One of the researchers whom I wholeheartedly respect shared a repo
with some code publish for reproducibility sake as an accompanying
resource for a publication. At first, it didn't run since they didn't
include any environment description. Then I realised that it expected
some particular datasets in a peculiar format. The author even
provided me with data after an additional request. But it still
didn't work failing with some weird error. The lesson is that
research code usually sucks, although it can contain some brilliant
ideas inside.
