CSS Corpus for Reproducible Analysis

Reproducibility of research heavily depends on the availability of the datasets from the experiments in the context of metaprogramming, the corpus of the code that was used to run the analyses and transformations. In the case of CSS, the problem is even more acute since the web is a constantly changing environment where the same address can refer to a frequently changing artefact. In this report, we explain how we created a corpus of CSS les as a part of our project of building a framework for analysing style sheets. We also include two case studies of explanatory nature showing how style sheets from various websites go about coding conventions and about code duplication. We believe this work will be useful for other CSS researchers to compare techniques they develop, on a uniform yet realistic dataset.