Round the numeric columns of a data frame

Simulation output is quoted to three or four significant figures anywhere it is reported and was stored at fifteen, which is most of the size of the committed summaries and none of the information in them.

Usage

round_sig(x, digits = 6)

Arguments

x

a data frame.

digits

significant figures, or NULL to leave x alone – which is what anything a tolerance is asserted against needs.

Value

x with its numeric columns rounded.

Details

This is exported and called at the point the files are written, not applied once by hand afterwards. It was applied by hand once: the next run of run.R wrote full precision again and rep_monthly.csv went back from 6.4 MB to 19.9 MB without anything noticing. A rule about stored precision has to live in the code that stores it.