15 results about "Layer (object-oriented design)" patented technology

The invention provides a software architecture governance system based on automatic modeling and holographic penetration analysis. The software architecture governance system comprises multiple modules: a project management module used for managing all analysis projects; the source code management module is used for pulling, updating and deleting project source codes; the analysis drive management module is used for uniformly managing and calling analysis drive programs; the multi-language analysis driving cluster is integrated with analysis driving programs corresponding to multiple programming languages and is responsible for analyzing source codes, generating a standardized software chart model and storing the standardized software chart model in a database; the rule engine module is used for supporting a complex code analysis service; the cross-project integrated analysis module generates a system architecture diagram according to the plurality of software models and stores the system architecture diagram into a diagram model library; the multi-dimensional data fusion interface module provides penetrating analysis interfaces of a code layer, a design layer and a framework layer; and the architecture model visualization module is responsible for displaying the system architecture in a multi-dimensional visualization manner. The method can be used for automatic modeling, tracing analysis and interactive visual presentation of a complex system architecture.

The invention discloses a code static review method and system based on a fine tuning large language model, and belongs to the technical field of software static testing. The method comprises the steps that a code static rule set is collected and preprocessed, standard rules are screened according to programming languages and project requirements, and grading classification and duplicate removal are carried out; the method comprises the following steps: constructing a positive and negative example program data set in three modes of open source code analysis, fault implantation and large language model generation, and integrating into structured training data in combination with a template; selecting a base model matched with a code reasoning task, optimizing model parameters by adopting a LoRA fine tuning method, and predicting code rule violation conditions through output layer design; and evaluating model performance based on the indexes. According to the method, the efficiency of static review is enhanced, including readability, friendly interaction and expandability of reports, false reports and missing reports are reduced by improving the ability of code context understanding, and cross-language and cross-platform portability is realized.

The application discloses a software development package compiling method and device, and relates to the technical field of computers. A specific implementation of the method comprises the following steps: performing encapsulation based on a second programming language on a first type of module of any core component in a software development package, extracting a header file of the first type of module and a header file of a second type of module after the encapsulation to generate a common umbrella header file of the core component; and after compiling core component source code and upper layer component source code in the software development package into an executable file, processing the common umbrella header file of any core component in the software development package, so that the common umbrella header file only contains declaration content configured in advance for the core component. The implementation can effectively reduce the risk of core code leakage.

The invention discloses an LLM software defect positioning and repairing method and system fusing multi-source information. The method comprises the following steps: firstly, performing automatic analysis and mining on multi-source information of software by adopting a unified mining framework; secondly, based on the LLM and the multi-source information obtained in the step 1, analyzing the failed test case to realize automatic positioning of software defects; and finally, automatically repairing the software defects in combination with the LLM, the multi-source information obtained by mining and the defect codes positioned in the step 2. The system comprises an analysis mining module, a software defect positioning module and a software defect repairing module, and is used for realizing the LLM software defect positioning and repairing method fusing the multi-source information. According to the method, the accuracy and the repairing effect of software defect positioning are improved, and the middle layer representation and the reserved interface of the multi-source information are designed, so that the method can be conveniently expanded to other languages.

A method develops a programming language model for machine learning tasks. The method includes adjusting a token list to include a language token used by a tokenizer for a pretrained language model. The pretrained language model includes a set of layers. The set of layers includes a set of initial layers, an embedding layer, and an output layer. The method further includes performing an output layer modification of the output layer to replace the output vector with the embedding vector. The method further includes freezing the set of initial layers to generate a set of frozen layers of the pretrained language model that do not update during training. The method further includes training the pretrained language model using the language token, the output layer modification, and the set of frozen layers to form a fine-tuned model from the pretrained language model.

A method for tuning a language model is provided. The method may be performed by a computing device, and may comprise: receiving, from a user, a coding requirement and a target programming language to be input into a first model, wherein the first model is a model trained to output a code snippet corresponding to the received coding requirement in the target programming language; determining whether the target programming language is in-domain or out-of-domain with respect to the first model; and updating pretrained parameters of the first model based on a result of the determination and a masking result of the pretrained parameters, wherein the updating the pretrained parameters comprises: updating pretrained parameters of all layers of the first model when the target programming language is in-domain; and updating pretrained parameters of a dense layer of the first model when the target programming language is out-of-domain.

The invention provides a multi-language cross-project system-oriented multi-layer architecture visualization model, which comprises a software element meta-model and a programming language exclusive domain model, and is characterized in that instantiation mapping is carried out on a specific programming language based on the software element meta-model to obtain the programming language exclusive domain model; the organization and expression mode of the multi-layer architecture visualization model comprises a link design diagram, an element organization diagram and a system architecture diagram; the multi-layer viewing view angle of the multi-layer architecture visualization model comprises a code layer, a link design and element organization layer and a system architecture layer; the applicable object of the multi-layer architecture visualization model is a cross-project composite software architecture comprising at least two programming languages and two independent code libraries. The model supports penetrating viewing of a calling link, element organization and an overall architecture in three progressive perspectives of a code layer, a link design and element organization layer and an architecture layer, and a standardized data base is provided for architecture analysis and visualization of a large-scale system.

The invention relates to a parallel programming unified interface architecture of a cross-platform heterogeneous system and computing equipment, and relates to the field of heterogeneous programming interfaces. The architecture comprises a language layer used for providing a standard grammar and an extended API based on a unified programming interface; the presentation layer is used for converting codes of the language layer into unified intermediate representations irrelevant to hardware; the scheduling layer is used for being responsible for the dependency relationship and the execution process of parallel tasks and supporting multiple scheduling paradigms; the scheduling strategy comprises a priority queue, batch submission and instant scheduling, and the scheduling strategy is used for allocating computing resources between different devices and tasks; managing dependency and an execution sequence among tasks by adopting an event chain mechanism; and the plug-in layer is used for being responsible for hardware back-end extension and cross-platform adaptation of the architecture. The computing device is loaded with the parallel programming unified interface architecture of the cross-platform heterogeneous system. According to the invention, the efficiency and universality of parallel computing are greatly improved.

The invention provides an IEC104 protocol processing method, system and device based on object-oriented program design, and the method comprises the steps: carrying out the objectification segmentation processing of a target IEC104 binary message, and obtaining a plurality of defined IEC104 protocol sub-objects; coding and calling are carried out based on the IEC104 protocol sub-objects, and information objects of the multiple IEC104 protocol sub-objects are obtained; the information object of the IEC104 protocol sub-object is integrated into a Netty network communication framework, and network communication is carried out; and constructing an intermediate service layer to package an IEC104 protocol and isolate the IEC104 protocol, accessing an IEC104 protocol sub-object through a service interface, and notifying an upper-layer application. According to the application, objectification-oriented program design is introduced, all parts agreed in the protocol are subjected to objectification splitting and combination, and program coding is carried out based on the object, so that the software development efficiency of the protocol can be remarkably improved; on the network communication level, the system is combined with a mature netty network framework, and landing implementation of the system is accelerated.

The application provides an optimization method for fault code (DTC) storage and readback, comprising an application software layer, a basic software layer and an intermediate software component, the basic software layer and the intermediate software component are arranged in a first code partition, the intermediate software component is used for receiving an instruction issued by the application software layer, then the intermediate software component calls a corresponding function of the basic software layer according to the instruction and returns an execution result to the application software layer. The application introduces the intermediate software component as an intermediate medium of the application software layer and the basic software layer, when a fault occurs, the application software layer only needs to send a detected fault state to the intermediate software component, the intermediate software component uniformly calls the function, and cross-code partition calling of the function is avoided. Meanwhile, after an interactive interface of the application software layer and the intermediate software component is determined, the application software layer can be decoupled from the basic software layer in functional development, and development work can be carried out in parallel, and development efficiency is greatly improved.

The invention provides a cross-language vulnerability detection method and system based on hierarchical heterogeneous modeling, and belongs to the technical field of software security detection. In order to solve the technical problem that an existing single-language vulnerability detection method cannot effectively process a multi-programming language collaborative development software project, a three-layer progressive modeling strategy is adopted, specifically, a language specific layer keeps independent boundary features, a language family layer achieves intra-family knowledge sharing through soft boundary partitioning, and a basic semantic layer captures a cross-language universal vulnerability mode; performing cross-language semantic alignment in combination with unified representation learning and a CWE standard; end-to-end detection is realized through a pre-training encoder, a cross-attention mechanism and a language family adapter network. The method has the advantages that the overall F1 score of 88.2% is achieved on eight programming languages and 82 vulnerability types, the problem of sample imbalance can be effectively solved, and a unified security detection solution is provided for modern multi-language software projects.

The invention discloses a cross-normal-form automatic conversion method and system for parameterized geometric modeling scripts and a medium, and provides a conversion architecture based on double-layer abstract syntax tree (AST) mapping and state stack management for a semantic gap between an implicit structure entity geometry (CSG) description language (such as OpenSCAD) and an explicit boundary representation (B-rep) programming language (such as CadQuery / Python). Deducing logic from an implicit functional list in the source script by adopting a recursive structure mapping mechanism, decoupling and reconstructing into an explicit implicit circulating control flow in a target language; a statement container stack is used for solving the isolation and recombination problems of hierarchical action domains; and identifying a specific geometric combination (such as Minkowski and cubes and spheres in operation) through semantic analysis, and reconstructing the specific geometric combination into efficient function call based on feature modeling (such as chamfer). According to the method, the parameterization design intention of the original script is reserved, and the calculation efficiency of the generative model in the B-rep kernel is remarkably improved.

On the basis of the design idea of DDD software, a four-layer architecture serves as a software design framework, a Modelica modeling tool serves as a professional universal language basis, and the software design method and system for achieving field drive design based on Modelica are provided. The core of the method is that software service items are classified according to domain functions and layered according to a DDD architecture, corresponding function modules and units are arranged in a domain layer and an application layer respectively, a Modelica language is applied to establish a domain model and an application model, and finally the models are converted into C / C + + application codes. The method can be used for reconstruction of an existing three-layer architecture-based software system and can also be used for design and development of a brand-new software system, so that basic codes or modules in a field layer can be adapted to wider reuse of an application layer, complex and changeable field requirements can be met, and the application layer can be widely applied to the field layer. The problems that a software system is bloated and inflexible to use due to repeated configuration are solved, and the operation efficiency of software and the efficiency of software design work can be improved.

The application belongs to the technical field of software development, and particularly relates to a software development system and a software development method. The BERT-base-based semantic understanding model can accurately extract function parameters, performance indexes and constraint conditions in natural language requirements, a three-layer verification mechanism of a technical entity design rule set is combined to ensure requirement compliance and integrity, priority calculation of the analytic hierarchy process makes development resources focus on core requirements, greatly reduces requirement deviation rate, through requirement hierarchical decomposition and intelligent matching of a component template library, combined with deep optimization of a technical entity design model, including design pattern adaptation, interface standardization and performance strategy customization, the generated core function components, auxiliary function components and interface components not only meet individualized requirements, but also meet performance requirements, and component reuse rate and development efficiency are significantly improved.

The embodiment of the invention provides a cross-language data scheduling method and device, equipment, a medium and a program product, which can relate to the computer technology, and the method comprises the following steps: converting target data under a second programming language layer into a target pointer under a third programming language layer through a first programming language layer; packaging the target pointer through the first programming language layer to obtain recognizable data under the second programming language layer; returning identifiable data to the second programming language layer through the first programming language layer; transmitting identifiable data to the first programming language layer through the second programming language layer under the condition that the target data is called by the third programming language layer; decapsulating the recognizable data through the first programming language layer to obtain a target pointer; transmitting a target pointer to a third programming language layer through the first programming language layer; and under the third programming language layer, calling the target data based on the target pointer. The cross-language data calling difficulty is reduced.