JP2012174212A - Window processing device, method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a window processing device capable of reducing computation time and a calculation area for window processing.SOLUTION: A window processing device 1 comprises: a data stream acquisition unit 12 that acquires a data stream including elements having the inclusion relation; a data stream analysis unit 13 that sequentially detects a position of a window element and a position of an arithmetic target element related to the window element from a data series, and stores a set of arithmetic processing values corresponding to the most recently detected arithmetic target element and a window element start symbol in a buffer 14 for every detection of the window element; a solution output unit 16 that performs window processing on the set of arithmetic processing values stored in the buffer 14.

Description

  The present invention relates to a window processing technique for a stream of data including elements having an inclusion relationship.

  In recent years, stream processing (Stream Processing) for the purpose of processing in real time a large amount of data that is constantly generated, such as sensor data, POS (Point of Sales) data, and web data such as blogs and Twitter. Has attracted attention.

  In stream processing, real-time performance is required, so a technique for calculating with as little processing as possible using a limited calculation area is essential.

  Until now, DBMS (database management system) has been the mainstream for processing large amounts of data. However, since DBMS requires a huge amount of processing time for data storage and auxiliary data creation, it is not suitable for real-time processing.

  In addition, the current commercial or open source stream processing engine assumes fixed-length data such as CSV as the data type, and cannot efficiently handle data in XML format that allows more flexible data representation. .

  One of the most basic processes in stream processing is window processing. In window processing, only data within a predetermined processing period is sequentially processed, so that an infinitely long data stream can be processed in real time using only a predetermined calculation area.

  In order to apply the above window processing to a stream of XML data, an Xquery extension language and its execution system have been proposed (first existing technology).

  FIG. 16 is a diagram for explaining an example of an XML data stream.

  In the example shown in FIG. 16, it is assumed that XML data including constituent elements having an inclusion relationship with a certain element as a root is generated one after another to form a stream.

  The XML data stream shown in FIG. 16 is data generated by the POS system. The XML data having the POS element as a root (hereinafter referred to as a POS tree) has a custom element and a shop element. Further, it is assumed that a new POS tree does not occur until the cashier person (store clerk) changes.

  The custom element represents customer shopping data. The “cust” element has a “pric” element representing the price of the purchased product as an essential element, and a “name” element representing the purchaser name as an optional element. The custom element may appear any number of times in a single POS tree.

  The shop element represents the clerk data in charge of the cash register. The shop element has an id element representing a clerk ID as an essential element. The shop element shall appear exactly once in the POS tree.

  Assume that the following XML window query is given as window processing for the POS tree shown in FIG.

for window $ w in / pos / cust
range 3
sliding 1
return SUM ($ w / pric)
In the above XML window query, a sliding window with window size (range) = 3, slide width (sliding) = 1 is set as window processing, and a subtree specified by the path expression “/ pos / cust” is set. This indicates that the sum of the calculation target element values corresponding to the calculation target element (pric element) specified by the path expression “$ w / pric” is output for each window $ w.

  FIG. 17 is a diagram for explaining window processing for an XML data series.

  As shown in FIG. 17, in the window processing, one window component (partial tree starting from a custom element) is cut out from the XML data stream in the first scan and stored in the window buffer B. Then, in the process of the second scan, the calculation target element value is acquired from the window constituent element in the window buffer B and stored in the buffer in the counting device M. Thereafter, the calculation target element values stored in the buffer are totaled by the totaling device M.

  As another existing technology, stream-type Xpath matching has been proposed (second existing technology).

  The stream type Xpath collation is realized by a stream type Xpath collation engine based on a method of collating an Xpath query by scanning XML data once.

  FIG. 18 is a diagram for explaining stream-type Xpath verification for an XML data series.

  In the stream type Xpath collation, the path (Xpath: / pos / cus) specifying the window component is collated at the first scan, and the path (Xpath: / pos) specifying the operation target element at the second scan. / Cust / pric) is collated, and the calculation target element value associated with the calculation target element is specified.

  If the calculation target element (prix) is simply detected from the XML data, the XML subtree specified by Xpath can be acquired in one scan. However, the XML data can be expressed flexibly, for example, the calculation target element included in the window component is unspecified. For this reason, two scans are required in order to acquire the calculation target element in the inclusive relation in association with the subtree by the window constituent element.

  Specifically, when it is desired to perform an operation in units of each window component (XP1) as in the case of the above XML window inquiry, as shown in FIG. 18, the operation target elements XP2_2 and XP2_3 are in an inclusion relationship. It had to be acquired in association with a certain window component XP1_2. That is, two scans are required for collating the window constituent elements from the XML data and collating the calculation target elements from the respective window constituent elements.

A. Arasu, et al, “The CQL Continuous Query Language: Semantic Foundations and Query Execution”, VLDB J. et al. , Vol. 15 (2), pp. 121-142, 2006 I. Botan, et al, “Extending XQuery with Window Functions”, Proc. VLDB 2007, pp. 75-86, 2007 I. Avila-Camillo, et al, “XMLTK: An XML Tool for Scalable XML Stream Processing”, Proc. PLANX'02, 2002 Y. Diao, et al, “Path Sharing and Predicate Evaluation for High-Performance XPath Filtering”, Proc. ACM TODS, 28 (4), 467-516, 2003

  In the window processing for the XML data series described above as the first existing technology, it is necessary to execute two scans in total, that is, one scan for acquiring the window component and one scan for acquiring the calculation target element value. There was a problem that the processing was inefficient.

  Also, in the collation processing of the stream type Xpath for the XML data series described as the second existing technology, in order to obtain the subtrees in an inclusive relationship in association with each other, the collation between the window constituent element and the calculation target element is performed. Two scans are required to be executed, and similarly, there is a problem that the processing is inefficient.

  An object of the present invention is to provide a processing technique capable of acquiring a calculation target element associated with a window constituent element in a single scan in a window process for a data series including elements having an inclusive relation such as XML data. It is.

  A window processing device disclosed as one aspect of the present invention includes 1) a buffer for storing data, 2) a data acquisition unit for acquiring a data series including elements having inclusion relations, and 3) a window from the data series. The position of the window component that defines the processing unit and the position of the operation target element included in the window component are sequentially detected in association with the window component, and are detected immediately before each window component is detected. A data stream analysis unit that stores in the buffer a set of operation target element values corresponding to the operation target element associated with the window component, and 4) an operation processing value associated with the window component stored in the buffer A window processing unit for performing window processing on the set of

  In addition, in a window processing method disclosed as another aspect of the present invention, a computer executes processing steps corresponding to each process realized by each processing unit of the window processing apparatus.

  A window processing program disclosed as another aspect of the present invention causes a computer to execute processing steps corresponding to each processing realized by each processing unit of the window processing device.

  According to the above-described means, the window constituent element and the operation target element can be simultaneously performed from the data series by one scan, and the efficiency of the calculation time and calculation area of the window processing can be realized.

It is a figure which shows the structural example in one Example of the window processing apparatus disclosed as 1 aspect of this invention. It is a figure which shows the structural example of a data stream analysis part. It is a figure which shows the example of expression of the data structure of XML stream XS. It is a figure which shows the example of the window query Q. It is a figure which shows the processing flow of a window processing apparatus. It is a figure which shows the example of a more detailed process flow of a window query acquisition process. It is a figure which shows the example of a processing flow of a data stream analysis process. It is a figure which shows the example of the data CX. It is a figure which shows the more detailed process flow of the stream analysis of step S22. It is a figure which shows the example of a relationship between the XML data of an XML stream, and the data output to a buffer by path | pass verification. It is a figure which shows the example of a processing flow of a solution output process. FIG. 6 is a diagram (part 1) schematically illustrating a relationship between buffer data and data output to an output sequence Seq by a solution output process. FIG. 10 is a diagram (part 2) schematically illustrating a relationship between buffer data and data output to an output sequence Seq by a solution output process. FIG. 11 is a diagram (No. 3) schematically illustrating a relationship between buffer data and data output to an output sequence Seq by a solution output process. It is a figure which shows typically the relationship between each XML data of the XML stream XS, and the data output to a buffer. It is a figure for demonstrating the example of the stream of XML data. It is a figure for demonstrating the window process with respect to an XML data series. It is a figure for demonstrating the stream type Xpath collation with respect to an XML data series.

  A window processing apparatus disclosed as one aspect of the present invention detects a calculation target element associated with a window constituent element from an XML stream using a known stream type XPath collation function, and performs a calculation corresponding to the detected calculation target element. A function that acquires a target element value and detects the start position of the window component, and sends a symbol representing the start of the window component and a set of calculation target values corresponding to the detected calculation target element to the buffer. Have.

  FIG. 1 is a diagram illustrating a configuration example in an embodiment of a window processing apparatus 1 disclosed as one aspect of the present invention.

  The window processing apparatus 1 executes window processing on a data series including elements having an inclusion relationship, for example, an XML data stream (hereinafter, XML stream). More specifically, the window processing device 1 receives a window inquiry (query) Q for the XML stream XS from the query device 2 used by the user, and performs window processing for the XML stream XS transmitted from the data transmission device 3 via the network N. Execute. As window processing, based on the window size w and slide width s set in the window query Q, the latest window total value a is output each time the window slides, and the output is the solution of the window query Q (total) Value a) Send as A.

  The window processing device 1 includes a window query reception unit 11, a data stream acquisition unit 12, a data stream analysis unit 13, a buffer 14, a buffer totaling unit 15, and a solution output unit 16 in order to realize the above functions.

  The window query reception unit 11 receives a window component element XP1, which is a window unit, an operation target element XP2, an operation target element XP2, a window size w (w ≧ 1), and a slide width s (s ≧ 1) from the user query device 2. A window query Q including designation is accepted.

  The data stream acquisition unit 12 receives the XML stream XS through the network N from the data transmission device 3 that is the generation source of the XML data.

  The data stream analyzer 13 sequentially detects the position of the window component XP1 from the received XML stream XS, and further associates the position of the operation target element XP2 included in the window component XP1 with the window component XP1. To detect. Then, the data stream analysis unit 13 holds a set of calculation target element values corresponding to the detected calculation target element XP2, and every time the start position of the window component element XP1 is detected, the data stream analysis unit 13 A set, that is, a set of calculation target element values corresponding to the calculation target element XP2 detected immediately before is stored in the buffer 14.

  The buffer 14 holds a value list V that is a set of element values to be calculated associated with the detection of the window component XP1 and the detected window component XP1.

  Based on the window size w and the slide width s, the buffer tabulation unit 15 tabulates a set of calculation target element values for the window size w stored in the buffer 14 to obtain a window tabulation value a. The buffer totalization unit 15 outputs the window totalization value a obtained most recently to the solution output unit 16 each time the window slides.

  The solution output unit 16 sends the output window total value a to the query device 2 as the solution A of the window query.

  The buffer totaling unit 15, the solution output unit 16, and the window query receiving unit 11 are examples of a window processing unit.

  FIG. 2 is a diagram illustrating a configuration example of the data stream analysis unit 13.

  The data stream analysis unit 13 includes a stream type path matching unit 131, a window component position detection unit 132, a calculation target element position detection unit 133, a calculation target element value detection unit 135, a calculation target element value acquisition unit 136, and a window component start A symbol transmission unit 137 and a calculation target element value list transmission unit 138 are provided.

  The stream type path verification unit 131 performs stream type Xpath verification on each XML data of the received XML stream XS, and detects elements from the XML data.

  The window component position detector 132 detects the start position or the end position of the window component XP1 from the XML data. Here, the start position and the end position of the window component XP1 are tags or events indicating the start / end of the window component XP1.

  The calculation target element position detection unit 133 detects the start position or end position of the calculation target element XP2 from the XML data. Here, the start position and end position of the calculation target element XP2 are a tag or event indicating the start / end indicating the calculation target element XP2.

  The calculation target element value detection unit 135 detects a calculation target element value corresponding to the detected calculation target element XP2.

  The calculation target element value acquisition unit 136 acquires a calculation target element value corresponding to the calculation target element XP2 detected in association with the window component from the XML data, and adds the calculation target element value to the value list V.

  The window component start symbol transmission unit 137 writes a flag ($) indicating the start position of the detected window component XP1 to the buffer 14 every time the start position of the window component XP1 is detected.

  The calculation target element value list transmission unit 138 stores the list V of detected calculation target element values in the buffer 14 when the end position of the calculation target element XP2 is detected after the start position of the window component element XP1 is detected. Write out.

  FIGS. 3A to 3C are diagrams illustrating examples of the representation of the data structure of the XML stream XS.

  Each XML data in the XML stream XS received by the data stream acquisition unit 12 is data generated by the POS system, and includes elements having an inclusion relationship with the POS element as a root. The XML data has a custom element and a shop element, and new XML data (POS element) is not generated until the cashier person (store clerk) changes.

  The custom element indicating the customer's shopping data has a price element indicating the price of the purchased item as an essential element and a name element indicating the purchaser name as an optional element. The custom element may appear any number of times in a piece of XML data.

  The shop element indicating the clerk data in charge of the cash register has an id element representing the clerk ID as an essential element, and appears only once in the XML data.

  FIGS. 3A to 3C are diagrams illustrating examples of the representation of the data structure of the XML stream XS.

  FIG. 3A shows the tree structure of the data structure of the XML data described above. The tree structure representation of the XML data shown in FIG. 3A can be converted into a text representation using tags as shown in FIG. Further, the text representation of the XML data in FIG. 3B can be treated as an event string representation as shown in FIG. As a representative example of the event string representation of XML data, there is the W3C standard SAX.

  In the present embodiment, the data stream acquisition unit 12 can acquire a text expression or event expression stream of XML data as the XML stream XS.

  In the present embodiment, the data stream analysis unit 13 operates on the event string representation of the XML data. However, the data stream analysis unit 13 directly inputs the text representation of the XML data and performs processing while sequentially detecting the event. It may be configured.

  FIG. 4 is a diagram illustrating an example of the window query Q.

  The window query Q shown in FIG. 4 includes a path expression “/ pos / cust” that specifies the window component XP1, a path expression “/ pos / cust / pric” that specifies the operation target element XP2, a window size range = 3, and The designation of slide width sliding = 1 is included.

  Hereinafter, the process flow of the window processing apparatus 1 will be described.

  FIG. 5 is a diagram showing a processing flow of the window processing apparatus 1.

Step S1: Window Query Acquisition Processing The window query reception unit 11 executes window query acquisition processing.

  FIG. 6 is a diagram illustrating a more detailed processing flow example of the window query acquisition processing.

  The window query reception unit 11 receives the window query Q from the user query device 2 (step S11).

  The window query reception unit 11 extracts path designations of the window component element XP1 and the calculation target element XP2 from the received window query Q. Then, the window query reception unit 11 specifies the specified path used for the Xpath collation to the data stream analysis unit 13, that is, Xpath XP1 = / pos / cus for specifying the window component element XP1, and Xpath XP2 = / for specifying the operation target element XP2. pos / cust / pric is transmitted (step S12).

  The window query accepting unit 11 extracts the window size w (= 3) and the slide width s (= 1) from the received window query Q, and sends them to the buffer totaling unit 15 (step S13).

Step S2: Data Stream Analysis Processing The data stream analysis unit 13 performs Xpath collation on the XML stream XS, and extracts the window component element XP1 and the operation target element XP2.

  FIG. 7 is a diagram illustrating a processing flow example of the data stream analysis processing.

  The data stream analysis unit 13 receives the XML stream XS. The data stream analysis unit 13 prepares CX indicating the XML data to be processed in the XML stream XS, and sets the first XML data of the received XML stream XS as data CX as initialization (step S21). FIG. 8 is a diagram illustrating an example of the data CX. The data stream analysis unit 13 receives an event string representation of XML data.

  Next, the data stream analysis unit 13 performs a path matching process on the data CX (step S22).

  The data stream analysis unit 13 determines whether XML data exists after the data CX of the XML stream XS (step S23). If there is XML data next to the data CX of the XML stream XS (Y in step S23), the data stream analysis unit 13 sets the next XML data as the data CX (step S24), and the next XML data in the XML stream XS. Is not present (N in step S23), the process is terminated.

  FIG. 9 is a diagram showing a more detailed processing flow of the path matching process in step S22.

  The stream type path matching unit 131 of the data stream analyzing unit 13 sets the cursor C indicating the scan position as the first event of the data CX (event sequence) as an initialization, the calculation target element flag FA = 0, the value list V Is an empty list (step S221). Then, the stream type path matching unit 131 performs path matching and determines the type of event where the cursor C of the data CX is present (step S222).

  In the process of step S222, when the window component position detection unit 132 determines that the event is “start of window component XP1”, the window component start symbol transmission unit 137 displays the window component start symbol. “$” Is output to the buffer 14 (step S223). When the window component position detecting unit 132 determines that the event is “end of window component XP1”, the calculation target element value list transmitting unit 138 outputs the value list V to the buffer 14 (step S224). ).

  When the calculation target element position detection unit 133 determines that the event is “start of calculation target element XP2”, the calculation target element position detection unit 133 sets the calculation target element flag FA = 1 (step S225). ). When it is determined that the event is “completion of the calculation target element XP2”, the calculation target element position detection unit 133 sets the calculation target element flag FA = 0 (step S226).

  When the calculation target element value detection unit 135 determines that the event is a “text” node, the calculation target element value acquisition unit 136 further determines whether the calculation target element flag FA = 1. (Step S227). If the calculation target element flag FA = 1 (Y in step S227), the calculation target element value acquisition unit 136 adds the text value of the event where the cursor C is located to the list V (step S228). On the other hand, if the calculation target element flag FA is not 1 (N in step S227), the process proceeds to step S229.

  After the processing of steps S223 to S226 and S228, or when the event where the cursor C is located in the processing of step S222 does not correspond to any of the above cases (other than that), or in the processing of step S227, the calculation target element flag When FA is not 1, the stream type path matching unit 131 determines whether there is an event after the event where the cursor C is located in the data CX (step S229). If there is an event next to the event where the cursor C is located (Y in step S229), the stream type path matching unit 131 moves the cursor C to the next event (step S2210), and returns to the processing in step S222.

  When there is no event next to the event where the cursor C is located (N in step S229), the calculation target element value list transmission unit 138 outputs the value list V to the buffer 14 (step S2211).

  FIG. 10 is a diagram illustrating an example of the relationship between XML data of an XML stream and data output to the buffer 14 by path verification.

In XML stream XS shown in FIG. 10 shows a data CX ₁ by a broken line rectangle. The data CX _{1 in} FIG. 10 corresponds to the event sequence shown in FIG.

When the data stream analysis unit 13 detects the path / pos / cus in the data CX ₁ by collating with XP1 = / pos / cus (FIG. 8A), the start position of the window component XP1 “cust” The window component start symbol “$” is output to the buffer 14.

  After that, when the path / pos / cus / pric is detected by collating with XP2 = / pos / cust / pric (FIG. 8 (b)), it is determined as the start position of the calculation target element “pric”, and FA = 1 is set, the corresponding element value (500) to be calculated is extracted and stored in the value list V (FIG. 8 (c)). Subsequently, when the end of the path / pos / cust / pric is detected (FIG. 8D), it is determined that the calculation target element “pric” ends, and FA = 0 is changed.

  Thereafter, when the end of the path / pos / cus is detected (FIG. 8 (e)), it is determined as the end position of the window component XP1 “cust”, and the value list V = {500} is output to the buffer 14 Is done.

In XML stream XS shown in FIG. 10 shows a data CX ₂ by a broken line rectangle.

Similarly to the case of the data CX ₁ , when the data stream analysis unit 13 detects the start position of the window component XP1 “cust” by path verification with XP1 = / pos / cus, the window component start symbol “$” "Is output to the buffer 14.

Thereafter, when the start position of the calculation target element “pric” is detected by path matching with XP2 = / pos / cust / pric, FA = 1 is set, and the corresponding calculation target element value (300) is displayed in the list V. Stored in When the end position of the calculation target element “pric” is detected, FA = 0 is changed. However, in the data X ₂ , the start of the calculation target element “prix” is subsequently detected and FA = 1 is reset. The calculation target element value (200) is set and added to the list V.

Thereafter, as in the case of the data CX ₁ , when the end of the window component XP 1 “cust” is detected, the list V = {300, 200} is output to the buffer 14.

Step S3: Solution Output Processing FIG. 11 is a diagram illustrating a processing flow example of the solution output processing.

  As an initialization, the buffer totaling unit 15 sets the cursor CB indicating the position of data in the buffer 14 to the first data in the buffer 14, sets the window size counter CW = 0, and sets the output sequence Seq as an empty string. (Step S31).

  The buffer totaling unit 15 determines the type of data for which the cursor CB is set in the buffer 14 (step S32). If the data type is the window component start symbol “$” (“symbol $” in step S32), the buffer tabulation unit 15 increments the window size counter CW by 1 (step S33), and proceeds to step S39. If the data type is the value list V (“value list” in step S32), the process proceeds to step S34. In the process of step S34, if CW ≧ 0, the buffer totaling unit 15 adds the value list V in which the cursor CB is set to the end of the output sequence Seq.

  Subsequently, the buffer totalization unit 15 determines whether the window size counter CW is smaller than the window size w (CW <w) (step S35), and the window size counter CW is not smaller than the window size w and is equal (CW = w) In the case (N in step S35), the total value a of the data (value list V) in the output sequence Seq is obtained and output to the solution output unit 16 (step S36). The solution output unit 16 sequentially receives the total value a by the processing of step S36 described above, and sends it to the query device 2 as the solution A of the window query Q.

  Furthermore, the buffer totalization unit 15 deletes s data from the head of the data stored in the output sequence Seq. When the slide width s is larger than the window size w (s> w), all data stored in the output sequence Seq is deleted. The buffer totalization unit 15 replaces the window size counter CW with the value of CW-s (step S37).

  Thereafter, the buffer totaling unit 15 determines whether data exists after the data for which the cursor CB is set in the buffer 14 (step S38). When the next data exists in the buffer 14 (Y in step S38), the buffer totalization unit 15 moves the cursor CB to the next data (value list) (step S39), and returns to the process in step S32. If the next data does not exist in the buffer 14 (N in step S38), the process is terminated.

  12 to 14 are diagrams schematically showing the relationship between the data in the buffer 14 and the data output to the output sequence Seq by the solution output process. The processing shown in FIGS. 12 to 14 is for the case where the window size w = 3 and the slide width s = 1.

  The buffer tabulation unit 15 sets the cursor CB to the first data in the buffer 14 and initializes the window size counter CW = 0.

  As shown in FIG. 12A, the data in which CB is located is the window element start symbol “$”, the window size counter CW is incremented by 1 (CW = 1), and the cursor CB is moved to the next data. . Then, as shown in FIG. 12B, if the data moved by CB is the value list V of the element values to be calculated, the value list V = {500} is stored in the output sequence Seq, and the cursor CB further Move to next data.

  Then, as shown in FIG. 13A, if the data that the cursor CB has moved is the window element start symbol “$”, the window size counter CW = 2, and the cursor CB is moved to the next data. Then, as shown in FIG. 13B, the data value list V = {300, 200} to which the cursor CB has moved is stored in the output sequence Seq, and the cursor CB is moved to the next data.

  Similarly, if the data to which the cursor CB has moved is the window component start symbol “$”, CW = 3, and the cursor CB is moved to the next data.

  Thereafter, as shown in FIG. 14A, if the data to which the cursor CB has moved is a value list, the value list V = {100} is stored in the output sequence Seq. Here, since the window size counter CW is equal to the window size w (= 3), the value list stored in the output sequence Seq is totaled to obtain the total value a (1100). Further, as shown in FIG. 14B, s = 1 data (value list) is deleted from the head of the data (value list) stored in the output sequence Seq, and the window size counter CW is set to CW-s. The value is changed to (= 2). The cursor CB is similarly moved to the next data.

  FIG. 15 is a diagram schematically illustrating a relationship between each XML data of the XML stream XS and data output to the buffer.

  Each time the window processing device 1 detects the start position of the window component XP1, the buffer 14 stores a set of calculation target element values corresponding to the calculation target element XP2 detected immediately before and the window component start symbol “$”. To store.

  The window processing apparatus 1 detects all the included calculation target elements XP2 in association with the window component XP1 even if the calculation target element XP2 included in one window component XP1 is an unspecified number. The calculation target values corresponding to XP2 are held together in a list.

  Therefore, in the buffer 14, the window component element start symbol “$” indicating the start position of the window component element XP1 and the calculation object element value corresponding to the operation element XP2 included in the window element element XP1 are stored in the buffer 14 once. Are stored alternately (value list V). Therefore, in the solution aggregation process, it is possible to aggregate the calculation target element values in units of window components.

  In the present embodiment, the window processing apparatus 1 executes the data stream analysis process of step S2 and the solution output process of step S3 for each XML data X of the XML stream XS, and loops according to the acquisition of the XML data X of the XML stream XS. And run. However, if the XML stream XS is a finite sequence, the window processing apparatus 1 may be configured to execute the data stream analysis process in step S2 and the solution output process in step S3 in batch for the XML stream XS. Good.

  The window processing device 1 can be realized by a computer system including hardware having a CPU, a memory, and the like, and a software program, or dedicated hardware.

  The window processing apparatus 1 includes a processing unit (CPU), a temporary storage device (DRAM, flash memory, etc.), and a permanent storage device (HDD, flash memory, etc.), and is implemented by a computer that inputs and outputs data to the outside. be able to. The window processing apparatus 1 can also be implemented by a program executable by this computer. In this case, a program describing the processing contents of the functions that the window processing apparatus 1 should have is provided. When the computer executes the provided program, the processing functions of the window processing apparatus 1 described above are realized on the computer. The computer can also read a program directly from a portable recording medium and execute processing according to the program. Furthermore, the program can be recorded on a computer-readable recording medium.

  As described above, the disclosed window processing apparatus 1 has the following effects.

  In other words, the window processing device 1 enables window processing by one scan even for a data stream in which an operation target element included in a window component is an unspecified number, and processing time and storage The area can be reduced.

  In the prior art, in order to detect the start position of the window constituent element and the calculation target element position, it is necessary to scan each XML data in the XML stream twice. However, according to the window processing apparatus 1, since the start position of the window constituent element and the calculation target element position are detected simultaneously in one scan, the calculation time can be reduced.

  Further, as a secondary effect, the window processing apparatus 1 obtains the minimum data (value list) necessary for the totaling processing and obtains the buffer, unlike the conventional technique in which all the window components are held in the buffer 14. 14. As a result, according to the window processing apparatus 1, the amount of data held in the buffer 14 can be reduced, so that the calculation time and the calculation area can be reduced.

DESCRIPTION OF SYMBOLS 1 Window processing apparatus 11 Window query reception part 12 Data stream acquisition part 13 Data stream analysis part 131 Stream type | mold path collation part 132 Window component element position detection part 133 Calculation object element position detection part 135 Calculation object element value detection part 136 Calculation object element Value acquisition unit 137 Window component element start symbol transmission unit 138 Operation target element value list transmission unit 14 Buffer 15 Buffer aggregation unit 16 Solution output unit 2 Query device 3 Data transmission device

Claims (4)

A window processing device for performing window processing on a data series,
A buffer for storing data,
A data acquisition unit for acquiring a data series including elements having inclusion relations;
From the data series, the position of the window component defining the window processing unit and the position of the operation target element included in the window component are sequentially detected in association with the window component, and the window component is detected. A data stream analysis unit that stores a set of calculation target element values corresponding to the calculation target element associated with the window component detected immediately before, in the buffer;
A window processing device, comprising: a window processing unit that performs window processing on a set of operation processing values associated with the window component stored in the buffer. The data stream analysis unit holds a set of calculation target element values corresponding to the detected calculation target element, and each time the start position of the window constituent element is detected, the held set of calculation target element values And the start position symbol of the detected window component in the buffer,
The window processing device according to claim 1, wherein the window processing unit performs the window processing based on a start position symbol of a window component stored in the buffer. A window processing method for a data series executed by a computer,
A processing step for obtaining a data series including elements having inclusion relations;
A step of sequentially detecting, from the data series, a position of a window component that defines a window processing unit and a position of an operation target element included in the window component in association with the window component;
A step of storing, in the buffer, a set of operation target element values corresponding to the operation target elements associated with the window configuration element detected immediately before each detection of the window component;
A window processing method comprising: performing a window process on a set of operation processing values associated with the window component stored in the buffer. A window processing program for causing a computer to perform window processing on a data series,
A process of obtaining a data series including elements having an inclusion relation in the computer;
A process of sequentially detecting, from the data series, a position of a window component that defines a window processing unit and a position of an operation target element included in the window component in association with the window component;
A process of storing, in the buffer, a set of operation target element values corresponding to the operation target element associated with the window component detected immediately before each detection of the window component;
A window processing program for executing a window processing on a set of operation processing values associated with the window component stored in the buffer.

Images