Testing framework for GRASS GIS: ensuring reproducibility of scientific geospatial computing

Abstract:

GRASS GIS, a free and open source GIS, is used by many scientists directly or through other projects such as R or QGIS to perform geoprocessing tasks. Thus, a large number of scientific geospatial computations depend on quality and correct functionality of GRASS GIS. Automatic functionality testing is therefore necessary to ensure software reliability. Here we present a testing framework for GRASS GIS which addresses different needs of GRASS GIS and geospatial software in general. It allows to test GRASS tools (referred to as GRASS modules) and examine outputs including large raster and vector maps as well as temporal datasets. Furthermore, it enables to test all levels of GRASS GIS architecture including C and Python application programming interface and GRASS modules invoked as subprocesses. Since GRASS GIS is used as a platform for development of geospatial algorithms and models, the testing framework allows not only to test GRASS GIS core functionality but also tools developed by scientists as a part of their research. Using testing framework we can test GRASS GIS and related tools automatically and repetitively and thus detect errors caused by code changes and new developments. Tools and code are then easier to maintain which results in preserving reproducibility of scientific results over time. Similarly to open source code, the test results are publicly accessible, so that all current and potential users can see them. The usage of testing framework will be presented on an example of a test suite for r.slope.aspect module, a tool for computation of terrain slope, aspect, curvatures and other terrain characteristics.