I discovered in #rstats that there is a lot of common ground between the new recode_values (in dplyr) and the old left_join and rows_update, which makes me wonder why they all exist. I wrote a blog post exploring the issues, at https://blog.ritsokiguess.site/posts/same-ground/