 2 times = 28 clocks) (division 34 clocks × 1 time = 34 clocks) CPU time in the search method according to the present invention = 30 clocks × the number of characters (15 clocks × 2 times = 30 clocks) Therefore, when the search method according to the present invention is used, the CPU time is reduced to about ２ and the character string is searched at high speed. Can be performed. Also, the more the number of hash functions and the number of operations increase, the more the difference between them increases.

[Brief description of the drawings]

FIG. 1 is a flowchart of a character string search method using a character string search table according to the present invention.

FIG. 2 is a flowchart of a method for creating a character string search table according to the present invention.

FIG. 3 is a flowchart of a character string search method using a character string search table according to the embodiment of the present invention.

FIG. 4 is a diagram showing a table size of a character string search table required for searching for a reserved word “NEC” in the embodiment of the present invention.

FIG. 5 is a diagram showing a character string search table for searching for a reserved word “NEC” in the embodiment of the present invention.

FIG. 6 is a diagram showing a character string search table for searching for reserved words “NEC” and “NET” in the embodiment of the present invention.

FIG. 7 shows reserved words “NEC” and “N” in the embodiment of the present invention.
It is a figure which shows the table for character string search for searching for "ET" and "NIMS".

FIG. 8 shows reserved words “NEC” and “N” in the embodiment of the present invention.
It is a figure which shows the table for character string search for searching for "ET", "NIMS", and "XYZ".

FIG. 9 is a flowchart of a hash table creation method in the hash search method.

FIG. 10 is a processing flowchart inside a hash function.

FIG. 11 is a flowchart of a hash search method.

[Explanation of symbols]

 100 initialization 101 read processing unit 102 jump 103 processing 0 104 processing 1 105 processing M + 1 106 processing N

Continuation of the front page (56) References JP-A-4-348469 (JP, A) JP-A-63-31530 (JP, A) JP-A-3-75869 (JP, A) JP-A-2-255985 (JP) , A) Junichi Aoe, "Tries and Applications", Information Processing Vol. 34, no. 2,244-251 (February 15, 1993) Tetsuro Ito, "Software Course 19 Information Retrieval," pp 99-100, Shokodo (August 10, 1986) (58) Fields surveyed (58) Int.Cl. ⁷ , DB name) G06F 17/30

Claims (5)

(57) [Claims] 1. All characters that can be input are defined as lines,
This is a table of a two-dimensional array structure in which status value types indicating a character string search processing state are set as columns, and a processing procedure for changing, in each array element, a state value determined in advance according to a search processing state of all reserved words A character string search table storing a branch destination address to branch to, or a branch destination address to branch to a processing procedure when the reserved word is searched, wherein a reserved word having a common halfway character is stored. In the method for creating a character string search table for creating a character having a common character string search processing state with respect to a common character, all character types that can be input are set to the size in the line direction of the character string search table. A second procedure of preparing a processing value “0” indicating an initial value of the character string search situation and a processing unit “processing 0” for changing the state value to “0”; Register A third procedure for sorting all reserved words, extracting a first reserved word “reserved word 1” from all the sorted reserved words, and extracting a first reserved word to an n-th character constituting the “reserved word 1” (N is the character string length of the reserved word 1), a fourth procedure for preparing state values “1” to “n” from the state values “1” to “n”, respectively. A fifth procedure for setting the size of the search table in the column direction, and a processing unit “processing 1” for changing the state value to “1” to a processing unit “processing n” for changing the state value to “n” a sixth procedure for preparing n processing units; a seventh procedure for preparing a process “processing reserved word 1” when a reserved word “reserved word 1” is searched; 1 of "reserved word 1"
The branch address from "Process 1" to "Process n" is set in each element in which each character from the character to the n-th character is a line, and the status value is from "0" to "n-1" as a column. 8th to store
A ninth procedure of storing the address of the processing “processing reserved word 1” in an element having a delimiter character as a row and a state value “n” as a column in the character string search table; If there is a subsequent reserved word “reserved word m”, it is retrieved and the tenth character number A of the reserved word “m” is obtained.
And an eleventh procedure in which the reserved word “reserved word m” and the reserved word “reserved word m−1” are compared to obtain the number B of the same characters consecutive from the first character to the n-th character of both reserved words. , AB, the state value “n + 1” is newly changed from the state value “n + 1”.
A twelfth procedure for preparing up to "+ AB"; and a thirteenth procedure for again setting the size from the state value "0" to the state value "n + AB" in the column direction of the character string search table.
And the processing unit “process n” that changes the state value to “n + AB” from the processing unit “process n + 1” that changes the state value to “n + 1”
+ A-B "to prepare AB processing units, respectively.
And a fifteenth procedure for preparing a process “process reserved word m” when the reserved word “reserved word m” is searched; and in the character string search table, the “reserved word m”
The first address stores the branch address to the processing unit “processing n + 1” in an element having the B + 1st character as a row and the status value “B” as a column.
6. In the character string search table, the "reserved word m"
From the (process n + 2) to the process (n + AB) in each element having the B + 2 to A-th characters as rows and the state values "n + 1" to "n + AB-1" as columns. And a seventeenth procedure for storing the branch address of the process. In the character string search table, the address of the process “processing reserved word m” is set in an element having a row of delimiter characters and a column of the state value “n + AB”. An eighteenth procedure for storing, and a nineteenth procedure for substituting the state value “n + AB” for the state value “n”
Steps 10 to 19 are repeated for all the remaining reserved words that have been sorted, and after all the reserved words have been registered, among the elements of the character string search table, A twentieth procedure of storing a branch address to the processing unit “processing 0” in an element storing nothing. 2. A computer-readable recording medium on which a program for causing a computer to execute the method for creating a character string search table according to claim 1 is recorded. 3. The first to twentieth procedures of the method for creating a character string search table according to claim 1, a twenty-first procedure for initializing a state value to 0, and a step of obtaining input characters one by one. A twenty-third procedure, referring to the character string search table using the obtained input character as a line and a state value as a column, and branching to an address stored in an element, A twenty-fourth procedure for performing a state value changing process or a reserved word matching process and returning to the twenty-second procedure again. 4. A computer-readable recording medium on which a program for causing a computer to execute the character string search method according to claim 3 is recorded. 5. A computer-readable recording medium on which a character string search table created by the method of claim 1 is recorded.