In #rstats #tidyverse + {targets} is the have your cake + eat it too combo.

High level expressive code, with freedom from much of the performance penalties associated with higher abstraction via caching computational results. IMHO this is the entry level toolkit for doing real data science work with R. But too often targets gets shunted from courses and books, leaving people to figure it out for themselves, and a missing discourse about best practices for building pipelines with graphs.