Skip to contents

This operates in different ways based on `undo`. Need to use both `dataset` and `trait` to identify data for a `trait`.

Usage

unite_datatraits(
  object,
  traitnames,
  undo = FALSE,
  sep = ": ",
  filters = NULL,
  key = FALSE
)

Arguments

object

data frame

traitnames

names as `dataset: trait` for subsetting (when `undo` = `TRUE`)

undo

logical flag on function use

sep

separator for `dataset: trait` (default ": ")

filters

optional list of columns to filter on (when `undo` = `FALSE`)

key

get key traits if `TRUE`

Value

either a subset `object` based on `traitnames` = `dataset: trait` (if `undo` = `TRUE`) or vector of `dataset: trait` names (if `undo` = `FALSE`, default)