Data transformation converts information between structures, representations, or languages while retaining the original values and intended logic. In testing, LLMs can transform test data, code, and raw testing notes to accelerate exploratory testing, debugging, and test automation. section 3.3.2 section 6.1.2
Useful transformation tasks include:
- Convert JSON or plain-text records into SQL table definitions and insert statements. section 6.1.2
- Translate code between languages while retaining its logic and flow, then test the result. section 3.3.2
- Turn exploratory or design-testing notes into concise summaries for sharing. section 3.3.2
- Use delimiters, explicit output requirements, and completeness checks to reduce altered values and missing fields. section 6.1.2
###
[{
"firstname": "Sally",
"lastname": "Wilson",
"totalprice": 229,
"depositpaid": false,
"bookingdates": {
"checkin": "2017-02-20",
"checkout": "2019-07-05"
},
"additionalneeds": "Breakfast"
},{
"firstname": "Jim",
"lastname": "Jones",
"totalprice": 871,
"depositpaid": false,
"bookingdates": {
"checkin": "2017-08-11",
"checkout": "2021-06-30"
},
"additionalneeds": "Breakfast"
},{
"firstname": "Mary",
"lastname": "Ericsson",
"totalprice": 558,
"depositpaid": true,
"bookingdates": {
"checkin": "2018-07-04",
"checkout": "2022-09-25"
}
}]
###
A concrete booking dataset shows the source JSON that can be transformed into a relational SQL structure and records. section 6.1.2
JSON to SQL
A reusable prompt can define the transformation, require a CREATE TABLE statement and INSERT statements, and ask the model to check every JSON field before responding. The generated SQL preserves names, prices, dates, Boolean values, and a missing additional-needs value as NULL; humans should still verify correctness and validity. section 6.1.2
