Dates and Times I

Storing, creating, displaying, and converting time.

Last Time

1. Shortcut classes name common sets like digits.
2. Anchors pin a pattern to the start or end.
3. Grouping sets the scope of a quantifier.
4. Alternation matches one pattern or another.

Key Ideas for Today

1. A date-time counts seconds since 1970 (Unix time).
2. Parsing turns text into a date-time.
3. Getters and setters read and change its parts.
4. Some parts, like months, are circular.

Concept Acquisition

1. Computers store a date-time as the number of seconds since the Unix epoch, midnight UTC on 1 January 1970.
2. A POSIXct vector is a double vector with a class, so it prints as a legible date-time but still supports arithmetic, rounding, and other formats.
3. A Date vector is a simpler class that stores dates without times.
4. Date-time strings come in many formats; parsing one requires knowing the order of its components.
5. The components of a date-time (year, month, weekday, hour) can be extracted and changed in place.
6. Some components are circular, so averaging them as plain numbers gives nonsense; average the date-times instead.

Tool Acquisition

1. Sys.time(), as.POSIXct() (with tz and format), as.Date(), and format().
2. lubridate parsers such as mdy_hm() and dmy_hm().
3. Getters and setters: year(), month(), week(), day(), wday(label =, abbr =), hour().
4. make_datetime().

Concept Application

1. Predict the result of arithmetic on an unquoted date like 2026-10-14+1.
2. Find the day of the week of New Year’s Eve and of Independence Day, and the corresponding Unix time.
3. Find the average month of observations that span a new year.

What are the results of the following operations?

t1 <- 2026-10-14+1
t2 <- 10/14-10/7

01:00

Agenda

1. How does a computer Store dates and times?
2. How do we Create dates and times?
3. How do we Get and Set them?

Storing

On the pinned board, draw the two october dates on a timeline and show the existence of a duration as well as absolute positioning relative to a milestone called an epoch (indicate with a star).

On the next slides, return to the pinned board after each row of the table and add that date to the timeline.

Unix Time

A method to measure time as the number of seconds¹ that have elapsed since midnight UTC on 1 January 1970, the Unix epoch.

Date Time (UTC)	Unix Time
12:01 am Jan 1, 1970	60

1. excluding leap-seconds

Unix Time

A method to measure time as the number of seconds¹ that have elapsed since midnight UTC on 1 January 1970, the Unix epoch.

Date Time (UTC)	Unix Time
12:01 am Jan 1, 1970	60
12:01 am Jan 2, 1970	86460

1. excluding leap-seconds

Unix Time

A method to measure time as the number of seconds¹ that have elapsed since midnight UTC on 1 January 1970, the Unix epoch.

Date Time (UTC)	Unix Time
12:01 am Jan 1, 1970	60
12:01 am Jan 2, 1970	86460
7 pm Oct 14, 2026 (today at noon)	1792004400

1. excluding leap-seconds

Unix Time

A method to measure time as the number of seconds¹ that have elapsed since midnight UTC on 1 January 1970, the Unix epoch.

Date Time (UTC)	Unix Time
12:01 am Jan 1, 1970	60
12:01 am Jan 2, 1970	86460
7 pm Oct 14, 2026 (today at noon)	1792004400
5:55 pm Dec 7, 1941	−8993325700

1. excluding leap-seconds

https://www.unixtimestamp.com/

POSIXct

POSIXct (Portable Operating System Interface Calendar Time) is an atomic vector to store date-times that is built on top of a double vector.

now <- Sys.time()
now

[1] "2026-10-26 15:44:22 UTC"

class(now)

[1] "POSIXct" "POSIXt" 

Start off with reminder about factors: a integer vector with the factor class that allows it to store the levels look up table.

Why not just use a double vector?

# it prints as a legible string
now

[1] "2026-10-26 15:44:22 UTC"

# while respecting artithmetic
now + 1

[1] "2026-10-26 15:44:23 UTC"

# and other sensible operations
round(now, units = "hours")

[1] "2026-10-26 16:00:00 UTC"

# and makes it easy to use other formats
format(now, "%m/%d/%y")

[1] "10/26/26"

Date

A simpler vector than POSIXct that stores date data without times. Created with as.Date().

halloween <- as.Date("2000-10-31")

class(halloween)

[1] "Date"

typeof(halloween)

[1] "double"

halloween

[1] "2000-10-31"

Creating

Creating date-time vectors

Use as.POSIXct() to create a date-time vector from a numeric (in Unix Time) or a string in a particular format.

Creating date-time vectors

Use as.POSIXct() to create a date-time vector from a numeric (in Unix Time) or a string in a particular format.

as.POSIXct(x = 60, tz = "UTC")

[1] "1970-01-01 00:01:00 UTC"

as.POSIXct(x = "1970-01-01 00:01:00", tz = "UTC")

[1] "1970-01-01 00:01:00 UTC"

as.POSIXct(x = "Jan 1, 1970 12:01 am",
           format = "%b %d, %Y %I:%M %p",
           tz = "UTC")

[1] "1970-01-01 00:01:00 UTC"

Formats: https://r4ds.hadley.nz/datetimes.html#tbl-date-formats

Annotate each format code and connect it to the matching piece of the string.

Call class() and typeof() on this object.

Creating with lubridate

as.POSIXct(x = "1970-01-01 00:01:00", tz = "UTC")

[1] "1970-01-01 00:01:00 UTC"

lubridate() has many functions that can parse the year (y), month (m), day (d), hour (h), minute (m), and second (s) components of a date-time string.

library(lubridate)
mdy_hm("Jan 1, 1970 12:01 am")

[1] "1970-01-01 00:01:00 UTC"

dmy_hm("1-1-1970 00:01")

[1] "1970-01-01 00:01:00 UTC"

Underline the letters in mdy_hm and draw arrows to the parts of the string they parse.

Call class() and typeof() on this object.

Your Turn

Question: Which lubridate function parses each string?

1. "14 October 2026"
2. "10/14/26 9:00"
3. "2026-10-14 12:00:00"

01:00

Pair up. Debrief, then run dmy(), mdy_hm(), and ymd_hms() on the next slide.

Parsing each string

dmy("14 October 2026") # question 1

[1] "2026-10-14"

mdy_hm("10/14/26 9:00") # question 2

[1] "2026-10-14 09:00:00 UTC"

ymd_hms("2026-10-14 12:00:00") # question 3

[1] "2026-10-14 12:00:00 UTC"

Getting and Setting

Getting

You can extract components of the date-time object using year(), month(), week(), day(), wday(), hour(), etc.

year(now)

[1] 2026

day(now)

[1] 26

wday(now)

[1] 2

Getting

You can extract components of the date-time object using year(), month(), week(), day(), wday(), hour(), etc.

year(now)

[1] 2026

day(now)

[1] 26

wday(now, label = TRUE)

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

Getting

You can extract components of the date-time object using year(), month(), week(), day(), wday(), hour(), etc.

year(now)

[1] 2026

day(now)

[1] 26

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

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

Setting

You can set those same components in an existing date-time object using change-in-place notation.

year(now) <- 2027
now

[1] "2027-10-26 15:44:22 UTC"

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

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

Your Turn

Question: Write the (lines) of code necessary to determine:

1. On which day of the week will New Years Eve fall this year?
2. What day of the week was US Independence Day?
3. Bonus What was the corresponding Unix time?

02:00

Your Turn: Solutions

wday(ymd("2026-12-31"), label = TRUE, abbr = FALSE) # question 1

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

wday(ymd("2026-07-04"), label = TRUE, abbr = FALSE) # question 2

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

Your Turn: Bonus Solution

as.numeric(ymd("2026-07-04")) # question 3: days, not seconds!

[1] 20638

as.numeric(ymd("2026-07-04", tz = "UTC")) # question 3: seconds

[1] 1783123200

ymd() returns a Date, which counts days since 1970. Supplying tz makes it return a POSIXct, which counts seconds.

Circular Data

Example: Circular Data

What is the average month of these observations?

# A tibble: 6 × 2
  month  year
  <dbl> <dbl>
1     2  2026
2     1  2026
3     2  2026
4    11  2025
5    12  2025
6    12  2025

df |>
  summarize(avg_month = mean(month))

# A tibble: 1 × 1
  avg_month
      <dbl>
1      6.67

Example: Circular Data

What is the month of the average of these dates?

df |>
  mutate(datetime = make_datetime(year = year,
                                  day = 15,
                                  month = month))

# A tibble: 6 × 3
  month  year datetime           
  <dbl> <dbl> <dttm>             
1     2  2026 2026-02-15 00:00:00
2     1  2026 2026-01-15 00:00:00
3     2  2026 2026-02-15 00:00:00
4    11  2025 2025-11-15 00:00:00
5    12  2025 2025-12-15 00:00:00
6    12  2025 2025-12-15 00:00:00

Example: Circular Data

What is the month of the average of these dates?

df |>
  mutate(datetime = make_datetime(year = year,
                                  day = 15,
                                  month = month)) |>
  summarize(avg_datetime = mean(datetime))

# A tibble: 1 × 1
  avg_datetime       
  <dttm>             
1 2026-01-04 20:00:00

Example: Circular Data

What is the month of the average of these dates?

df |>
  mutate(datetime = make_datetime(year = year,
                                  day = 15,
                                  month = month)) |>
  summarize(avg_datetime = mean(datetime)) |>
  mutate(mon_of_avg = month(avg_datetime))

# A tibble: 1 × 2
  avg_datetime        mon_of_avg
  <dttm>                   <dbl>
1 2026-01-04 20:00:00          1

On the board, draw the months as a clock face, mark the six observations, then the mean month (July) and the month of the mean date.

Guess the Metaphor

A date-time is like a giant sundial: underneath it’s a single measurement, how far the shadow has travelled from a fixed origin, but we read it off the dial as hours and days. And like the dial, some parts wrap around: December sits right next to January.

Key Ideas for Today

1. A date-time counts seconds since 1970 (Unix time).
2. Parsing turns text into a date-time.
3. Getters and setters read and change its parts.
4. Some parts, like months, are circular.

For Next Time

1. Sign up for Quiz 4 (10/19-10/20) on PrairieTest.
2. Work on PS 4, due Saturday 10/17 at 11:59 pm.
3. Do Project 3 now and ask for help.