CODEWITHSUNDEEP
r
Sebastian Zeki
Sebastian Zeki

Reputation: 6874

How to add to a date with ifelse

I have a dataframe with a column for the date and a text column called Rule. If the text in the Rule Column is "Rule_1" I want to add 3 years to the date in that row, but if "Rule_2" then two years.

Date        Rule     NewDate
2010-01-01  Rule_1   2013-01-01
2012-01-01  Rule_2   2014-01-01
2015-01-01  No_Rule  2015-01-01

I seem to have a problem with adding the year using lubridate

 ifelse(df$Rule=="Rule1",df$Date %m+% years(3),df$Date)

It converts the added date to a number which I think is the number of days from an origin (?1970). However if I run the expression df$Date %m+% years(3) without ifelse then it works fine. What to do?

Upvotes: 1

Views: 923

Answers (2)

Greg
Greg

Reputation: 3670

Since you're already using one tidyverse package (lubridate) how about case_when from dplyr? It's a vectorized if - and basically allows you to avoid a bunch of nested ifelse statements if you have multiple conditions.

library(dplyr)
df %>% 
  mutate(NewDate_Demo = case_when(Rule == "Rule_1" ~ Date %m+% years(3),
                          Rule == "Rule_2" ~ Date %m+% years(2),
                          TRUE ~ Date))
# A tibble: 3 x 4
  Date       Rule    NewDate    NewDate_Demo
  <date>     <chr>   <date>     <date>      
1 2010-01-01 Rule_1  2013-01-01 2013-01-01  
2 2012-01-01 Rule_2  2014-01-01 2014-01-01  
3 2015-01-01 No_Rule 2015-01-01 2015-01-01

dplyr does also offer if_else, which has the same general syntax as base::ifelse but won't break with lubridate https://github.com/tidyverse/lubridate/issues/644

Upvotes: 2

mnist
mnist

Reputation: 6954

The base ifelse function is not type stable. Rather make use if the dplyr implementation if_else(). This automatically checks and retains the desired output type. From its documentation:

Compared to the base ifelse(), this function is more strict. It checks that true and false are the same type. This strictness makes the output type more predictable, and makes it somewhat faster.

library(dplyr)
library(lubridate)

# never mix up tidyverse and data.table but damn, fread() is awesome :D 
df <- data.table::fread("
Date        Rule     NewDate
2010-01-01  Rule_1   2013-01-01
2012-01-01  Rule_2   2014-01-01
2015-01-01  No_Rule  2015-01-01") %>% 
  as_tibble() %>% 
  mutate_at(.vars = c(1, 3), as_date)

# more type stable ifelse
dplyr::if_else(df$Rule=="Rule_1", df$Date %m+% years(3), df$Date)

# desired output with nested if_else
# faster than case_when()
dplyr::if_else(df$Rule=="Rule_1", df$Date %m+% years(3), 
               if_else(df$Rule=="Rule_2", df$Date %m+% years(2), df$Date))

# piped dplyr way
df %>% 
  mutate(desired_date = if_else(Rule == "Rule_1",
                                true = Date %m+% years(3), 
                                false = if_else(Rule == "Rule_2", 
                                                true = Date %m+% years(2),
                                                false = Date)))

Upvotes: 1

Related Questions