The "rows" JSON / YAML gives the ARD back: its columns with their
types, each variable's levels as factors in their order, the
statistics as numbers or text as written (the formatting functions are
not kept; stat_fmt is). An XPT or CSV is read as far as a flat table
allows: the level columns become list columns again, with no factors.
An rds is read as it is.
Arguments
- path
The file.
- format
As
tfl_write_ard();NULLfrom the extension (.csvtoo).