Nicholas Cross (Institute for Astronomy, Edinburgh), Nicholas Cross (IfA, Edinburgh), Albert Heyrovsky (IfA, Edinburgh)
Gaia is an ESA cornerstone mission which will launch in 2013 to improve our understanding of the Milky Way, primarily through the measurement of accurate distances for 1 billion stars through extremely accurate astrometry. The processing of Gaia data is extremely complex and is done through 3 main pipelines. The first of these is Initial Data Processing and First Look (IDT-FL), which processes the unsorted incoming data stream on a daily basis. First Look consists of One Day Calibrations, Monitors and Evaluators which test the output of IDT to ensure the quality control.
The FL monitor infrastructure is written in Java and designed to handle a wide range of different object types. Each method is tested using JUnit tests to ensure that it is working correctly and the whole of IDT-FL is tested using simulations that try to encapsulate as carefully as possible the expected behaviour of Gaia data. However, unit tests can only test the method in question, and strange output from an end-to-end test of IDT-FL from a simulation can be caused by problems in the simulations, in IDT or in an earlier part of FL, and not necessarily the part that seems to fail.
We have designed an intermediate testing environment that allows us to describe a set of automated tests that produce data from simple algorithms that can be set up to produce a range of test data with expectation values that can test each diagnostic class all the way through. A new diagnostic class can be tested by simply adding some properties which outline the datatypes to be tested, creating the new data and running the diagnostic tasks tester.
Poster in PDF format
Paper ID: P026