parse string like "now-1h" in R - r

Question

what about experience by parsing/converting strings like "now-1h", "today", "now-3d", "today+30m" in R?
how to recognize and convert string (for example as function's argument) to date_time?

Answer 1

I do not recommend using this code, which is brittle. This is just to demonstrate this is possible:
library(lubridate)
library(stringr)
library(dplyr)
data <- c("now-1h", "today", "now-3d", "today+30m", "-3d")
We're going to use lubridate functions now(), today(), days(), hours(), ... which resemble your input data:
str(now())
# POSIXct[1:1], format: "2017-03-24 12:26:18"
str(today())
# Date[1:1], format: "2017-03-24"
str(now() - days(3))
# POSIXct[1:1], format: "2017-03-21 12:27:11"
str(- days(3))
# Formal class 'Period' [package "lubridate"] with 6 slots
# ..# .Data : num 0
# ..# year : num 0
# ..# month : num 0
# ..# day : num -3
# ..# hour : num 0
# ..# minute: num 0
We're going to have to parse() them as strings, to be able to actually use them, like that:
eval(parse(text = "now() + days(3)"))
# [1] "2017-03-27 12:41:34 CEST"
Now let's parse the input strings with a regex, manipulate them a bit to match lubridate syntax, then eval()uate them:
res <-
str_match(data, "(today|now)?([+-])?(\\d+)?([dhms])?")[, - 1] %>%
apply(1, function(x) {
time_ <- if (is.na(x[1])) NULL else paste0(x[1], "()")
offset_ <- if (any(is.na(x[2:4]))) NULL else paste(x[2],
recode(x[4], "d" = "days(", "h" = "hours(", "m" = "minutes(", "s" = "seconds("),
x[3],
")")
parse(text = paste(time_, offset_))
}) %>%
lapply(eval)
Notice that you get a variety of classes as output (either POSIXct or Date or lubridate::Period):
invisible(lapply(res, function(x) { print(x) ; str(x) }))
# [1] "2017-03-24 11:57:52 CET"
# POSIXct[1:1], format: "2017-03-24 11:57:52"
# [1] "2017-03-24"
# Date[1:1], format: "2017-03-24"
# [1] "2017-03-21 12:57:52 CET"
# POSIXct[1:1], format: "2017-03-21 12:57:52"
# [1] "2017-03-24 00:30:00 UTC"
# POSIXlt[1:1], format: "2017-03-24 00:30:00"
# [1] "-3d 0H 0M 0S"
# Formal class 'Period' [package "lubridate"] with 6 slots
# ..# .Data : num 0
# ..# year : num 0
# ..# month : num 0
# ..# day : num -3
# ..# hour : num 0
# ..# minute: num 0
(What I recommend instead is to pre-process the data with the language that produced it and possesses the right tools for the job, which appears to be Perl).