US Patent Application 18353529. REGRESSION TESTING FOR WEB APPLICATIONS simplified abstract
Contents
REGRESSION TESTING FOR WEB APPLICATIONS
Organization Name
INTERNATIONAL BUSINESS MACHINES CORPORATION
Inventor(s)
REGRESSION TESTING FOR WEB APPLICATIONS - A simplified explanation of the abstract
This abstract first appeared for US patent application 18353529 titled 'REGRESSION TESTING FOR WEB APPLICATIONS
Simplified Explanation
- This patent application describes a method for training a predictive model using network traffic and data change messages from a web application in a production environment. - The predictive model is trained to predict data changes resulting from API calls made through network traffic. - The method involves replaying a stream of network traffic from the existing web application with an upgraded version of the same application to generate real data changes. - The stream of network traffic is then applied to the predictive model to generate predicted data change messages. - These predicted data change messages are compared with real data change messages that represent the actual data changes. - Based on any inconsistencies between the predicted and real data change messages, one or more existing APIs are identified as potentially functionally degraded.
Original Abstract Submitted
Training a predict model with network traffic and data change messages generated by an existing web application running in a production environment. The predict model being is trained to predict data changes resulted from API calls embodied in network traffic. A stream of network traffic of the existing web application is replayed with an upgraded version of the existing web application to generate real data changes. The stream of network traffic is applied to the predict model to generate predicted data change messages. The predicted data change messages are comparing with real data change messages representing the real data changes. One or more existing APIs is identified as being possibly functionally degraded based on any inconsistency of the predicted data change messages with the real data change messages.
