![]() |
![]() |
|
![]() |
![]() |
![]() ![]() ![]() |
![]() |
The AI Accelerated version of FmPro Migrator Platinum Edition 11.39 - Adds a new COBOL import feature enabling the automated conversion of COBOL to over 50 programming languages. The use of locally running models with FmPro Migrator Custom Dev Edition and FmPro Migrator Site License Edition is designed to support secure computing environments where the computer may be "air gapped" - unable to access the internet for security reasons. Automated COBOL Conversion with the Code Conversion Workbench The Code Conversion Workbench enables developers to manage the automated import and conversion of hundreds of COBOL program files in a single window. Individual programs can be selected by name for single-click submission to the selected AI provider and machine learning model. Output files are saved automatically as the results are received into a folder of converted scripts. The Code Conversion Workbench supports conversions into the top 50 most popular programming languages as found in the TIOBE index. Machine learning prompts are generated automatically, but the prompt text and source code fields are fully editable. Training records can be created to "train" the LLM and fine-tune the results, improving the conversion process to nearly 100%. Top 10 Features
Top 10 Features - Code Conversion Workbench
For more info email, call or chat. Description The COBOL Conversion Feature of the Code Conversion Workbench, integrated into FmPro Migrator AI Accelerated Edition, offers an economical migration or conversion solution for COBOL projects. This feature facilitates a nearly 100% conversion of all code functionality when utilized in conjunction with the Training feature. Scale-up & Transition COBOL to Cloud Hosted Apps Scale up legacy mainframe applications to run on cloud infrastructure with modern APIs, save money on hardware & software leasing costs, and own all of the source code for your application, enabling customization at your convenience. FmPro Migrator is a cross platform tool available for macOS and Windows platforms. The Code Conversion Workbench facilitates the automated conversion of COBOL scripts for mission critical migration projects.
Code Conversion Workbench Training [Video] The Code Conversion Workbench Training feature empowers developers to tailor the automated conversion process to their specific codebase requirements. Are you seeking to utilize user-defined functions from modern development environments to replace the original COBOL procedure throughout your codebase? This feature allows you to define the unique features you desire the machine learning model to perform using plain English text examples, as demonstrated in the video linked above. This functionality is compatible with nearly all machine learning models from various vendors and provides precise instructions on how you want the converted code to be structured. You can add as many training records as necessary and modify them as required during the conversion process. C# example training records are shown on the .NET Conversion page.
COBOL to Java Conversion [Sample File] The COBOL code being migrated here implements a LISP interpreter. The DeepSeek-R1 7B model created the 105 lines of Java code (as shown in the screenshot) within 2 minutes. Calls are made to methods referenced as separate files in the original code. As a “thinking” model, DeepSeek has analyzed the referenced code files and generated methods to represent each of those files. This can be considered an advantage if you can’t locate those referenced code files. For this project, those files are available and can be converted separately and then used to replace the placeholder methods shown here.
COBOL to C# Conversion [Sample File] The Microsoft Phi4 14B model created the 243 lines of C# .NET code shown in this screenshot in about 6 minutes. The C# code is approximately 50% smaller than the original COBOL code. In the TokenizeLispProcedure method, the IsNullOrWhiteSpace() method is being used to check for the end of a token - which is an effective technique. The variable names are appropriate for a C# program. The original WS-FLAG-YES from COBOL has been rewritten as flagYes in C#. Additionally, the logging of the text adheres to the case of the original text. The remainder of the C# code also looks very reasonable from an automated conversion perspective. C# example training records are shown on the .NET Conversion page.
COBOL to Python Conversion [Sample File] This example demonstrates the output generated by the cloud-hosted Google Gemini 2.0 Flash-Thinking model in comparison to the DeepSeek-R1 and Microsoft Phi4 models running locally. While publicly available cloud-hosted models may not be accessible to all customers, the Google Gemini 2.0 Flash Thinking model generated the 209 lines of Python code depicted in the screenshot within approximately 16 seconds. This is significantly faster than the Phi4 model, which took over twice as long to produce the same code. The format_list_procedure class is 100 lines of code with multiple inner classes inside the main procedure. The format_list_procedure class alone comprises 100 lines of code, including multiple inner classes within the main procedure. This is substantially more extensive than the 39 lines of code generated by the DeepSeek-R1 model or the 12 lines generated by the Phi4 model. Overall, the code is well-commented and easily readable.
There are multiple editions of FmPro Migrator available for testing and production:
|
![]() |
![]() |
|
![]() |
![]() |
|||||||||||||||
|
|||||||||||||||||||||||
![]() |
||
![]() |
![]() |
|
![]() |