Client Success Story  > Auto Migrate Test Cases

Auto Migrate Test Cases

The objective is to develop tools to auto-migrate test cases written for the source test device bed to equivalent test cases reported for a targeted test device bed.

 

Test Cases

The objective is to develop tools to auto-migrate test cases written for the source test device bed to equivalent test cases reported for a targeted test device bed.

 

Test Method

A test device testbed is either a physical or a virtual device. It runs the flavours of Android and other applications specified by the customer. A combination of device+os+app is called a test bed. The test bed has some capabilities that are exposed through UI widgets. Test cases exercise the capabilities of the test bed by invoking these widgets in a prescribed sequence.

 

Proposed method and timeline

The proposed auto-correction method was to use existing automated tools to extract the UI widget hierarchy of the source and target test bed. Tools were also built to infer a difference map between the source and target test bed. These different maps were applied to the source to generate target test cases.

The proposed timeline for the exploration and feasibility phase was around 8 weeks - with periodic progress and issue review. Telaverge identified and explored existing tools to extract the UI widget hierarchy of source and target test bed.

Google App Scraper and similar tools

  • Build a parse tree of source test cases.
  • Modify the source to the target parse tree using the difference map.
  • Generate target test cases from the source target parse tree.

The following tools are being considered:

  • Robot Framework Parsing - for parsing test cases.
  • Google Crawler - Jetpack - for extracting UI widget hierarchy.
  • Explore ML and non-ML techniques to infer the difference map based on the UI widget hierarchy.

Acceptance Criteria

Best outcome – The tool auto migrates all the sample test cases. The resulting test cases run unmodified on the target test bed.

Acceptance criteria – Less than 15% of the lines generated by the tool need to be human-corrected to get the target test case running.   This limit applies at a single and gross test case level.