A multi-index horizontal expansion data processing method and system

Abstract

The present application relates to the technical field of data query, in particular to a multi-index horizontal expansion data processing method and system, comprising the following steps: collecting time series and dimension attributes to generate a business detail wide table, matching non-empty measure fields to form independent index logic, combining rule nodes to build a general materialized architecture, and outputting multi-index processing results through scheduling mapping and window aggregation. In the present application, a wide table is constructed in the model layer and atomic and derived indexes are centrally managed, realizing a unified index definition method independent of dimension combination. When a user queries, dimension parameters are dynamically combined by the query engine to automatically build an aggregation statement and calculate the result, avoiding the problem of derived index explosion caused by the expansion of the number of dimensions. Meanwhile, multi-index joint analysis is completed in the same materialized table without the need for multiple table joins, improving query execution efficiency, having high adaptability to index updates and dimension changes, and significantly reducing data development and operation and maintenance burden.

Description

Technical Field

[0001] This invention relates to the field of data query technology, and in particular to a data processing method and system with multi-indicator horizontal expansion. Background Technology

[0002] The field of data query technology involves the organization, extraction, and aggregation analysis of structured big data. It supports various applications such as business intelligence analysis, operational monitoring, and decision support through the construction of indicator systems. Typically, query models are built around the relationship between indicators and dimensions to achieve efficient filtering and statistical aggregation of multidimensional data. Indicator data processing methods involve defining atomic indicators and pre-binding dimension combinations to generate derived indicators based on fact tables or wide tables. Further, composite indicators are constructed according to computational relationships. This involves extracting measure fields from business data detail tables, using aggregation functions such as summarization and counting to generate basic atomic indicators, and combining them with fixed specific filtering conditions and dimensions to generate a large number of predefined derived indicators. To achieve more complex business expressions, computational logic is established between multiple atomic or derived indicators to generate composite indicators. However, when facing frequently changing analytical needs, the explosion in the number of dimension combinations often leads to system overload. Furthermore, multi-indicator joint queries rely on multi-table joins and cross-indicator view integration, resulting in complex queries, slow response times, and high maintenance difficulty.

[0003] The existing data query model binds dimensions and indicators during the design phase, which leads to an exponential increase in the number of derived indicators as the number of dimensions increases. This drastically increases the complexity of indicator maintenance. Querying multiple indicators requires multiple join queries of different pre-aggregated tables or views, which reduces query performance. Furthermore, new analysis requirements require redefining indicators and rebuilding views. The lack of a flexible adaptation mechanism results in untimely system response. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a data processing method with multi-index horizontal expansion.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a data processing method with multi-indicator horizontal expansion, comprising the following steps: S1: Collect timestamp columns and dimension attribute columns from the enterprise data warehouse, perform field assembly operations on the timestamp columns and dimension attribute columns, and obtain the underlying business detail wide table; S2: Read the preset measurement field set, match and filter the non-empty fields in the preset measurement field set with all columns in the underlying business detail wide table, and generate independent indicator definition logic; S3: Read the built-in logic operator library of the analysis front end, extract the conditional filtering expression, perform a full sorting combination mapping between the conditional filtering expression and the business attributes in the definition logic of the independent indicator, remove the table structure disk statement, retain the business rule association instruction and format it as a rule node, and splice all the rule nodes according to the logical dependency order to obtain the composite calculation rule node matrix. S4: Read the automated scheduling instruction list in the data platform and perform data structure format mapping and conversion processing with the rule nodes in the composite calculation rule node matrix to establish a general indicator materialization architecture; S5: Obtain the snapshot node and window start and end node in the SQL query engine, perform condition mapping in the general indicator materialization architecture, extract the grouping clause set and aggregation field set under the corresponding mapping branch, and obtain the data processing results of horizontal expansion of multiple indicators.

[0006] As a further embodiment of the present invention, the underlying business detail wide table includes a basic transaction record set, an entity feature matrix, and a time-series state snapshot; the independent indicator definition logic includes a basic statistical algorithm, a single-dimensional evaluation formula, and an original value quantization script; the composite calculation rule node matrix includes a multi-dimensional derived topology graph, a cascaded operation grid, and a dependency derivation flow graph; the general indicator materialization architecture includes a physical storage engine model, a task orchestration skeleton, and a computing resource allocation template; and the data processing results of the horizontal expansion of multiple indicators include a periodic summary set, a grouping clause set, and an aggregation field set.

[0007] As a further aspect of the present invention, the steps for obtaining the underlying business detail wide table are as follows: S111: Collect sample timestamp columns and dimension attribute columns from the enterprise data warehouse, extract discrete time nodes in the timestamp column, extract business feature characters in the dimension attribute column, perform field assembly operations on the timestamp column and the dimension attribute column, merge discrete time nodes and business feature characters, concatenate same-origin attribute entries, and generate an initial assembled character matrix. S112: Based on the initial assembled character matrix, parse the primary key identifier and related foreign key identifier inside the matrix, compare the primary key identifier and related foreign key identifier, filter out target items that are completely consistent, perform vertical alignment operation based on the timestamp column involved in the target item and the dimension attribute column, arrange the feature items corresponding to the same key value, and obtain the vertical alignment feature spectrum. S113: Based on the vertical alignment feature spectrum, identify the internal horizontal correspondence, perform horizontal splicing and association operations on each business field with horizontal correspondence, integrate multi-source dimensional attributes and temporal evolution status, reshape the underlying entity architecture, and obtain the underlying business detail wide table.

[0008] As a further aspect of the present invention, the steps for obtaining the independent indicator definition logic are specifically as follows: S211: Collect a preset set of measurement fields in local storage, extract non-empty fields in the preset set of measurement fields, obtain the full list header attribute names in the underlying business detail wide table, perform string sequence verification operation on non-empty fields and full list header attribute names, filter target items with completely identical characters, aggregate target items, and establish a set of feature names with the same name. S212: Based on the same feature name set, locate the corresponding column storage type in the underlying business detail wide table, extract the original definition type of the associated metric field in the preset metric field set, compare the corresponding column storage type with the original definition type of the associated metric field, remove columns with mismatched types, retain the same name and type of metric attribute, and generate a list of matching attribute types. S213: Based on the list of matching attribute types, extract all the same-name and same-type measurement attributes, map each of the same-name and same-type measurement attributes to the blank indicator definition template, establish a topological network of association mapping between multi-dimensional attribute nodes, and obtain independent indicator definition logic.

[0009] As a further aspect of the present invention, the step of obtaining the composite calculation rule node matrix specifically includes: S311: Obtain the built-in logic operator library, extract the internal conditional filtering expression, perform full-out combination mapping with the business attributes in the independent indicator definition logic, establish a mapping relationship, and generate a conditional attribute mapping set; S312: Based on the conditional attribute mapping set, parse the internal statement type, compare the statement type with the preset disk write identifier, remove the table structure disk write statement, retain the business related instructions and perform format conversion operation, unify the instruction architecture standard, and obtain standardized related nodes. S313: Based on the standardized associated nodes, calculate the topology priority value of each node, analyze the dependency hierarchy between nodes, extract the hierarchy order, perform an ordered splicing operation on all standardized associated nodes according to the order, construct a coherent execution path topology, and establish a composite calculation rule node matrix.

[0010] As a further aspect of the present invention, the formula for calculating the topology priority value is as follows: ; in, Representative node The topology priority value, Represents a pointer to a node The set consisting of all predecessor nodes, Represents the predecessor node The computational complexity factor, Represents the node To the node Dependency weights Representative node Its own adjustment parameters Representative node The number of successor nodes it points to. Represents the global scaling factor.

[0011] As a further aspect of the present invention, the steps for obtaining the materialized architecture of the general indicator are as follows: S411: Obtain the automated scheduling instruction list within the data platform, extract the trigger cycle status and execution frequency identifier within the automated scheduling instruction list, perform a structure splitting and decoding operation on the trigger cycle status and execution frequency identifier, strip away the underlying control parameter level framework of the instruction, and generate a scheduling instruction parameter set; S412: Based on the scheduling instruction parameter set, extract the core entity business control nodes inside the composite calculation rule node matrix, compare the data type differences between the core entity business control nodes and the control parameter hierarchical framework, perform data structure format conversion and alignment operations, and obtain a unified format logical matrix. S413: Based on the unified format logic matrix, the core entity business control nodes and control parameter hierarchical framework are structurally integrated, the structure mapping result items are written to the blank physical table, the underlying real physical addressing path of the physical table is supplemented, the corresponding connection of the entire process of business data flow is solidified, and a general indicator materialization architecture is established.

[0012] As a further aspect of the present invention, the steps for obtaining the data processing results of the multi-indicator horizontal expansion are specifically as follows: S511: Obtain the snapshot node and window start and end node in the local SQL query engine, combine the general indicator materialization architecture, extract the underlying storage addressing path, map the snapshot node and window start and end node to the underlying storage addressing path according to the time alignment dimension, establish the time-series state correspondence, and generate a time-series mapping branch set. S512: Based on the time-series mapping branch set, analyze the associated business scenarios within the underlying storage addressing path, locate and extract the grouping clause set and aggregation field set under the mapping branch, and perform assembly and combination operations on the grouping clause set and aggregation field set according to the business association dimension to obtain the grouping aggregation query framework. S513: Based on the grouping and aggregation query framework, read the internal multi-dimensional analysis hierarchy, combine the snapshot node to trigger the multi-indicator horizontal expansion instruction, perform the same-granularity horizontal extension and splicing operation on the aggregation field set according to the grouping clause set, summarize the full table header indicator dimension attributes, and obtain the data processing results of the multi-indicator horizontal expansion.

[0013] A data processing system with multi-indicator horizontal scaling includes: The business wide table integration module collects timestamp columns and dimension attribute columns from the enterprise data warehouse, performs field assembly operations on the timestamp columns and dimension attribute columns, and obtains the underlying business detail wide table. The dimension constraint decoupling module reads a preset set of measurement fields, matches and filters the non-empty fields in the preset set of measurement fields with all columns in the underlying business detail wide table, and generates independent indicator definition logic. The composite indicator assembly module reads the built-in logic operator library of the analysis front end, extracts the conditional filtering expression, performs a full-sort combination mapping between the conditional filtering expression and the business attributes in the definition logic of the independent indicator, removes the table structure disk statements, retains the business rule association instructions and formats them as rule nodes, and assembles all rule nodes according to the logical dependency order to obtain the composite calculation rule node matrix. The script materialization scheduling module reads the automated scheduling instruction list in the data middle platform, performs data structure format mapping and conversion processing with the rule nodes in the composite calculation rule node matrix, and establishes a general indicator materialization architecture. The engine view construction module obtains the snapshot node and window start and end node within the SQL query engine, performs condition mapping within the general indicator materialization architecture, extracts the grouping clause set and aggregation field set under the corresponding mapping branch, and obtains the data processing results of horizontal expansion of multiple indicators.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by constructing a wide table at the model layer and centrally managing atomic and derived indicators, a unified indicator definition method that does not depend on dimension combinations is achieved. When users query, the dimension parameters are dynamically combined and the query engine automatically constructs aggregation statements and calculates the results, avoiding the problem of derived indicator explosion caused by the expansion of the number of dimensions. At the same time, the joint analysis of multiple indicators is completed in the same materialized table without the need for multiple table joins, which improves query execution efficiency, has high adaptability to indicator updates and dimension changes, and significantly reduces the burden of data development and maintenance. Attached Figure Description

[0015] Figure 1 This is a flowchart of the main steps of the present invention; Figure 2 This is a flowchart illustrating the process of obtaining the underlying business detail wide table in this invention. Figure 3 This is a flowchart illustrating the logic for obtaining the independent indicator definition in this invention. Figure 4 This is a flowchart of the process for obtaining the composite calculation rule node matrix of the present invention; Figure 5 This is a flowchart illustrating the process of obtaining the general indicator materialization architecture of this invention. Figure 6 This is a flowchart illustrating the data processing results acquisition process of the present invention, which involves horizontal expansion of multiple indicators. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0017] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0018] Please see Figure 1 A data processing method with multi-indicator horizontal scaling includes the following steps: S1: Collect timestamp columns and dimension attribute columns from the enterprise data warehouse, perform field assembly operations on the timestamp columns and dimension attribute columns, and perform vertical alignment and horizontal splicing association on the timestamp columns and dimension attribute columns according to the primary key and foreign key relationship to obtain the underlying business detail wide table; S2: Read the preset measurement field set, match and filter the non-empty fields in the preset measurement field set with all columns in the underlying business detail wide table, perform name verification and data type verification and comparison, remove mismatched columns and retain the same name and type measurement attributes, and generate independent indicator definition logic. S3: Read the built-in logic operator library of the analysis front end, extract the conditional filtering expressions in the built-in logic operator library, perform full sorting combination mapping between the conditional filtering expressions and the business attributes in the independent indicator definition logic, remove the table structure disk statements, retain the business rule association instructions and format them as rule nodes, and splice all rule nodes according to the logical dependency order to obtain the composite calculation rule node matrix. S4: Read the automated scheduling instruction list in the data platform, perform data structure format mapping and conversion processing between the automated scheduling instruction list and the rule nodes in the composite calculation rule node matrix, and establish a general indicator materialization architecture; S5: Obtain the snapshot node and window start and end node in the SQL query engine, perform condition mapping on the snapshot node and window start and end node in the general indicator materialization architecture, extract the grouping clause set and aggregation field set under the corresponding mapping branch, and obtain the data processing results of horizontal expansion of multiple indicators.

[0019] The underlying business detail wide table includes a basic transaction record set, an entity feature matrix, and a time-series state snapshot. The independent indicator definition logic includes basic statistical algorithms, single-dimensional evaluation formulas, and original value quantification scripts. The composite calculation rule node matrix includes a multi-dimensional derived topology graph, a cascaded operation grid, and a dependency derivation flow graph. The general indicator materialization architecture includes a physical storage engine model, a task orchestration skeleton, and a computing resource allocation template. The data processing results of horizontal expansion of multiple indicators include a periodic summary set, a grouping clause set, and an aggregation field set.

[0020] Please see Figure 2 Step S1 is as follows: S111: Collect sample timestamp columns and dimension attribute columns from the enterprise data warehouse, extract discrete time nodes from the timestamp columns, extract business feature characters from the dimension attribute columns, perform field assembly operations on the timestamp columns and dimension attribute columns, merge discrete time nodes and business feature characters, concatenate same-origin attribute entries, and generate an initial assembled character matrix. Establish a connection and access the underlying distributed file system of the enterprise-grade distributed Hadoop data warehouse. Use a Java database connection driver to read the original order transaction details table containing 10,000,000 historical transaction records and the product dimension mapping table containing 50,000 product information entries. Scan all column field attributes in the original order transaction details table, identify target columns defined as date / time data types, and lock them as sample timestamp columns. Simultaneously scan the business label columns defined as variable-length strings in the product dimension mapping table and lock them as dimension attribute columns. Perform null value removal and format normalization preprocessing operations on the sample timestamp columns. Traverse 10,000,000 timestamp data, uniformly converting non-standard time strings to the standard Universal Time format. Extract year, month, day, and hour data items from the standard Universal Time format, discarding minute and second-level minor differences, and generate discrete time nodes composed of 10 digits in the format 2026031810, thereby removing noise interference from the continuous timeline. Text data within the collected dimension attribute columns is segmented into Chinese and stop words are removed using a pre-defined industry dictionary. Function words and conjunctions without substantive business meaning are eliminated, while nouns and verbs representing entity characteristics are retained. The first letter of the pinyin of the retained words is extracted and concatenated with a pre-defined first-level industry code value to generate a fixed-length 8-character business feature string. 10,000,000 processed sample records are traversed, and each record is assigned an independent auto-incrementing integer row index. The discrete time node corresponding to the same row index is placed at the beginning of the string, double underscores are added as delimiters, and the business feature string is appended to the end of the delimiter. The concatenated character sequence is converted into a one-dimensional character array structure in memory. The character arrays of all samples are merged to generate an initial assembled character matrix with 10,000,000 rows and 1 column.

[0021] S112: Based on the initial assembled character matrix, parse the primary key identifier and related foreign key identifier inside the matrix, compare the primary key identifier and related foreign key identifier, filter out target items that are completely consistent, perform vertical alignment operation based on the timestamp column and dimension attribute column involved in the target item, arrange the feature items corresponding to the same key value, and obtain the vertical alignment feature spectrum. The system reads metadata information from the initial assembled character matrix memory block and parses the relational database constraint definition dictionary attached to the record rows. It iterates through all dictionary entries, identifying the 12-digit primary key identifier representing the unique identity of an entity based on uniqueness and non-null constraints. Simultaneously, it extracts the 12-digit foreign key identifiers pointing to entities in other dimensions based on the foreign key association mapping table. Two independent hash mapping sets are created, placing the primary key identifier in the key position of the first hash mapping set and the foreign key identifier in the key position of the second hash mapping set. A double-loop comparison logic is initiated, with the outer loop iterating through the first hash mapping set and the inner loop iterating through the second hash mapping set, comparing the absolute value structure of the primary key identifier and the foreign key identifier character by character. The character structure difference is recorded; if and only if the character difference equals 0, the current comparison item is determined to be a completely identical target item. Matches with a difference greater than 0 are filtered out, retaining the 2,500,000 successfully filtered completely identical target items. Retrieve the complete row records corresponding to these 2,500,000 target items from the original database, and extract the sample timestamp column content and dimension attribute column content carried in the records. Establish a two-dimensional aligned grid model, with timestamps as the vertical axis and dimension attribute features as the horizontal data positions. Traverse the target item data, moving data items within the same timestamp range to the same physical row block on the vertical axis, and pushing data belonging to the same type of business feature characters into the corresponding column blocks. When a target item is detected to have missing values ​​in a specific column block, use the median of the values ​​corresponding to the five adjacent time nodes within that column block for interpolation to fill the missing values, ensuring the matrix is ​​full. After completing the arrangement operation of all 2,500,000 target items, solidify the data distribution state of the two-dimensional aligned grid model and obtain a hole-free vertical alignment feature spectrum.

[0022] S113: Based on the vertical alignment feature spectrum, identify the internal horizontal correspondence, perform horizontal splicing and association operations on each business field with horizontal correspondence, integrate multi-source dimensional attributes and time-series evolution status, reshape the underlying entity architecture, and obtain the underlying business detail wide table. Extract the row block sequence features of the vertically aligned feature spectrum. For a set of multi-dimensional business fields at the same horizontal coordinate, calculate the Pearson correlation coefficient between every two business fields. Determine the interval position of the Pearson correlation coefficient value. When the coefficient value is greater than or equal to 0.8, it is considered that there is a significant horizontal correlation between the two business fields; when the coefficient value is less than 0.8, it is considered a weak correlation or no direct horizontal correlation. For the set of business fields with horizontal correlation, establish a field merging mapping pointer. Using the basic user account field as the concatenation leader, read heterogeneous field data such as user status dimension, historical purchase frequency dimension, and most recent login time attribute. Perform a memory-level horizontal concatenation and association operation, appending the multi-source dimension attribute fields column by column to the right of the leader, and synchronously reading the status change log records on the timeline. Serialize the evolution status values ​​of each dimension attribute at different time points into a comma-separated dynamic array, and embed it as a new column into the concatenation result. The internal dependencies of entities were streamlined, redundant nested subqueries were eliminated, and fragmented information originally scattered across five different business tables was flattened and reorganized. A new globally unified wide table primary key was assigned to solidify the reshaped flat entity underlying architecture, ultimately generating a low-level business detail wide table with 250 feature columns and 2,500,000 rows in the distributed storage system.

[0023] Please see Figure 3 Step S2 is as follows: S211: Collect a preset set of measurement fields in local storage, extract non-empty fields from the preset set of measurement fields, obtain the full list header attribute names in the underlying business detail wide table, perform string sequence verification operation on non-empty fields and full list header attribute names, filter target items with completely identical characters, aggregate target items, and establish a set of feature names with the same name. The system reads the extensible markup language (EXP) format metric field configuration document pre-deployed in the absolute physical path ` / opt / data / config.xml`, parses the tag tree structure within the document, and collects a preset metric field set containing 300 preset basic metrics. It iterates through this preset metric field set, checking the internal node attribute values ​​of each metric item definition. The system calculates the character length of each node attribute value; if and only if the character length is greater than 0 and does not contain entirely whitespace strings, the field is considered to have actual defined content, thus extracting 285 non-empty fields. It accesses the underlying business detail wide table persistently stored in the distributed file system, retrieves its table header metadata description file, and reads and extracts the 250 full list header attribute names declared in the first row of the wide table structure. The extracted 285 non-empty fields are placed in the left-hand matching sequence, and the 250 full list header attribute names are placed in the right-hand baseline sequence. A character-by-character string sequence verification operation is performed, converting the left-hand sequence elements to standard lowercase format and removing underscores; similarly, the right-hand sequence elements undergo equivalent conversion preprocessing. The similarity score between the two pairs is calculated using the longest common substring matching algorithm. A similarity benchmark of 1.0 is set, requiring absolute alignment of the converted characters. Target pairings with a similarity score of 1.0 are selected, identifying a total of 180 corresponding matching groups. These selected target pairings are then grouped together, redundant records pointing to the same right-hand list header in the left-hand sequence are removed, and uniquely identified correspondences are registered in a dictionary cache, creating a set of 150 independent mapping relationships for the same-name feature.

[0024] S212: Based on the same feature name set, locate the storage type of the corresponding column in the underlying business detail wide table, extract the original definition type of the associated measurement field in the preset measurement field set, compare the storage type of the corresponding column with the original definition type of the associated measurement field, remove the corresponding column with mismatched type, retain the same name and type of measurement attribute, and generate a list of matching attribute types. The system queries the memory dictionary cache for the set of feature names with the same name, extracts the right-hand column header names from the mapping relationships, and re-scans the column definition dictionary of the underlying business detail wide table. For each column header name, it reads the physical storage type code allocated at the database level to locate the corresponding column storage type within the underlying business detail wide table. Simultaneously, it extracts the left-hand measure field names from the mapping relationships of the feature name set with the same name, traces back to the measure field configuration document in Extensible Markup Language (XML) format, reads the primitive data type tags declared within the measure node, and extracts the primitive definition type of the associated measure field. A type compatibility mapping matrix is ​​established, with rules such as: integers marked at the database level and integers declared in the configuration document are considered type matches, and double-precision floating-point numbers and decimals are considered matches. Each of these 150 mapping relationships is compared and judged. When the corresponding column storage type cannot match the primitive definition type of the associated measure field in the compatibility mapping matrix, an elimination mechanism is triggered, the mapping relationship is severed, and the corresponding column with the type mismatch is removed from the valid queue. After filtering, 12 incompatible fields caused by deviations in the initial table type design were eliminated, leaving 138 identical metric attributes with the same name and compatible data structure. The names, underlying types, original definition types, and wide table column index numbers of these 138 attributes were structured and encapsulated, and written into a unified standard lightweight data exchange format file to generate a list of matching attribute types.

[0025] S213: Based on the list of matching attribute types, extract all identical metric attributes of the same name and type, map each identical metric attribute of the same name and type to the blank indicator definition template, establish a topological network of association mapping between multi-dimensional attribute nodes, and obtain the independent indicator definition logic. Load the list of matching attribute types from the lightweight data exchange format file, parse the internally encapsulated array object, and iteratively extract the basic information of all 138 identically named and typed metric attributes. Read the pre-set blank metric definition template file located at the absolute path / usr / local / templates / blank_metric.tpl in the local template library. This template file contains standard dimension definition areas, metric definition areas, and blank areas for calculation logic mounting points. Categorize the 138 metric attributes according to their business tags, insert them one by one into the metric definition area of ​​the blank metric definition template, and automatically assign them globally unique metric codes. Based on the entity relationship graph of the underlying business detail wide table, extract the business dimension level to which these 138 metric attributes belong. Establish a topological network of association mapping between multi-dimensional attribute nodes, connecting metric attribute nodes with the same or similar dimension levels (such as all belonging to the product category level or all belonging to the user region level) using directed edges. The data drill-down correlation values ​​between attributes are recorded on the edge weights of the topology network. All node connection paths in the topology network are traversed using a depth-first search algorithm, and the metrics, dimensions, and aggregation methods involved in the paths are automatically filled into the computational logic mounting points of the template. The memory state of the filled template file is then fixed, resulting in 138 independent indicator definition logics that are physically independent of the original table and have complete input / output declarations and operation rules.

[0026] Please see Figure 4 Step S3 is as follows: S311: Obtain the built-in logic operator library, extract the internal conditional filtering expression, perform full-out combination mapping with the business attributes within the independent indicator definition logic, establish mapping relationship, and generate a conditional attribute mapping set; The system reads the logic operator library file pre-installed in the system kernel's absolute service path, ` / usr / lib / operators.dll`. This operator library file is encapsulated as a dynamic link library. The application interface is called to traverse the exported function table of the operator library, identifying and extracting a set of internal conditional filtering expressions marked as filtering types. This set includes 50 basic and combinational logic operators, such as greater than a specific value, within a time interval, and containing a specific string. A set containing 138 independent indicator definition logics generated by preprocessing is obtained, and the business attribute identifiers bound to each indicator definition logic are extracted. The 50 internal conditional filtering expressions are placed in a horizontal set, and the 138 business attribute identifiers are placed in a vertical set. A two-dimensional Cartesian product operation is performed, and a full permutation combination mapping is performed, forcibly pairing each business attribute identifier with each conditional filtering expression to generate an initial combination pool containing 6900 candidate options. The initial combination pool is traversed, and the type compatibility verification interface within the operator library is called to check whether the data type of the business attribute matches the data type required by the filtering expression. If a business attribute is text and the matching expression is a numerical comparison range, the candidate item is destroyed; otherwise, the successfully matched item is retained. After strict data type compatibility filtering and hard mutual exclusion rule verification for business scenarios, combination items without business meaning are eliminated, legal and valid mapping relationships are established, and finally the remaining 1200 reasonable configuration items are serialized to generate a conditional attribute mapping set.

[0027] S312: Based on the conditional attribute mapping set, parse the internal statement type, compare the statement type with the preset disk write identifier, remove the table structure disk write statement, retain the business related instructions and perform format conversion operation, unify the instruction architecture standard, and obtain standardized related nodes. The system reads the condition attribute mapping set maintained in memory and parses the control logic instruction hierarchy of each record. It extracts the structured query language verb tags and execution action classification codes contained within the instructions to clarify the actual functional intent of the statements. The extracted statement types are matched against a pre-defined disk write identifier set in the global environment variables. This disk write identifier set strictly defines all instruction feature strings that trigger substantial write operations to the physical storage medium, including table creation commands, data insertion commands, data update commands, and table structure modification commands. Each statement type is compared to see if its verb tags exist in the disk write identifier set. If a match is found, the system immediately triggers the removal of the table structure write statement from the disk, intercepting its flow into subsequent calculations and preventing unexpected disk overwrites and space consumption during execution. This interception operation excluded 45 instructions containing persistent write characteristics. The remaining 1155 business association instructions that did not match the disk write identifiers and only performed in-memory data filtering and dimension table association calculations are retained. The kernel's underlying abstract syntax tree conversion engine is invoked to perform format conversion operations on the retained business association instructions. Function calls and syntax structures that originally contained dialectal features specific to each database vendor were replaced with lexical substitutions and reconstructed according to the general paradigm of standard database query languages. A unified output architecture standard for all instructions was achieved, eliminating underlying dialectal differences and resulting in 1155 standardized related nodes that fully conform to the standard execution specifications.

[0028] S313: Based on standardized associated nodes, the formula is used: ; Calculate the topology priority value of each node, analyze the dependency hierarchy between nodes, extract the hierarchy order, perform ordered concatenation operations on all standardized associated nodes according to the order, construct a coherent execution path topology, and establish a composite computational rule node matrix. Representative node The topology priority value, Represents a pointer to a node The set consisting of all predecessor nodes, Represents the predecessor node The computational complexity factor, Represents the node To the node Dependency weights Representative node Its own adjustment parameters Representative node The number of successor nodes it points to. Represents the global scaling factor; The runtime graph metadata of the 1155 standardized associated nodes was collected to initialize the node priority calculation engine. To calculate the topology priority value of each node, multiple metrics parameters of each node in the real execution environment were extracted from historical monitoring logs. The parameter acquisition channels and quantification standards were clearly defined: the set pointed to by the predecessor node was obtained by parsing the current node's input data source graph; the computational complexity factor of the predecessor node was calculated using a multinomial scoring method based on the number of nested query layers and the number of join tables involved in the node; the dependency weight was set based on the frequency and throughput of upstream and downstream data interactions; the self-adjustment parameter was obtained by the system through Boolean variable mapping based on whether the node is in the core business main link; the number of successor nodes was directly counted by traversing the downstream branches of the directed graph; the global scaling factor was given based on the load rate fluctuation of the current system's available computing resource pool. Substituting the obtained computational complexity factor (value 2.5), dependency weight (value 1.2), self-adjustment parameter (value 0.5), number of successor nodes (value 3), and global scaling factor (value 0.9) into the formula, the calculation process expressed in standard mathematical form is as follows: ; The result indicates that the current node's topology priority value is 1.575, placing it in the medium-to-high priority range. The topology priority values ​​of all nodes are calculated and compared with a preset priority benchmark (0.0 to 3.0). Nodes are sorted from largest to smallest based on their calculated values ​​to analyze the dependency hierarchy and establish an absolute execution order. An ordered concatenation operation is then performed on all standardized associated nodes according to this order, constructing a directed acyclic continuous execution path topology. The topology structure is then solidified, establishing a composite calculation rule node matrix with a row width of 1155 and a column height of 1155. The advantage of this formula lies in its effective suppression of irrational preemption of overall scheduling resources by excessively dispersed nodes through the introduction of a square root-based decay mechanism for the number of successor nodes.

[0029] Please see Figure 5 Step S4 is as follows: S411: Obtain the automated scheduling instruction list within the data platform, extract the trigger cycle status and execution frequency identifier within the automated scheduling instruction list, perform structural splitting and decoding operations on the trigger cycle status and execution frequency identifier, strip away the underlying control parameter level framework of the instruction, and generate a scheduling instruction parameter set; A dedicated enterprise-grade internal network security tunnel connects to the core control plane of the data platform. The API provided by the control plane, representing the concrete state transmission, is invoked to paginate and read the automated scheduling instruction table data stored in the platform configuration database, retrieving a total of 500 active automated scheduling instruction lists. These 500 instruction lists are traversed, and the binary control packet header of each instruction is identified. The control segment data from byte 4 to byte 8 within the header is read to extract the trigger cycle status and execution frequency identifier within the automated scheduling instruction list. The instruction decoder is invoked, and a pre-configured structure splitting mask is used to perform structure splitting and decoding operations on the trigger cycle status and execution frequency identifier. The aggregated control instructions are separated according to specific delimiters, removing the redundant outer communication handshake protocol shell. A deep analysis of the stripped-down underlying data reveals key parameter pairs in the instruction's underlying control parameter hierarchy framework, including the Cron expression for the time trigger, the timeout retry threshold (range 1 to 5, actually set to 3), and the dependency timeout waiting upper limit (set to 120 minutes). The stripped and cleaned full-cycle control attributes and frequency restriction rules are reconstructed and encapsulated in the form of standard key-value pairs to generate a set of scheduling instruction parameters that includes an independent scheduling schedule and fault tolerance configuration mechanism.

[0030] S412: Based on the scheduling instruction parameter set, extract the core entity business control nodes inside the composite calculation rule node matrix, compare the data type differences between the core entity business control nodes and the control parameter hierarchical framework, perform data structure format conversion and alignment operations, and obtain a unified format logical matrix. The system retrieves the resident scheduling instruction parameter set and the previously constructed composite calculation rule node matrix. It scans the primary attribute feature markers on the diagonal of this matrix and extracts core entity business control nodes with values ​​greater than or equal to 2.0 and more than 5 successor nodes based on the node priority value distribution. A total of 120 key nodes directly controlling the core data flow and possessing high priority are selected. These 120 core entity business control nodes are traversed, and the trigger time variable type and retry count type declared in the node metadata are read. Simultaneously, the corresponding control parameter hierarchical data types are extracted from the scheduling instruction parameter set. A bidirectional comparison verification is performed to detect data type differences between the data structures required by the core business nodes and the external scheduling parameter structures. 45 nodes are found to have incompatible type definitions. Data structure format conversion and alignment operations are performed on these 45 differences. String-type timing expressions are converted to long integer timestamp ranges in real time using a date parsing library, and text-type numbers are converted to corresponding short integer values. After precise type casting and alignment, all type definition conflicts are eliminated, and the input ports of business logic nodes and the output ports of scheduling parameters are merged to obtain a unified format logic matrix of 120 rows with completely consistent data types and format alignment. Table 1, the data alignment parameter table for the core entity business control node, shows the data alignment processing records of this example.

[0031] Table 1 Data Alignment Parameters for Core Entity Business Control Nodes As shown in Table 1, by mapping the required format of each core control node input terminal to the original scheduling parameters, alignment at the absolute numerical level has been achieved.

[0032] S413: Based on a unified format logical matrix, the core entity business control nodes and control parameter hierarchical framework are structured and integrated. The structure mapping result items are written to the blank physical table, the underlying real physical addressing path of the physical table is supplemented, the corresponding relationship of the entire process of business data flow is solidified, and a general indicator materialization architecture is established. Extract the transformation result data from the unified format logical matrix. Using a data modeling engine, structurally integrate the unified formatted scheduling execution rules with the underlying computational logic instruction execution structure of 120 core entity business control nodes. Request allocation of an independent storage space in the underlying relational database of the data platform, and request the creation and disk creation of a blank physical table. Read the integrated structure mapping result items and write the integrated instruction data into the blank physical table using standard batch insertion. Read the current block distribution status table of the distributed file system, calculate and supplement the underlying true absolute physical addressing path of the blank physical table on the data node (e.g., / var / data / hadoop / hdfs / blocks / blk_1001). Reverse write the physical addressing path information and bind it to the console configuration center, solidifying the entire process of business data flow from triggering computation to result disk creation. Finally, a general indicator materialization architecture with both scheduling control and underlying execution capabilities is established within the platform system.

[0033] Please see Figure 6 The S5 steps are as follows: S511: Obtain the snapshot node and window start and end node in the local SQL query engine, combine the general indicator materialization architecture, extract the underlying storage addressing path, map the snapshot node and window start and end node to the underlying storage addressing path according to the time alignment dimension, establish the time-series state correspondence, and generate the time-series mapping branch set. By remotely calling the service port of the local structured query language query engine, the abstract syntax tree of the execution plan inside the engine is intercepted. The snapshot node time parameters used to record historical data time slices (e.g., the data snapshot boundary at midnight on March 18, 2025) are read from the execution plan, and the window start and end node parameters defined in streaming computing and micro-batch processing (e.g., a rolling window boundary with a window length of 60 minutes and a step size of 10 minutes) are extracted simultaneously. The metadata center of the previously built general indicator materialization architecture is invoked, and the business target table name involved in the snapshot node is input to extract the underlying storage absolute addressing path information (e.g., / var / data / hadoop / hdfs / partitions / ). A time-aligned coordinate system is established, and the time slice parameters of the snapshot node and the time boundary parameters of the window start and end nodes are standardized and preprocessed, uniformly converted into long integer timestamps in milliseconds. These long integer timestamps are directly mapped to the partition field filtering conditions within the extracted underlying storage addressing path according to the time alignment dimension. By forcibly injecting the time parameters into the path partition query conditions, a temporal correspondence between the snapshot time and window time and the physical location of the underlying data blocks is established. Exclude data block loading instructions outside the time range and generate a time-series mapping branch set containing only the data required for the current time window and slice.

[0034] S512: Based on the time-series mapping branch set, it parses the related business scenarios within the underlying storage addressing path, locates and extracts the grouping clause set and aggregation field set under the mapping branch, and performs assembly and combination operations on the grouping clause set and aggregation field set according to the business association dimension to obtain the grouping aggregation query framework. The process involves parsing the time-series mapping branch set residing in memory and reading the execution context information bound to each branch. The Uniform Resource Identifier (URI) parameters within the context information are extracted, matched against the built-in business scenario dictionary, and the associated business scenarios within the underlying storage addressing path are parsed. For clearly defined business scenario characteristics, the set of grouping clauses and aggregation fields generated by the query engine's execution plan for that mapping branch are located. The extracted grouping clause set is used as the main logic framework, and the aggregation field set is used as the attached leaf nodes. Assembly and combination operations are performed according to the business association dimension rules defined in the business scenario dictionary. The compatibility between the aggregation logic and the grouping granularity is determined to ensure that each output of the grouping clause receives the correct measurement result corresponding to the aggregation field. The main logic and leaf nodes are merged to obtain a structurally complete and syntactically compliant grouping aggregation query framework. Table 2, the detailed breakdown of the grouping aggregation query architecture, lists the data acquisition process results for the corresponding business scenarios.

[0035] Table 2. Detailed Analysis of Grouping Aggregation Query Architecture As shown in Table 2, the parsed grouping aggregation query framework accurately integrates the grouping requirements and calculated fields under different business scenarios, and the compatibility scores are all at the highest level.

[0036] S513: Based on the grouping and aggregation query framework, read the internal multi-dimensional analysis hierarchy, combine the snapshot node to trigger the multi-indicator horizontal expansion instruction, perform the same-granularity horizontal extension and splicing operation on the aggregation field set according to the grouping clause set, summarize the full table header indicator dimension attributes, and obtain the data processing results of multi-indicator horizontal expansion. The system reads the established grouping and aggregation query framework, sends a request to the data warehouse's multidimensional analysis engine, and reads the drill-up and drill-down hierarchy tree defined in the internal multidimensional analysis hierarchy. It captures the trigger flags for multi-metric horizontal expansion instructions attached to snapshot nodes. The data processing pipeline is started, performing horizontal expansion and concatenation operations on different metrics defined within the aggregation field set (such as different aggregation results like order volume, refund volume, and repeat purchase volume under the same dimension) based on the business granularity defined by the grouping clause set. In the result set matrix, the uniqueness of row dimension grouping identifiers is maintained, and the result values ​​of different calculated metrics are appended column-wise, summarizing all header metric dimension attributes. During the concatenation execution, intermediate calculation results from multi-source heterogeneous data streams, such as real-time streaming data shards and offline data warehouse partitions, are integrated to eliminate barriers caused by data silos. The final metric values ​​are calculated and written, obtaining the data processing results of multi-metric horizontal expansion.

[0037] A data processing system with multi-indicator horizontal scaling includes: The business wide table integration module is used to execute S1: collect timestamp columns and dimension attribute columns in the enterprise data warehouse, perform field assembly operations on the timestamp columns and dimension attribute columns, and obtain the underlying business detail wide table; The dimension constraint decoupling module is used to execute S2: read the preset measurement field set, match and filter the non-empty fields in the preset measurement field set with all columns in the underlying business detail wide table, and generate independent indicator definition logic; The composite indicator assembly module is used to execute S3: read the built-in logic operator library of the analysis front end, extract the conditional filtering expression, perform a full sorting combination mapping between the conditional filtering expression and the business attributes in the independent indicator definition logic, remove the table structure disk statements, retain the business rule association instructions and format them as rule nodes, and assemble all rule nodes according to the logical dependency order to obtain the composite calculation rule node matrix. The script materialization scheduling module is used to execute S4: read the automated scheduling instruction list in the data middle platform, perform data structure format mapping and conversion processing with the rule nodes in the composite calculation rule node matrix, and establish a general indicator materialization architecture; The engine view building module is used to execute S5: obtain snapshot nodes and window start and end nodes within the SQL query engine, perform condition mapping within the general indicator materialization architecture, extract the grouping clause set and aggregation field set under the corresponding mapping branch, and obtain the data processing results of horizontal expansion of multiple indicators.

[0038] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A data processing method with multi-indicator horizontal expansion, characterized in that, Includes the following steps: S1: Collect timestamp columns and dimension attribute columns from the enterprise data warehouse, perform field assembly operations on the timestamp columns and dimension attribute columns, and obtain the underlying business detail wide table; S2: Read the preset measurement field set, match and filter the non-empty fields in the preset measurement field set with all columns in the underlying business detail wide table, and generate independent indicator definition logic; S3: Read the built-in logic operator library of the analysis front end, extract the conditional filtering expression, perform a full sorting combination mapping between the conditional filtering expression and the business attributes in the definition logic of the independent indicator, remove the table structure disk statement, retain the business rule association instruction and format it as a rule node, and splice all the rule nodes according to the logical dependency order to obtain the composite calculation rule node matrix. S4: Read the automated scheduling instruction list in the data platform and perform data structure format mapping and conversion processing with the rule nodes in the composite calculation rule node matrix to establish a general indicator materialization architecture; S5: Obtain the snapshot node and window start and end node in the SQL query engine, perform condition mapping in the general indicator materialization architecture, extract the grouping clause set and aggregation field set under the corresponding mapping branch, and obtain the data processing results of horizontal expansion of multiple indicators.

2. The data processing method for horizontal expansion of multiple indicators according to claim 1, characterized in that: The underlying business detail wide table includes a basic transaction record set, an entity feature matrix, and a time-series state snapshot. The independent indicator definition logic includes a basic statistical algorithm, a single-dimensional evaluation formula, and an original value quantification script. The composite calculation rule node matrix includes a multi-dimensional derived topology graph, a cascaded operation grid, and a dependency derivation flow graph. The general indicator materialization architecture includes a physical storage engine model, a task orchestration skeleton, and a computing resource allocation template. The data processing results of the multi-indicator horizontal expansion include a periodic summary set, a grouping clause set, and an aggregation field set.

3. The data processing method for horizontal expansion of multiple indicators according to claim 1, characterized in that, The specific steps for obtaining the underlying business detail wide table are as follows: S111: Collect sample timestamp columns and dimension attribute columns from the enterprise data warehouse, extract discrete time nodes in the timestamp column, extract business feature characters in the dimension attribute column, perform field assembly operations on the timestamp column and the dimension attribute column, merge discrete time nodes and business feature characters, concatenate same-origin attribute entries, and generate an initial assembled character matrix. S112: Based on the initial assembled character matrix, parse the primary key identifier and related foreign key identifier inside the matrix, compare the primary key identifier and related foreign key identifier, filter out target items that are completely consistent, perform vertical alignment operation based on the timestamp column involved in the target item and the dimension attribute column, arrange the feature items corresponding to the same key value, and obtain the vertical alignment feature spectrum. S113: Based on the vertical alignment feature spectrum, identify the internal horizontal correspondence, perform horizontal splicing and association operations on each business field with horizontal correspondence, integrate multi-source dimensional attributes and temporal evolution status, reshape the underlying entity architecture, and obtain the underlying business detail wide table.

4. The data processing method for horizontal expansion of multiple indicators according to claim 1, characterized in that, The specific steps for obtaining the definition logic of the independent indicator are as follows: S211: Collect a preset set of measurement fields in local storage, extract non-empty fields in the preset set of measurement fields, obtain the full list header attribute names in the underlying business detail wide table, perform string sequence verification operation on non-empty fields and full list header attribute names, filter target items with completely identical characters, aggregate target items, and establish a set of feature names with the same name. S212: Based on the same feature name set, locate the corresponding column storage type in the underlying business detail wide table, extract the original definition type of the associated metric field in the preset metric field set, compare the corresponding column storage type with the original definition type of the associated metric field, remove columns with mismatched types, retain the same name and type of metric attribute, and generate a list of matching attribute types. S213: Based on the list of matching attribute types, extract all the same-name and same-type measurement attributes, map each of the same-name and same-type measurement attributes to the blank indicator definition template, establish a topological network of association mapping between multi-dimensional attribute nodes, and obtain independent indicator definition logic.

5. The data processing method for horizontal expansion of multiple indicators according to claim 1, characterized in that, The specific steps for obtaining the composite calculation rule node matrix are as follows: S311: Obtain the built-in logic operator library, extract the internal conditional filtering expression, perform full-out combination mapping with the business attributes in the independent indicator definition logic, establish a mapping relationship, and generate a conditional attribute mapping set; S312: Based on the conditional attribute mapping set, parse the internal statement type, compare the statement type with the preset disk write identifier, remove the table structure disk write statement, retain the business related instructions and perform format conversion operation, unify the instruction architecture standard, and obtain standardized related nodes. S313: Based on the standardized associated nodes, calculate the topology priority value of each node, analyze the dependency hierarchy between nodes, extract the hierarchy order, perform an ordered splicing operation on all standardized associated nodes according to the order, construct a coherent execution path topology, and establish a composite calculation rule node matrix.

6. The data processing method for horizontal expansion of multiple indicators according to claim 5, characterized in that, The formula for calculating the topology priority value is as follows: ； in, Representative node The topology priority value, Represents a pointer to a node The set consisting of all predecessor nodes, Represents the predecessor node The computational complexity factor, Represents the node To the node Dependency weights Representative node Its own adjustment parameters Representative node The number of successor nodes it points to. Represents the global scaling factor.

7. The data processing method for horizontal expansion of multiple indicators according to claim 1, characterized in that, The specific steps for obtaining the materialized architecture of the general indicator are as follows: S411: Obtain the automated scheduling instruction list within the data platform, extract the trigger cycle status and execution frequency identifier within the automated scheduling instruction list, perform a structure splitting and decoding operation on the trigger cycle status and execution frequency identifier, strip away the underlying control parameter level framework of the instruction, and generate a scheduling instruction parameter set; S412: Based on the scheduling instruction parameter set, extract the core entity business control nodes inside the composite calculation rule node matrix, compare the data type differences between the core entity business control nodes and the control parameter hierarchical framework, perform data structure format conversion and alignment operations, and obtain a unified format logical matrix. S413: Based on the unified format logic matrix, the core entity business control nodes and control parameter hierarchical framework are structurally integrated, the structure mapping result items are written to the blank physical table, the underlying real physical addressing path of the physical table is supplemented, the corresponding connection of the entire process of business data flow is solidified, and a general indicator materialization architecture is established.

8. The data processing method for horizontal expansion of multiple indicators according to claim 1, characterized in that, The specific steps for obtaining the data processing results of the multi-indicator horizontal expansion are as follows: S511: Obtain the snapshot node and window start and end node in the local SQL query engine, combine the general indicator materialization architecture, extract the underlying storage addressing path, map the snapshot node and window start and end node to the underlying storage addressing path according to the time alignment dimension, establish the time-series state correspondence, and generate a time-series mapping branch set. S512: Based on the time-series mapping branch set, analyze the associated business scenarios within the underlying storage addressing path, locate and extract the grouping clause set and aggregation field set under the mapping branch, and perform assembly and combination operations on the grouping clause set and aggregation field set according to the business association dimension to obtain the grouping aggregation query framework. S513: Based on the grouping and aggregation query framework, read the internal multi-dimensional analysis hierarchy, combine the snapshot node to trigger the multi-indicator horizontal expansion instruction, perform the same-granularity horizontal extension and splicing operation on the aggregation field set according to the grouping clause set, summarize the full table header indicator dimension attributes, and obtain the data processing results of the multi-indicator horizontal expansion.

9. A data processing system with multi-index horizontal scaling, characterized in that, The system is used to implement the data processing method for multi-indicator horizontal expansion as described in any one of claims 1-8, including: The business wide table integration module collects timestamp columns and dimension attribute columns from the enterprise data warehouse, performs field assembly operations on the timestamp columns and dimension attribute columns, and obtains the underlying business detail wide table. The dimension constraint decoupling module reads a preset set of measurement fields, matches and filters the non-empty fields in the preset set of measurement fields with all columns in the underlying business detail wide table, and generates independent indicator definition logic. The composite indicator assembly module reads the built-in logic operator library of the analysis front end, extracts the conditional filtering expression, performs a full-sort combination mapping between the conditional filtering expression and the business attributes in the definition logic of the independent indicator, removes the table structure disk statements, retains the business rule association instructions and formats them as rule nodes, and assembles all rule nodes according to the logical dependency order to obtain the composite calculation rule node matrix. The script materialization scheduling module reads the automated scheduling instruction list in the data middle platform, performs data structure format mapping and conversion processing with the rule nodes in the composite calculation rule node matrix, and establishes a general indicator materialization architecture. The engine view construction module obtains the snapshot node and window start and end node within the SQL query engine, performs condition mapping within the general indicator materialization architecture, extracts the grouping clause set and aggregation field set under the corresponding mapping branch, and obtains the data processing results of horizontal expansion of multiple indicators.

Images