CN-122019334-A - Method, system, equipment and medium for incremental verification of low-code meta-model

CN122019334ACN 122019334 ACN122019334 ACN 122019334ACN-122019334-A

Abstract

The invention discloses a method, a system, equipment and a medium for incremental verification of a low-code meta-model, and belongs to the technical field of low-code application verification. The system comprises a graph structure construction module, an index construction module, an increment sub-graph generation module, a characteristic reasoning mapping module and a verification analysis module. According to the invention, by constructing the physical and virtual semantic fusion graph and extracting the incremental subgraph, the accurate positioning and the quick verification of the change area are realized, and the invalid full-quantity calculation is avoided. Meanwhile, by utilizing the structural feature multiplexing and task perception attention mechanism of the graphic neural network, the reasoning expense is obviously reduced, the depth detection of the interface structure and the business logic can be adaptively considered, and the development efficiency and the stability of the low-code application are greatly improved.

Inventors

MA YANJIE
SHU YU
BAO CHENYI
YI YE
JI YUAN
LU XIUCHANG
QIAN JUNFENG
YANG JIAN
Yang Erman

Assignees

贵州电网有限责任公司

Dates

Publication Date: 20260512
Application Date: 20251222

Claims (10)

1. A method for incremental verification of a low-code meta-model is characterized by comprising the following steps of, Analyzing a low-code application model, and constructing graph structure data for representing component relationships, wherein the graph structure data comprises physical connection relationships reflecting component physical hierarchical relationships and virtual semantic connection relationships constructed based on semantic dependency relationships among components; Constructing a metadata dependent index based on the virtual semantic connection relationship; Responding to the change operation of at least one component in the low-code application model, acquiring an inverse dependency closure associated with the dependency of the at least one component based on the metadata dependency index, extracting a neighborhood context by combining the physical connection relation, and generating an increment to-be-verified subgraph limiting a verification range; Based on the increment to-be-verified subgraphs, aggregating component nodes with equivalent structures, selecting a representative node input graph inference model, generating a high-dimensional feature vector representing service logic, and mapping the high-dimensional feature vector to the component nodes with equivalent structures; and calculating the similarity between the high-dimensional feature vector and a preset defect feature vector, outputting a verification conclusion based on a similarity calculation result, and associating a repair strategy.
2. A method of incremental verification of a low code metamodel according to claim 1, wherein: the construction of graph structure data for characterizing component relationships includes: Defining a node type range of the graph structure data, the node type range including a multi-level metadata entity constituting a low-code application, the multi-level including presentation layer elements and business function elements; Extracting a hierarchical nested structure of the presentation layer element, and directly mapping the hierarchical nested structure into the physical connection relation; Analyzing the configuration attribute of the presentation layer element, extracting the reference association pointing to the business function element in the configuration attribute, ignoring the topological distance in the physical hierarchy structure, and directly establishing directed connection between the two as the virtual semantic connection relation.
3. A method of incremental verification of a low code metamodel according to claim 1, wherein: The generating the incremental to-be-verified subgraph defining the verification range comprises: Configuring the metadata dependency index as a reverse mapping structure for recording the reference relation; retrieving components for which explicit attribute binding exists for the at least one component and components for which implicit data flow dependencies exist based on the reverse mapping structure to form the reverse dependent closure, and defining components in the reverse dependent closure as core nodes; expanding at least a first-order physical neighborhood by taking the core node as a center based on the physical connection relation, and taking a component in the physical neighborhood as a neighborhood context; And intercepting a local topological structure formed by the core node, the neighborhood context and the connection relation connecting the core node and the neighborhood context as the increment sub-graph to be verified.
4. A method of incremental verification of a low code metamodel according to claim 1, wherein: the aggregate structure equivalent component node comprises: Extracting component attribute features and topological connection features of the constituent nodes contained in the partial increment to-be-verified sub-graph model, and serializing the component attribute features and the topological connection features into structural feature fingerprints of unique identification nodes; Dividing the constituent nodes based on the structural feature fingerprints to construct at least one topological equivalence set, and determining the representative nodes from the topological equivalence set; after the high-dimensional feature vector of the representative node is obtained through the graph inference model, feature multiplexing mapping, pointed to other nodes in the topological equivalence set, by the representative node is established, so that batch feature updating of the topological equivalence set is completed.
5. A method of incremental verification of a low code metamodel according to claim 1, wherein: the generating the high-dimensional feature vector for characterizing the business logic comprises: Analyzing the verification task instruction to identify a target dimension which is focused on by the current verification task, wherein the target dimension covers an interface structure dimension and a business logic dimension; The weight parameters of the task perception attention mechanism integrated by the graph inference model are configured based on the target dimension differentiation; Responding to the target dimension as the interface structure dimension, and improving the aggregation weight ratio of the physical connection relation to exceed the virtual semantic connection relation; and in response to the target dimension being the service logic dimension, increasing the aggregation weight ratio of the virtual semantic connection relationship to exceed the physical connection relationship.
6. A method of incremental verification of a low code metamodel according to claim 1, wherein: and calculating the similarity between the high-dimensional feature vector and a preset defect feature vector, outputting a verification conclusion based on a similarity calculation result, and associating a repair strategy comprises: Constructing a defect mode library, wherein the defect mode library records the corresponding relation between the defect characteristic vector and defect type identification and repair suggestion; calculating similarity values of the high-dimensional feature vectors and the defect feature vectors in the defect mode library; Screening target defect feature vectors with similarity degree values exceeding a preset judging threshold, and determining the defect type identifiers with corresponding relations with the target defect feature vectors as verification conclusion; And extracting the repair suggestion based on the corresponding relation, and outputting a repair strategy.
7. The method for incremental verification of a low-code metamodel of claim 1, further comprising the training step of constructing the graph inference model: collecting a historical low-code application model to construct a sample graph dataset; Performing random image data enhancement transformation on the sample images in the sample image data set, generating an enhancement view with consistent semantics, and forming positive sample pairs by the enhancement view and the sample images; randomly sampling non-topologically related heterogeneous nodes from the sample graph dataset to form negative sample pairs; And optimizing parameters of the graph inference model based on a contrast learning objective function, driving the graph inference model to map the feature vectors of the positive sample pairs to an adjacent space, and mapping the feature vectors of the negative sample pairs to a distant space.
8. A system for incremental verification of a low code metamodel, applying a method of incremental verification of a low code metamodel according to any one of claims 1 to 7, comprising: The diagram structure construction module is used for analyzing the low-code application model and constructing diagram structure data for representing the component relation, wherein the diagram structure data comprises physical connection relations reflecting the component physical hierarchy relation and virtual semantic connection relations constructed based on the semantic dependency relation among the components; the index construction module is used for constructing a metadata dependent index based on the virtual semantic connection relation; The incremental sub-graph generation module is used for responding to the change operation of at least one component in the low-code application model, acquiring an inverse dependency closure with dependency association with the at least one component based on the metadata dependency index, extracting a neighborhood context in combination with the physical connection relation, and generating an incremental sub-graph to be verified, wherein the incremental sub-graph defines a verification range; The feature reasoning mapping module is used for aggregating the component nodes with equivalent structures based on the increment to-be-verified subgraphs, selecting a representative node input graph reasoning model, generating a high-dimensional feature vector representing service logic, and mapping the high-dimensional feature vector to the component nodes with equivalent structures; The verification analysis module is used for calculating the similarity between the high-dimensional feature vector and a preset defect feature vector, outputting a verification conclusion based on a similarity calculation result and associating a repair strategy; The model training module is used for collecting a historical low-code application model to construct a sample graph data set.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of a method of low code metamodel incremental verification according to any one of claims 1 to 7.
10. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program when executed by a processor implements the steps of a method of incremental verification of a low code metamodel according to any one of claims 1 to 7.

Description

Method, system, equipment and medium for incremental verification of low-code meta-model Technical Field The invention relates to the technical field of low-code application verification, in particular to a method, a system, equipment and a medium for incremental verification of a low-code meta-model. Background With acceleration of enterprise digital transformation, a low-code development platform has become an important tool for constructing enterprise-level applications by virtue of visual drag interaction and model-driven development concepts thereof. In a typical low code platform architecture, application pages, business logic and data models are generally abstracted into metadata and stored and run in the form of a massive JSON tree structure or abstract syntax tree. To ensure the correctness of the application, the platform typically incorporates a static checker or verification engine that checks the user's configuration in real-time in the design state, such as checking for variable naming conflicts, component attribute type matching, and referential integrity. However, as low code applications evolve toward large-scale core business systems such as ERP, CRM, etc., application complexity is rising exponentially, and currently, the challenges facing developers have shifted from single grammar correctness to complex logical consistency and architecture compliance. When facing deep business logic verification, the existing verification technology based on tree structures and hard coding rules exposes some defects: First, the full-scale traversal mechanism has performance bottleneck in deep dependency analysis, so that currently, the existing engine is based on a tree traversal algorithm of O (N) complexity, such as a common DFS/BFS traversal algorithm, and although processing single-point grammar checking is feasible, the efficiency of analyzing cross-page and cross-component complex dependency chains is extremely low, and in large-scale application of thousands of components, some tiny logic changes can trigger global dependency recalculation, invalid computation is easy to increase, operation delay is caused at a browser end, and development efficiency is affected. Second, the lack of structured deduplication machines results in wasted complex logic inference computation effort, there are a large number of structurally equivalent components in low code applications, such as list cells, for simple regular verification, repeated computation is acceptable, but for complex logic reasoning involving context awareness, performing equal overhead computation on each repeated instance results in significant wasted computation effort, failing to take advantage of the optimization space brought by structural repeatability. Third, physical isolation results in cross-plane semantic dead zones, failing to identify hidden risks. The UI, data and logic in the traditional metadata are physically separated, and the existing tool can only verify whether the references of the grammar layer exist through ID matching, but cannot understand whether the references of the semantic layer are reasonable, for example, the high-density fields are bound to the public components, or the non-batch interface is wrongly referred in the loop logic. The above operations are fully syntactically compliant, but conventional tools are unable to identify such compliant but erroneous implicit business risks. Fourth, hard-coded rules are difficult to exhaust from non-deterministic architectural flaws, traditional engines rely on predefined If-Then rules, however complex logic flaws tend to exhibit non-deterministic topological features such as dead-loops caused by specific combinations of data flow directions. The dead-end rule engine is difficult to exhaust all variants, lacking heuristic reasoning ability based on pattern recognition by similarly senior engineers for CodeReview. In summary, although the prior art can effectively solve the basic grammar checking, there is a significant short board in the aspect of coping with deep semantic understanding and complex architecture defect recognition, and a new architecture capable of having incremental reasoning and semantic perception capabilities is needed. Disclosure of Invention In view of the foregoing, the present invention provides a method, system, apparatus, and medium for incremental verification of low-code metamodel. Therefore, the invention solves the technical problems that the conventional low-code verification is dependent on full static scanning, frequent change of applications cannot be realized efficiently, implicit business logic dependence among components is difficult to capture, verification efficiency is low, omission ratio is high, and real-time feedback requirements of agile delivery cannot be met. The invention provides a method for incremental verification of a low-code meta-model, which comprises the following steps of analyzing a low-code application model, const