CN114756553A - Columnar database and method for retrieving sequence of events - Google Patents

Abstract

The present invention provides a column database and a method for retrieving event sequences for storing and retrieving event sequences. The columnar database processes the data as follows: step a: input the data into the columnar database; step b: set an index on the time field of the data; step c: determine the data type of the event field, if it is text, and the number of characters is greater than 8 Step d: deduplicate the event field; step e: perform hash coding according to the deduplication result, and construct a dictionary of event data serial numbers; step f: traverse the columnar database, The data in all event fields whose data type is text is assigned a corresponding hash code value to generate an event index column; step g: sort the column database in ascending time order according to the index. The invention solves the problems that the existing column database is inconvenient to sort and retrieve the input data, and the retrieval is complicated when the data type of the event field is text.

Description

Columnar database and method for retrieving event sequences

technical field

The invention relates to a column database and a method for retrieving event sequences, belonging to the field of data storage.

Background technique

In the database, a large amount of data is stored, including a time, a subject and an event that occurred. When the amount of data is small, events in the database can be evaluated to detect problems or trends in advance. In today's increasingly large amount of data, it is impossible to intuitively see the occurrence and development trend through the retrieval of a single event, and it is impossible to evaluate the impact on the subject.

Event sequence refers to the sequence of multiple events that occur in sequence. The event sequence can accurately evaluate the occurrence and development trend, but the existing columnar database can only retrieve a single event, and cannot retrieve the event sequence. It is even more impossible to perform fuzzy retrieval of event sequences. At the same time, the existing retrieval method can only directly retrieve the text content when the data type of the event is text, and the retrieval is complicated and inconvenient.

In view of this, it is indeed necessary to provide a columnar database and a method for retrieving event sequences to solve the above problems.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a columnar database and a method for retrieving event sequences, which solves the problem that the existing columnar database is inconvenient for sorting and retrieving input data, and the retrieval is complex and efficient when the data type of the event field is text. lower problem.

To achieve the above purpose, the present invention provides a columnar database for storing and retrieving event sequences, and the columnar database can process data as follows:

Step a: input data into a columnar database, the data is time series data, including at least a time field and an event field;

Step b: set an index to the time field of the data entered in step a;

Step c: judge the data type of the event field in the data, if the data type of the event field is text, and there is data with the number of characters greater than 8, then enter step d, otherwise enter step g;

Step d: de-duplicate the content of the event field;

Step e: perform hash coding according to the deduplication result of step d, and construct a dictionary of event data serial numbers;

Step f: traverse the column database, assign the corresponding hash code value to the data whose data type is text in all event fields according to the event data serial number dictionary, and generate the corresponding event index column in the column database;

Step g: Sort the columnar database in ascending time order according to the index set on the time field in step b.

As a further improvement of the present invention, the data in the columnar database further includes any attribute field except the event field and the time field.

The present invention also provides a method for retrieving event sequences, which is applied to the aforementioned columnar database, and specifically includes the following steps:

Step 1: Input the event sequence to be retrieved and convert it into a query statement;

Step 2: Judging the data type of the event sequence to be retrieved, when the data type of the event sequence to be retrieved is a numerical value or a single character or text with a number of characters less than or equal to 8, go to step 4; when the data type of the event sequence to be retrieved is text, and there is data with more than 8 characters, go to step 3;

Step 3: perform hash coding on the sequence of events to be retrieved, and enter step 4 after the data type of the sequence of events to be retrieved is converted into a numerical value;

Step 4: Select the query type. If it is an exact query, use the sequence of events to be retrieved as a whole to search for exact matching; if it is a fuzzy query, perform a fuzzy matching retrieval based on the preset fuzzy query distance. As an array as a whole, calculate the minimum distance from the data composition sequence in the event sequence, and the minimum distance is less than or equal to the preset fuzzy query distance;

Step 5: Build a result dictionary, add all the matched corresponding sequences to the result dictionary, until all data in the event field is traversed;

Step 6: Output the result dictionary as the result.

As a further improvement of the present invention, the to-be-retrieved event sequence is a sequence composed of multiple events arranged in sequence.

As a further improvement of the present invention, the query statement in step 1 is an SQL-like statement.

As a further improvement of the present invention, the query statement includes a first parameter, a second parameter and a third parameter set in sequence, the first parameter is configured as a field for filtering the query, and the second parameter is configured as a storage limit for the query address, and the third parameter is configured to limit retrieval conditions.

As a further improvement of the present invention, the third parameter includes the sequence of events to be retrieved.

As a further improvement of the present invention, in the step 4, the fuzzy query is performed to calculate the minimum distance by using the Levenshtein algorithm. When the calculated data type is a string, any string in the to-be-retrieved event sequence is added, deleted, The minimum number of steps required to convert the way characters are replaced into any string in the event field.

As a further improvement of the present invention, the calculation of the minimum steps is specifically:

Step 41: Calculate the length n of any character string strA in the sequence of events to be retrieved, and calculate the length m of any character string strB in the sequence of events to be retrieved;

Step 42: If n=0, the minimum edit distance is m, if m=0, the minimum edit distance is n, if not, then go to step 43;

Step 43: Construct a matrix Arr of (m+1)*(n+1), and initialize the first row and first column of the matrix to be 0～n, 0～m respectively;

Step 44: Double loop, traverse strA, and traverse strB on this basis, if strA[i]=strB[j], then cost=0, otherwise cost=1, judge Arr[j-1][i]+1 , Arr[j][i-1]+1, the minimum value of Arr[j-1][i-1]+cost, and assign the minimum value to Arr[j][i];

Step 45: After the loop is over, the last element of the matrix is the minimum edit distance, that is, the minimum steps.

As a further improvement of the present invention, the result dictionary output in step 6 includes all matching sequences in the database that meet the conditions, and the matching sequences output at least the time field and the event field.

The beneficial effects of the present invention are: compared with the prior art, the columnar database of the present invention solves the problem that the existing columnar database is inconvenient for sorting and retrieving the input data by setting an index on the time field. Hash coding is performed on the event field whose type is text, which solves the problem of complex retrieval and low efficiency in the existing columnar database when the data type of the event field is text.

Description of drawings

FIG. 1 is a step diagram of processing data in the columnar database of the present invention.

FIG. 2 is a flowchart of FIG. 1 .

FIG. 3 is a schematic diagram of the time series data of step 1 in FIG. 1 .

FIG. 4 is a step diagram of the event sequence retrieval method of the present invention.

FIG. 5 is a flowchart of the Levenshtein algorithm of step 4 in FIG. 4 .

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

Here, it should be noted that, in order to avoid obscuring the present invention due to unnecessary details, only the structures and/or processing steps closely related to the solution of the present invention are shown in the drawings, and the drawings related to the present invention are omitted. Other details that don't matter much.

In addition, it should be noted that the terms "comprising", "comprising" or any other variation thereof are intended to encompass a non-exclusive inclusion such that a process, method, article or device comprising a series of elements includes not only those elements, but also Also included are other elements not expressly listed or inherent to such a process, method, article or apparatus.

Referring to Figures 1 and 2, the present invention provides a columnar database for storing and retrieving event sequences. The columnar database processes data as follows:

Step a: Input the data into the columnar database, the data is time series data, including at least the time field and the event field;

Step b: set an index to the time field of the data entered in step a;

Step c: Judging the data type of the event field in the data, if the data type of the event field is text, and there is data with more than 8 characters, then go to step d, otherwise go to step g;

Step d: De-duplicate the content of the event field;

Step e: perform hash coding according to the deduplication result of step d, and construct a dictionary of event data serial numbers;

Step f: traverse the columnar database, assign the corresponding hash code value to the data whose data type is text in all event fields according to the event data serial number dictionary, and generate the corresponding event index column in the columnar database;

Step g: Sort the columnar database in ascending time order.

Please refer to Figure 3. When data is input in step a, the order of the input data may be out of order, and the data type of the event field is the case of long text, for example, as shown in Table 1:

Table 1

time event 2022-01-04 engine maintenance 2022-01-01 come on 2022-01-03 Refill glass water 2022-01-02 Wiper Maintenance 2022-01-06 CAR WASH 2022-02-04 come on 2022-02-06 Refill glass water 2022-02-05 come on

In Table 1, several pieces of data are entered into the columnar database, and these data are randomly ordered, so certain processing is required. In the exemplary data in Table 1, the data type of the event field is text, and the number of characters in the data "engine maintenance" in the event field is 2*5 or 10, which is greater than 8. Therefore, after judging according to step c, it is necessary to Perform steps d to g in sequence, first press step d to de-duplicate the content of the event field, and obtain the de-duplication set as [engine maintenance, refueling, replenishing glass water, wiper maintenance, car wash], and then proceed to step e to remove The text content of the re-event field is hashed, and the sequence number of the constructed event data dictionary is {engine maintenance: 1, refueling: 2, glass water replenishment: 3, wiper maintenance: 4, car wash: 5}, and then according to step f, Traverse the columnar database, assign the corresponding hash code value to the data whose data type is text for all event fields according to the event data serial number dictionary, and generate the corresponding event index column in the columnar database. The form of Table 1 is transformed into the form of Table 2:

Table 2

time event event index 2022-01-04 engine maintenance 1 2022-01-01 come on 2 2022-01-03 Refill glass water 3 2022-01-02 Wiper Maintenance 4 2022-01-06 CAR WASH 5 2022-02-04 come on 2 2022-02-06 Refill glass water 3 2022-02-05 come on 2

The input column database for storage includes at least an event field and a time field, and the event sequence to be retrieved is a sequence composed of multiple events arranged in sequence. The same or different events are arranged in the order of time from early to late to be called an event sequence, which is also a further analysis method for the development of things on the basis of time series, which can be realized by retrieving and analyzing the events that occurred in sequence. The evaluation of the development trend or the influence of the main body, so the column database must include the event field and the time field, and the column database should be sorted in ascending time order at the beginning, so as to facilitate retrieval. Therefore, it is necessary to perform step g at the end to sort the columnar database in ascending time order. At this time, the data in the columnar database will be changed from the form of Table 2 to the form of Table 3:

table 3

A large amount of chaotic data originally input into the column-type database is processed and sorted by the column-type database of the present application, and data in ascending order of time is obtained, thereby facilitating retrieval of the sequence of events to be retrieved. Before that, the columnar database judged the data in the event field whose data type is text. When the number of characters is less than or equal to 8, that is, when the 64-bit operating system can be read and judged at one time, it is considered that this is the case. The data below will not affect the retrieval speed and will not be processed. However, when the data type is text and the number of characters is greater than 8, it is considered that the data of long text type is not easy to be judged and processed, and the retrieval of the original text will make the retrieval system too expensive. The operation is slow, so in the process of data input into the columnar database, the data content of the event field whose data type is text and the number of characters is greater than 8 is processed and hash encoded, so as to facilitate the retrieval of subsequent event sequences.

It should be noted that the content in the event field here is only exemplary, and not limited thereto.

In step b, setting an index on the time field of the input data can also improve the retrieval speed.

Referring to Figure 4, the present invention provides a method for retrieving event sequences, which is applied to a columnar database, and specifically includes the following steps:

Step 1: Input the event sequence to be retrieved and convert it into a query statement;

Step 2: Judging the data type of the event sequence to be retrieved, when the data type of the event sequence to be retrieved is a numerical value or a single character or text with a number of characters less than or equal to 8, go to step 4; when the data type of the event sequence to be retrieved is text, and there is data with more than 8 characters, go to step 3;

Step 3: perform hash coding on the sequence of events to be retrieved, and enter step 4 after the data type of the sequence of events to be retrieved is converted into a numerical value;

Step 4: Select the query type. If it is an exact query, use the sequence of events to be retrieved as a whole to search for exact matching; if it is a fuzzy query, perform a fuzzy matching retrieval based on the preset fuzzy query distance. As an array as a whole, calculate the minimum distance from the data composition sequence in the event sequence, and the minimum distance is less than or equal to the preset fuzzy query distance;

Step 5: Build a result dictionary, add all the matched corresponding sequences to the result dictionary, until all data in the event field is traversed;

Step 6: Output the result dictionary as the result.

In step 1, the input event sequence to be retrieved needs to be converted into a query statement that can identify applications and perform query retrieval in the column database. Statements are simpler and more convenient. The query statement includes a first parameter, a second parameter, and a third parameter that are set in sequence. The first parameter is configured as a field for filtering the query, the second parameter is configured as a storage address for limiting the query, the third parameter is configured as a limited retrieval condition, and the third parameter is configured as a limited search condition. Parameters include the sequence of events to retrieve. In the embodiment of the present application, the query statement is select...from...where..., the first parameter is the select parameter, the input parameter value is the time, and the event, that is, the field of the filter query, is the time in the database Field and event field; the second parameter is the from parameter, and the input parameter value is mydb.mydata, that is, the storage address of the database to be queried; the third parameter is the where parameter, and the input parameter value includes the event &list='****' , where **** is the sequence of events to be retrieved, which is the retrieval condition.

The result dictionary output in step 6 includes all matching sequences in the database that meet the conditions, and the matching sequences output at least the time field and the event field.

When the data type of the event sequence to be retrieved in step 3 is numeric, as shown in Table 4, there are two columns of data fields, time field and event field in the columnar database, and the data type of the event field is also numeric, and no processing is required. . Take the event sequence to be retrieved composed of the four lines of "2, 0, 2, 2" as an example, to query all event sequences of "2, 0, 2, 2" in the columnar database, use The sequence of events to be retrieved is converted into a SQL-like query statement: select time, event frommydb.mydata where event &list='2,0,2,2', where event &list='2,0,2,2' means that '2, 0, 2, 2' are retrieved as an event sequence. In this embodiment of the present application, there are two sets of matching results, which are marked with underscores in Table 4. Finally, according to the event sequence retrieval, the result of the returned result dictionary is: [ {group 1:[{time:01:01:02,event:2},{time:01:01:03,event:0},{time:01:01:04,event:2},{time :01:01:05,event:2}]},{group 2:[{time:01:01:07,event:2},{time:01:01:08,event:0},{time :01:01:09, event: 2}, {time: 01:01:10, event: 2}]}].

Table 4

time event 01:01:01 0 <u>01:01:02</u> <u>2</u> <u>01:01:03</u> <u>0</u> <u>01:01:04</u> <u>2</u> <u>01:01:05</u> <u>2</u> 01:01:06 3 <u>01:01:07</u> <u>2</u> <u>01:01:08</u> <u>0</u> <u>01:01:09</u> <u>2</u> <u>01:01:10</u> <u>2</u> 01:01:11 0

When the data type of the event sequence to be retrieved in step 3 is a single character, as shown in Table 5, there are two columns of data fields, a time field and an event field, and the data type of the event field is also a single character, and no processing is required. . Taking the sequence of events to be retrieved composed of the four lines of "CACC" as a whole as an example, to query all event sequences of "CACC" in the columnar database, convert the sequence of events to be retrieved into a SQL-like query statement: select time , event frommydb.mydata where event&list='C,A,C,C', where event&list='C,A,C,C' means to retrieve 'C,A,C,C' as an event sequence, In this embodiment of the present application, there are two sets of matching results, which are marked with underlines in Table 5. Finally, the results of the returned result dictionary are retrieved according to the sequence of events: [{Group 1: [{Time: 01:01: 02, event: C}, {time: 01:01:03, event: A}, {time: 01:01:04, event: C}, {time: 01:01:05, event: C}]} , {group 2: [{time:01:01:07,event:C},{time:01:01:08,event:A},{time:01:01:09,event:C},{ Time: 01:01:10, Event: C}]}].

table 5

time character 01:01:01 A <u>01:01:02</u> <u>C</u> <u>01:01:03</u> <u>A</u> <u>01:01:04</u> <u>C</u> <u>01:01:05</u> <u>C</u> 01:01:06 D <u>01:01:07</u> <u>C</u> <u>01:01:08</u> <u>A</u> <u>01:01:09</u> <u>C</u> <u>01:01:10</u> <u>C</u> 01:01:11 A

The columnar database also includes any attribute fields other than the event field and the time field, which are configured to limit the query scope of the event sequence to be retrieved.

When the data type of the event sequence to be retrieved in step 3 is a string, as shown in Table 6, there are three data fields of time field, device name field and event field in the columnar database, and the device name field is a columnar database Any attribute field except the event field and time field in the field is used to limit the query scope of the event sequence to be retrieved. The data in the event field is a character string of 2 characters, which does not need to be processed. Taking the event sequence composed of the four lines of "AA, AA, BB, BB" of device 2 as a whole as an example, to query all the "AA, AA, BB, BB" sequences of device 2 in the database, wait The retrieval event sequence is converted into a SQL-like query statement: select time, event frommydb.mydata where device name = 'device two' and event &list = 'AA, AA, BB, BB', in the where parameter of the third parameter for additional pairs The equipment field is limited, and the query condition is limited to the equipment two in the equipment field, and then the event &list='AA, AA, BB, BB' in the third parameter indicates that 'AA, AA, BB, BB' is used as an event In this embodiment of the present application, there is a set of matching results, which are marked with underscores in Table 6. Finally, the results returned to the dictionary according to the event sequence retrieval are: [{Group 1: [{Time: 01 :01:05, event: AA}, {time: 01:01:08, event: AA}, {time: 01:01:11, event: BB}, {time: 01:01:14, event: BB }]}].

Table 6

time device name event 01:01:01 equipment one AA 01:01:02 equipment two BB 01:01:03 equipment three CC 01:01:04 equipment one DD <u>01:01:05</u> equipment two <u>AA</u> 01:01:06 equipment three BB 01:01:07 equipment one CC <u>01:01:08</u> <u>Device Two</u> <u>AA</u> 01:01:09 equipment three CC 01:01:10 equipment one CC <u>01:01:11</u> <u>Device Two</u> <u>BB</u> 01:01:12 equipment three CC 01:01:13 equipment one AA <u>01:01:14</u> <u>Device Two</u> <u>BB</u> 01:01:15 equipment three CC

When the data type of the event sequence to be retrieved in step 3 is text, the data originally entered into the column database only has two columns of data fields, the time field and the event field. After the column database processes the data from steps a to g, According to the event data serial number dictionary obtained in step e, the data of the event field in the columnar database whose data type is text is given a corresponding hash code value. In the embodiment of this application, the text in the event field: I am Long text one, I am long text two, I am long text three, I am long text four, I am long text five, I am long text six are texts with more than 8 characters, so they are hash-coded as 1, 2, 3, 4, 5, and 6, the column database generates event index columns after coding, as shown in Table 7. At this time, if the three lines of "I am long text 1, I am long text 2, and I am long text 4" are used as an example of an event sequence as a whole, if you want to query the database to find all "I am long text 1" , I am long text two, I am long text four" event sequence, convert the event sequence to be retrieved into a SQL-like query statement: select time, event from mydb.mydatawhere event &list='I am long text one, I am long text Text two, I am long text four', where event &list='I am long text one, I am long text two, I am long text four' means that 'I am long text one, I am long text two, I am The long text 4' is retrieved as an event sequence, and according to the event data serial number dictionary obtained in step e, the event sequence to be retrieved is given the corresponding hash code value "1, 2, 4", that is, the actual event sequence to be retrieved is The matching in the columnar database uses "1, 2, 4" as the to-be-retrieved sequence to perform matching retrieval in the event index column. In this embodiment of the present application, there is a set of matching results, which are used in Table 7. Underlined, the result of the result dictionary returned according to the sequence of events to be retrieved is: [{Group 1: [{Time: 01:01:04, Event: I am a long text}, {Time: 01:01: 05, event: I am long text two}, {time: 01:01:06 event: I am long text four}]}].

Table 7

When the data type of the event sequence to be retrieved is text, if the text data is retrieved directly in the database, the entire retrieval process will be expensive, and the retrieval process cannot be efficiently retrieved. In the columnar database of the present application, steps d and f are performed when the data is stored in the columnar database, and hash coding is performed on the data whose data type is text and whose number of characters is greater than 8. When the event sequence to be retrieved is input, the retrieval is performed During the process, the data type of the to-be-retrieved event sequence is judged. When the data type of the to-be-retrieved event sequence is judged to be text, the text-type to-be-retrieved event sequence is hash-coded according to the event data serial number dictionary in the columnar database, and the encoded The subsequent event sequence to be retrieved is compared with the event index column of the encoded columnar database, and the data retrieval of the text type is changed to the data retrieval of the numerical type, thereby reducing the retrieval complexity and improving the retrieval speed.

The columnar database and the method for retrieving event sequences of the present application can also perform precise retrieval and fuzzy retrieval. If fuzzy retrieval is selected in step 4, as shown in Table 8, the columnar database has a time field, a device name field, an event Field Three-column data fields, take the event sequence composed of the four lines of "A, A, B, C" of device two as an example, the preset fuzzy query distance is 1, if you want to query the device in the column database All the "A, A, B, C" sequences of error distance 1 of the second, convert the sequence of events to be retrieved into SQL-like query statements: select time, event frommydb.mydata where device name='device two' event&list='A , A, B, C'&dist=1, where device name='device two' means that the retrieval condition is limited to the device two in the device name field, event &list='A, A, B, C'&dist=1 means Take 'A, A, B, C' as an event sequence and perform a fuzzy search with an error distance of 1. In this embodiment of the application, "A, A, B, B" is the time sequence that matches the time sequence to be retrieved" A, A, B, C" with an error distance of 1" is marked with an underline in Table 8, and finally retrieved according to the event sequence, the returned result is: [{Group 1:[{Time:01:01:05 , event: A}, {time: 01:01:08, event: A}, {time: 01:01:11, event: B}, {time: 01:01:14, event: B}]}] .

Table 8

time device name event 01:01:01 equipment one A 01:01:02 equipment two B 01:01:03 equipment three C 01:01:04 equipment one D <u>01:01:05</u> <u>Device Two</u> <u>A</u> 01:01:06 equipment three B 01:01:07 equipment one C <u>01:01:08</u> <u>Device Two</u> <u>A</u> 01:01:09 equipment three C 01:01:10 equipment one C <u>01:01:11</u> <u>Device Two</u> <u>B</u> 01:01:12 equipment three C 01:01:13 equipment one A <u>01:01:14</u> <u>Device Two</u> <u>B</u> 01:01:15 equipment three C

Take Table 3 as an example again. At this time, there are three columns of data fields in the columnar database: time field, event field and event index field. Take the event sequence composed of the two behaviors of "engine maintenance, refueling" as a whole as an example, The preset fuzzy query distance is 1. If you want to query the column database for all event sequences with an error distance of 1 from "engine maintenance, refueling", convert the to-be-retrieved event sequence into an SQL-like query statement: select time, event from mydb .mydatawhere event &list='engine maintenance, refueling'&dist=1, event &list='engine maintenance, refueling'&dist=1 means that 'engine maintenance, refueling' is taken as an event sequence and a fuzzy search with an error distance of 1 is performed. In this embodiment of the present application, "engine maintenance, car wash, refueling" means that the corresponding event index is "1, 5, 2", which is consistent with the to-be-retrieved event sequence "engine maintenance, refueling", that is, the corresponding retrieval content is "1, 2" ”, the search results with an error distance of 1 are marked with underlines in Table 9. Finally, according to the event sequence retrieval, the returned results are: [{Group 1: [{Time: 2022-01-04, Event: Engine Maintenance}, {time: 2022-01-06, event: car wash}, {time: 2022-02-04, event: refueling}]}].

Table 9

time event event index 2022-01-01 come on 2 2022-01-02 Wiper Maintenance 4 2022-01-03 Refill glass water 3 <u>2022-01-04</u> <u>Engine Maintenance</u> <u>1</u> 2022-01-06 CAR WASH <u>5</u> <u>2022-02-04</u> <u>Come on</u> <u>2</u> 2022-02-05 come on 2 2022-02-06 Refill glass water 3

Please refer to Fig. 5. In step 4, the minimum distance calculated by the fuzzy query is to use the Levenshtein algorithm. When the calculated data type is a character string, any character string in the event sequence to be retrieved is added, deleted, or replaced by characters. The minimum steps required to convert into any string in the event field. The minimum steps are the minimum distances. The calculation of the minimum steps is as follows:

Step 41: Calculate the length n of any character string strA in the sequence of events to be retrieved, and calculate the length m of any character string strB in the sequence of events to be retrieved;

Step 42: If n=0, the minimum edit distance is m, if m=0, then the minimum edit distance is n, if not, go to step 43;

Step 43: Construct a matrix Arr of (m+1)*(n+1), and initialize the first row and first column of the matrix to be 0～n, 0～m respectively;

Step 44: Double loop, traverse strA, and traverse strB on this basis, if strA[i]=strB[j], then cost=0, otherwise cost=1, judge Arr[j-1][i]+1 , Arr[j][i-1]+1, the minimum value of Arr[j-1][i-1]+cost, and assign the minimum value to Arr[j][i];

Step 45: After the loop is over, the last element of the matrix is the minimum edit distance, that is, the minimum steps.

In the search for Table 8 in this application, "A, A, B, B" can be converted into "A, A, B, C" in at least one step, which is the sequence of events to be searched "A, A, B, C" with an error distance of 1. In the retrieval of Table 9 in this application, the event sequence to be retrieved "engine maintenance, refueling" is assigned a hash code value of "1, 2", and the event sequence to be retrieved "engine maintenance, car wash, refueling" is assigned a hash code value of "1, 2". The minimum step of "1, 5, 2" after the encoded value is 1, which is the retrieval result that meets the requirement that the preset fuzzy query distance is 1.

To sum up, the columnar database of the present invention solves the problem that the existing columnar database is inconvenient for sorting and retrieving the input data by setting an index on the time field. The event field is hashed and encoded, which solves the problem of complex retrieval and low efficiency in the existing columnar database when the data type of the event field is text. The method for retrieving event sequences of the present invention converts the to-be-retrieved event sequence into a query statement, and hash codes the text type according to the data type to perform retrieval query. The event sequence to be retrieved is calculated as an array as a whole to calculate the minimum distance from the sequence of data in the event sequence, and the minimum distance is less than or equal to the preset fuzzy query distance, so that fuzzy matching retrieval is performed, which solves the problem that the existing columnar database is inconvenient for event sequences. Perform precise and fuzzy searches and retrieve complex questions when the data type is text.

The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced. without departing from the spirit and scope of the technical solutions of the present invention.

Claims (10)

1. A columnar database for storing and retrieving event sequences, wherein the columnar database can process data as follows: Step a: input data into a columnar database, the data is time series data, including at least a time field and an event field; Step b: set an index to the time field of the data entered in step a; Step c: judge the data type of the event field in the data, if the data type of the event field is text, and there is data with the number of characters greater than 8, then enter step d, otherwise enter step g; Step d: de-duplicate the content of the event field; Step e: perform hash coding according to the deduplication result of step d, and construct a dictionary of event data serial numbers; Step f: traverse the column database, assign the corresponding hash code value to the data whose data type is text in all event fields according to the event data serial number dictionary, and generate the corresponding event index column in the column database; Step g: Sort the columnar database in ascending time order according to the index set on the time field in step b. 2 . The columnar database according to claim 1 , wherein the data in the columnar database further comprises any attribute field except the event field and the time field. 3 . 3. A method for retrieving event sequences, applied in the columnar database according to any one of claims 1 to 2, characterized in that it specifically comprises the following steps: Step 1: Enter the event sequence to be retrieved and convert it into a query statement; Step 2: Judging the data type of the event sequence to be retrieved, when the data type of the event sequence to be retrieved is a numerical value or a single character or text with a number of characters less than or equal to 8, go to step 4; when the data type of the event sequence to be retrieved is text, and there is data with more than 8 characters, go to step 3; Step 3: perform hash coding on the sequence of events to be retrieved, and enter step 4 after the data type of the sequence of events to be retrieved is converted into a numerical value; Step 4: Select the query type. If it is an exact query, use the sequence of events to be retrieved as a whole to search for exact matching; if it is a fuzzy query, perform a fuzzy matching retrieval based on the preset fuzzy query distance. As an array as a whole, calculate the minimum distance from the data composition sequence in the event sequence, and the minimum distance is less than or equal to the preset fuzzy query distance; Step 5: Build a result dictionary, add all the matched corresponding sequences to the result dictionary, until all data in the event field is traversed; Step 6: Output the result dictionary as the result. 4. The method for retrieving an event sequence according to claim 3, wherein the to-be-retrieved event sequence is a sequence composed of multiple events arranged in sequence. 5 . The method for retrieving event sequences according to claim 3 , wherein the query statement in step 1 is an SQL-like statement. 6 . 6 . The method for retrieving an event sequence according to claim 5 , wherein the query statement comprises a first parameter, a second parameter and a third parameter set in sequence, and the first parameter is configured as a field to filter the query. 7 . , the second parameter is configured to limit the storage address of the query, and the third parameter is configured to limit the retrieval condition. 7 . The method for retrieving an event sequence according to claim 6 , wherein the third parameter comprises the to-be-retrieved event sequence. 8 . 8. the method for retrieving event sequence according to claim 3, is characterized in that: in described step 4, carrying out fuzzy query and calculating minimum distance is specifically adopting Levenshtein algorithm, when calculating data type is character string, described event to be retrieved The minimum number of steps required to convert any string in the sequence into any string in the event field by adding, deleting, or replacing characters. 9. The method for retrieving event sequences according to claim 8, wherein the calculation of the minimum steps is specifically: Step 41: Calculate the length n of any character string strA in the sequence of events to be retrieved, and calculate the length m of any character string strB in the sequence of events to be retrieved; Step 42: If n=0, the minimum edit distance is m, if m=0, the minimum edit distance is n, if not, then go to step 43; Step 43: Construct a matrix Arr of (m+1)*(n+1), and initialize the first row and first column of the matrix to be 0～n, 0～m respectively; Step 44: Double loop, traverse strA, and traverse strB on this basis, if strA[i]=strB[j], then cost=0, otherwise cost=1, judge Arr[j-1][i]+1 , Arr[j][i-1]+1, the minimum value of Arr[j-1][i-1]+cost, and assign the minimum value to Arr[j][i]; Step 45: After the loop is over, the last element of the matrix is the minimum edit distance, that is, the minimum steps. 10. The method for retrieving event sequences according to claim 3, wherein the result dictionary output in step 6 includes all matching sequences that meet the conditions in the database, and the matching sequences output at least the time field and all the matching sequences. the event field.

Images