I've been using readr for reading/writing data frames. Obviously, as you add more data, the process is slower. However, it seems that adding columns is more problematic than adding rows. For example, a 100k x 100 table is noticeably faster to read/write than a 100 x 100k table. I expected additional columns to add extra overhead, but not to such a degree.
There are a lot of benchmarks for reading/writing text files, but they focus on long tables. Are there any tips or suggestions for wide tables?