Skip to contents

Split a date column into year, month, and day

Source: R/raw_data.R
split_dates.Rd

The raw ECOMON export carries one DATE column, while the rest of this package works in separate year, month, and day columns — that is what columns.year/month/day name, and what the covariate matching joins on. This is the conversion, taken from original/create_database.R:27-29.

Usage

split_dates(
  dat,
  column = "DATE",
  formats = c("%d-%b-%y", "%d-%b-%Y", "%Y-%m-%d", "%m/%d/%Y", "%m/%d/%y"),
  keep = FALSE
)

Arguments

dat

a data frame with a date column

column

name of the date column

formats

candidate strptime formats, tried in order; the first that parses every non-missing value is used. The default is the ECOMON export's 05-JAN-03, followed by the usual four-digit and slash-separated variants.

keep

whether to keep the original date column

Value

dat with integer year, month, and day columns added

Details

Two things the original left implicit are handled here, because both fail quietly rather than loudly:

  • %b matches an abbreviated month name, which is locale-dependent. Under a non-English LC_TIME every 05-JAN-03 parses to NA and the rows are silently dropped later as incomplete. Parsing runs under the C locale.

  • %y is a two-digit year, which R maps 00–68 to the 2000s and 69–99 to the 1900s. That is correct for an ECOMON record running 1977 to the present, but it means a date can only be read as within roughly the last century — so a parsed date in the future is reported rather than accepted.

Examples

raw <- data.frame(STATION = 1:3, DATE = c("05-JAN-03", "17-JUN-11", "28-SEP-19"))
split_dates(raw)
#>   STATION year month day
#> 1       1 2003     1   5
#> 2       2 2011     6  17
#> 3       3 2019     9  28