Do Research, not Data Visualization! How to Create More Consistent Plots for Experimental Research Papers in Less Time
Abstract: In the database community, research is accompanied by extensive experimental evaluation. This evaluation produces large amounts of result data which is typically visualized in form of various plots. Unfortunately, the path from the data to the corresponding plots is currently long, cumbersome, and frustrating. In our experience, this largely stems from the challenges that arise in keeping all plots of a paper consistent with each other: Ensuring that all prop- erties of all plots, such as labeling, coloring, scales, or order of the presented results actually match each other requires a constant revisiting and readjusting of the redundant portions of a large number of plotting scripts. This unnecessarily eats away time from doing meaningful research. To address this problem, we present our Python framework ChartGallery, which we already successfully used in our group for previous research papers. It sets itself apart by not being yet another plotting library, but a framework that focuses on managing all plotting setups and styles of a paper in an organized way in order to compose various different plots with minimal effort and code redundancy.
Repository: https://gitlab.rlp.net/chartgallery/chartgallery