CROPTEST : data-drive test automation for crop modeling systems

Abstract

The present work describes the development of a new version of CROPTEST that address the problem to compare different versions of DSSAT-CSM using automated procedures. The new version expands the current one adding more statistical analysis applied to output files and, using a methodology called data-driven test (DDT) from software engineering that use a fixed set of input files and compare the output results – in this case, the batch files and experiments settings. Agile scrum was adopted as the development methodology. The new user interface and all operation pattern were designed to maximize the user experience and improve their productivity. The final code was written using modern JavaScript web technologies which evidence useful and responsive to desktop development, using libraries and frameworks combined for this goal. The CROPTEST API, implemented and design to be the core source for parsing and calculations procedures, is described in detail and it showed how to set up a configuration to attached it to a PaaS (Platform as a Service) like Travis-CI or Circle-CI. In this setup is possible to run a subset of the all crop tests to check the consistency and improvement of the proposed code in relation with last stable version stored on Github master branch. The desktop version was tested and have automatic installation procedures already defined for the three major OS – Windows, Linux, and MacOS – making the CROPTEST a multi-platform tool for test CSM models