CN115756477A - Low-code integrated development system based on multi-path intelligent recommendation - Google Patents
Low-code integrated development system based on multi-path intelligent recommendation Download PDFInfo
- Publication number
- CN115756477A CN115756477A CN202211371028.7A CN202211371028A CN115756477A CN 115756477 A CN115756477 A CN 115756477A CN 202211371028 A CN202211371028 A CN 202211371028A CN 115756477 A CN115756477 A CN 115756477A
- Authority
- CN
- China
- Prior art keywords
- code
- module
- code module
- information
- development
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Stored Programmes (AREA)
Abstract
The invention discloses a low-code integrated development system based on multi-path intelligent recommendation, which stores a code module through a code module warehouse subsystem, and records quality information, version information and user permission information of the code module, thereby reducing the dependence of the system on a service scene and improving the availability and flexibility of the system. The modification and updating of the code module are realized through the low-code development subsystem; the code module is retrieved and recommended for the low-code development subsystem through the multi-channel intelligent recommendation algorithm of the code module warehouse subsystem, the user-defined code modules can be selectively eliminated, hierarchical clustering of the code modules is achieved, retrieval and recommendation of the code modules are conducted according to the quality level coefficient, and the code modules with higher quality can be provided for users. The process tracking and quality assurance document is established by the automatic process management subsystem, so that the quality assurance of the self-defined code module is improved, and the process management overhead caused by introducing a low-code development platform is reduced.
Description
Technical Field
The invention relates to the technical field of low-code platforms, in particular to a low-code integrated development system based on multi-path intelligent recommendation.
Background
The Low-Code Development Platform (LCDP) is Platform software mainly written in graphical interfaces and configurations, and can effectively reduce the amount of traditional Code and accelerate the completion time of business application software. Although the low-code development platform can effectively reduce the program development period and the development cost in most application scenarios, the following three limitations still exist:
(1) The existing low-code development platform is oriented to a certain service scene, code multiplexing is highly dependent on platform predefinition, namely, a code multiplexing module is mainly predefined by a platform software designer, a user-defined code segment/code module cannot be shared among platform users, and the multiplexing of the user-defined code segment/code module plays a vital role in the usability and flexibility of the low-code development platform.
(2) The low-code development platform lacks effective treatment on a custom module, an auditing and evaluating method for code quality is lacked in the development process, the factors such as the effectiveness, the availability, the code normalization and the version information of a custom multiplexing code segment are not or rarely effectively controlled, and the instability of the code quality limits the wide application of the low-code development platform in different service scenes.
(3) When the existing low-code development platform is used for multi-user collaborative development of a complex system, effective project management is difficult to achieve, and when a process management tool provided by the platform is used for development of the complex system, an automatic full-life-cycle control and management scheme is difficult to provide, so that higher process management cost can be brought while the low-code development platform is introduced.
Disclosure of Invention
In view of this, the invention provides a low-code integrated development system based on multi-channel intelligent recommendation, which can efficiently and flexibly create and modify code modules and realize continuous multiplexing of the code modules.
The invention adopts the following specific technical scheme:
a low-code integrated development system based on multi-path intelligent recommendation comprises: a low code development subsystem, a code module warehouse subsystem, and an automation process management subsystem;
the low-code development subsystem is used for creating and updating a code module and storing the code module to the code module warehouse subsystem;
the code module warehouse subsystem is used for storing the code module, including the quality information, the version information and the user authority information of the code module, and retrieving a recommended code module for the low code development subsystem by adopting a multi-path intelligent recommendation algorithm;
the automatic process management subsystem is used for automatically arranging the development process according to the service flow information and the development organization structure information from the low-code development subsystem and generating a module-level process tracking and quality assurance document according to the code development process of the low-code development subsystem.
Further, the low code development subsystem includes: the system comprises a modular process definition unit, a module definition and automation filling unit, a modular code development unit, a code module unit test unit and a system packaging deployment unit;
the modular process definition unit is used for defining a system service process and a development organization structure and sending the service process information and the development organization structure information to the automatic process management subsystem;
the module definition and automation filling unit is used for coding and mapping the code modules, and each code module has a unique identification number; the system is used for automatically initializing module interface documents according to the interface configuration information and automatically filling test samples;
the modular code development unit is used for developing and realizing the creation and the update of the code module;
the code module unit testing unit is used for testing the created and updated code module, generating quality information of the code module according to a testing result and integrating the quality information on the version information of the code module, wherein the testing is carried out through a pre-configured module interface test case or an automatic testing tool;
and the system packaging and deploying unit is used for generating the running environment configuration of the code module according to the independent service environment of the code module and the public running environment of the code module.
Further, in the modular code development unit, the code module is created and updated in a visible module migration, multi-path intelligent recommendation and manual creation mode of the code module warehouse subsystem, or rapid system development is performed based on the code module.
Further, the code module repository subsystem comprises: the system comprises a code warehouse construction unit and a multi-path intelligent recommendation unit;
the code warehouse building unit is used for storing the code modules and sending the statistical information of the code modules to the multi-path intelligent recommending unit, wherein the code modules comprise code modules pre-built by a platform and code modules created or modified by a user;
the multi-path intelligent recommendation unit is used for retrieving the code module in the code warehouse construction unit according to the requirement retrieval information of the user and combining the statistical information, and providing the code module for the modular code development unit of the low-code development subsystem to directly use or modify and update.
Further, the code repository constructing unit includes: the code module creates and modifies the subunit, code module authority control subunit and version tracking control subunit;
the code module creating and modifying subunit is used for storing the pre-constructed code module and the code module created or modified by the user;
the code module authority control subunit is used for carrying out authority control on users of different code modules, wherein the authority control at least comprises all users visible, only organization visible, only item visible and only self visible;
the version tracking control subunit is used for automatically recording code module configuration information, version change information and quality information aiming at each code module and carrying out intra-authority domain control on different code modules;
and the authority control information of the code module authority control subunit and the code module configuration information, the version change information and the quality information of the version tracking control subunit form statistical information of the code module.
Further, the multi-path intelligent recommendation unit searches the code module in the code warehouse building unit according to the requirement retrieval information of the user and by combining the statistical information, and the method comprises the following steps:
s1, constructing a keyword vector for each code module and calculating a module vector of the code module;
s2, selecting module vectors of k code modules as a first cluster center, and recording an initial cluster classification level S as 1;
s3, traversing other code modules except the code module corresponding to the first cluster center, classifying the current code module to the cluster where the first cluster center is located and with the smallest vector cosine distance with the current code module, recording the result as a real center classification result, taking the mean value of the S-th keyword vectors of all code modules corresponding to the cluster where the first cluster center is located as the mean value vector of the cluster, taking the mean value vector as the second cluster center, re-classifying all code modules to the cluster where the second cluster center is located and with the smallest vector cosine distance with the corresponding code module, and recording the result as an estimated center classification result;
s4, aiming at the real center classification result and the estimated center classification result obtained in the step S3, if the classification of a code module changes, the code module with the changed classification is used as a core path set, the mean value of the S-th keyword vectors corresponding to the module vectors of all the code modules in the core path set is calculated, and the mean value is correspondingly added to the second cluster center; aiming at the second cluster center, finding the module vector of the code module with the minimum cosine distance corresponding to the second cluster center, taking the module vector as a new first cluster center, repeating the steps S3 and S4 until no classification of the code module is changed, and ending the circulation;
step S5, aiming at the m clusters obtained in the step S4, if the number of the keywords of the module vector of the code module corresponding to the first cluster center is equal to the cluster classification level S, taking the S-th keyword of the code module corresponding to the first cluster center at the moment as the classification path of the code module of the cluster, and finishing the layering of the cluster; if the number of the key words of the module vectors of the code modules corresponding to the first cluster center is smaller than the classification level S of the cluster, adding one to the classification level of the cluster, namely S +1, randomly selecting the module vectors of k code modules for the cluster as the first cluster center of the sub-cluster of the cluster, repeating the steps S3, S4 and S5, ending the circulation when the number of the key words of the module vectors of the code modules corresponding to all the first cluster centers is equal to the classification level S of the cluster, and obtaining the classification path tree of the code modules according to the hierarchy relationship of the cluster;
s6, calculating a quality level coefficient of each code module according to the statistical information and the code use information of each code module, and performing quality level sequencing on the code modules in each cluster according to the quality level coefficient;
and S7, retrieving information according to the requirements of the user, determining a classification path hierarchy corresponding to the key words input by the user, and taking a code module at the beginning of the hierarchy of the classification path as an initial recommendation code module.
Further, in step S5, the method further includes: and updating the classification path tree periodically, including updating according to the statistical information of the code module or updating according to a new keyword vector of the code module.
Further, in step S6, the code usage information at least includes: the calling/enabling frequency of the code module, the description correlation of the code module, the modification calling frequency of the code module and the artificial scoring evaluation of the code module.
Further, in step S7, the method further includes: and if the keywords input by the user do not have corresponding classification path levels, converting the keywords into new word vectors, calculating the cosine distance between the new word vectors and the keyword vectors on the classification path tree, and selecting a code module corresponding to the keyword vector with the smallest cosine distance as the initial recommendation code module.
Further, the automated process management subsystem includes: the automatic module task disassembling and distributing tracking unit and the module level management document automatic generation unit;
the automatic module task disassembling and distributing tracking unit is used for distributing a code module and/or a business flow chart to be realized for each developer according to the business flow information and the development organization structure information of the modular process definition unit of the low-code development subsystem, and tracking and recording the process management information of each code module in the development process;
and the module-level management document automatic generation unit is used for establishing a process tracking and quality assurance document of each code module according to the statistical information, the process management information and the test information of the code modules in time sequence.
Has the advantages that:
(1) A low-code integrated development system based on multi-path intelligent recommendation stores code modules through a code module warehouse subsystem and records quality information, version information and user permission information of the code modules, dependence of the system on service scenes is reduced, and usability and flexibility of the system are improved. The creation, modification and updating of the code module are realized through the low-code development subsystem; and the recommended code modules are retrieved for the low-code development subsystem through a multi-path intelligent recommendation algorithm of the code module warehouse subsystem, so that the user-defined code modules can be selectively eliminated. The process tracking and quality assurance document is established through the automated process management subsystem to improve the quality assurance of the custom code module.
(2) The code warehouse construction unit comprises a code module authority control subunit, and the flexibility of the use of the system can be improved through different authority control. The code warehouse construction unit comprises a version tracking control subunit, and performs authority intra-domain control on different code modules according to the code module configuration information, the version change information and the quality information, namely, controls aiming at different user authorities, so that the modules among similar software projects can be effectively migrated and shared.
(3) The code module warehouse subsystem adopts a multi-channel intelligent recommendation algorithm, so that hierarchical clustering of the code modules can be realized, the code modules are classified into different clusters, the different clusters are further divided into a plurality of categories of sub-clusters according to the number of keywords, retrieval recommendation of the code modules is carried out according to the quality level coefficient, the code modules with higher quality can be provided for users, and irrelevant available code modules can be prevented from being provided for the users when development tasks of complex service scenes are faced.
(4) The automatic process management subsystem establishes a process tracking and quality assurance document of each code module, can realize the unification of the system configuration document generation, the process management document generation and the system development process, achieves the module-level process management tracking and quality assurance with low cost, and reduces the process management overhead brought by introducing a low-code development platform.
Drawings
FIG. 1 is a schematic diagram of the framework of the multi-path intelligent recommendation-based low-code integrated development system of the present invention.
Detailed Description
The low-code development platform is platform software which is convenient for generating application programs, does not use the traditional programming method, and provides a development environment for users to write programs in a graphical interface and configuration mode. Low code development platforms may be designed and developed for a particular class of applications, such as databases, business processes, and user interfaces (e.g., web applications). While such platforms may produce complete and executable applications, in some special cases programs still need to be written, low code development platforms may reduce the amount of legacy code, speeding up the completion time of commercial applications. The low-code development platform has the greatest advantages that more people can participate in the software development, not only people with programming skills, but also the initial cost of the software development can be reduced.
The method mainly comprises the following steps that a mode of predefining code segments by a platform is adopted in most of existing commercial low-code development platforms to provide a low-code development means under a certain specific service scene for users, the JNPF low-code development platform introducing step information is taken as an example, the platform is preset with code frames of various common controls, the user introduces the predefined code segments in a control dragging mode in the using process, and in the mode, the range of the reusable code segments is limited by a technical stack adopted by the platform; according to the scheme, a code segment multiplexing scheme depending on users and communities is adopted, users can create own reusable code segments (namely code modules in the scheme) at any time and share the code segments to other platform users, and the code reuse scheme enables the reusable code segments on the platform not to be limited by platform developers and certain technology stacks or service scenes.
The invention provides a low-code integrated development system based on multi-path intelligent recommendation, which can provide a high-quality low-code development process for WEB application, and the specific improvement comprises the following steps:
aiming at the limitation (1), a multi-level code warehouse of the self-defined code module is constructed, and a controllable authority sharing method of the user code module is provided, so that the dependence on a service scene is reduced, and the availability and the flexibility of the platform are improved.
Aiming at the limitation (2), three main improvements are provided: firstly, a user-defined module unit is adopted for testing, and the function perfection of the module is guaranteed; secondly, selecting and eliminating low-quality custom code modules by using a multi-path intelligent recommendation algorithm; thirdly, the version tracking control technology is adopted to carry out domain control on the version information and the quality information of the self-defined code module. And the improvement of three aspects is integrated, and the quality guarantee of the self-defined code module is improved.
Aiming at the limitation (3), the platform provides a full-life-cycle module-level management document automatic generation and automatic module task disassembling and distributing tracking method, so that the additional process management overhead can be effectively avoided.
Specifically, the invention supports continuous low-code development with high quality and low scene dependence on WEB application, mainly adopts a multi-level code warehouse technology to construct user code module warehouses with different authority domains, and realizes effective sharing of user code modules, thereby reducing the dependence of platform predefined code modules on service scenes; the user-defined unit test is combined to ensure the complete function of the module and generate the quality information of the module; in addition, by means of a multi-channel intelligent recommendation algorithm, module quality information, module calling information and module artificial evaluation are referred to, and comprehensive evaluation and recommendation are carried out on the quality of the code module; the quality information and the module version information are bound by adopting a version tracking control technology, and tracking control is carried out among different authority domains so as to realize high-quality guarantee of the user-defined code; in addition, a full-life-cycle module-level management document automatic generation and module task automatic disassembly distribution tracking method is adopted, so that the system configuration document generation, the process management document generation and the system development process are coupled, and the module-level process management tracking and quality assurance are realized without introducing extra overhead.
The invention is described in detail below by way of example with reference to the accompanying drawings.
The terms noun in the embodiments of the present invention are explained first:
low code development platform: the LCDP is a platform software which is convenient for generating application programs, and the LCDP provides a development environment for users to write programs by graphical interfaces and configuration instead of the traditional programming method. Most low-code development platforms are capable of developing some specific applications in several business scenarios, such as databases, business processes, and user interfaces (e.g., web applications). Such platforms may produce complete and executable applications, but in some special cases still require the writing of programs. The low-code development platform can reduce the quantity of traditional codes and accelerate the completion time of business application software. It is a common advantage to have more people involved in software development than just those with programming skills, and low code development can also reduce the initial cost of software development.
The multi-path intelligent recommendation algorithm comprises the following steps: according to the recommendation algorithm based on the clustering algorithm and the traditional recommendation algorithm, provided by the invention, the multi-path intelligent recommendation algorithm is used for dividing all articles to be recommended into different recommendation ways, namely different categories. And calculating related recommendation ways according to the search words and the distances between the centers of different categories, and recommending the articles according to the ways according to the quality indexes.
A code module: the combination body which combines the code segment, the version information, the interface test information, the process management information, the quality assurance information, the visual authority information, the user evaluation information and other factors for realizing a certain specific function can be repeatedly utilized in a low-code platform.
Fig. 1 is a schematic diagram of a framework of a multi-channel intelligent recommendation-based low-code integrated development system of the present invention (with "subsystems" and "units" omitted in fig. 1), as shown in fig. 1, the low-code integrated development system includes: a low code development subsystem, a code module repository subsystem, and an automated process management subsystem.
The embodiment of the invention provides a low-code integrated development platform based on multi-path intelligent recommendation, aiming at the defects that the existing low-code development platform is highly dependent on a specific service scene and the problems of development process and quality guarantee segmentation in the use of the low-code development platform.
The low-code development subsystem is a main body of a low-code development platform and comprises the whole process consisting of modular process definition, module definition and automatic filling, modular code development, code module unit testing and system packaging deployment. The code module warehouse subsystem is a key module for realizing low-scene dependence continuous development of the low-code development platform, the code warehouse module is used for completing coupling of quality information and version information of the code module, completing construction of views among different permission domains and realizing visibility and consistency of the code module to authorized users. The automatic process management subsystem realizes the automatic arrangement of process management and the alignment of the quality process and the development process so as to ensure the consistency of the development process and the quality process.
In the running process of a system platform, code fragments created and updated at each stage of the low-code development module form the basis of a code module warehouse. The code module warehouse controls and utilizes the quality information and the version information of the code module on the basis, and provides assistance for modular code development through a multi-path intelligent recommendation algorithm. The automatic process management module is used as an auxiliary unit, a series of system business process information and development organization structure information generated by the modularized process definition are used for automatically arranging the development process, the full life cycle of low-code development is tracked, and the automatic generation of the module-level management document is realized.
The following description is made for each subsystem:
1. the low-code development subsystem is used for creating and updating the code module and storing the code module to the code module warehouse subsystem;
in a particular embodiment, a low code development subsystem includes: the system comprises a modular process defining unit, a module defining and automatic filling unit, a modular code development unit, a code module unit testing unit and a system packaging deployment unit;
(1.1) a modular process definition unit, which is used for defining the system business process and development organization structure and sending the business process information and the development organization structure information to an automatic process management subsystem;
in actual implementation, the modular process defines: the modular process definition consists of two sub-processes of system business process definition and development organization structure definition.
And (3) defining a system service process:
when low code development is carried out, firstly, a system designer needs to carry out hierarchical definition on the service flow of the whole system, and the system service flow is divided into four stages from top to bottom: system level, component level, file level, module level. Each level of system service flow chart is composed of one or more next level of system service flow charts, and system designers can define the processing flow and the nesting relation of each level of service flow chart in a dragging mode in a front-end interface of a development platform. The system level business flow chart is composed of one or more component level business flow charts and defines the range of system business; the component level business flow chart is composed of one or more file level business flow charts, and defines communication modes and processing logics among different functional components; the file-level business flow chart is composed of one or more module-level flow charts and defines the context environment for the operation of the modules; for a module level flow diagram, its internal implementation is by one or more code modules; for all the four-level flow charts, interface configuration options of each flow chart are provided, including but not limited to interface parameter configuration, interface test configuration, interface quality management configuration and the like, and each flow chart automatically carries out interface configuration verification on the context interface when connection operation is completed.
Developing an organizational structure definition:
after the business flow diagram is defined, a development organization structure of the system is defined, wherein the development organization structure is composed of a development process definition and a development member definition.
For system development process definition, a platform calculates a service flow dependency graph according to the dependency relationship of a module-level system service flow graph, performs topological sequencing, decomposes the module-level system service flow graph into a plurality of parallel linear workflows, automatically divides the system development flow into a plurality of development stages by dividing the number of modules on the workflows in proportion, allocates the development stages to each module-level system service flow, and a system development organizer can configure development cycles required by different development stages on the platform;
for the development member definition, the system development organization personnel can authorize the project development permission and the project code module warehouse view permission for the platform registered user in an invitation mode and the like, and can configure different permission domains for the system service flow chart, so that different views of the system service flow chart by the system development personnel with different levels and functions are realized.
(1.2) a module definition and automation filling unit, which is used for coding and mapping code modules, wherein each code module has a unique identification number; the automatic initialization module is used for automatically initializing module interface documents according to the interface configuration information and automatically filling test samples;
in the actual implementation process, the module definition and the automatic filling are as follows: after the system designer and the system development organization personnel complete the modular process definition, the platform uniformly codes and maps all modules after checking that all configurations are correct, generates a unique ID in the project space, and simultaneously initializes the module interface document automatically according to the interface configuration information. Subsequent developers can supplement interface description and test cases of the modules, in the subsequent development process, for the modules which are not realized, configured test cases can be directly selected to serve as input and output occupation, the platform automatically generates corresponding occupation codes, and test development of other modules is not influenced.
(1.3) a modular code development unit, which is used for developing and realizing the creation and the update of a code module;
in a specific embodiment, in the modular code development unit, the code module is created and updated in a visible module migration, multi-path intelligent recommendation and manual creation mode of the code module warehouse subsystem, or rapid system development is performed based on the code module.
In the actual implementation process, after module definition and automatic filling are completed, a modular code development stage is entered, developers only need to implement a defined but not yet implemented module flow chart visible under self account authority, and the implementation modes include three types: module migration in a visible project space, multi-channel intelligent recommendation in a public code module warehouse and manual code implementation. Visible module migration in the project space refers to a code module in the project space which can be commonly authorized by a developer through the account and the project, the code module comprises the project and the project space of the existing project of the organization, and the platform provides a corresponding cross-project space retrieval tool; the multi-path intelligent recommendation in the public code warehouse means that developers can pull open code modules in the code module warehouse, can input module names or function descriptions for associative recommendation when the modules are realized, and introduce the modules into the current project for use; the manual code implementation means that a developer can implement the code module in the module-level flow chart by writing the code manually. The code modules obtained by the three modes can be manually modified and stored, and the modified code modules are stored in the current project space for subsequent development and reuse.
(1.4) a code module unit testing unit, which is used for testing the created and updated code module, generating quality information of the code module according to a test result and integrating the quality information on the version information of the code module, wherein the testing is carried out through a pre-configured module interface test case or an automatic testing tool;
in the actual implementation process, the code module unit tests: for the implemented code module, the platform provides a code module unit test function. The code module unit test supports two automatic test schemes: under default configuration, the platform tests the code module by using the configured module interface test case; meanwhile, the platform supports the automatic testing tools integrating open sources, such as AutoRunner and AutoTst. The test results generated by the two test schemes are converted into a uniform format and integrated into the current version of the code module along with version tracking control information as part of the quality information of the code module.
And (1.5) the system packaging deployment unit is used for generating the running environment configuration of the code module according to the independent service environment of the code module and the common running environment of the code module.
In the actual implementation process, the system is packaged and deployed: after the modularized development of the whole system is completed, a system packing and deployment link is entered, the common operation environment dependency is configured on the platform, the dependency set of each code module is automatically calculated, the system operation environment configuration is exported, a system developer can modify the configuration on the platform, and then the platform automatically arranges different docker packages according to project configuration so as to provide for subsequent system deployment.
2. The code module warehouse subsystem is used for storing the code modules, comprises quality information, version information and user authority information of the code modules and adopts a multi-path intelligent recommendation algorithm to retrieve recommended code modules for the low-code development subsystem;
in a particular embodiment, the code module repository subsystem comprises: the system comprises a code warehouse construction unit and a multi-path intelligent recommendation unit;
(2.1) a code warehouse construction unit, which is used for storing the code modules and sending the statistical information of the code modules to a multi-path intelligent recommendation unit, wherein the code modules comprise pre-constructed code modules and code modules created or modified by users;
in a particular embodiment, the code repository construction unit comprises: the code module creates and modifies the subunit, code module authority control subunit and version tracking control subunit; the code module creating and modifying subunit is used for storing a pre-constructed code module and the code module created or modified by a user; the code module authority control subunit is used for carrying out authority control on users of different code modules, wherein the authority control at least comprises all user visibility, organization visibility, project visibility and self visibility; the version tracking control subunit is used for automatically recording the configuration information, the version change information and the quality information of the code modules aiming at each code module and carrying out intra-authority domain control on different code modules; and the authority control information of the code module authority control subunit, the code module configuration information, the version change information and the quality information of the version tracking control subunit form the statistical information of the code module.
In the actual implementation process, the code warehouse construction is realized by three parts, namely code module creation and modification, code module authority control and version tracking control, and a usable, flexible and high-quality code module is provided for a platform, so that the dependence of the platform on a service scene is eliminated.
Creation and modification of code modules:
a part of general code modules constructed in advance by the platform are stored in the code module warehouse, the general code modules comprise common functions in development, the general code modules are modularized in modes of constructing corresponding module description, extracting parameter settings to serve as module configuration and the like, the part of pre-constructed code modules are stored in an open-source core library, all platform users can see the code modules, and versions of the code modules are maintained by platform developers; the code module warehouse also stores all code modules created or modified by the platform user, and the platform user can save the user code segments as the code modules. Platform pre-construction is combined with user creation, so that the requirement on a universal code segment is met, and a flexible self-defining method is provided, so that the flexibility and the usability of a code module are improved.
Code module authority control:
for the code module library, different units, development teams, development groups and development members can have the code module libraries with different visual levels, and owners of the code module library can select to pull code modules from other sources or share the code modules to other code module libraries, so that sharing multiplexing and permission control of user-defined modules are realized.
For a code module created by a platform user, the visible authority of the code module is determined by a creator and a project space in which the code module is created, a smaller visible range is taken between the creator and the project space, and on the basis, the code module can be divided into different types of open sources (visible for all users), visible for only organization, visible for only project, private (visible for only creator) and the like.
Based on the authority control of the code module and the code module library, the module effective migration and sharing among similar software projects can be realized, and platform-independent available code module support is provided for platform users when the development task of a complex service scene is faced, so that the dependence of the platform on the service scene is reduced.
Version tracking control:
the modification of each code module creator to the code module is automatically recorded with version change information by the platform, the version tracking control technology is adopted to bind the code module configuration information, the version change information and the quality information, the project directly referring to the code module can be selected to be pulled and updated, and all the code module users can evaluate and feed back loopholes in the visible range of the code module, thereby realizing the domain control of the version information and the quality information of the custom code module and improving the quality of the multiplexed code segment.
For the code module with an open source in the code module warehouse, the invention also designs a multi-path intelligent recommendation algorithm to help developers to quickly search the target code module. The multi-path intelligent recommendation algorithm combines the statistical information of the multi-path code modules to construct a high-quality open source code recommendation system, clusters the open source code modules according to the similarity described by the keywords of the code modules, and then calculates the quality level coefficient of each open source code module by referring to the module quality information obtained by multiple paths, such as the calling/starting frequency, the description correlation, the modification calling frequency, the quality information, the artificial scoring evaluation and other factors of each code module, and serves as a recommendation ranking basis in the correlation class. When a user searches a public code module warehouse, intelligent association matching is carried out according to the module name or the function description keyword input by the user, and multi-path intelligent recommendation of high-quality code modules is provided.
And (2.2) the multi-path intelligent recommendation unit is used for retrieving the code module in the code warehouse construction unit according to the requirement retrieval information of the user and combining the statistical information, and providing the code module for the modular code development unit of the low-code development subsystem to directly use or modify and update.
In the practical implementation process, the multi-path intelligent recommendation algorithm provides high-quality recommendation service of the code module for platform users on the basis of the construction of the code module warehouse.
In a specific embodiment, the multi-channel intelligent recommendation unit retrieves information according to the requirement of the user and combines with statistical information to retrieve the code module in the code warehouse construction unit, and the method includes:
s1, constructing a keyword vector for each code module and calculating a module vector of the code module;
s2, selecting module vectors of k code modules as a first cluster center, and recording an initial cluster classification level S as 1;
s3, traversing other code modules except the code module corresponding to the first cluster center, classifying the current code module to the cluster where the first cluster center is located and with the smallest vector cosine distance with the current code module, recording the result as a real center classification result, taking the mean value of the S-th keyword vectors of all code modules corresponding to the cluster where the first cluster center is located as the mean value vector of the cluster, taking the mean value vector as the second cluster center, re-classifying all code modules to the cluster where the second cluster center is located and with the smallest vector cosine distance with the corresponding code module, and recording the result as an estimated center classification result;
s4, aiming at the real center classification result and the estimated center classification result obtained in the step S3, if the classification of a code module changes, the code module with the changed classification is used as a core path set, the mean value of the S-th keyword vectors corresponding to the module vectors of all the code modules in the core path set is calculated, and the mean value is correspondingly added to the second cluster center; aiming at the second cluster center, finding the module vector of the code module with the minimum cosine distance corresponding to the second cluster center, taking the module vector as a new first cluster center, repeating the steps S3 and S4 until no classification of the code module is changed, and ending the circulation;
step S5, aiming at the m clusters obtained in the step S4, if the number of the keywords of the module vector of the code module corresponding to the first cluster center is equal to the cluster classification level S, taking the S-th keyword of the code module corresponding to the first cluster center at the moment as the classification path of the code module of the cluster, and finishing the layering of the cluster; if the number of the key words of the module vectors of the code modules corresponding to the first cluster center is smaller than the classification level S of the cluster, adding one to the classification level of the cluster, namely S +1, randomly selecting the module vectors of k code modules for the cluster as the first cluster center of the sub-cluster of the cluster, repeating the steps S3, S4 and S5, ending the circulation when the number of the key words of the module vectors of the code modules corresponding to all the first cluster centers is equal to the classification level S of the cluster, and obtaining the classification path tree of the code modules according to the hierarchy relationship of the cluster;
in a specific embodiment, in step S5, the method further includes: and updating the classification path tree periodically, including updating according to the statistical information of the code module or updating according to a new keyword vector of the code module.
S6, calculating a quality level coefficient of each code module according to the statistical information and the code use information of each code module, and performing quality level sequencing on the code modules in each cluster according to the quality level coefficient;
in a specific embodiment, in step S6, the code usage information at least includes: the calling/enabling frequency of the code module, the description correlation of the code module, the modification calling frequency of the code module and the artificial scoring evaluation of the code module.
And S7, retrieving information according to the requirements of the user, determining a classification path hierarchy corresponding to the key words input by the user, and taking a code module at the beginning of the hierarchy of the classification path as an initial recommendation code module.
In a specific embodiment, in step S7, the method further includes: and if the keywords input by the user do not have corresponding classification path levels, converting the keywords into new word vectors, calculating the cosine distance between the new word vectors and the keyword vectors on the classification path tree, and selecting a code module corresponding to the keyword vector with the smallest cosine distance as the initial recommendation code module.
3. And the automatic process management subsystem is used for automatically arranging the development process according to the business process information and the development organization structure information from the low-code development subsystem and generating a module-level process tracking and quality assurance document according to the code development process of the low-code development subsystem.
In the actual implementation process, the automatic process management subsystem is composed of an automatic module task disassembling, distributing and tracking method and a full-life-cycle module-level management document automatic generation part, so that the system configuration document generation process, the process management document generation process and the system development process are unified, the low-cost module-level process management tracking and quality assurance are realized, and the process management overhead brought by introducing a low-code development platform is reduced.
In one embodiment, the automated process management subsystem comprises: the automatic module task disassembling and distributing tracking module unit and the module level management document automatic generation unit;
(3.1) an automatic module task disassembling and distributing tracking unit, which is used for distributing a code module and/or a business flow chart to be realized for each developer according to the business flow information and the development organization structure information of the modular process definition unit of the low-code development subsystem, and tracking and recording the process management information of each code module in the development process;
in the actual implementation process, the automatic module task disassembling, distributing and tracking method comprises the following steps:
for the development process, the developers and the system business flow chart determined in the modularization process definition, the platform automatically allocates the corresponding module to be realized or the corresponding business flow chart to each developer, and in the subsequent development process, the platform can track the development and audit information of each module to form a process management information record following the module version information, thereby providing an effective process management approach for the multi-user collaborative development scene of the complex system.
And (3.2) a module-level management document automatic generation unit, which is used for establishing a process tracking and quality assurance document of each code module according to the statistical information, the process management information and the test information of the code modules in time sequence.
In the actual implementation process, the module level management document is automatically generated: in the invention, the construction method of the code module, the automatic module task disassembling, distributing and tracking method and the unit test scheme of the code module jointly determine the process management information and the quality information generated by version alternation in the development process, and the platform leads out the information in a time sequence arrangement and formatting manner to form a module-level process tracking and quality assurance document in the whole software development period and reduce the process management overhead of introducing a low-code development platform.
In order to make those skilled in the art better understand the technical solution of the present invention, the following describes the technical solution of the present invention with reference to a specific scenario embodiment.
Scene embodiment
In the first embodiment of the present scenario, a description is mainly given for a multi-path intelligent recommendation algorithm in combination with a specific formula operation.
The multi-path intelligent recommendation algorithm comprises the following steps:
step one, constructing a keyword vector.
Before constructing the multi-path intelligent recommendation algorithm, the keywords of the code module are required to be vectorized firstly,
here, word embedding (e.g., word2vec, etc.) is used for vectorization. Firstly, every word W in a set W containing o key words i Conversion into corresponding o-dimensional one-hot vectors
Then, taking the one-hot vector as an input training word2vec model, and obtaining a corresponding p-dimensional word vector v through the word2vec model i . Wherein o and p are both positive integers.
And step two, calculating a code module vector.
In the scheme, each code module is defined by a keyword Key i K, when calculating the code module vector, the influence level of a plurality of keywords on the code module classification needs to be comprehensively considered, and then the jth code module vector
The calculation formula is as follows:
where σ is the keyword weight coefficient matrix, σ i Is the ith keyword key i The weight coefficient of (a); v. of j Is the set of all keyword vectors corresponding to the jth code module;
is the ith keyword vector corresponding to the jth code module.
And step three, hierarchically clustering the code modules according to the code module vectors.
After the second step, each code module is calculated into a p-dimensional vector according to the key words of the code module
Next, the code module vectors are multi-way clustered. The multi-path clustering algorithm is an improved k-means algorithm, firstly, k code module vectors are selected as cluster centers, and all the code module vectors are traversed
And calculates the cosine distance d between the center of each cluster and the center of each cluster k The calculation formula is as follows:
And then respectively taking the obtained l clusters as the input of a multi-path clustering algorithm, carrying out hierarchical clustering on the l clusters, recording s-th key words corresponding to the central code module vector of each sub-cluster as a classification path, and so on until the current sub-cluster level s is equal to the number of key words of the central module vector of the sub-cluster, and ending the hierarchical clustering of the sub-cluster.
And step four, maintaining the classification path tree.
And recording the classification paths solved by the multi-path clustering algorithm in the third step, constructing a classification path tree according to the hierarchical relationship among the clusters, periodically recalculating the first step, the second step and the third step, and maintaining the classification path tree. Each node in the classification path tree records the basic information of the code module in the corresponding cluster and the classification path (i.e. the classification key word) of the cluster.
And step five, sorting the quality in the class.
In order to provide high-quality code module recommendation, the quality information of the code modules needs to be measured, the quality ranking of the code modules refers to the module quality information obtained by various ways, such as the calling/starting frequency, description correlation, modification calling frequency, quality information, artificial scoring evaluation and other factors of each code module, and the quality level coefficient of each open source code module is calculated by the following calculation formula:
wherein alpha is k Is the weight coefficient of the kth quality information, normal (info) k ) And the quality level of each code module in the cluster is sorted in a descending order according to the cluster information recorded by each node in the classification path tree after the quality level coefficient of the code module is obtained.
And step six, intelligently associating and recommending keywords.
Keyword association intelligent recommendation is the last step of a multi-path intelligent recommendation algorithm and occurs when a user retrieves a code module.
When a user searches the module name, a target code module is searched in a code warehouse by adopting a direct searching mode; when a user searches keywords, the multi-path intelligent recommendation algorithm searches a classification path tree from top to bottom, and if the currently input keywords appear on the classification path, corresponding nodes are used as initial recommendation nodes; if the keyword does not exist in the classification path, the word vectors corresponding to the keyword are recalculated in the step four, the cosine similarity is respectively calculated with the word vectors corresponding to all the keywords in the classification path, and the node with the minimum cosine distance is selected as the initial recommendation node.
And after the initial recommendation node is obtained, if the user continues to input the keywords, searching the initial recommendation node again on the subtree taking the initial recommendation node as the root node according to the steps, otherwise, selecting the top n from the clusters corresponding to the initial recommendation node as recommendation results according to the quality sorting results in the class in the step five, and displaying the recommendation results to the user.
In conclusion, the invention provides the low-code integrated development system based on multi-path intelligent recommendation, and solves the defects that the conventional low-code development platform is highly dependent on a specific service scene and the problems of the development process and the quality process segmentation in the use of the low-code development platform. Compared with other schemes, the invention has smaller dependence on the scale and quality of the code segment created by a platform maintainer, the compiling and maintenance of the reusable code are mainly completed by a platform user, the reusability and quality level of the code segment can rise along with the increase of the scale and the service time of the platform user, and the invention is suitable for developing and organizing similar services on the platform for a long time; the dependency of other low-code development platforms is strong, the writing and maintenance cost of the multiplex code is borne by a platform creator, the multiplex rate and the quality level are related to the code level of the platform creator, and the system is suitable for teams and individuals pursuing rapid low-cost low-customization development.
The above embodiments only describe the design principle of the present invention, and the shapes and names of the components in the description may be different without limitation. Therefore, a person skilled in the art of the present invention can modify or substitute the technical solutions described in the foregoing embodiments; such modifications and substitutions do not depart from the spirit and scope of the present invention.
Claims (10)
1. A low-code integrated development system based on multi-path intelligent recommendation is characterized by comprising the following components: a low code development subsystem, a code module warehouse subsystem, and an automation process management subsystem;
the low-code development subsystem is used for creating and updating a code module and storing the code module to the code module warehouse subsystem;
the code module warehouse subsystem is used for storing the code modules, including the quality information, the version information and the user authority information of the code modules, and adopting a multi-path intelligent recommendation algorithm to retrieve recommended code modules for the low-code development subsystem;
the automatic process management subsystem is used for automatically arranging the development process according to the business process information and the development organization structure information from the low-code development subsystem and generating a module-level process tracking and quality assurance document according to the code development process of the low-code development subsystem.
2. The low-code integrated development system of claim 1, wherein the low-code development subsystem comprises: the system comprises a modular process defining unit, a module defining and automatic filling unit, a modular code development unit, a code module unit testing unit and a system packaging deployment unit;
the modular process definition unit is used for defining a system service process and a development organization structure and sending the service process information and the development organization structure information to the automatic process management subsystem;
the module definition and automation filling unit is used for coding and mapping the code modules, and each code module has a unique identification number; the automatic initialization module is used for automatically initializing module interface documents according to the interface configuration information and automatically filling test samples;
the modular code development unit is used for developing and realizing the creation and the update of the code module;
the code module unit testing unit is used for testing the created and updated code module, generating quality information of the code module according to a testing result and integrating the quality information on the version information of the code module, wherein the testing is carried out through a pre-configured module interface test case or an automatic testing tool;
the system packaging deployment unit is used for generating the running environment configuration of the code module according to the independent service environment of the code module and the public running environment of the code module.
3. The low-code integrated development system of claim 2, wherein in the modular code development unit, the code module is created and updated by visible module migration, multi-path intelligent recommendation of the code module warehouse subsystem, manual creation or rapid system development based on the code module.
4. The low-code integration development system of claim 1, wherein the code module repository subsystem comprises: the system comprises a code warehouse construction unit and a multi-path intelligent recommendation unit;
the code warehouse building unit is used for storing the code modules and sending statistical information of the code modules to the multi-path intelligent recommendation unit, wherein the code modules comprise code modules which are built in advance by a platform and code modules which are created or modified by a user;
the multi-path intelligent recommending unit is used for retrieving the code module in the code warehouse constructing unit according to the requirement retrieval information of the user and combining the statistical information, and providing the code module for the modular code developing unit of the low-code developing subsystem to directly use or modify and update.
5. The low-code integration development system of claim 4, wherein the code warehouse building unit comprises: the code module creates and modifies the subunit, code module authority control subunit and version tracking control subunit;
the code module creating and modifying subunit is used for storing the pre-constructed code module and the code module created or modified by the user;
the code module authority control subunit is used for carrying out authority control on users of different code modules, wherein the authority control at least comprises all users visible, only organization visible, only item visible and only self visible;
the version tracking control subunit is used for automatically recording code module configuration information, version change information and quality information aiming at each code module and carrying out intra-authority domain control on different code modules;
and the authority control information of the code module authority control subunit, and the code module configuration information, the version change information and the quality information of the version tracking control subunit form the statistical information of the code module.
6. The system of claim 4, wherein the multi-channel intelligent recommendation unit retrieves the code module in the code repository construction unit according to the requirement retrieval information of the user in combination with the statistical information, comprising:
s1, constructing a keyword vector for each code module and calculating a module vector of the code module;
s2, selecting module vectors of k code modules as a first cluster center, and recording an initial cluster classification level S as 1;
s3, traversing other code modules except the code module corresponding to the first cluster center, classifying the current code module to the cluster where the first cluster center is located and with the smallest vector cosine distance with the current code module, recording the result as a real center classification result, taking the mean value of the S-th keyword vectors of all code modules corresponding to the cluster where the first cluster center is located as the mean value vector of the cluster, taking the mean value vector as the second cluster center, re-classifying all code modules to the cluster where the second cluster center is located and with the smallest vector cosine distance with the corresponding code module, and recording the result as an estimated center classification result;
s4, aiming at the real center classification result and the estimated center classification result obtained in the step S3, if the classification of the code module is changed, the code module with the changed classification is used as a core path set, the mean value of the S-th keyword vectors corresponding to the module vectors of all the code modules in the core path set is calculated, and the mean value is correspondingly added into the second cluster center; aiming at the second cluster center, finding the module vector of the code module with the minimum cosine distance corresponding to the second cluster center, taking the module vector as a new first cluster center, repeating the steps S3 and S4 until no classification of the code module is changed, and ending the circulation;
step S5, aiming at the m clusters obtained in the step S4, if the number of the keywords of the module vector of the code module corresponding to the first cluster center is equal to the cluster classification level S, taking the S-th keyword of the code module corresponding to the first cluster center at the moment as the classification path of the code module of the cluster, and finishing the layering of the cluster; if the number of the key words of the module vectors of the code modules corresponding to the first cluster center is smaller than the classification level S of the cluster, adding one to the classification level of the cluster, namely S +1, randomly selecting the module vectors of k code modules for the cluster as the first cluster center of the sub-cluster of the cluster, repeating the steps S3, S4 and S5, ending the circulation when the number of the key words of the module vectors of the code modules corresponding to all the first cluster centers is equal to the classification level S of the cluster, and obtaining the classification path tree of the code modules according to the hierarchy relationship of the cluster;
s6, calculating a quality level coefficient of each code module according to the statistical information and the code use information of each code module, and performing quality level sequencing on the code modules in each cluster according to the quality level coefficient;
and S7, retrieving information according to the requirements of the user, determining a classification path hierarchy corresponding to the key words input by the user, and taking a code module at the beginning of the hierarchy of the classification path as an initial recommendation code module.
7. The low-code integrated development system according to claim 6, further comprising, in step S5: and updating the classification path tree periodically, including updating according to the statistical information of the code module or updating according to a new keyword vector of the code module.
8. The low-code integrated development system according to claim 6, wherein in step S6, the code usage information includes at least: the calling/enabling frequency of the code module, the description correlation of the code module, the modification calling frequency of the code module and the artificial scoring evaluation of the code module.
9. The low-code integrated development system according to claim 6, further comprising, in step S7: and if the keywords input by the user do not have corresponding classification path levels, converting the keywords into new word vectors, calculating the cosine distance between the new word vectors and the keyword vectors on the classification path tree, and selecting a code module corresponding to the keyword vector with the smallest cosine distance as the initial recommendation code module.
10. The low-code integrated development system of claim 1, wherein the automated process management subsystem comprises: the automatic module task disassembling and distributing tracking unit and the module level management document automatic generation unit;
the automatic module task disassembling and distributing tracking unit is used for distributing a code module and/or a business flow chart to be realized for each developer according to the business flow information and the development organization structure information of the modular process definition unit of the low-code development subsystem, and tracking and recording the process management information of each code module in the development process;
and the module-level management document automatic generation unit is used for establishing a process tracking and quality assurance document of each code module according to the statistical information, the process management information and the test information of the code modules in time sequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211371028.7A CN115756477A (en) | 2022-11-03 | 2022-11-03 | Low-code integrated development system based on multi-path intelligent recommendation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211371028.7A CN115756477A (en) | 2022-11-03 | 2022-11-03 | Low-code integrated development system based on multi-path intelligent recommendation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115756477A true CN115756477A (en) | 2023-03-07 |
Family
ID=85357785
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211371028.7A Pending CN115756477A (en) | 2022-11-03 | 2022-11-03 | Low-code integrated development system based on multi-path intelligent recommendation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115756477A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117369787A (en) * | 2023-12-08 | 2024-01-09 | 泰德网聚(北京)科技股份有限公司 | System for automatically developing low codes based on AIGC technology |
-
2022
- 2022-11-03 CN CN202211371028.7A patent/CN115756477A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117369787A (en) * | 2023-12-08 | 2024-01-09 | 泰德网聚(北京)科技股份有限公司 | System for automatically developing low codes based on AIGC technology |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cohen et al. | Learning to match and cluster large high-dimensional data sets for data integration | |
| US5870735A (en) | Method and system for generating a decision-tree classifier in parallel in a multi-processor system | |
| CA2906232C (en) | Domain-specific syntax tagging in a functional information system | |
| US7152065B2 (en) | Information retrieval and text mining using distributed latent semantic indexing | |
| CN102341781B (en) | Software test bed generation | |
| CN108701254A (en) | System and method for the tracking of dynamic family, reconstruction and life cycle management | |
| CN109815343A (en) | Obtain method, apparatus, equipment and the medium of the data model in knowledge mapping | |
| Reichardt | Structure in complex networks | |
| CN115756477A (en) | Low-code integrated development system based on multi-path intelligent recommendation | |
| KR102153259B1 (en) | Data domain recommendation method and method for constructing integrated data repository management system using recommended domain | |
| CN108280176A (en) | Data mining optimization method based on MapReduce | |
| CN108256086A (en) | Data characteristics statistical analysis technique | |
| CN110083339A (en) | A kind of novel computer programming language | |
| Calégari | Parallelization of population-based evolutionary algorithms for combinatorial optimization problems | |
| Hirchoua et al. | Topic hierarchies for knowledge capitalization using hierarchical Dirichlet processes in big data context | |
| CN109299381A (en) | A kind of software defect retrieval and analysis system and method based on semantic concept | |
| US11775267B2 (en) | Identification and application of related source code edits | |
| CN114676258B (en) | Disease classification method based on symptom description text and not aiming at diagnosis | |
| Jangra et al. | Efficient Utilization of Cloud Based IoT Services Using Ontology System | |
| Skotnik et al. | An Approach towards Reusability in Hybrid Avionic Software Development by Using a Unified Graph Representation of the Software System | |
| CN116450935A (en) | Request parameter conversion method for Web API call | |
| Nezlek et al. | An Integrated Approach to Defining Enterprise Computing Architectures | |
| Wong | A methodology for the definition of data base workloads: an extension to the IPSS methodology | |
| CN113505600A (en) | Distributed indexing method of industrial chain based on semantic concept space | |
| Smirnov et al. | Object model of DELPHI Data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |