Data Wrangling with R

Aravind Hebbali

12 Date & Time: Parsing and Components

13 Date & Time: Parsing and Components

Continued in part B

This chapter was split for readability. Arithmetic, intervals, time zones and rounding continue in @date-arithmetic-time-zones.

This chapter is the first of two on dates and times. Here we parse character data into dates and extract components (year, month, weekday, quarter). @date-arithmetic-time-zones continues with arithmetic, intervals, time zones, rounding and rollback.

13.1 Introduction

Let us begin by looking at the current date and time.

13.1.1 Date

Sys.Date() and today() will return the current date.

Sys.Date()

[1] "2026-10-01"

lubridate::today()

[1] "2026-10-01"

13.1.2 Time

Sys.time() and now() return the date, time and the timezone. In now(), we can specify the timezone using the tzone argument.

Sys.time()

[1] "2026-10-01 17:42:57 IST"

lubridate::now()

[1] "2026-10-01 17:42:57 IST"

lubridate::now(tzone = "UTC")

[1] "2026-10-01 12:12:57 UTC"

13.1.3 AM or PM?

am() and pm() allow us to check whether date/time occur in the am or pm? They return a logical value i.e. TRUE or FALSE

lubridate::am(now())

[1] FALSE

lubridate::pm(now())

[1] TRUE

13.1.4 Leap Year

We can also check if the current year is a leap year using leap_year().

lubridate::leap_year(Sys.Date())

[1] FALSE

13.1.5 Summary

Function	Description
`Sys.Date()`	Current Date
`lubridate::today()`	Current Date
`Sys.time()`	Current Time
`lubridate::now()`	Current Time
`lubridate::am()`	Whether time occurs in am?
`lubridate::pm()`	Whether time occurs in pm?
`lubridate::leap_year()`	Check if the year is a leap year?

13.1.6 Your Turn

get current date
get current time
check whether the time occurs in am or pm?
check whether the following years were leap years
- 2018
- 2016

13.2 Case Study

Throughout the tutorial, we will work on a case study related to transactions of a imaginary company. The data set includes information about invoice and payment dates.

13.2.1 Data

transact <- readr::read_csv('https://raw.githubusercontent.com/rsquaredacademy/datasets/master/transact.csv')

# A tibble: 2,466 × 3
   Invoice    Due        Payment   
   <date>     <date>     <date>    
 1 2013-01-02 2013-02-01 2013-01-15
 2 2013-01-26 2013-02-25 2013-03-03
 3 2013-07-03 2013-08-02 2013-07-08
 4 2013-02-10 2013-03-12 2013-03-17
 5 2012-10-25 2012-11-24 2012-11-28
 6 2012-01-27 2012-02-26 2012-02-22
 7 2013-08-13 2013-09-12 2013-09-09
 8 2012-12-16 2013-01-15 2013-01-12
 9 2012-05-14 2012-06-13 2012-07-01
10 2013-07-01 2013-07-31 2013-07-26
# ℹ 2,456 more rows

13.2.2 Data Dictionary

The data set has 3 columns. All the dates are in the format (yyyy-mm-dd).

Column	Description
Invoice	Invoice Date
Due	Due Date
Payment	Payment Date

In the case study, we will try to answer a few questions we have about the transact data.

extract date, month and year from Due
compute the number of days to settle invoice
compute days over due
check if due year is a leap year
check when due day in february is 29, whether it is a leap year
how many invoices were settled within due date
how many invoices are due in each quarter

13.3 Date & Time Classes

13.3.1 Introduction

In this section, we will look at two things. First, how to create date/time data in R, and second, how to convert other data types to date/time. Let us begin by creating the latest R release date manually.

release_date <- 2019-12-12
release_date

[1] 1995

Okay! Why do we see 1995 when we call the date? What is happening here? Let us quickly check the data type of release_date.

class(release_date)

[1] "numeric"

The data type is numeric i.e. R has subtracted 12 twice from 2019 to return 1995. Clearly, the above method is not the right way to store date/time. Let us see if we can get some hints from the builtin R functions we used in the previous section. If you observe the output, all of them returned date/time wrapped in quotes. Hmmm… let us wrap our date in quotes and see what happens.

release_date <- "2019-12-12"
release_date

[1] "2019-12-12"

Alright, now R does not do any arithmetic and returns the date as we specified. Great! Is this the right format to store date/time? No. Why? What is the problem if date/time is saved as character/string? The problem is the nature or type of operations done on date or time is different when compared to string/character, number or logical values.

how do we add/subtract dates?
how do we extract components such as year, month, day etc.

To answer the above questions, we will first check the data type of Sys.Date() and now().

class(Sys.Date())

[1] "Date"

class(lubridate::now())

[1] "POSIXct" "POSIXt"

class(release_date)

[1] "character"

As you can see from the above output, there are 3 different classes for storing date/time in R

Date
POSIXct
POSIXlt

Let us explore each of the above classes one by one.

13.3.2 Date

13.3.2.1 Introduction

The Date class represents calendar dates. Let us go back to Sys.Date(). If you check the class of Sys.Date(), it is Date. Internally, this date is a number i.e. an integer. The unclass() function will show dates are stored internally.

unclass(Sys.Date())

[1] 20727

What does this integer represent? Why has R stored the date as an integer? Before we answer this question, we need to know something else. In R, dates are represented as the number of days since 1970-01-01. All the dates in R are internally stored in this way. Before we explore this concept further, let us learn to create Date objects in R. We will continue to use the latest R release date, 2019-12-12.

Until now, we have stored the above date as character/string but now we will use as.Date() to save it as a Date object. as.Date() is the easiest and simplest way to create dates in R.

release_date <- as.Date("2019-12-12")
release_date

[1] "2019-12-12"

The as_date() function from the lubridate package is similar to as.Date().

release_date <- lubridate::as_date("2019-12-12")
release_date

[1] "2019-12-12"

If you look at the difference between release_date and 1970-01-01, it will be the same as unclass(release_date).

release_date - as.Date("1970-01-01")

Time difference of 18242 days

unclass(release_date)

[1] 18242

Let us come back to 1970-01-01 i.e. the origin for dates in R.

lubridate::origin

[1] "1970-01-01 UTC"

From the previous examples, we know that dates are internally stored as number of days since 1970-01-01. How about dates older than the origin? How are they stored? Let us look at that briefly.

unclass(as.Date("1963-08-28"))

[1] -2318

Dates older than the origin are stored as negative integers. For those who are not aware, Martin Luther King, Jr. delivered his famous I Have a Dream speech on 1963-08-28. Let us move on and learn how to convert numbers into dates.

13.3.2.2 Convert Numeric

The as.Date() function can be used to convert any of the following to a Date object

character/string
number
factor (categorical/qualitative)

We have explored how to convert strings to date. How about converting numbers to date? Sure, we can create date from numbers by specifying the origin and number of days since it.

as.Date(18242, origin = "1970-01-01")

[1] "2019-12-12"

The origin can be changed to another date (while changing the number as well.)

as.Date(7285, origin = "2000-01-01")

[1] "2019-12-12"

13.3.3 ISO 8601

If you have carefully observed, the format in which we have been specifying the dates as well as of those returned by functions such as Sys.Date() or Sys.time() is the same i.e. YYYY-MM-DD. It includes

the year including the century
the month
the date

The month and date separated by -. This default format used in R is the ISO 8601 standard for date/time. ISO 8601 is the internationally accepted way to represent dates and times and uses the 24 hour clock system. Let us create the release date using another function ISOdate().

ISOdate(year  = 2019,
        month = 12,
        day   = 12,
        hour  = 8,
        min   = 5, 
        sec   = 3,
        tz    = "UTC")

[1] "2019-12-12 08:05:03 UTC"

We will look at all the different weird ways in which date/time are specified in the real world in the Date & Time Formats section. For the time being, let us continue exploring date/time classes in R. The next class we are going to look at is POSIXct/POSIXlt.

13.3.4 POSIX

You might be wondering what is this POSIX thing? POSIX stands for Portable Operating System Interface. It is a family of standards specified f or maintaining compatibility between different operating systems. Before we learn to create POSIX objects, let us look at now() from lubridate.

class(lubridate::now())

[1] "POSIXct" "POSIXt"

now() returns current date/time as a POSIXct object. Let us look at its internal representation using unclass()

unclass(lubridate::now())

[1] 1790856789
attr(,"tzone")
[1] ""

The output you see is the number of seconds since January 1, 1970.

13.3.4.1 POSIXct

POSIXct represents the number of seconds since the beginning of 1970 (UTC) and ct stands for calendar time. To store date/time as POSIXct objects, use as.POSIXct(). Let us now store the latest R release date as POSIXct as shown below

release_date <- as.POSIXct("2019-12-12 08:05:03")
class(release_date)

[1] "POSIXct" "POSIXt"

unclass(release_date)

[1] 1576118103
attr(,"tzone")
[1] ""

13.3.4.2 POSIXlt

POSIXlt represents the following information in a list

seconds
minutes
hour
day of the month
month
year
day of week
day of year
daylight saving time flag
time zone
offset in seconds from GMT

The lt in POSIXlt stands for local time. Use as.POSIXlt() to store date/time as POSIXlt objects. Let us store the release date as a POSIXlt object as shown below

release_date <- as.POSIXlt("2019-12-12 08:05:03")
release_date

[1] "2019-12-12 08:05:03 IST"

As we said earlier, POSIXlt stores date/time components in a list and these can be extracted. Let us look at the date/time components returned by POSIXlt using unclass().

release_date <- as.POSIXlt("2019-12-12 08:05:03")
unclass(release_date)

$sec
[1] 3

$min
[1] 5

$hour
[1] 8

$mday
[1] 12

$mon
[1] 11

$year
[1] 119

$wday
[1] 4

$yday
[1] 345

$isdst
[1] 0

$zone
[1] "IST"

$gmtoff
[1] NA

attr(,"tzone")
[1] ""      "IST"   "+0630"
attr(,"balanced")
[1] TRUE

Use unlist() if you want the components returned as a vector.

release_date <- as.POSIXlt("2019-12-12 08:05:03")
unlist(release_date)

   sec    min   hour   mday    mon   year   wday   yday  isdst   zone gmtoff 
   "3"    "5"    "8"   "12"   "11"  "119"    "4"  "345"    "0"  "IST"     NA

To extract specific components, use $.

release_date <- as.POSIXlt("2019-12-12 08:05:03")
release_date$hour

[1] 8

release_date$mon

[1] 11

release_date$zone

[1] "IST"

Now, let us look at the components returned by POSIXlt. Some of them are intuitive

Component	Description
`sec`	Second
`min`	Minute
`hour`	Hour of the day
`mon`	Month of the year (0-11
`zone`	Timezone
`wday`	Day of week
`mday`	Day of month
`year`	Years since 1900
`yday`	Day of year
`isdst`	Daylight saving flag
`gmtoff`	Offset is seconds from GMT

Great! We will end this section with a few tips/suggestions on when to use Date or POSIXct/POSIXlt.

use Date when there is no time component
use POSIX when dealing with time and timezones
use POSIXlt when you want to access/extract the different components

13.3.5 Your Turn

R 1.0.0 was released on 2000-02-29 08:55:23 UTC. Save it as

Date using character
Date using origin and number
POSIXct
POSIXlt and extract
- month day
- day of year
- month
- zone
ISODate

13.4 Date & Time Formats

13.4.1 Introduction

After the timezones and daylight savings detour, let us get back on path and explore another important aspect, date & time formats. Although it is a good practice to adher to ISO 8601 format, not all date/time data will comply with it. In real world, date/time data may come in all types of weird formats. Below is a sample

Format
December 12, 2019
12th Dec, 2019
Dec 12th, 19
12-Dec-19
2019 December
12.12.19

When the data is not in the default ISO 8601 format, we need to explicitly specify the format in R. We do this using conversion specifications. A conversion specification is introduced by %, usually followed by a single letter or O or E and then a single letter.

13.4.2 Conversion Specifications

Specification	Description	Example
`%d`	Day of the month (decimal number)	12
`%m`	Month (decimal number)	12
`%b`	Month (abbreviated)	Dec
`%B`	Month (full name)	December
`%y`	Year (2 digit)	19
`%Y`	Year (4 digit)	2019
%H	Hour	8
%M	Minute	5
%S	Second	3

Time to work through a few examples. Let us say you are dealing with dates in the format 19/12/12. In this format, the year comes first followed by month and the date; each separated by a backslash (/). The year consists of only 2 digits i.e. it does not include the century. Let us now map each component of the date to the format table shown at the beginning.

Date	Specification
19	`%y`
12	`%m`
12	`%d`

Using the format argument, we will specify the date format as a character vector i.e. enclosed in quotes.

as.Date("19/12/12", format = "%y/%m/%d")

[1] "2019-12-12"

Another way in which the release data can be written is 2019-Dec-12. We still have the year followed by the month and the date but there are a few changes here:

the components are separated by a - instead of /
year has 4 digits i.e. includes the century
the month is specified using abbreviation instead of digits

Let us map the components to the format table:

Date	Specification
2019	`%Y`
Dec	`%b`
12	`%d`

Let us specify the format for the date using the above mapping.

as.Date("2019-Dec-12", format = "%Y-%b-%d")

[1] "2019-12-12"

In both the above examples, we have not dealt with time components. Let us include the time of the latest R release in the next one i.e. 19/12/12 08:05:03.

Date	Specification
19	`%y`
12	`%m`
12	`%d`
08	`%H`
05	`%M`
03	`%S`

Since we are dealing with time, we will use as.POSIXct() instead of as.Date().

as.POSIXct("19/12/12 08:05:03", tz = "UTC", format = "%y/%m/%d %H:%M:%S")

[1] "2019-12-12 08:05:03 UTC"

In the below table, we look at some of the most widely used conversion specifications. You can learn more about these specifications by running ?strptime or help(strptime).

Specification	Description
`%a`	Abbreviated weekday
`%A`	Full weekday
`%C`	Century (00-99)
`%D`	Same as `%m/%d/%y`
`%e`	Day of month [1 - 31]
`%F`	Same as `%Y-%m-%d`
`%h`	Same as `%b`
`%I`	Hours as decimal [01 - 12]
`%j`	Day of year [001 - 366]
`%R`	Same as `%H:%M`
`%t`	Tab
`%T`	Same as `%H:%M:%S`
`%u`	Weekday [1 - 7](Monday is 1
`%U`	Week of year [00 - 53]
`%V`	Week of year [01 - 53]
`%w`	Weekday [0 - 6](sunday is 0
`%W`	Week of year [00 - 53]

We have included a lot of practice questions for you to explore the different date/time formats. The solutions are available in the Learning Management system as well as in our GitHub repo. Try them and let us know if you have any doubts.

13.4.3 Guess Format

guess_formats() from lubridate is a very useful function. It will guess the date/time format if you specify the order in which year, month, date, hour, minute and second appear.

release_date_formats <- c("December 12th 2019",
                        "Dec 12th 19",
                        "dec 12 2019")

guess_formats(release_date_formats, 
              orders = "mdy", 
              print_matches = TRUE)

                          Omdy          mdy         
[1,] "December 12th 2019" "%Om %dth %Y" "%B %dth %Y"
[2,] "Dec 12th 19"        "%Om %dth %y" "%b %dth %y"
[3,] "dec 12 2019"        "%Om %d %Y"   "%b %d %Y"

         Omdy          Omdy          Omdy           mdy           mdy 
"%Om %dth %Y" "%Om %dth %y"   "%Om %d %Y"  "%B %dth %Y"  "%b %dth %y" 
          mdy 
   "%b %d %Y"

13.4.4 Your Turn

Below, we have specified July 5th, 2019 in different ways. Create the date using as.Date() while specifying the correct format for each of them.

July-05-19
JUL-05-19
05.07.19
5-July 2019
July 5th, 2019
July 05, 2019
2019-July- 05
05/07/2019
07/05/2019
7/5/2019
07/5/19
2019-07-05

13.5 Parse Date & Time

While creating date-time objects, we specified different formats using the conversion specification but most often you will not create date/time and instead deal with data thay comes your way from a system or colleague/collaborator. In such cases, we need to be able to parse date/time from the data provided to us. In this section, we will focus on parsing date/time from character data. Both base R and the lubridate package offer functions to parse date and time and we will explore a few of them in this section. We will initially use functions from base R and later on explore those from lubridate which will give us an opportunity to compare and contrast. It will also allow us to choose the functions based on the data we are dealing with.

strptime() will convert character data to POSIXlt. You will use this when converting from character data to date/time. On the other hand, if you want to convert date/time to character data, use any of the following:

strftime()
format()
as.character()

The above functions will convert POSIXct/POSIXlt to character. Let us start with a simple example. The data we have been supplied has date/time as character data and in the format YYYYMMDD i.e. nothing separates the year, month and date from each other. We will use strptime() to convert this to an object of class POSIXlt.

rel_date <- strptime("20191212", format = "%Y%m%d")
class(rel_date)

[1] "POSIXlt" "POSIXt"

If you have a basic knowledge of conversion specifications, you can use strptime() to convert character data to POSIXlt. Let us quickly explore the functions to convert date/time to character data before moving on to the functions from lubridate.

rel_date_strf <- strftime(rel_date)
class(rel_date_strf)

[1] "character"

rel_date_format <- format(rel_date)
class(rel_date_format)

[1] "character"

rel_date_char <- as.character(rel_date)
class(rel_date_char)

[1] "character"

As you can see, all the 3 functions converted date/time to character. Time to move on and explore the lubridate package. We will start with an example in which the release date is formatted in 3 different ways but they have one thing in common i.e. the order in which the components appear. In all the 3 formats, the year is followed by the month and then the date.

To parse the release date, we will use parse_date_time() from lubridate which parses the input into POSIXct objects.

release_date <- c("19-12-12", "20191212", "19-12 12")
parse_date_time(release_date, "ymd")

[1] "2019-12-12 UTC" "2019-12-12 UTC" "2019-12-12 UTC"

parse_date_time(release_date, "y m d")

[1] "2019-12-12 UTC" "2019-12-12 UTC" "2019-12-12 UTC"

parse_date_time(release_date, "%y%m%d")

[1] "2019-12-12 UTC" "2019-12-12 UTC" "2019-12-12 UTC"

Try to use strptime() in the above example and see what happens. Now, let us look at another data set.

release_date <- c("19-07-05", "2019-07-05", "05-07-2019", "07-05-2019")

What happens in the below case? The same date appears in multiple formats. How do we parse them? parse_date_time() allows us to specify mutiple date-time formats. Let us first map the dates to their formats.

Date	Specification
19-07-05	`ymd`
2019-07-05	`ymd`
05-07-2019	`dmy`
07-05-2019	`mdy`

The above specifications can be supplied as a character vector.

parse_date_time(release_date, c("ymd", "ymd", "dmy", "mdy"))

[1] "2019-07-05 UTC" "2019-07-05 UTC" "2019-07-05 UTC" "2019-05-07 UTC"

Great! We have used both strptime() and parse_date_time() now. Can you tell what differentiates parse_date_time() when compared to strptime()? We summarize it in the points below:

no need to include % prefix or separator
specify several date/time formats

There are other helper functions that can be used to

parse dates with year, month, day components
parse dates with year, month, day, hour, minute, seconds components
parse period with hour, minute, second components

and are explored in the below examples.

# year/month/date
ymd("2019-12-12")

[1] "2019-12-12"

# year/month/date
ymd("19/12/12")

[1] "2019-12-12"

# date/month/year
dmy(121219)

[1] "2019-12-12"

# year/month/date/hour/minute/second
ymd_hms(191212080503)

[1] "2019-12-12 08:05:03 UTC"

# hour/minute/second
hms("8, 5, 3")

[1] "8H 5M 3S"

# hour/minute/second
hms("08:05:03")

[1] "8H 5M 3S"

# minute/second
ms("5,3")

[1] "5M 3S"

# hour/minute
hm("8, 5")

[1] "8H 5M 0S"

Note, in a couple of cases where the components are not separated by /, - or space, we have not enclosed the values in quotes.

13.5.1 Your Turn

Below, we have specified July 5th, 2019 in different ways. Parse the dates using strptime() or parse_date_time() or any other helper function.

July-05-19
JUL-05-19
05.07.19
5-July 2019
July 5th, 2019
July 05, 2019
2019-July- 05
05/07/2019
07/05/2019
7/5/2019
07/5/19
2019-07-05

13.6 Date & Time Components

In the second section, we discussed the downside of saving date/time as character/string in R. One of the points we discussed was that we can’t extract components such as year, month, day etc. In this section, we will learn to extract date/time components such as

year
month
date
week
day
quarter
semester
hour
minute
second
timezone

The below table outlines the functions we will explore in the first part of this section.

Function	Description
`year()`	Get year
`month()`	Get month (number)
`month(label = TRUE)`	Get month (abbreviated name)
`month(abbr = FALSE)`	Get month (full name)
`months()`	Get month
`week()`	Get week

13.6.1 Year

release_date <- ymd_hms("2019-12-12 08:05:03")
year(release_date)

[1] 2019

13.6.2 Month

month() will return the month as a number i.e. 07 for July.

month(release_date)

[1] 12

If you want the name of the month instead, use the label argument and set it to TRUE. Now it returns Jul instead of 07.

month(release_date, label = TRUE)

[1] Dec
12 Levels: Jan < Feb < Mar < Apr < May < Jun < Jul < Aug < Sep < ... < Dec

But this is the abbreviated name and not the full name. How do we get the full name of the month? Set the abbr argument to FALSE.

month(release_date, label = TRUE, abbr  = FALSE)

[1] December
12 Levels: January < February < March < April < May < June < ... < December

Ah! now we can see the full name of the month. months() from base R will return the full name of the month by default. If you want the abbreviated name, use the abbreviate argument and set it to TRUE.

months(release_date)

[1] "December"

13.6.3 Week

week() returns the number of complete 7 day periods between the date and 1st January plus one.

week(release_date)

[1] 50

13.6.4 Day

Use day() to extract the date component. There are other variations such as

Function	Description
`day`	Get day
`mday()`	Day of the month
`wday()`	Day of the week
`qday()`	Day of quarter
`yday()`	Day of year
`weekdays()`	Day of week
`days_in_month()`	Days in the month

day(release_date)

[1] 12

mday(release_date)

[1] 12

qday(release_date)

[1] 73

yday(release_date)

[1] 346

wday can return

a number
abbreviation of the weekday
full name of the weekday

wday(release_date)

[1] 5

wday(release_date, label = TRUE)

[1] Thu
Levels: Sun < Mon < Tue < Wed < Thu < Fri < Sat

wday(release_date, label = TRUE, abbr  = FALSE)

[1] Thursday
7 Levels: Sunday < Monday < Tuesday < Wednesday < Thursday < ... < Saturday

weekdays() from base R also returns the day of the week (the name and not the number). If you want the abbreviated name, use the abbreviate argument.

weekdays(release_date)

[1] "Thursday"

weekdays(release_date, abbreviate = TRUE)

[1] "Thu"

13.6.5 Days in Month

If you want to know the number of days in the month, use days_in_month(). In our example, the month is December and it has 31 days.

days_in_month(release_date)

Dec 
 31

13.6.6 Hour, Minute & Seconds

Function	Description
`hour()`	Get hour
`minute()`	Get minute
`second()`	Get second
`seconds()`	Number of seconds since `1970-01-01`

So far we have been looking at date components. Now, let us look at time components.

hour(release_date)

[1] 8

minute(release_date)

[1] 5

second(release_date)

[1] 3

seconds() returns the number of seconds since 1970-01-01.

seconds(release_date)

[1] "1576137903S"

13.6.7 Quarter & Semester

quarter() will return the quarter from the date. December is in the 4th quarter and hence it returns 4.

quarter(release_date)

[1] 4

If you want the year along with the quarter, set the with_year argument to TRUE.

quarter(release_date, with_year = TRUE)

[1] 2019.4

In India, the fiscal starts in April and December falls in the 3rd quarter. How can we accommodate this change? The fiscal_start argument allows us to set the month in which the fiscal begins. We will set it to 4 for April. Now it returns 3 instead of 4.

quarter(release_date, fiscal_start = 4)

[1] 3

quarters() from base R also returns the quarter.

quarters(release_date)

[1] "Q4"

Function	Description
`quarter()`	Get quarter
`quarter(with_year = TRUE)`	Quarter with year
`quarter(fiscal_start = 4)`	Fiscal starts in April
`quarters()`	Get quarter
`semester()`	Get semester

13.6.8 Case Study

13.6.8.1 Extract Date, Month & Year from Due Date

Let us now extract the date, month and year from the Due column.

transact |>
  mutate(
    due_day   = day(Due),
    due_month = month(Due),
    due_year  = year(Due)
  )

# A tibble: 2,466 × 6
   Invoice    Due        Payment    due_day due_month due_year
   <date>     <date>     <date>       <int>     <dbl>    <dbl>
 1 2013-01-02 2013-02-01 2013-01-15       1         2     2013
 2 2013-01-26 2013-02-25 2013-03-03      25         2     2013
 3 2013-07-03 2013-08-02 2013-07-08       2         8     2013
 4 2013-02-10 2013-03-12 2013-03-17      12         3     2013
 5 2012-10-25 2012-11-24 2012-11-28      24        11     2012
 6 2012-01-27 2012-02-26 2012-02-22      26         2     2012
 7 2013-08-13 2013-09-12 2013-09-09      12         9     2013
 8 2012-12-16 2013-01-15 2013-01-12      15         1     2013
 9 2012-05-14 2012-06-13 2012-07-01      13         6     2012
10 2013-07-01 2013-07-31 2013-07-26      31         7     2013
# ℹ 2,456 more rows

13.6.8.2 Data Sanitization

Let us do some data sanitization. If the due day happens to be February 29, let us ensure that the due year is a leap year. Below are the steps to check if the due year is a leap year:

we will extract the following from the due date:
- day
- month
- year
we will then create a new column is_leap which will have be set to TRUE if the year is a leap year else it will be set to FALSE
filter all the payments due on 29th Feb
select the following columns:
- Due
- is_leap

transact |>
  mutate(
    due_day   = day(Due),
    due_month = month(Due),
    due_year  = year(Due),
    is_leap   = leap_year(due_year)
  ) |>
  filter(due_month == 2 & due_day == 29) |>
  select(Due, is_leap)

# A tibble: 4 × 2
  Due        is_leap
  <date>     <lgl>  
1 2012-02-29 TRUE   
2 2012-02-29 TRUE   
3 2012-02-29 TRUE   
4 2012-02-29 TRUE

13.6.8.3 Invoices Distribution by Quarter

Let us count the invoices due for each quarter.

transact |>
  mutate(
    quarter_due = quarter(Due)
  ) |>
  count(quarter_due)

# A tibble: 4 × 2
  quarter_due     n
        <int> <int>
1           1   521
2           2   661
3           3   618
4           4   666

13.6.9 Your Turn

Using the transact data, tabulate the following from the Due column:

year
month with label
weekday with label
hour
and quarter

13.7 Create, Update & Verify

In the second section, we learnt to create date-time objects using as.Date(), as.POSIXct() etc. In this section, we will explore a few other functions that will allow us to do the same

make_date()
make_datetime()

13.7.1 Create

To create date without time components, use make_date() and specify the following:

year
month
date

We need to specify all the components in numbers i.e. we cannot use Jul or July for the month. It has to be 7.

make_date(year  = 2019,
          month = 12,
          day   = 12)

[1] "2019-12-12"

When you need to include time components, use make_datetime().

make_datetime(year  = 2019,
              month = 12,
              day   = 12,
              hour  = 08,
              min   = 05,
              sec   = 03,
              tz    = "UTC")

[1] "2019-12-12 08:05:03 UTC"

13.7.2 Update

Let us look at another scenario. You have a date-time object and want to change one of its components i.e. any of the following

year
month
date

Instead of creating another date-time object, you can change any of the components using update(). In the below example, we will start with the date of release of R version 3.6.1 and using update(), we will change it to 2019-12-12.

prev_release <- ymd("2019-07-05")
prev_release |> 
  update(year  = 2019,
         month = 12,
         mday  = 12)

[1] "2019-12-12"

13.7.3 Date Sequence

So far we have created a single date-time instance. How about creating a sequence of dates? We can do that using seq.Date(). We need to specify the from date as the bare minimum input. If the end date is not specified, it will create the sequence uptil the current date.

The interval of the sequence can be specified in any of the following units:

day
week
month
quarter
year

We can add the following to the interval units

integer
+ / - (increment or decrement)

Using the integer, we can specify multiples of the units mentioned and using the sign, we can specify whether to increment or decrement.

The below table displays the main arguments used in seq.Date():

Function	Description
`from`	Starting date of the sequence
`by`	End date of the sequence
`to`	Date increment of the sequence
`length.out`	Length of the sequence
`along.with`	Use length of this value as length of sequence

In the first example, we will create a sequence of dates from 2010-01-01 to 2019-12-31. The unit of increment should be a year i.e. the difference between the dates in the sequence should be 1 year, specified using the by argument.

seq.Date(from = as.Date("2010-01-01"), to = as.Date("2019-12-31"), by = "year")

 [1] "2010-01-01" "2011-01-01" "2012-01-01" "2013-01-01" "2014-01-01"
 [6] "2015-01-01" "2016-01-01" "2017-01-01" "2018-01-01" "2019-01-01"

In the next example, we change the unit of increment to a quarter i.e. the difference between the dates in the sequence should be a quarter or 3 months.

seq.Date(from = as.Date("2009-12-12"), to = as.Date("2019-12-12"), by = "quarter")

 [1] "2009-12-12" "2010-03-12" "2010-06-12" "2010-09-12" "2010-12-12"
 [6] "2011-03-12" "2011-06-12" "2011-09-12" "2011-12-12" "2012-03-12"
[11] "2012-06-12" "2012-09-12" "2012-12-12" "2013-03-12" "2013-06-12"
[16] "2013-09-12" "2013-12-12" "2014-03-12" "2014-06-12" "2014-09-12"
[21] "2014-12-12" "2015-03-12" "2015-06-12" "2015-09-12" "2015-12-12"
[26] "2016-03-12" "2016-06-12" "2016-09-12" "2016-12-12" "2017-03-12"
[31] "2017-06-12" "2017-09-12" "2017-12-12" "2018-03-12" "2018-06-12"
[36] "2018-09-12" "2018-12-12" "2019-03-12" "2019-06-12" "2019-09-12"
[41] "2019-12-12"

We will now create a sequence of dates but instead of specifying the unit of increment, we specify the number of dates in the sequence i.e. the length of the sequence. We do this using the length.out argument which specifies the desired length of the sequence. We want the sequence to have 10 dates including the start and end date, and hence we supply the value 10 for the length.out argument.

seq.Date(from = as.Date("2010-01-01"), to = as.Date("2019-12-31"), length.out = 10)

 [1] "2010-01-01" "2011-02-10" "2012-03-22" "2013-05-02" "2014-06-11"
 [6] "2015-07-22" "2016-08-31" "2017-10-10" "2018-11-20" "2019-12-31"

In all of the previous examples, we have specified both the start and the end date. Let us look at a few examples where we create a sequence of dates where we only specify the start date. In the below example, we want to create a sequence of dates starting from 2010-01-01. The unit of increment should be 1 year i.e. the difference between the dates in the sequence should be 1 year and the length of the sequence should be 10 i.e. the number of dates including the start date should be 10.

seq.Date(from = as.Date("2010-01-01"), by = "year", length.out = 10)

 [1] "2010-01-01" "2011-01-01" "2012-01-01" "2013-01-01" "2014-01-01"
 [6] "2015-01-01" "2016-01-01" "2017-01-01" "2018-01-01" "2019-01-01"

The unit of increment can include multiples and +/- sign i.e. it can be an unit of increment or decrement. In the next example, we can increment the dates in the sequence by 2 i.e. the difference between the dates should be 2 instead of 1. This is achieved by specifying the unit of increment (multiple) first followed by a space and then the unit. In our example, it is 2 year. As you can see, the sequence now goes all the way till 2028 and the gap between the dates is 2 years.

seq.Date(from = as.Date("2010-01-01"), by = "2 year", length.out = 10)

 [1] "2010-01-01" "2012-01-01" "2014-01-01" "2016-01-01" "2018-01-01"
 [6] "2020-01-01" "2022-01-01" "2024-01-01" "2026-01-01" "2028-01-01"

Let us say instead of increment we want to decrement the dates i.e. the sequence of dates will go backwards as shown in the next example. We achieve this by using the - sign along with the unit of decrement. The sequence of dates in next example starts from 2010 and goes back upto 1992 and the difference between the dates in 2 years.

seq.Date(from = as.Date("2010-01-01"), by = "-2 year", length.out = 10)

 [1] "2010-01-01" "2008-01-01" "2006-01-01" "2004-01-01" "2002-01-01"
 [6] "2000-01-01" "1998-01-01" "1996-01-01" "1994-01-01" "1992-01-01"

In the last example, we will explore the along.with argument. Here we have supplied a vector which is a sequence of numbers from 1 to 10. The length of this vector is 10 and the same length is used as the length of the sequence i.e. the length of value supplied to along.with is also the length of the sequence.

seq.Date(from = as.Date("2010-01-01"), by = "-2 year", along.with = 1:10)

 [1] "2010-01-01" "2008-01-01" "2006-01-01" "2004-01-01" "2002-01-01"
 [6] "2000-01-01" "1998-01-01" "1996-01-01" "1994-01-01" "1992-01-01"

13.7.4 Verify Type

How do you check if the data is a date-time object? You can do that using any of the following from the lubridate package.

is.Date()
is.POSIXct()
is.POSIXlt()

is.Date(release_date)

[1] FALSE

is.POSIXct(release_date)

[1] TRUE

is.POSIXlt(release_date)

[1] FALSE

13.7.5 Your Turn

R 2.0.0 was released on 2004-10-04 14:24:38. Create this date using both make_date() and make_datetime()
R 3.0.0 was released on 2013-04-03 07:12:36. Update the date created in the previous step to the above using update()

Try it live

Try it in your browser

Edit the code below and press Run. It executes entirely in your browser via WebR — no R installation needed. The runtime downloads once, on this page only.

--- title: "Date & Time: Parsing and Components" description: "Data Wrangling in R: parse dates with ymd/mdy/dmy, extract year, month and weekday components." --- # Date & Time: Parsing and Components {#date-and-time-in-r} :::{.callout-note} ## Continued in part B This chapter was split for readability. Arithmetic, intervals, time zones and rounding continue in @date-arithmetic-time-zones. ::: This chapter is the first of two on dates and times. Here we parse character data into dates and extract components (year, month, weekday, quarter). @date-arithmetic-time-zones continues with arithmetic, intervals, time zones, rounding and rollback. ## Introduction ```{r pkg_load, echo=FALSE, message=FALSE, warning=FALSE} library(knitr) library(kableExtra) library(magrittr) library(lubridate) library(readr) library(dplyr) ``` Let us begin by looking at the current date and time. ### Date `Sys.Date()` and `today()` will return the current date. ```{r sys_date} Sys.Date() lubridate::today() ``` ### Time `Sys.time()` and `now()` return the date, time and the timezone. In `now()`, we can specify the timezone using the `tzone` argument. ```{r sys_time} Sys.time() lubridate::now() lubridate::now(tzone = "UTC") ``` ### AM or PM? `am()` and `pm()` allow us to check whether date/time occur in the am or pm? They return a logical value i.e. `TRUE` or `FALSE` ```{r lub_am} lubridate::am(now()) lubridate::pm(now()) ``` ### Leap Year We can also check if the current year is a leap year using `leap_year()`. ```{r leap_year} lubridate::leap_year(Sys.Date()) ``` ### Summary ```{r table_current_date_time, echo=FALSE} cname <- c("`Sys.Date()`", "`lubridate::today()`", "`Sys.time()`", "`lubridate::now()`", "`lubridate::am()`", "`lubridate::pm()`", "`lubridate::leap_year()`") descrip <- c("Current Date", "Current Date", "Current Time", "Current Time", "Whether time occurs in am?", "Whether time occurs in pm?", "Check if the year is a leap year?") data.frame(Function = cname, Description = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` ### Your Turn - get current date - get current time - check whether the time occurs in am or pm? - check whether the following years were leap years - 2018 - 2016 ## Case Study Throughout the tutorial, we will work on a case study related to transactions of a imaginary company. The data set includes information about invoice and payment dates. ### Data ```{r import, eval=FALSE} transact <- readr::read_csv('https://raw.githubusercontent.com/rsquaredacademy/datasets/master/transact.csv') ``` ```{r show, echo=FALSE, eval=TRUE, message=FALSE} transact <- readr::read_csv('https://raw.githubusercontent.com/rsquaredacademy/datasets/master/transact.csv') transact ``` ### Data Dictionary The data set has 3 columns. All the dates are in the format (yyyy-mm-dd). <br> <br> ```{r table_data_dict, echo=FALSE} cname <- c("Invoice", "Due", "Payment") descrip <- c("Invoice Date", "Due Date", "Payment Date") data.frame(Column = cname, Description = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` In the case study, we will try to answer a few questions we have about the `transact` data. - extract date, month and year from Due - compute the number of days to settle invoice - compute days over due - check if due year is a leap year - check when due day in february is 29, whether it is a leap year - how many invoices were settled within due date - how many invoices are due in each quarter ## Date & Time Classes ### Introduction In this section, we will look at two things. First, how to create date/time data in R, and second, how to convert other data types to date/time. Let us begin by creating the latest R release date manually. ```{r date_manual} release_date <- 2019-12-12 release_date ``` Okay! Why do we see `1995` when we call the date? What is happening here? Let us quickly check the data type of `release_date`. ```{r date_type} class(release_date) ``` The data type is `numeric` i.e. R has subtracted `12` twice from `2019` to return `1995`. Clearly, the above method is not the right way to store date/time. Let us see if we can get some hints from the builtin R functions we used in the previous section. If you observe the output, all of them returned date/time wrapped in quotes. Hmmm... let us wrap our date in quotes and see what happens. ```{r date_wrap} release_date <- "2019-12-12" release_date ``` Alright, now R does not do any arithmetic and returns the date as we specified. Great! Is this the right format to store date/time? No. Why? What is the problem if date/time is saved as character/string? The problem is the nature or type of operations done on date or time is different when compared to string/character, number or logical values. - how do we add/subtract dates? - how do we extract components such as year, month, day etc. To answer the above questions, we will first check the data type of `Sys.Date()` and `now()`. ```{r check_date_type} class(Sys.Date()) class(lubridate::now()) class(release_date) ``` As you can see from the above output, there are 3 different classes for storing date/time in R - `Date` - `POSIXct` - `POSIXlt` Let us explore each of the above classes one by one. ### Date #### Introduction The `Date` class represents calendar dates. Let us go back to `Sys.Date()`. If you check the class of `Sys.Date()`, it is `Date`. Internally, this date is a number i.e. an integer. The `unclass()` function will show dates are stored internally. ```{r unclass_sys_date} unclass(Sys.Date()) ``` What does this integer represent? Why has R stored the date as an integer? Before we answer this question, we need to know something else. In R, dates are represented as the number of days since `1970-01-01`. All the dates in R are internally stored in this way. Before we explore this concept further, let us learn to create `Date` objects in R. We will continue to use the latest R release date, `2019-12-12`. Until now, we have stored the above date as character/string but now we will use `as.Date()` to save it as a `Date` object. `as.Date()` is the easiest and simplest way to create dates in R. ```{r release_as_date} release_date <- as.Date("2019-12-12") release_date ``` The `as_date()` function from the lubridate package is similar to `as.Date()`. ```{r release_date_lubridate} release_date <- lubridate::as_date("2019-12-12") release_date ``` If you look at the difference between `release_date` and `1970-01-01`, it will be the same as `unclass(release_date)`. ```{r release_date_unclass} release_date - as.Date("1970-01-01") unclass(release_date) ``` Let us come back to `1970-01-01` i.e. the origin for dates in R. ```{r origin} lubridate::origin ``` From the previous examples, we know that dates are internally stored as number of days since `1970-01-01`. How about dates older than the origin? How are they stored? Let us look at that briefly. ```{r older_than_origin} unclass(as.Date("1963-08-28")) ``` Dates older than the origin are stored as negative integers. For those who are not aware, Martin Luther King, Jr. delivered his famous **I Have a Dream** speech on `1963-08-28`. Let us move on and learn how to convert numbers into dates. #### Convert Numeric The `as.Date()` function can be used to convert any of the following to a `Date` object - character/string - number - factor (categorical/qualitative) We have explored how to convert strings to date. How about converting numbers to date? Sure, we can create date from numbers by specifying the origin and number of days since it. ```{r release_date_origin} as.Date(18242, origin = "1970-01-01") ``` The origin can be changed to another date (while changing the number as well.) ```{r release_date_diff_origin} as.Date(7285, origin = "2000-01-01") ``` ### ISO 8601 ```{r img1lub, fig.align='center', echo=FALSE} knitr::include_graphics('img/iso.png') ``` If you have carefully observed, the format in which we have been specifying the dates as well as of those returned by functions such as `Sys.Date()` or `Sys.time()` is the same i.e. `YYYY-MM-DD`. It includes - the year including the century - the month - the date The month and date separated by `-`. This default format used in R is the ISO 8601 standard for date/time. ISO 8601 is the internationally accepted way to represent dates and times and uses the 24 hour clock system. Let us create the release date using another function `ISOdate()`. ```{r iso_date} ISOdate(year = 2019, month = 12, day = 12, hour = 8, min = 5, sec = 3, tz = "UTC") ``` We will look at all the different weird ways in which date/time are specified in the real world in the [Date & Time Formats] section. For the time being, let us continue exploring date/time classes in R. The next class we are going to look at is `POSIXct/POSIXlt`. ### POSIX You might be wondering what is this POSIX thing? POSIX stands for **P**ortable **O**perating **S**ystem **I**nterface. It is a family of standards specified f or maintaining compatibility between different operating systems. Before we learn to create POSIX objects, let us look at `now()` from lubridate. ```{r now_class} class(lubridate::now()) ``` `now()` returns current date/time as a POSIXct object. Let us look at its internal representation using `unclass()` ```{r now_unclass} unclass(lubridate::now()) ``` The output you see is the number of seconds since January 1, 1970. #### POSIXct `POSIXct` represents the number of seconds since the beginning of 1970 (UTC) and `ct` stands for calendar time. To store date/time as `POSIXct` objects, use `as.POSIXct()`. Let us now store the latest R release date as `POSIXct` as shown below ```{r release_date_posixct} release_date <- as.POSIXct("2019-12-12 08:05:03") class(release_date) unclass(release_date) ``` #### POSIXlt `POSIXlt` represents the following information in a list - seconds - minutes - hour - day of the month - month - year - day of week - day of year - daylight saving time flag - time zone - offset in seconds from GMT The `lt` in `POSIXlt` stands for local time. Use `as.POSIXlt()` to store date/time as `POSIXlt` objects. Let us store the release date as a `POSIXlt` object as shown below ```{r release_date_posixlt_1} release_date <- as.POSIXlt("2019-12-12 08:05:03") release_date ``` As we said earlier, `POSIXlt` stores date/time components in a list and these can be extracted. Let us look at the date/time components returned by `POSIXlt` using `unclass()`. ```{r release_date_posixlt_2} release_date <- as.POSIXlt("2019-12-12 08:05:03") unclass(release_date) ``` Use `unlist()` if you want the components returned as a `vector`. ```{r release_date_posixlt_3} release_date <- as.POSIXlt("2019-12-12 08:05:03") unlist(release_date) ``` To extract specific components, use `$`. ```{r release_date_posixlt_4} release_date <- as.POSIXlt("2019-12-12 08:05:03") release_date$hour release_date$mon release_date$zone ``` Now, let us look at the components returned by `POSIXlt`. Some of them are intuitive ```{r table_posixlt_1, echo=FALSE} cname <- c("`sec`", "`min`", "`hour`", "`mon`", "`zone`", "`wday`", "`mday`", "`year`", "`yday`", "`isdst`", "`gmtoff`") descrip <- c("Second", "Minute", "Hour of the day", "Month of the year (0-11", "Timezone", "Day of week", "Day of month","Years since 1900", "Day of year", "Daylight saving flag", "Offset is seconds from GMT") data.frame(Component = cname, Description = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` Great! We will end this section with a few tips/suggestions on when to use `Date` or `POSIXct/POSIXlt`. - use `Date` when there is no time component - use `POSIX` when dealing with time and timezones - use `POSIXlt` when you want to access/extract the different components ### Your Turn R 1.0.0 was released on `2000-02-29 08:55:23 UTC`. Save it as - `Date` using character - `Date` using origin and number - `POSIXct` - `POSIXlt` and extract - month day - day of year - month - zone - ISODate ## Date & Time Formats ### Introduction After the timezones and daylight savings detour, let us get back on path and explore another important aspect, date & time formats. Although it is a good practice to adher to ISO 8601 format, not all date/time data will comply with it. In real world, date/time data may come in all types of weird formats. Below is a sample ```{r table_formats_sample, echo=FALSE} cname <- c("December 12, 2019", "12th Dec, 2019", "Dec 12th, 19", "12-Dec-19", "2019 December", "12.12.19") data.frame(Format = cname) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` When the data is not in the default ISO 8601 format, we need to explicitly specify the format in R. We do this using conversion specifications. A conversion specification is introduced by %, usually followed by a single letter or O or E and then a single letter. ### Conversion Specifications ```{r table_formats_1, echo=FALSE} cname <- c("`%d`", "`%m`", "`%b`", "`%B`", "`%y`", "`%Y`", "%H", "%M", "%S") descrip <- c("Day of the month (decimal number)", "Month (decimal number)", "Month (abbreviated)", "Month (full name)", "Year (2 digit)", "Year (4 digit)", "Hour", "Minute", "Second") example <- c(12, 12, "Dec", "December", 19, 2019, 08, 05, 03) data.frame(Specification = cname, Description = descrip, Example = example) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` Time to work through a few examples. Let us say you are dealing with dates in the format `19/12/12`. In this format, the year comes first followed by month and the date; each separated by a backslash (`/`). The year consists of only 2 digits i.e. it does not include the century. Let us now map each component of the date to the format table shown at the beginning. ```{r table_formats_ex_1, echo=FALSE} cname <- c("19", "12", "12") descrip <- c("`%y`", "`%m`", "`%d`") data.frame(Date = cname, Specification = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` Using the format argument, we will specify the date format as a character vector i.e. enclosed in quotes. ```{r format_examples_1} as.Date("19/12/12", format = "%y/%m/%d") ``` Another way in which the release data can be written is `2019-Dec-12`. We still have the year followed by the month and the date but there are a few changes here: - the components are separated by a `-` instead of `/` - year has 4 digits i.e. includes the century - the month is specified using abbreviation instead of digits Let us map the components to the format table: ```{r table_formats_ex_2, echo=FALSE} cname <- c("2019", "Dec", "12") descrip <- c("`%Y`", "`%b`", "`%d`") data.frame(Date = cname, Specification = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` Let us specify the format for the date using the above mapping. ```{r format_examples_2} as.Date("2019-Dec-12", format = "%Y-%b-%d") ``` In both the above examples, we have not dealt with time components. Let us include the time of the latest R release in the next one i.e. `19/12/12 08:05:03`. ```{r table_formats_ex_3, echo=FALSE} cname <- c("19", "12", "12", "08", "05", "03") descrip <- c("`%y`", "`%m`", "`%d`", "`%H`", "`%M`", "`%S`") data.frame(Date = cname, Specification = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` Since we are dealing with time, we will use `as.POSIXct()` instead of `as.Date()`. ```{r format_examples_3} as.POSIXct("19/12/12 08:05:03", tz = "UTC", format = "%y/%m/%d %H:%M:%S") ``` In the below table, we look at some of the most widely used conversion specifications. You can learn more about these specifications by running `?strptime` or `help(strptime)`. ```{r table_formats_2, echo=FALSE} cname <- c("`%a`", "`%A`", "`%C`", "`%D`", "`%e`", "`%F`", "`%h`", "`%I`", "`%j`", "`%R`", "`%t`", "`%T`", "`%u`", "`%U`", "`%V`", "`%w`", "`%W`") descrip <- c("Abbreviated weekday", "Full weekday", "Century (00-99)", "Same as `%m/%d/%y`", "Day of month [1 - 31]", "Same as `%Y-%m-%d`", "Same as `%b`", "Hours as decimal [01 - 12]", "Day of year [001 - 366]", "Same as `%H:%M`", "Tab", "Same as `%H:%M:%S`", "Weekday [1 - 7](Monday is 1)", "Week of year [00 - 53]", "Week of year [01 - 53]", "Weekday [0 - 6](sunday is 0)", "Week of year [00 - 53]") data.frame(Specification = cname, Description = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` We have included a lot of practice questions for you to explore the different date/time formats. The solutions are available in the Learning Management system as well as in our GitHub repo. Try them and let us know if you have any doubts. ### Guess Format `guess_formats()` from lubridate is a very useful function. It will guess the date/time format if you specify the order in which year, month, date, hour, minute and second appear. ```{r guess_formats} release_date_formats <- c("December 12th 2019", "Dec 12th 19", "dec 12 2019") guess_formats(release_date_formats, orders = "mdy", print_matches = TRUE) ``` ### Your Turn Below, we have specified `July 5th, 2019` in different ways. Create the date using `as.Date()` while specifying the correct format for each of them. - `July-05-19` - `JUL-05-19` - `05.07.19` - `5-July 2019` - `July 5th, 2019` - `July 05, 2019` - `2019-July- 05` - `05/07/2019` - `07/05/2019` - `7/5/2019` - `07/5/19` - `2019-07-05` ## Parse Date & Time While creating date-time objects, we specified different formats using the conversion specification but most often you will not create date/time and instead deal with data thay comes your way from a system or colleague/collaborator. In such cases, we need to be able to parse date/time from the data provided to us. In this section, we will focus on parsing date/time from character data. Both base R and the lubridate package offer functions to parse date and time and we will explore a few of them in this section. We will initially use functions from base R and later on explore those from lubridate which will give us an opportunity to compare and contrast. It will also allow us to choose the functions based on the data we are dealing with. `strptime()` will convert character data to `POSIXlt`. You will use this when converting from character data to date/time. On the other hand, if you want to convert date/time to character data, use any of the following: - `strftime()` - `format()` - `as.character()` The above functions will convert `POSIXct/POSIXlt` to character. Let us start with a simple example. The data we have been supplied has date/time as character data and in the format `YYYYMMDD` i.e. nothing separates the year, month and date from each other. We will use `strptime()` to convert this to an object of class `POSIXlt`. ```{r strptime_1} rel_date <- strptime("20191212", format = "%Y%m%d") class(rel_date) ``` If you have a basic knowledge of conversion specifications, you can use `strptime()` to convert character data to `POSIXlt`. Let us quickly explore the functions to convert date/time to character data before moving on to the functions from lubridate. ```{r} rel_date_strf <- strftime(rel_date) class(rel_date_strf) rel_date_format <- format(rel_date) class(rel_date_format) rel_date_char <- as.character(rel_date) class(rel_date_char) ``` As you can see, all the 3 functions converted date/time to character. Time to move on and explore the lubridate package. We will start with an example in which the release date is formatted in 3 different ways but they have one thing in common i.e. the order in which the components appear. In all the 3 formats, the year is followed by the month and then the date. To parse the release date, we will use `parse_date_time()` from lubridate which parses the input into `POSIXct` objects. ```{r parse_date_time} release_date <- c("19-12-12", "20191212", "19-12 12") parse_date_time(release_date, "ymd") parse_date_time(release_date, "y m d") parse_date_time(release_date, "%y%m%d") ``` Try to use `strptime()` in the above example and see what happens. Now, let us look at another data set. ```{r parse_date_time_multiple} release_date <- c("19-07-05", "2019-07-05", "05-07-2019", "07-05-2019") ``` What happens in the below case? The same date appears in multiple formats. How do we parse them? `parse_date_time()` allows us to specify mutiple date-time formats. Let us first map the dates to their formats. ```{r table_formats_pdt, echo=FALSE} cname <- c("19-07-05", "2019-07-05", "05-07-2019", "07-05-2019") descrip <- c("`ymd`", "`ymd`", "`dmy`", "`mdy`") data.frame(Date = cname, Specification = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` The above specifications can be supplied as a character vector. ```{r parse_date_2} parse_date_time(release_date, c("ymd", "ymd", "dmy", "mdy")) ``` Great! We have used both `strptime()` and `parse_date_time()` now. Can you tell what differentiates `parse_date_time()` when compared to `strptime()`? We summarize it in the points below: - no need to include `%` prefix or separator - specify several date/time formats There are other helper functions that can be used to - parse dates with year, month, day components - parse dates with year, month, day, hour, minute, seconds components - parse period with hour, minute, second components and are explored in the below examples. ```{r parse_ymd} # year/month/date ymd("2019-12-12") # year/month/date ymd("19/12/12") # date/month/year dmy(121219) # year/month/date/hour/minute/second ymd_hms(191212080503) # hour/minute/second hms("8, 5, 3") # hour/minute/second hms("08:05:03") # minute/second ms("5,3") # hour/minute hm("8, 5") ``` Note, in a couple of cases where the components are not separated by `/`, `-` or space, we have not enclosed the values in quotes. ### Your Turn Below, we have specified `July 5th, 2019` in different ways. Parse the dates using `strptime()` or `parse_date_time()` or any other helper function. - `July-05-19` - `JUL-05-19` - `05.07.19` - `5-July 2019` - `July 5th, 2019` - `July 05, 2019` - `2019-July- 05` - `05/07/2019` - `07/05/2019` - `7/5/2019` - `07/5/19` - `2019-07-05` ## Date & Time Components In the second section, we discussed the downside of saving date/time as character/string in R. One of the points we discussed was that we can't extract components such as year, month, day etc. In this section, we will learn to extract date/time components such as - year - month - date - week - day - quarter - semester - hour - minute - second - timezone ```{r img_day_week_month, fig.align='center', echo=FALSE} knitr::include_graphics('img/day_week_month.png') ``` The below table outlines the functions we will explore in the first part of this section. ```{r table_components_1, echo=FALSE} cname <- c("`year()`", "`month()`", "`month(label = TRUE)`", "`month(abbr = FALSE)`", "`months()`", "`week()`") descrip <- c("Get year", "Get month (number)", "Get month (abbreviated name)", "Get month (full name)", "Get month", "Get week") data.frame(Function = cname, Description = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` ### Year ```{r year} release_date <- ymd_hms("2019-12-12 08:05:03") year(release_date) ``` ### Month `month()` will return the month as a number i.e. `07` for July. ```{r month_1} month(release_date) ``` If you want the name of the month instead, use the `label` argument and set it to `TRUE`. Now it returns `Jul` instead of `07`. ```{r month_2} month(release_date, label = TRUE) ``` But this is the abbreviated name and not the full name. How do we get the full name of the month? Set the `abbr` argument to `FALSE`. ```{r month_3} month(release_date, label = TRUE, abbr = FALSE) ``` Ah! now we can see the full name of the month. `months()` from base R will return the full name of the month by default. If you want the abbreviated name, use the `abbreviate` argument and set it to `TRUE`. ```{r month_4} months(release_date) ``` ### Week `week()` returns the number of complete 7 day periods between the date and 1st January plus one. ```{r week} week(release_date) ``` ### Day Use `day()` to extract the date component. There are other variations such as ```{r table_components_2, echo=FALSE} cname <- c("`day`", "`mday()`", "`wday()`", "`qday()`", "`yday()`", "`weekdays()`", "`days_in_month()`") descrip <- c("Get day", "Day of the month", "Day of the week", "Day of quarter", "Day of year", "Day of week", "Days in the month") data.frame(Function = cname, Description = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` ```{r day_1} day(release_date) mday(release_date) qday(release_date) yday(release_date) ``` `wday` can return - a number - abbreviation of the weekday - full name of the weekday ```{r day_2} wday(release_date) wday(release_date, label = TRUE) wday(release_date, label = TRUE, abbr = FALSE) ``` `weekdays()` from base R also returns the day of the week (the name and not the number). If you want the abbreviated name, use the `abbreviate` argument. ```{r day_3} weekdays(release_date) weekdays(release_date, abbreviate = TRUE) ``` ### Days in Month If you want to know the number of days in the month, use `days_in_month()`. In our example, the month is December and it has 31 days. ```{r days_in_month} days_in_month(release_date) ``` ### Hour, Minute & Seconds ```{r table_components_3, echo=FALSE} cname <- c("`hour()`", "`minute()`", "`second()`", "`seconds()`") descrip <- c("Get hour", "Get minute", "Get second", "Number of seconds since `1970-01-01`") data.frame(Function = cname, Description = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` So far we have been looking at date components. Now, let us look at time components. ```{r hour_min_sec_tz} hour(release_date) minute(release_date) second(release_date) ``` `seconds()` returns the number of seconds since `1970-01-01`. ```{r hour_min_seconds} seconds(release_date) ``` ### Quarter & Semester ```{r img_quarter_semester, fig.align='center', echo=FALSE} knitr::include_graphics('img/quarter_semester.png') ``` `quarter()` will return the quarter from the date. December is in the 4th quarter and hence it returns 4. ```{r quarter_1} quarter(release_date) ``` If you want the year along with the quarter, set the `with_year` argument to `TRUE`. ```{r quarter_2} quarter(release_date, with_year = TRUE) ``` In India, the fiscal starts in April and December falls in the 3rd quarter. How can we accommodate this change? The `fiscal_start` argument allows us to set the month in which the fiscal begins. We will set it to 4 for April. Now it returns 3 instead of 4. ```{r quarter_3} quarter(release_date, fiscal_start = 4) ``` `quarters()` from base R also returns the quarter. ```{r quarter_4} quarters(release_date) ``` ```{r table_components_4, echo=FALSE} cname <- c("`quarter()`", "`quarter(with_year = TRUE)`", "`quarter(fiscal_start = 4)`", "`quarters()`", "`semester()`") descrip <- c("Get quarter", "Quarter with year", "Fiscal starts in April", "Get quarter", "Get semester") data.frame(Function = cname, Description = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` ### Case Study #### Extract Date, Month & Year from Due Date Let us now extract the date, month and year from the `Due` column. ```{r lub2} transact |> mutate( due_day = day(Due), due_month = month(Due), due_year = year(Due) ) ``` #### Data Sanitization Let us do some data sanitization. If the due day happens to be February 29, let us ensure that the due year is a leap year. Below are the steps to check if the due year is a leap year: - we will extract the following from the due date: - day - month - year - we will then create a new column `is_leap` which will have be set to `TRUE` if the year is a leap year else it will be set to `FALSE` - filter all the payments due on 29th Feb - select the following columns: - `Due` - `is_leap` ```{r lub6} transact |> mutate( due_day = day(Due), due_month = month(Due), due_year = year(Due), is_leap = leap_year(due_year) ) |> filter(due_month == 2 & due_day == 29) |> select(Due, is_leap) ``` #### Invoices Distribution by Quarter Let us count the invoices due for each quarter. ```{r lub8} transact |> mutate( quarter_due = quarter(Due) ) |> count(quarter_due) ``` ### Your Turn Using the `transact` data, tabulate the following from the `Due` column: - year - month with label - weekday with label - hour - and quarter ## Create, Update & Verify In the second section, we learnt to create date-time objects using `as.Date()`, `as.POSIXct()` etc. In this section, we will explore a few other functions that will allow us to do the same - `make_date()` - `make_datetime()` ### Create To create date without time components, use `make_date()` and specify the following: - year - month - date We need to specify all the components in numbers i.e. we cannot use `Jul` or `July` for the month. It has to be `7`. ```{r make_date} make_date(year = 2019, month = 12, day = 12) ``` When you need to include time components, use `make_datetime()`. ```{r make_datetime} make_datetime(year = 2019, month = 12, day = 12, hour = 08, min = 05, sec = 03, tz = "UTC") ``` ### Update Let us look at another scenario. You have a date-time object and want to change one of its components i.e. any of the following - year - month - date Instead of creating another date-time object, you can change any of the components using `update()`. In the below example, we will start with the date of release of R version 3.6.1 and using `update()`, we will change it to `2019-12-12`. ```{r update} prev_release <- ymd("2019-07-05") prev_release |> update(year = 2019, month = 12, mday = 12) ``` ### Date Sequence So far we have created a single date-time instance. How about creating a sequence of dates? We can do that using `seq.Date()`. We need to specify the from date as the bare minimum input. If the end date is not specified, it will create the sequence uptil the current date. The interval of the sequence can be specified in any of the following units: - day - week - month - quarter - year We can add the following to the interval units - integer - `+` / `-` (increment or decrement) Using the integer, we can specify multiples of the units mentioned and using the sign, we can specify whether to increment or decrement. The below table displays the main arguments used in `seq.Date()`: ```{r table_seq_date, echo=FALSE} cname <- c("`from`", "`by`", "`to`", "`length.out`", "`along.with`") descrip <- c("Starting date of the sequence", "End date of the sequence", "Date increment of the sequence", "Length of the sequence", "Use length of this value as length of sequence") data.frame(Function = cname, Description = descrip) |> kable() |> kable_styling( bootstrap_options = c("striped", "hover", "condensed", "responsive") ) ``` In the first example, we will create a sequence of dates from `2010-01-01` to `2019-12-31`. The unit of increment should be a year i.e. the difference between the dates in the sequence should be 1 year, specified using the `by` argument. ```{r seq_date_1} seq.Date(from = as.Date("2010-01-01"), to = as.Date("2019-12-31"), by = "year") ``` In the next example, we change the unit of increment to a quarter i.e. the difference between the dates in the sequence should be a quarter or 3 months. ```{r seq_date_2} seq.Date(from = as.Date("2009-12-12"), to = as.Date("2019-12-12"), by = "quarter") ``` We will now create a sequence of dates but instead of specifying the unit of increment, we specify the number of dates in the sequence i.e. the length of the sequence. We do this using the `length.out` argument which specifies the desired length of the sequence. We want the sequence to have 10 dates including the start and end date, and hence we supply the value `10` for the `length.out` argument. ```{r seq_date_3} seq.Date(from = as.Date("2010-01-01"), to = as.Date("2019-12-31"), length.out = 10) ``` In all of the previous examples, we have specified both the start and the end date. Let us look at a few examples where we create a sequence of dates where we only specify the start date. In the below example, we want to create a sequence of dates starting from `2010-01-01`. The unit of increment should be 1 year i.e. the difference between the dates in the sequence should be 1 year and the length of the sequence should be `10` i.e. the number of dates including the start date should be `10`. ```{r seq_date_4} seq.Date(from = as.Date("2010-01-01"), by = "year", length.out = 10) ``` The unit of increment can include multiples and `+/-` sign i.e. it can be an unit of increment or decrement. In the next example, we can increment the dates in the sequence by `2` i.e. the difference between the dates should be `2` instead of `1`. This is achieved by specifying the unit of increment (multiple) first followed by a space and then the unit. In our example, it is `2 year`. As you can see, the sequence now goes all the way till `2028` and the gap between the dates is 2 years. ```{r seq_date_5} seq.Date(from = as.Date("2010-01-01"), by = "2 year", length.out = 10) ``` Let us say instead of increment we want to decrement the dates i.e. the sequence of dates will go backwards as shown in the next example. We achieve this by using the `-` sign along with the unit of decrement. The sequence of dates in next example starts from `2010` and goes back upto `1992` and the difference between the dates in 2 years. ```{r seq_date_6} seq.Date(from = as.Date("2010-01-01"), by = "-2 year", length.out = 10) ``` In the last example, we will explore the `along.with` argument. Here we have supplied a vector which is a sequence of numbers from 1 to 10. The length of this vector is 10 and the same length is used as the length of the sequence i.e. the length of value supplied to `along.with` is also the length of the sequence. ```{r seq_date_7} seq.Date(from = as.Date("2010-01-01"), by = "-2 year", along.with = 1:10) ``` ### Verify Type How do you check if the data is a date-time object? You can do that using any of the following from the lubridate package. - `is.Date()` - `is.POSIXct()` - `is.POSIXlt()` ```{r check_type} is.Date(release_date) is.POSIXct(release_date) is.POSIXlt(release_date) ``` ### Your Turn - R 2.0.0 was released on `2004-10-04 14:24:38`. Create this date using both `make_date()` and `make_datetime()` - R 3.0.0 was released on `2013-04-03 07:12:36`. Update the date created in the previous step to the above using `update()` ## Try it live {.unnumbered} :::{.callout-tip} ## Try it in your browser Edit the code below and press **Run**. It executes entirely in your browser via [WebR](https://docs.r-wasm.org/webr/latest/) — no R installation needed. The runtime downloads once, on this page only. ::: ```{webr-r} library(dplyr) library(lubridate) library(readr) transact <- read_csv("data-webr/transact_sample.csv", show_col_types = FALSE) transact |> mutate(due_year = year(Due), due_month = month(Due, label = TRUE)) |> head() ```