Parse the dates argument into a sorted, unique vector of dates

YYYYMMDD is what a survey database usually holds, and YYYY-MM-DD is what R prints, so both are accepted, along with Date objects and a mixture of the three. Numbers are taken as YYYYMMDD too, since an unquoted 20150402 is an easy thing to type.

Usage

parse_dates(dates)

Arguments

dates

the dates argument as the caller wrote it

Value

a sorted, deduplicated Date vector

Why invalid dates are an error

"20150230" parses to NA rather than to a date. Dropping it would fetch fewer days than were asked for and say nothing, leaving the gap to be found much later as a missing row. The offending values are named instead.