Converting a character string into a date in R

问题

The data I'm trying to convert is supposed to be a date, however it is formatted as mmddyyyy with no separation by dashes or slashes. In order to work with dates in R, I would like to have this formatted as mm-dd-yyyy or mm/dd/yyyy.

I think I might need to use grep(), but I'm not sure how to use it to reformat all of the dates that are in the mmddyyyy format.

回答1:

Have a look at lubridate mdy function

require(lubridate)
a <- "10281994"
mdy(a)

gives you

[1] "1994-10-28 UTC"

of class "POSIXct" "POSIXt" so a datetime in R. (thanks Joshua Ulrich for the correction)

You could use as.Date(mdy(a)) = 1994-10-28 to get a Object of class Date.

There are mutations like ymd and dmy within lubridate as well.

回答2:

Updated: Improved with @Richard Scriven's colClasses and simpler as.Date() suggestions

Here are two similar methods that worked for me, going from a csv containing mmddyyyy format date, to getting it recognized by R as a date object.

Starting first with a simple file tv.csv:

Series,FirstAir
Quantico,09272015
Muppets,09222015

Method 1: All as string

Once within R,

> t = read.csv('tv.csv', colClasses = 'character')

imports tv.csv as a data frame named t
colClasses = 'character') option causes all the data to be considered the character data type (instead of being Factor, int types)

Examine its initial structure:

> str(t)
'data.frame':   2 obs. of  2 variables:
 $ Series  : chr  "Quantico" "Muppets"
 $ FirstAir: chr  "09272015" "09222015"

R has imported all as strings of characters, indicated here as type chr

The chr or string of characters are then easily converted into a date:

> t$FirstAir = as.Date(t$FirstAir, "%m%d%Y")

as.Date() performs string to date conversion
%m%d%Y specifies how to interpret the input in t$FirstAir. These format codes, at least on Linux, can be found with running $ man date which brings up the manual on the date program, where there is a list of formatting codes. For example it says %m month (01..12)

Method 2: Import then fix only the date

If for some reason you don't want a blanket import conversion to all characters, for example a file with many variables and wish to leave R's auto type recognition in use but merely "fix" the one date variable, follow this method.

Once within R,

> t = read.csv('tv.csv')

imports tv.csv as a data frame named t

Examine its initial structure:

> str(t)
'data.frame':   2 obs. of  2 variables:
 $ Series  : Factor w/ 2 levels "Muppets","Quantico": 2 1
 $ FirstAir: int  9272015 9222015
>

R tries its best to guess the variable type per variable
As you can see an immediate problem is, for FirstAir variable R has imported 09272015 as int meaning integer, and dropped off the leading zero padding , the 0 in 09 is important later for date conversion yet R has imported it without. So we need to fix this.

This can be done in a single command but for clarity I have broken this into two steps. First,

> t$FirstAir = sprintf("%08d", t$FirstAir)

sprintf is a formatting function
0 means pad with zeroes
8 means ensure 8 characters, because mmddyyyy is total 8 characters
d is used when the input is a number, which currently it is, recall str() output claimed the t$FirstAir is an int meaning integer
t$FirstAir is the variable we are both setting and using as input

Check the result:

> str(t$FirstAir)
 chr [1:2] "09272015" "09222015"

it successfully converted from an int to a chr type, for example 9272015 became "09272015"

Now it is a string or chr type we can then convert, same as method 1.

> t$FirstAir = as.Date(strptime(t$FirstAir, "%m%d%Y"))

Result

We do a final check:

> str(t$FirstAir)
 Date[1:2], format: "2015-09-27" "2015-09-22"

In both cases, what were original values in a text file are have now been successfully converted into R date objects.

来源：https://stackoverflow.com/questions/32854538/converting-a-character-string-into-a-date-in-r

标签

string

date

grep

format