JPH06162083A - Character-string retrieving device - Google Patents

Abstract

PURPOSE:To collate a character-string with a high throughput by operating in parallel plural filtering circuits as a pre-processing for collating the characterstring. CONSTITUTION:A parallel filtering means 3000 is constituted of a distributing means 3100 for fetching simultaneously plural characters of a text, dividing them into one character each and sending it out, plural filtering means 3200a, 3200b for deciding whether its character code is a character code contained in a designated retrieval term or not in parallel, and a collecting means 3300 for aligning the character code outputted therefrom and sending it out to a character-string collating means 102.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information processing system, and more particularly to a method of collating search terms in an information retrieval system, which collectively checks whether or not a plurality of partial character strings designated as a search term from a text character string exist. The present invention relates to a character string search device for searching.

[0002]

2. Description of the Related Art In the field of information processing systems, a word that a searcher wants to search, that is, a specific character string (hereinafter referred to as a search term) is selected from a document (hereinafter referred to as a text) that is a collection of character string data. It is one of the important processes to find all the documents including.

Several character string search devices for realizing such a search system have been proposed. The structure of a typical character string search device (LAE Horror:
"Text Retrieval Computers", Computers, March 1979, LA Hollaar: "Text
Retrieval Computers ", COMPUTER, March 1979)) is shown in FIG.

As shown in FIG. 2, in the character string search device 1, the search control means 101 controls the search device as a whole.
Communicates with the host computer. That is, the search request 201 sent from the host computer is accepted, analyzed, and sent as search control information 202 to the character string collating means 102 and the compound condition judging means 103. Further, the search control means 101 controls the storage device control means 104 to read the character string data 204 stored in the character string storage means 105 to the character string collating means 102.

The character string collating means 102 checks whether or not there is a character string that matches the search request 201, that is, a search term, in the input character string data 204, and if there is a corresponding character string, identifies the character string. The information 205 is output to the composite condition determination means 103. The compound condition determination means 103 checks whether or not the character string identification information 205 satisfies the logical condition or the like constituted by AND or OR specified in the search request 201. When the specified composite condition is satisfied, the identification information of the corresponding document and the text data of the document content are returned to the host computer as the search result 206.

As a character string collating method in the character string collating means 102, which is an essential part of the character string retrieving apparatus 1, a method using a finite automaton for searching a plurality of character strings by one text scanning is known. Further, hardware for executing this finite state automaton at high speed is disclosed in Japanese Patent Laid-Open No. 3-9
Japanese Patent No. 5672 is disclosed.

In this conventional example, the state of the automaton to be matched with the input character is indicated by placing a mark called a token. Ie 1 from the input text
When a character is entered, each state in which the token is placed is verified. Also, the token is always generated in the initial state when the input character code is input. However, if there is no transition state as a result of the collation,
Extinguish the token. The transition destination state is determined by referring to the state transition table using the state number of the state in which the token is placed and the input character code as an address. Therefore, when there are a plurality of tokens in the automaton, the state transition table is referred to a plurality of times for one character input. As a result, there arises a problem that the matching throughput drops to a fraction.

The collating operation when a plurality of tokens are present in the one character collating process will be described with reference to the automaton shown in FIG. This automaton is called "interface" and this different expression "interface",
"Interface", "interface", "interface", "interface", "interface", "interface", and "interface" for collating collectively It is a thing.

When "interface" is input as the input text, the token moves as shown in FIG. First, when "a" is input, a new token 1 is generated in the initial state of state 0. "A" in state 0
Since the transition due to is described (see FIG. 3), the matching is established, and token 1 moves to state 1.

When the next "n" is input, a new token 2 is generated in state 0. However, since the transition by "n" is not described here, the collation fails and the token 2 disappears. . Further, the token 1 that has moved to the state 1 is collated with “n” in the state 1. In this case, since the matching is established, the state moves to state 2. In this case, the collation operation is performed twice for each character in this way.

Similarly, "ta", "fu", and "e"
Is input, the token 1 moves from the state 3 to the states 5 and 6. During this period, tokens 3 to 5 are also generated in the initial state, but they disappear because the verification is not established.

In this way, the same processing is performed for "i" and "s" that are successively input.

As a result, in this example, in the above collation operation process, the collation is performed 14 times by the text input of 7 characters.

FIG. 5 shows the conventional character string collating means 102 for carrying out the above-mentioned character string collating process. The character string collating means 102 includes registers 211 and 250, 251, a state transition table 220, a collation result table 260, a selector 261, a gate 262, a multiplexer 263, buffers 280, 281, and a comparator 252.

An outline of the collating operation of the character string collating means 102 will be described below.

The input text 204 is stored in the register 211 character by character. The character code 302 output from the register 211 is input to the state transition table 220 as address information. The state transition table 220 is referred to by using the current state number 305 and the character code 302 as addresses, and the transition destination state number 303 to which the next transition is to be made
(Hereinafter referred to as the next state number) is read, and the register 2
Stored in 50.

In this conventional example, the next state number 303 is used as the token identifier. The next state number 303, which is the identifier of this token, is stored via the gate 262 and the multiplexer 263 in the buffer 280 or the buffer 281 that is selected as information indicating the location of the token. State transition table 22
When the next state number 303 output from 0 is 0 (zero), that is, the initial state number, it means that there is no token move destination. Therefore, when the next state number 303 is the initial state number 0, it is necessary to erase the token. This control is performed by the comparator 252 and the gate 262.

That is, the comparator 252 determines whether or not to erase the token, and the gate 262 controls whether or not to erase the token.

Specifically, when the next state number 303 is the initial state number 0 (zero), the comparator 252
Since the result of comparison with the state number 0 (initial state number) stored in the register 251 becomes equal, the gate 262 is closed, and the next state number 303 disappears in the gate 262 without being sent to the multiplexer 263. Become.
On the contrary, when the next state number 303 is not the initial state number 0 (zero), the next state number 303 is sent from the gate 262 to the multiplexer 263 and stored as a token.

The buffers 280 and 281 store the initial state number as the initial value at the head address, and the next state number 303 sent through the multiplexer 263.
Is stored from the next address in the initial state. By doing this, the token always exists in the initial state.

The next state number 303 stored in either the buffer 280 or the buffer 281 is read out as the current state number 305 at the next character code matching.

The selector 261 selects one of the buffer 280 or the buffer 281 in which the token, that is, the next state number 303 is stored, and the current state number 305 is sequentially read therefrom. The read end signal 307 is generated when all the data are read. The multiplexer 263 and the selector 261 are synchronized with each other, and the multiplexer 263 causes the buffer 280 to operate.
Is selected, the selector 261 selects the buffer 2
Select 81. Further, when the multiplexer 263 is selecting the buffer 281, the selector 261 selects the buffer 280. That is, the token to be transferred to the transition destination state is set as the next state number 303 in a buffer different from the buffer in which the token in the transition source state (stored in one buffer as the current state number) is stored. I am trying to store it.

Switching of the buffers 280 and 281 is performed by switching the buffer 280 selected by the selector 261.
Alternatively, it is performed at the time when the reading of any one of 281 or 281 is completed, that is, at the timing when the reading end signal 307 is generated. The character code is normally fetched from the text in the register 211 in synchronization with the register 250, but the character code is held until the read end signal 307 is generated, and the token of the transition destination, that is, the current state number is all stored in the buffer. It waits for the next input until it is read. The collation result table 260 stores a predetermined search term number for identifying each search term corresponding to a state that will be the end of the search term (hereinafter, referred to as an end state).
0 (zero) is stored in the other states. That is, the matching result 205 is meaningful only when the search term number output from the matching result table 260 corresponding to the state number is other than 0.

The character string collating process is realized by repeating the above series of operations for each character constituting the input text.

As described above, in this reference, the state transition table is referenced once when matching one token. Therefore, in the collation operation example of FIG. 4, the collation is performed on the token 14 times in the text input of 7 characters.
The state transition table is referenced twice. That is, the state transition table is referred to twice for each character on average. Therefore, there is a problem that the matching throughput is reduced to about half as compared with the case where only one matching is performed for one character.

As described above, Japanese Patent Application No. 4-63067 has been proposed as a method for preventing the collation throughput of the character string search device from decreasing due to the increased load of the character string collating means 102.

FIG. 6 shows the configuration of the character string search device in the above conventional example.

In this example, the filtering means 300 is provided between the character string storage means 105 and the character string collating means 102, and character codes not included in the search term are cut off from the text read from the character string storage means 105. hand,
Only the character code included in the search term is sent to the character string collating means 102, and unnecessary collating operation in the character string collating means 102 is omitted, so that the collating throughput is equivalently improved. That is, an object of the present invention is to provide an inexpensive character string collating device in which the search speed as a character string searching device does not decrease even if the character string collating means 102 using a low-speed memory is used.

FIG. 7 shows an embodiment in which a single filtering circuit is used as the filtering means in this conventional example.
Shown in.

The single filtering circuit 300 includes a character code register 320 and a single filtering table 3.
30 and a single character output circuit 340 including an output gate 360.

The single filtering table 330 is a one-dimensional memory which is accessed using a character code as an address. Here, "1" is stored as a match flag corresponding to the character code included in the search term.

For example, when "BUS" is given as the search term, the contents shown in FIG. 8 are set in the single filtering table 330. That is, 1 is set in the slots corresponding to the respective character codes of "B", "U" and "S" that form the search term "BUS".

The filtering operation is performed on the input text 20.
Character code register 32
It starts by taking in 0.

The single filtering table 330 is accessed using the character code 310 output from the character code register 320 as a reference address, and the coincidence signal 350 is read. In the case of the character code forming the search term, 1 is output as the matching signal 350. At this time, the output gate 360 is opened, and the character code 310 is output to the output line 207 and sent to the character string collating means 102 in the next stage. If the character code is not included in the search term,
0 is output as the coincidence signal 350. At this time, the output gate is closed and the character code 310 changes to the output line 20.
7 is not output and is not sent to the character string collating means 102 in the next stage.

For example, "BUS" is given as the search term, and "HIGH-SPEED CMOS SCSI B as the input text.
A specific operation of the single filtering circuit when "US CONTROLLER" is input will be described with reference to the timing chart of FIG.

First, "H" is input from the input text 204 into the character code register 320, and the character code 31
It is output as 0. Then, the single filtering table 330 is accessed at “H”, and 0 is output as the value of the coincidence signal 350. Since the value of the coincidence signal 350 is 0, the character code "H" is not output from the output gate 360.

The next "I", "G", "H", and "-" are similarly fetched in the character code register 320, but the coincidence signal 350 output from the single filtering table 330 is 0. There is no output from the output gate 360.

The next "S" is the character code register 3
Single filtering table 33 when loaded into 20
The value of the coincidence signal 350 output from 0 becomes 1, and the output gate 360 outputs “S” of the character code 310 to the output line 207.

During the next "P", "E", "E", "D", and "" (space), the coincidence signal 3350 output from the single filtering table 3300 becomes 0, so the output line 207. Nothing is output to.

Thereafter, by similarly performing the filtering process, "S", "S", "B" from the output gate,
“U” and “S” are output to the output line 207.

In this way, in this example, 6 characters of "SSSBUS" included in the search term are extracted from the input text of 30 characters "HIGH-SPEED SCSI BUS CONTROLLER" and output as the output text 207. That is, 23/30 of the input text 204, that is, about 4/5 is truncated as an unnecessary character, so only about 1/5 is sent to the character string matching unit 102, and the load on the character string matching unit 102 is reduced. It is possible to reduce it to 1/5.

[0042]

In the above-mentioned Japanese Patent Application No. 4-63067, a filtering means is provided between the character string storing means and the character string collating means, and a search term is selected from the text read from the character string storing means. By truncating the character codes not included in the above, and outputting only the character codes included in the search term to the character string matching means, the load of the character string matching means is reduced, and the matching throughput is improved equivalently.
As a result, in the character string collating means, "HIGHSPEED SCSI BUS
What was necessary to perform collation processing on the input text of 30 characters "CONTROLLER" can now be completed by collation processing for 6 characters of "SSSBUS", and the load of the character string collating means is reduced to about 1 /. It can be set to 5.

For example, when the processing speed of the filtering circuit is 50 nanoseconds / time and the processing speed of the character string collating means is 100 nanoseconds / time, the processing time when the filtering means is not provided and the case where the filtering means is provided. The processing time of is as follows.

First, if no filtering means is provided, it is necessary to collate all 30-character text "HIGH SPEED SCSI BUS CONTROLLER" with a character string, so the processing time of the character string collating means is 100 nanoseconds / times. 30 times = 3,000 nanoseconds are required.

Next, when the filtering means is provided, the 30-character text "HIGH SPEED SCSI BUS CONTROLLER" is converted to "SSSBUS" by the filtering means.
Will be deleted to 6 characters. In the filtering process, it is necessary to refer to the filtering table once for each input character code, so "HIGH SPEED SCSI BUSCONTROL"
It is necessary to process 30 times to filter the text of 30 characters "LER". As a result, the processing time of the filtering means is 50 nanoseconds / time × 30 times = 1,500 nanoseconds.

Further, as shown in FIG. 10, 6 of "SSSBUS"
In order to collate the character, the character string collating process needs to be performed eight times. Therefore, the processing time of the character string collating means needs to be 100 nanoseconds / time × 8 times = 800 nanoseconds.

Thus, by installing the filtering means between the character string storing means and the character string collating means,
The character string search processing that required 3,000 nanoseconds was 1,50
It became possible to perform processing in 0 nanoseconds, and it was possible to improve search throughput by a factor of two.

However, the character string retrieval apparatus according to this conventional example has the following problems.

That is, although the character string collating means is actually operating for 800 nanoseconds, the filtering processing requires a processing time of 1,500 nanoseconds. That is, a situation arises in which the filtering process provided for improving the throughput becomes a bottom neck.

As described above, in the character string search device according to the conventional example, the speed of the filtering process becomes a bottom neck, so that the character string collating means is idle and the collating throughput is improved up to the limit performance of the character string collating means. There is a problem that it cannot be done.

The object of the present invention is to further improve the throughput of the filtering process by operating a plurality of filtering means in parallel, avoid the bottom neck of the filtering process, and maximize the performance of the character string collating means. That is, to bring out, that is, to increase the search throughput of the entire character string search device to the limit.

[0052]

These problems are solved in a character string search device for collectively searching whether or not a plurality of designated search terms are present in a text composed of code-represented characters. This is achieved by providing the following parallel filtering means between the character string storage means and the character string collating means.

That is, as the parallel filtering means, a distributing means for fetching a plurality of characters at a time from the text read from the character string storing means and dividing and sending the divided characters one by one, and a character code output from the distributing means are used. Capturing one character at a time, determining whether or not the character code is included in the search term, a plurality of filtering means arranged in parallel, and a character code output from the plurality of filtering means It is composed of collecting means for arranging and sending to the character string collating means.

[0054]

The principle of the present invention will be described below with reference to FIG. In the parallel filtering unit 3000, the distribution unit 3100 reads the text from the character string storage unit 105 every two characters, and the filtering unit 3 at every one character.
200a and 3200b. The filtering means 3200a and 3200b determine whether or not one input character is a character included in the search term, and output the result to the collecting means 3300. In the collection means 3300, as the processing results in the filtering means 3200a and 3200b, only the character codes included in the search term are aligned and the character string matching means 102 is arranged.
Output to.

By doing so, it is possible to equivalently increase the processing throughput of the filtering means by a factor of two and avoid the filtering means from becoming a bottom neck. As a result, it is possible to further improve the search throughput as the character string search device by a factor of two.

The principle will be described below using a specific example. For example, "BUS" is specified as the search term, and "HIGH-SPEED SCSI BUS CONTRO" is displayed as text from the character string storage means.
Consider the case where "LLER" is loaded.

In the distributing means 3100, first, the first operation takes in the first and second characters "HI" of the text, and the first character "H" is filtered by the filtering circuit 3200.
The second character “I” is added to a and the filtering circuit 320
Output to 0b. In the filtering circuits 3200a and 3200b, a signal indicating that the input character codes "H" and "I" are not included in the search term is added and sent to the collecting means 3300. The collecting means 3300 collectively performs filtering processing based on these pieces of information. That is, "H" and "I" are deleted and not sent to the character string collating means 102. Next, even when the next two characters "GH" are input, they are not sent to the character string collating means 102 in the same manner. Further, when the following two characters "-S" are input to the distribution means 3100, "-" is input to the filtering circuit 3200a and "S" is input to the filtering circuit 3200.
It is output to 200b. Filtering circuit 3200a
Then, a signal that the input character code "-" is not the character code included in the search term is added to collect means 3
300, and in the filtering circuit 3200b,
A signal that "S" is a character code included in the search term is added and sent to the collecting means 3300. The collecting means 3300 collectively performs filtering processing based on these pieces of information, and sends "S" to the character string collating means 102.

Even when the following two characters "PE" are input, they are not sent to the character string collating means because they are not the character codes included in the search term.

Hereinafter, similarly, "ED", "S", "C"
S "," I "," BU "," S "," CO "," N "
When "T", "RO", "LL", and "ER" are input, the characters "S", "S", "BU", and "S" included in the search terms are output.

That is, as a result of such filtering processing, the 30-character text "HIGH-SPEED SCSI
"SSSB" composed only of the characters included in the search term by 15 times of filtering operations for "BUS CONTROLLER"
The character string "US" is sent to the character string collating means.

Thus, the 30-character text "HIGH-SPE
The ED SCSI BUS CONTROLLER "is filtered by the parallel filtering means into the 6-character text" SSSBUS "in 15 filtering operations, that is, the time required for the filtering process is 50 nanoseconds / time × 15 times = 750 nanoseconds. Becomes

Further, as described above, in order to collate 6 characters of "SSSBUS", the character string collating process is required 8 times. Therefore, the processing time of the character string collating means is 100 nanoseconds / time × 8 times. = 800 nanoseconds is required.

That is, the 30-character text "HIGH-SPE
ED SCSI BUS CONTROLLER ”can be verified in 800 nanoseconds.

As described above, according to the present invention, the filtering process, which requires 1,500 nanoseconds in the conventional method, can be performed by the present invention.
It can be processed in 50 nanoseconds. This allows
In the conventional method, the search time as a character string search device is 1,5
It was possible to reduce what was required from 00 nanoseconds to 800 nanoseconds. That is, it becomes possible to avoid the filtering process from becoming a bottom bottleneck and equivalently improve the search throughput of the character string search device by a factor of about two.

[0065]

EXAMPLE A first example using the principle of the present invention will be described below with reference to FIG. This figure is a block diagram showing the configuration of the present embodiment. In the present embodiment, the parallel filtering means 3000 is provided between the character string storage means 105 and the character string collating means 102 of the character string search device 1, and the character string storage means 10 is provided.
In order to improve the throughput of the filtering process and equivalently improve the matching throughput, a plurality of filtering circuits that cut out the character codes that are not included in the search term from the text read from 5 are operated in parallel. To do.

FIG. 1 shows an example of the configuration of the parallel filtering means 3000 used in this embodiment.

The parallel filtering means 3000 includes a distributing means 3100, filtering means 3200a and 3200.
200b and collecting means 3300.

First, the parallel filtering means 3000
The outline of the operation will be described.

The distribution means 3100 is the character string storage means 10
Two filtering means 3200a and 3200, each taking 2 bytes of text from 5 at a time
Send to b.

Filtering means 32 in this embodiment
00a and 3200b, the distribution means 3100 to 1
Each time a character is input, this is compared with the search term, it is determined whether there is the same character as the character included in the search term, and the result is output as a match flag together with the character code to the collecting means 3300. That is, when the input character code is a character included in the search term, 1 is output as a match flag for the one character and 0 is output for the other character codes.

The collecting means 3300 collectively performs output control processing for two characters based on the character code and the match flag output from the filtering means 3200a and 3200b.

The above is the parallel filtering means 3000.
Is a general operation of.

Next, the configurations and operations of the distributing means 3100, the filtering means 3200a and 3200b, and the collecting means 3300 will be specifically described.

FIG. 12 shows the structure of the distributing means 3100 in this embodiment. Distributing means 3100 in this embodiment
Is composed of a character code fetch register 3110 and input character code registers 3120a and 3120b.

The distributing means 3100 is the character string storing means 10
The text from 5 is fetched 2 bytes at a time and stored in the character code fetch register 3110. Then, in the next step, of the 2 bytes stored in the character code fetch register 3110, the upper 1 byte is stored in the input character code register 3120a and the lower 1 byte is stored in the input character code register 3120b. It outputs to the filtering means 3200a and 3200b via 3010b. And
The following 2 bytes are newly added to the character code acquisition register 31
Take in 10.

FIG. 13 shows the text "HIGH-SPEED" as an example.
Distribution means 3 when SCSI BUS CONTROLLER ”is input
100 operation is shown.

First, at the first input, the first 2 bytes "H"
I "is stored in the character code fetch register 3110. Next, one of the two bytes" HI "stored in the character code fetch register 3110 immediately before the second input.
"H" at the byte is the input character code register 3120a
The second byte "I" is the input character code register 31
20b and output to the filtering means 3200a and 3200b via the data lines 3010a and 3010b, respectively. Then, in the second input, the following 2 bytes “GH” are newly fetched in the character code fetch register 3110. Hereinafter, similarly, the distribution means 3
In 100, the text is taken in every 2 bytes and the filtering means 3200a and 3200 are taken in order of every 1 byte.
Output to b. The above is the configuration and operation of the distribution unit 3100.

Next, the configuration and operation of the filtering means 3200a and 3200b will be described.

Filtering means 3200a and 32
00b is the same circuit, and the configuration and operation of the filtering means 3200a will be described here as a representative.

FIG. 14 shows the structure of the filtering means 3200a. That is, the filtering means 3200a
Is a register 3210a and a single filtering table 32
20a, a match flag register 3230a and a character code register 3240a.

First, as an initial setting, 1 is set as a match flag in the slot of the single filtering table 3220a corresponding to the character code included in the search term, and 0 is set in the other slots. For example, when “BUS” is given as the search term, the contents shown in FIG. 8 are set in the single filtering table 3220a. That is, "B" that composes "BUS",
1 is set as a match flag in the slots corresponding to the character codes of "U" and "S".

The filtering operation is performed by the distribution means 3100.
It starts by fetching the character code into the register 3210a byte by byte.

The character code 3211a which is the output of the register 3210a is as it is, the character code register 3240a.
Is stored in the collecting means 33 via the data line 3020a.
00 is output. Also, the single filtering table 32
20a is accessed using the character code 3211a as a reference address, and the match flag is stored in the match flag register 3230a. That is, if the character code 3020a is a character code forming a search term, 1 is displayed. If the character code 3020a is any other character code, 0 is displayed.
It is read from a and output to the collection means 3300 via the data line 3030a.

In FIG. 15, "BUS" is given as the search term and "HIGH-SPEED SCSI BUS CONT" is given as the text.
The specific operation of the filtering means when "ROLLER" is input will be described.

First, in the first input, the distribution means 310
“H” is fetched from 0 to the filtering means 3200a and stored in the register 3210a. Further, “I” is fetched by the filtering means 3200b and stored in the register 3210b. Immediately before the second input, the filtering unit 3200a registers 32
“H” is output from 10a as the character code 3211a. Therefore, the single filtering table 3220a is accessed at "H". However, since "H" is a character code that is not included in the search term, 0 is stored in the match flag register 3230a as the value of the match flag, and is output to the collection means 3300 via the data line 3030a. Further, as the character code, “H” is stored in the character code register 3240a, and the data line 3020
It is output to the collection means 3300 via a.

Similarly, in the filtering means 3200b, the character code 321 from the register 3210b is "I".
It is output as 1b. Therefore, the single filtering table 3220b is accessed by "I". However, since "I" is also a character code not included in the search term, 0 is stored in the match flag register 3230a and output to the collecting means 3300 via the data line 3030b. As the character code, "I" is stored in the character code register 3240b and output to the collecting means 3300 via the data line 3020b.

At the second input, the distribution means 31
"G" is fetched from 00 to the filtering means 3200a and stored in the register 3210a. Further, “H” is fetched by the filtering means 3200b and stored in the register 3210b.

Hereinafter, similarly, the filtering means 320
At 0a and 3200b, the character code is fetched from the distribution means 3100 byte by byte and output to the collection means 3300 as character codes 3020a and 3020b. In addition, these are compared with the search term to determine whether or not there is the same character as the character included in the search term. If there is the same character, 1 is set as the match flags 3030a and 3030b for that one character, and other 0 is output to the collecting means 3300 for the character code of.

The above is the configuration and operation of the filtering means 3200a and 3200b.

FIG. 16 shows the collecting means 330 in this embodiment.
The structure of 0 is shown. That is, the collecting means 3300 in the present embodiment uses the flag registers 3310a and 3310.
b, character code registers 3320a and 3320b,
OR circuit 3330, flag buffers 3340a and 3
340b, character code buffers 3350a and 335
0b, a filtering control circuit 3360 and a character code selector 3370.

The match flags 3030a and 3030b output from the filtering means are stored in flag registers 3310a and 3310b, respectively. The character codes 3020a and 3020b are temporarily stored in the character code registers 3320a and 3320b. When at least one of the two fetched characters passes the filtering, that is, when the logical sum of the two match flags is 1, the flag register 3310
The match flags stored in a and 3310b are stored in flag buffers 3340a and 3340b, respectively,
The character codes stored in the output character code registers 3320a and 3320b are stored in the character code buffer 3350a.
And 3350b respectively. Captured 2
When both characters are deleted by filtering, that is, when the logical sum of the two match flags is 0, the match flags and the character codes corresponding to the two characters are set to the flag buffers 3340a and 3340b and the character code buffers 3350a and 3350b. The process of not importing into.

The filtering control circuit 3360 receives the values of the match flags 3303a and 3303b read from the flag buffers 3340a and 3340b as input, and the read enable (RE) signals 3301 of the flag buffers 3340a and 3340b and the character code buffers 3350a and 3350b, and The select signal 3302 of the character code selector 3370 is output.
That is, flag buffers 3340a and 3340b
And 1 is output as a read enable signal (RE) 3301 at the timing of reading the match flag and the character code from the character code buffers 3350a and 3350b.
In addition, the selection signal 3302 of the character code selector 3370
As 0, when selecting the character code 3304a (X port side) output from the character code buffer 3350a, and when selecting the character code (Y port side) 3304a output from the character code buffer 3350b. 1 is output.

The character code selector 3370 selects the character code 207 to be output from filtering according to the value of the select signal 3302 output from the filtering control circuit 3360. That is, the select signal 33
When the value of 02 is 0, the character code buffer 3350a
The character code 3304a (X port side) output from is selected, and when the value of the select signal 3302 is 1, the character code 330 output from the character code buffer 3350b.
4b (Y port side) is selected.

FIG. 17 shows the filtering control circuit 3360.
Shows the configuration of. In this embodiment, the filtering control circuit 3360 is composed of a filtering control decoder 3361 and a two-character passing flag register 3362.

FIG. 18 shows the filtering control circuit 3360.
The general operation of is shown. Filtering control decoder 336
1 receives the two-character passage flag 3363 of the previous step output from the two-character passage flag register 3362, the output 3303a from the flag buffer 3340a, and the output 3303b from the flag buffer 3340b as input, and the two-character passage flag 3364 of this step, The character code selector 3370 generates and outputs a select signal 3302 and a read enable (RE) signal 3301. The 2-character passage flag register 3362 stores 0 as an initial value.

First, when the output 3303a from the flag buffer 3340a is 0 and the output 3303b from the flag buffer 3340b is 1, only the character code 3304b output from the character code buffer 3350b is output to the character string collating means 102. . That is, 1 is set as the select signal 3302 of the character code selector 3370.
Is output as a read enable (RE) signal.

When the output 3303a from the flag buffer 3340a is 1 and the output 3303b from the flag buffer 3340b is 0, only the character code 3304a output from the character code buffer 3350a is output to the character string collating means 102. That is, 0 is set as the select signal 3302 of the character code selector 3370.
Is output as a read enable (RE) signal.

Further, the output 3303a from the flag buffer 3340a is 1, and the flag buffer 3340b is
When the output 3303b from is also 1, the character code 3304a first output from the character code buffer 3350a
Is output to the character string collating means 102. Then, in the next step, the character code 3304b output from the character code buffer 3350b is output to the character string collating means 102. That is, 0 is set as the select signal 3302 of the character code selector 3370 and read enable (RE) is performed.
In addition to outputting 0 as a signal, 1 is output as the step 2 character passage flag 3364. In addition, this step 2
The character passage flag 3364 is the 2 character passage flag register 3
Once stored in 362, 1 is output as the previous step 2 character passage flag 3363 in the next step. That is, when the value of the previous step 2 character passage flag 3363 is 1, the select signal 33 of the character code selector 3370 is displayed.
1 as 02 and 1 as read enable (RE) signal
Is output. In addition, this step 2 character passage flag 336
0 is output as 4.

Note that the output 3303a from the flag buffer 3340a and the output 3303b from the flag buffer 3340b cannot both be 0. That is, this is because the write conditions for the flag buffers 3340a and 3340b and the character code buffers 3350a and 3350b are inconsistent with (the logical sum of the two match flags output from the flag buffers 3340a and 3340b is 1).

Next, the specific operation of the collecting means 3300 will be described by taking as an example the case where "BUS" is given as the search term and "HIGH-SPEED SCSI BUS CONTROLLER" is inputted as the input text.

FIG. 19 shows flag registers 3310a and 3310b and character code registers 3320a and 3320.
320b, OR circuit 3330, flag buffer 3340
a and 3340b and character code buffers 3350a and 3350b are shown as operation examples.

First, in the first input, "H" and "I" are fetched as character codes from the filtering means 3200a and 3200b, and are stored in the character code registers 3320a and 3320b, respectively. Further, since these two characters are character codes that are not included in the search term, both 0 are stored in the flag registers 3310a and 3310b as match flags.

Immediately before the second input, 0 is output from the flag registers 3310a and 3310b.
At this time, the output from the OR circuit 3330 becomes 0, that is, since the two characters stored in the character code registers 3320a and 3320b are not included in the search term, they match the character codes corresponding to these two characters. Flags to character code buffers 3350a and 3350
b and flag buffers 3340a and 3340b.

At the time of the second input, "G" and "H" are stored in the character code registers 3320a and 3320b. In addition, a flag register 331 is used as a match flag.
0 is stored in both 0a and 3310b. However, since the output from the OR circuit 3330 is 0 at this time as well,
The character codes and match flags corresponding to these two characters are not taken into the character code buffers 3350a and 3350b and the flag buffers 3340a and 3340b.

At the time of the third input, "-" and "S" are stored in the character code registers 3320a and 3320b. At this time, "-" is a character code that is not included in the search term "BUS", but "S" is a character code that is included. Therefore, the flag register 331 is used as a match flag.
0 is stored in 0a and 1 is stored in 3310b. That is, since the output from the OR circuit 3330 at this time is 1, the character codes and the match flags corresponding to these two characters are fetched into the character code buffers 3350a and 3350b and the flag buffers 3340a and 3340b, respectively.

Similarly, the filtering means 3200 will be described below.
The 2-byte character code and the coincidence flag output from are temporarily stored in character code registers 3320a and 3320b and flag registers 3310a and 3310b.
When there is at least one character code included in the search term among these character codes, that is, when the output of the OR circuit 3330 is 1, the matching flag and the character code corresponding thereto are stored in the character code buffer 335.
0a and 3350b and flag buffers 3340a and 3340
Take it into 340b.

FIG. 20 shows the filtering control circuit 3360.
The specific operation of the character code selector 3370 will be described.

On the first input, the flag buffer 334
0a and 3340b and character code buffer 3350a
And 3350b are no longer empty, that is, immediately after "-" and "S" are written in the character code buffer, "-" and "S" are flagged as character codes from the character code buffers 3350a and 3350b. Buffers 3340a and 3340b
From 0, 0 is read as the match flag 3303a and 1 is read as 3303b. Also, 0 is read from the two-character passage flag register 3362 as the previous-step two-character passage flag 3363.

In response to the above input, the filtering control decoder 3361 has the present step 2 character passage flag 336.
4 is output as 0, 1 is output as the select signal 3302 of the character code selector 3370, and 1 is output as the read enable (RE) signal 3301 of the flag buffers 3340a and 3340b and the character code buffers 3350a and 3350b (see FIG. 18). Further, 0 is stored in the two-character passage flag register 3362 as the two-character passage flag 3364 in this step.

The character code selector 3370 selects the character code to be output from the fill ring according to the value of the select signal 3302. That is, the select signal 33
Since the value of 02 is 1, the character code buffer 3350
"S" (Y port side) which is the output from b is selected and output to the character string collating means 102. In addition, the two-character passage flag register 3362 stores this step two-character passage flag 3
0 is stored as 364.

On the second input, the flag buffer 334
0a and 3340b and character code buffer 3350a
And read enable (RE) signal 33 of 3350b
Since 01 is 1, the value corresponding to the next two characters from each buffer, that is, "" and "S" from the character code buffers 3350a and 3350b, and the match flag 3303a from the flag buffers 3340a and 3340b, respectively.
0, and 3303b is read as 1. Similarly to the previous case, the filtering control decoder 3361 sets 0 as the step 2 character passage flag 3364 to the above input and selects the selection signal 33 of the character code selector 3370.
1 as 02 is set to the flag buffers 3340a and 33
40b and character code buffers 3350a and 3350
1 is output as the read enable (RE) signal 3301 of b (see FIG. 18). That is, since the value of the select signal 3302 is 1 in the character code selector 3370, “S” (Y port side) which is the output from the character code buffer 3350b is selected. Further, 0 is stored in the two-character passage flag register 3362 as the two-character passage flag 3364 in this step.

Further, in the third input, the flag buffers 3340a and 3340b and the character code buffer 3 are input.
Read enable (RE) for 350a and 3350b
Since the signal 3301 is 1, the value corresponding to the next two characters from each buffer, that is, the character code buffer 3350.
a and 3350b to "C" and "S", and flag buffers 3340a and 3340b to match flag 3
0 is read as 303a and 1 is read as 3303b. The filtering control decoder 3361, in the same manner as described above, responds to the above input with the step 2 character passage flag 336.
4, 0, 1 as the select signal 3302 of the character code selector 3370, flag buffers 3340a and 3340b, and character code buffer 3350a and
1 is output as the read enable (RE) signal 3301 of the 3350b (see FIG. 18). That is, since the value of the select signal 3302 is 1 in the character code selector 3370, “S” (Y port side) which is the output from the character code buffer 3350b is selected. Further, 0 is stored in the two-character passage flag register 3362 as the two-character passage flag 3364 in this step.

Further, in the fourth input, the flag buffers 3340a and 3340b and the character code buffer 3 are input.
Read enable (RE) for 350a and 3350b
Since the signal 3301 is 1, the value corresponding to the next two characters from each buffer, that is, the character code buffer 3350.
a and 3350b to "B" and "U", and flag buffers 3340a and 3340b to match flag 3
1 is read out as 303a and 1 as 3303b. The filtering control decoder 3361 sets this step 2 character passage flag 3364 to 1 for the above input.
To the select signal 330 of the character code selector 3370.
2 as 0, flag buffers 3340a and 3340b
And 0 is output as the read enable (RE) signal 3301 of the character code buffers 3350a and 3350b (see FIG. 18). That is, the character code selector 33
At 70, since the value of the select signal 3302 is 0,
"B" which is the output from the character code buffer 3350a
Select (X port side). In addition, the 2-character passage flag register 3362 stores the 2-character passage flag 336 in this step.
1 is stored as 4.

At the fifth input, the flag buffer 334
0a and 3340b and character code buffer 3350a
And read enable (RE) signal 33 of 3350b
Since 01 is 0 and the value corresponding to the next two characters is not read from each buffer, the character code buffer 3350a
And 3350b from "B" and "U" from flag buffers 3340a and 3340b.
The state in which 1 is output as 03a and 1 as 3303b is maintained. Also, the previous step 2 character passage flag 33
In 64, 1 is output from the two-character passage flag 3362. The filtering control decoder 3361 sets 0 as the step 2 character passage flag 3364 for these inputs, and outputs the select signal of the character code selector 3370.
1 as 3302, flag buffers 3340a and 33
40b and character code buffers 3350a and 3350
1 is output as the read enable (RE) signal 3301 of b (see FIG. 18). That is, since the value of the select signal 3302 is 1 in the character code selector 3370, “U” (Y port side) which is the output from the character code buffer 3350b is selected. Further, 0 is stored in the two-character passage flag register 3362 as the two-character passage flag 3364 in this step.

Finally, in the sixth input, the flag buffers 3340a and 3340b and the character code buffer 3 are
Read enable (RE) for 350a and 3350b
Since the signal 3301 is 1, the value corresponding to the next two characters from each buffer, that is, the character code buffer 335.
"S" and "" are read from 0a and 3350b, and 1 is read as the match flag 3303a and 0 as 3303b from the flag buffers 3340a and 3340b.
The filtering control decoder 3361 sets the step 2 character passage flag 3364 to 0 for the above input.
To the select signal 330 of the character code selector 3370.
2 as 0, flag buffers 3340a and 3340b
And 1 is output as the read enable (RE) signal 3301 of the character code buffers 3350a and 3350b (see FIG. 18). That is, the character code selector 33
At 70, since the value of the select signal 3302 is 0,
"S" which is the output from the character code buffer 3350a
Select (Y port side).

The above is the configuration and operation of the parallel filtering means 3000 in this embodiment.

As described above, according to this embodiment, by operating two filtering circuits in parallel, the input text of 30 characters "HIGH-SPEED SCSI BUS CONTROLLER" is included in the search term in 15 processing cycles. "SSSB
It is possible to filter to only 6 characters of "US". That is, the processing cycle of filtering can be reduced to 15/30, that is, 1/2 as compared with the case of not parallelizing. The collation throughput can be doubled.

In this embodiment, among the filtering circuits described in Japanese Patent Application No. 4-93067 “Character string search device equipped with filtering circuit”, a delimiter (delimiter) indicating the appearance of discontinuous characters is used. ) Has been described as an example in which the single filtering circuits are operated in parallel. However, it is obvious that other filtering circuits described in the above-mentioned prior invention, for example, a single filtering circuit when a delimiter is inserted or a single head filtering circuit can be operated in parallel, can be similarly realized. is there.

Further, although the case where two filtering means are operated in parallel has been described in the present embodiment, the same method as in the above embodiment is applied to the case where the number of filtering means operated in parallel is three or more. It is also clear that can be realized with.

Further, although the case where the input text is represented by 1-byte characters has been described in the present embodiment, it is clear that the case where the input text is represented by 2-byte characters can be realized in the same manner as the above-mentioned embodiment. .

[0121]

As described above, according to the present invention, by operating two filtering circuits in parallel, the input text "HIGH-SPEED SCSI BUS CONTRO" of, for example, 30 characters can be obtained.
It is possible to filter "LLER" to only 6 characters of "SSSBUS" included in the search term in 15 processing cycles, that is, the processing cycle of filtering is 15/30, that is, compared to the case without parallelization. 1/2
Therefore, the matching throughput of the character string search device can be doubled.

[0122]

[Brief description of drawings]

FIG. 1 is an explanatory diagram of a character string search device of the present invention.

FIG. 2 is an explanatory diagram of a conventional character string search device.

FIG. 3 is a state transition diagram of a conventional automaton.

FIG. 4 is an explanatory diagram of a conventional token control method.

FIG. 5 is a block diagram showing a configuration of a conventional character string matching circuit.

FIG. 6 is an explanatory diagram of a conventional character string search device.

FIG. 7 is a block diagram showing the configuration of a conventional filtering means.

FIG. 8 is an explanatory diagram of a conventional filtering table.

FIG. 9 is a timing chart of conventional filtering means.

FIG. 10 is a state transition diagram of a conventional automaton and an explanatory diagram of a conventional token control method.

FIG. 11 is an explanatory diagram of a character string search device of the present invention.

FIG. 12 is a block diagram showing a configuration of a distribution unit of the present invention.

FIG. 13 is an explanatory diagram of an operation example of the distribution unit of the present invention.

FIG. 14 is a block diagram showing the configuration of the filtering means of the present invention.

FIG. 15 is an explanatory diagram of an operation example of the filtering means of the present invention.

FIG. 16 is a block diagram showing the configuration of a collection unit of the present invention.

FIG. 17 is a block diagram showing the configuration of a filtering control circuit according to the present invention.

FIG. 18 is an explanatory diagram of input / output relationships of the filtering control decoder of the present invention.

FIG. 19 is an explanatory diagram of an operation example of a character code register, a flag register, a character code buffer, and a flag buffer of the present invention.

FIG. 20 is an explanatory diagram of an operation example of the filtering control circuit and the character code selector of the present invention.

[Explanation of symbols]

102 ... Character string collating means, 105 ... Character string storing means, 3
000 ... Filtering means, 3200 ... Character code register, 3300 ... Single filtering table.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Masatsugu Shinozaki 4-6, Surugadai Kanda, Chiyoda-ku, Tokyo Inside Hitachi, Ltd.

Claims (6)

[Claims] 1. A character string search device for collectively determining whether or not a plurality of designated search terms exist in a text composed of code-represented characters, and a character string for storing the text. The storage means and the text read from the character string storage means are collectively fetched in a plurality of characters, it is judged in parallel whether or not the character code is included in the search term, and only when it is included, it is output. A character string search device comprising: parallel filtering means; and character string matching means for collectively checking whether or not the search term is present in the character code string output from the parallel filtering means. 2. A distribution means for fetching a plurality of characters from a text read from the character string storage means at a time and dividing and sending one character at a time, as the parallel filtering means, according to claim 1. A plurality of filtering means arranged in parallel for determining whether the output character code is a character code included in the search term, and the character codes output from the plurality of filtering means are fetched and aligned. A character string search device including a collecting means for outputting the information. 3. The filtering means according to claim 2, wherein the filtering means fetches the characters one by one from the text read out from the character string storing means, and if the character code is included in the designated search term, this is used. Is output to the above character string collating means and is not output if it is not included, and a specific delimiter code is output only when the character code output immediately before is included in the above search term. A character string search device having filtering means for outputting to. 4. The filtering means according to claim 2, wherein the slot corresponding to a character code of a character included in a search term designated in advance is set to 1 and the slots corresponding to other character codes are set to 0. The flag storage means and the flag storage means are referred to in correspondence with the character code input from the character string storage means, and when the read flag is 1, the character code is output to the character string collating means. However, the character string search device constituted by the output selection means for selecting not to output when 0. 5. The filtering means as claimed in claim 2, wherein each character is fetched from the text read from the character string storage means one by one, and when the character code is the first character code of the search term, the character string collating means. In addition to outputting, it does not output except for the first character code,
Once the first character code is output to the character string collating means, the input character code is output to the character string collating means only when it is included in the search term, and is not output when it is not included. A character string search device having. 6. The filtering means according to claim 2, wherein the character code is fetched from the text read from the character string storage means one by one, and when the character code is the first character code of the search term, the character string collation is performed. The above character string collation is performed only when the input character code is included in the above search term after outputting the above first character code to the above character string collating means while outputting to the means The filtering means that outputs the specified delimiter code to the character string collating means only when the character code output immediately before is output to the character string collating means while not outputting the character code to the character string collating means. A character string search device having.