Advanced R by Hadley WickhamChapter 3: VectorsTony ElHabr@TonyElHabr2020-04-161 / 17

What's in Chapter 32 / 17

What's in Chapter 3

Section 3.2: atomic vectors
Section 3.3: attributes
Section 3.4: "special" vectors (S3 atomic vectors)
Section 3.5: lists
Section 3.6: data frames and tibbles
Section 3.7: NULL

2 / 17

What's in Chapter 3

Section 3.2: atomic vectors
Section 3.3: attributes
Section 3.4: "special" vectors (S3 atomic vectors)
Section 3.5: lists
Section 3.6: data frames and tibbles
Section 3.7: NULL

2 / 17

Vectors3 / 17

Vectors

2 types: atomic and list

The difference is that elements of atomic vectors have the same "type", while the elements of list vectors can be of different types. In general, the word "atomic" describes things that are irreducible units or components of a system. Here, the irreducible units are the data types, and the system is the R programming language. So we can see why its appropriate to call vectors of a single type "atomic". Lists have their own special properties.

atomic

list

3 / 17

Vectors

2 types: atomic and list

atomic

list

... and there is also NULL

Bringing

$NULL$ into the picture, we have our complete set of vector types.

3 / 17

Atomic Vectors4 / 17

Atomic Vectors

4 primary types: logical, integer, double, character (i.e. strings)

c(TRUE, FALSE, T, F)
c(1234L, 42L)
c(3.14, .314e1, 0xbada55)
c('single quote', "double quote")

I think everyone is probably familiar with these types, so I won't spend too much time on them. When we create individual instances of these "types", we have what we call "scalars". In other programming languages, such a form---a single value of a single type---may have different properties. But with R, everything is a vector, even "scalars".

4 / 17

Atomic Vectors

4 primary types: logical, integer, double, character (i.e. strings)

c(TRUE, FALSE, T, F)
c(1234L, 42L)
c(3.14, .314e1, 0xbada55)
c('single quote', "double quote")

... also raw and complex

raw(42)
complex(real = 0, imaginary = -1)

4 / 17

Atomic Vectors

4 primary types: logical, integer, double, character (i.e. strings)

c(TRUE, FALSE, T, F)
c(1234L, 42L)
c(3.14, .314e1, 0xbada55)
c('single quote', "double quote")

... also raw and complex

raw(42)
complex(real = 0, imaginary = -1)

Check type with typeof()

Use

$typeof()$ to identify the type of a variable.

4 / 17

Coercion5 / 17

CoercionCoercion happens when you attempt to combine vectors with elements of different types

Coercion often happens automatically. Most mathematical functions (+, log, abs, etc.) will coerce to numeric.
5 / 17

Coercion

Coercion happens when you attempt to combine vectors with elements of different types

Coercion often happens automatically. Most mathematical functions (+, log, abs, etc.) will coerce to numeric.

Coercion order: character → double → integer → logical

c(1, 1.01) # to double

## [1] 1.00 1.01

c(1, '1') # to character

## [1] "1" "1"

c(1, TRUE) # to integer

## [1] 1 1

5 / 17

Coercion

Coercion happens when you attempt to combine vectors with elements of different types

Coercion often happens automatically. Most mathematical functions (+, log, abs, etc.) will coerce to numeric.

Coercion order: character → double → integer → logical

c(1, 1.01) # to double

## [1] 1.00 1.01

c(1, '1') # to character

## [1] "1" "1"

c(1, TRUE) # to integer

## [1] 1 1

Explicity coerce with as.*() functions

as.integer(c(1, 1.01))

## [1] 1 1

5 / 17

Coercion

Coercion happens when you attempt to combine vectors with elements of different types

Coercion often happens automatically. Most mathematical functions (+, log, abs, etc.) will coerce to numeric.

Coercion order: character → double → integer → logical

c(1, 1.01) # to double

## [1] 1.00 1.01

c(1, '1') # to character

## [1] "1" "1"

c(1, TRUE) # to integer

## [1] 1 1

Explicity coerce with as.*() functions

as.integer(c(1, 1.01))

## [1] 1 1

Failed coercion leads to warnings and NA

as.integer(c('1', '1.01', 'a'))

## Warning: NAs introduced by coercion

## [1]  1  1 NA

5 / 17

NA and NULL6 / 17

`NA` and `NULL`

NA is a "sentinel" value for explicit missingness
NA can be of any type, e.g. NA_integer_, NA_character_, etc.

Default is logical, which is achieved by just

$NA$ .

Calculations involving NAs usually result in more NAs

1 + NA

## [1] NA

...although not always

1 | NA

## [1] TRUE

Test with is.na()

6 / 17

`NA` and `NULL`

NA is a "sentinel" value for explicit missingness
NA can be of any type, e.g. NA_integer_, NA_character_, etc.

Default is logical, which is achieved by just

$NA$ .

Calculations involving NAs usually result in more NAs

1 + NA

## [1] NA

...although not always

1 | NA

## [1] TRUE

Test with is.na()

NULL is its own vector type

typeof(NULL)

## [1] "NULL"

Zero-length

length(NULL)

## [1] 0

Cannot have attributes

x <- NULL
attr(x, 'y') <- 1 # error

A vector with

$NA$ can have attributes.

Test with is.null()

Speaking of attributes...

6 / 17

Attributes7 / 17

AttributesName-value pairs of metadata for R objects
7 / 17

Attributes

Name-value pairs of metadata for R objects
Get and set a single attribute with attr()

x <- 'a'
attr(x, 'what') <- 'apple'
attr(x, 'what')

## [1] "apple"

7 / 17

Attributes

Name-value pairs of metadata for R objects
Get and set a single attribute with attr()

x <- 'a'
attr(x, 'what') <- 'apple'
attr(x, 'what')

## [1] "apple"

Get and set multiple attributes with attributes() and structure()

7 / 17

Attributes

Name-value pairs of metadata for R objects
Get and set a single attribute with attr()

x <- 'a'
attr(x, 'what') <- 'apple'
attr(x, 'what')

## [1] "apple"

Get and set multiple attributes with attributes() and structure()

x <- structure('a', what = 'apple', type = 'fruit')
attributes(x)

## $what
## [1] "apple"
## 
## $type
## [1] "fruit"

7 / 17

Attributes

Name-value pairs of metadata for R objects
Get and set a single attribute with attr()

x <- 'a'
attr(x, 'what') <- 'apple'
attr(x, 'what')

## [1] "apple"

Get and set multiple attributes with attributes() and structure()

x <- structure('a', what = 'apple', type = 'fruit')
attributes(x)

## $what
## [1] "apple"
## 
## $type
## [1] "fruit"

With the exception of names() and dim(), most attributes are lost with calculations

attributes(x[1])

## NULL

7 / 17

names()8 / 17

`names()`

names() can be assigned in multiple ways

x <- c(apple = 'a', banana = 'b') # 1
x
y <- c('a', 'b')
names(y) <- c('apple', 'banana') # 2
y
setNames(y, c('apple', 'banana')) # 3

##  apple banana 
##    "a"    "b" 
##  apple banana 
##    "a"    "b" 
##  apple banana 
##    "a"    "b"

8 / 17

dim()9 / 17

`dim()`

dim() has the capability of turning a 1-d vector into a 2-d matrix or an n-d array

a <- matrix(1:6, nrow = 2, ncol = 3)
a

##      [,1] [,2] [,3]
## [1,]    1    3    5
## [2,]    2    4    6

b <- array(1:6, dim = c(1, 3, 2))
b

## , , 1
## 
##      [,1] [,2] [,3]
## [1,]    1    2    3
## 
## , , 2
## 
##      [,1] [,2] [,3]
## [1,]    4    5    6

9 / 17

`dim()`

dim() has the capability of turning a 1-d vector into a 2-d matrix or an n-d array

a <- matrix(1:6, nrow = 2, ncol = 3)
a

##      [,1] [,2] [,3]
## [1,]    1    3    5
## [2,]    2    4    6

b <- array(1:6, dim = c(1, 3, 2))
b

## , , 1
## 
##      [,1] [,2] [,3]
## [1,]    1    2    3
## 
## , , 2
## 
##      [,1] [,2] [,3]
## [1,]    4    5    6

Weird things

9 / 17

`dim()`

dim() has the capability of turning a 1-d vector into a 2-d matrix or an n-d array

a <- matrix(1:6, nrow = 2, ncol = 3)
a

##      [,1] [,2] [,3]
## [1,]    1    3    5
## [2,]    2    4    6

b <- array(1:6, dim = c(1, 3, 2))
b

## , , 1
## 
##      [,1] [,2] [,3]
## [1,]    1    2    3
## 
## , , 2
## 
##      [,1] [,2] [,3]
## [1,]    4    5    6

Weird things
- 1-d vector without a dim attribute has NULL dimension

9 / 17

`dim()`

dim() has the capability of turning a 1-d vector into a 2-d matrix or an n-d array

a <- matrix(1:6, nrow = 2, ncol = 3)
a

##      [,1] [,2] [,3]
## [1,]    1    3    5
## [2,]    2    4    6

b <- array(1:6, dim = c(1, 3, 2))
b

## , , 1
## 
##      [,1] [,2] [,3]
## [1,]    1    2    3
## 
## , , 2
## 
##      [,1] [,2] [,3]
## [1,]    4    5    6

Weird things
- 1-d vector without a dim attribute has NULL dimension
- Matrices and arrays can be a single column or row vector

9 / 17

S3 atomic vectors10 / 17

S3 atomic vectorsObjects with a class attribute, making them S3 objects

This means that the vector will behave different with a generic function such as pr∫() or even mathematical operators such as +
10 / 17

S3 atomic vectors

Objects with a class attribute, making them S3 objects

This means that the vector will behave different with a generic function such as

$print()$ or even mathematical operators such as

$+$

4 important S3 vector types in base R: factor (categorical), Date (Date), POSIXct (date-time), duration (difftime).

10 / 17

Factors11 / 17

FactorsVector that can only contain pre-defined values

Thus, it is commonly used to store categorical data.
11 / 17

FactorsVector that can only contain pre-defined values

Thus, it is commonly used to store categorical data.
Has two attributes: class and levels
11 / 17

Factors

Vector that can only contain pre-defined values

Thus, it is commonly used to store categorical data.

Has two attributes: class and levels
Built on top of integers, not characters

fruits <- factor(c('banana', 'apple', 'carrot'))
fruits

## [1] banana apple  carrot
## Levels: apple banana carrot

Note that it is ordered alphabetically by default.

11 / 17

Factors

Vector that can only contain pre-defined values

Thus, it is commonly used to store categorical data.

Has two attributes: class and levels
Built on top of integers, not characters

fruits <- factor(c('banana', 'apple', 'carrot'))
fruits

## [1] banana apple  carrot
## Levels: apple banana carrot

Note that it is ordered alphabetically by default.

Variation: ordered factors

x <- ordered(c('two', 'three', 'one'), levels = c('one', 'two', 'three'))
x

## [1] two   three one  
## Levels: one < two < three

[Interesting story about

$stringsAsFactors$ ](https://simplystatistics.org/2015/07/24/stringsasfactors-an-unauthorized-biography/)

11 / 17

Date, POSIXct, and duration12 / 17

Date, POSIXct, and durationAll built on top of doubles
12 / 17

`Date`, `POSIXct`, and `duration`

All built on top of doubles
Dates have class = "Date"

12 / 17

`Date`, `POSIXct`, and `duration`

All built on top of doubles
Dates have class = "Date"
Date-times are trickier...

12 / 17

`Date`, `POSIXct`, and `duration`

All built on top of doubles
Dates have class = "Date"
Date-times are trickier...
- Represent seconds since Jan. 1, 1970

12 / 17

`Date`, `POSIXct`, and `duration`

All built on top of doubles
Dates have class = "Date"
Date-times are trickier...
- Represent seconds since Jan. 1, 1970
- POSIXct isn't the only possible class; there's also POSIXlt

12 / 17

`Date`, `POSIXct`, and `duration`

All built on top of doubles
Dates have class = "Date"
Date-times are trickier...
- Represent seconds since Jan. 1, 1970
- POSIXct isn't the only possible class; there's also POSIXlt
- Also have a "parent" class of POSIXt

12 / 17

`Date`, `POSIXct`, and `duration`

All built on top of doubles
Dates have class = "Date"
Date-times are trickier...
- Represent seconds since Jan. 1, 1970
- POSIXct isn't the only possible class; there's also POSIXlt
- Also have a "parent" class of POSIXt
- Have a tzone attribute

c is for "calendar" time and "l" is for local time

12 / 17

`Date`, `POSIXct`, and `duration`

All built on top of doubles
Dates have class = "Date"
Date-times are trickier...
- Represent seconds since Jan. 1, 1970
- POSIXct isn't the only possible class; there's also POSIXlt
- Also have a "parent" class of POSIXt
- Have a tzone attribute

c is for "calendar" time and "l" is for local time

Durations have 2 attributes: class = "difftime" and units corresponding to a temporal unit, e.g. "day"

This temporal unit is purely for pretty printing.

12 / 17

Lists13 / 17

Lists

Each element can by of any atomic type, or even another list

Lists are sometimes called "recursive" vectors because a list can contain other lists. Atomic vectors don't have the same capability

13 / 17

Lists

Each element can by of any atomic type, or even another list

Lists are sometimes called "recursive" vectors because a list can contain other lists. Atomic vectors don't have the same capability

Each element is really a reference

As shown in the last chapter, lists are references. (Pointer may be a more familiar term.) This means that the size of a list does not really scale as you might expect.)

x <- 1L
lobstr::obj_size(x)

## 56 B

lobstr::obj_size(rep(x, 3L))

## 64 B

13 / 17

Lists

Each element can by of any atomic type, or even another list

Lists are sometimes called "recursive" vectors because a list can contain other lists. Atomic vectors don't have the same capability

Each element is really a reference

As shown in the last chapter, lists are references. (Pointer may be a more familiar term.) This means that the size of a list does not really scale as you might expect.)

x <- 1L
lobstr::obj_size(x)

## 56 B

lobstr::obj_size(rep(x, 3L))

## 64 B

Combining with c is different than wrapping with list()

This follows from the capability of lists to store other lists.

x <- list(a = 1, b = 2)
y <- list(c = -1, d = -2)
length(list(x, y))

## [1] 2

length(c(x, y))

## [1] 4

13 / 17

Lists

Each element can by of any atomic type, or even another list

Lists are sometimes called "recursive" vectors because a list can contain other lists. Atomic vectors don't have the same capability

Each element is really a reference

As shown in the last chapter, lists are references. (Pointer may be a more familiar term.) This means that the size of a list does not really scale as you might expect.)

x <- 1L
lobstr::obj_size(x)

## 56 B

lobstr::obj_size(rep(x, 3L))

## 64 B

Combining with c is different than wrapping with list()

This follows from the capability of lists to store other lists.

x <- list(a = 1, b = 2)
y <- list(c = -1, d = -2)
length(list(x, y))

## [1] 2

length(c(x, y))

## [1] 4

+ Coercing to a list may not result in what you'd expect

$r as.list(letters[1:2])$

$## [[1]] ## [1] "a" ## ## [[2]] ## [1] "b"$

13 / 17

Data frames14 / 17

Data frames

S3 vectors built on top of lists

14 / 17

Data frames

S3 vectors built on top of lists

df <- data.frame(col1 = 1:2, col2 = c('a', 'b'))
df

##   col1 col2
## 1    1    a
## 2    2    b

14 / 17

Data frames

S3 vectors built on top of lists

df <- data.frame(col1 = 1:2, col2 = c('a', 'b'))
df

##   col1 col2
## 1    1    a
## 2    2    b

Data frames have some undesireable default behavior

class(df$col2)

## [1] "character"

14 / 17

Data frames

S3 vectors built on top of lists

df <- data.frame(col1 = 1:2, col2 = c('a', 'b'))
df

##   col1 col2
## 1    1    a
## 2    2    b

Data frames have some undesireable default behavior

class(df$col2)

## [1] "character"

... which spawned tibbles (with the {tibble} package)

tbl <- tibble::tibble(col1 = 1:2, col2 = c('a', 'b'))
class(tbl$col2)

## [1] "character"

14 / 17

Data frame vs tibble behavior15 / 17

Data frame vs tibble behaviorTibble don't coerce strings to factors by default
15 / 17

Data frame vs tibble behavior

Tibble don't coerce strings to factors by default
Tibbles discourage rownames, which are generally "bad"

Rownames are "bad" because: (1) storing metadata in a different way than the rest of the data is generally not a good idea; (2) only work if a row can be identified by a single string; (3) must be unique.

15 / 17

Data frame vs tibble behavior

Tibble don't coerce strings to factors by default
Tibbles discourage rownames, which are generally "bad"

Tibbles have a "prettier" print method

15 / 17

Data frame vs tibble behavior

Tibble don't coerce strings to factors by default
Tibbles discourage rownames, which are generally "bad"

Tibbles have a "prettier" print method
Tibbles have stricter subsetting rules

tibbles always return a tibble and don't allow partial matching with

$$$

15 / 17

Non-your-typical column16 / 17

Non-your-typical column

Data frame columns can be lists

Need to wrap with

$I()$ here. (

$I$ is for identity.)

data.frame(x = 1:2, y = I(list(1:3, 1:4)))

##   x          y
## 1 1    1, 2, 3
## 2 2 1, 2, 3, 4

16 / 17

Non-your-typical column

Data frame columns can be lists

Need to wrap with

$I()$ here. (

$I$ is for identity.)

data.frame(x = 1:2, y = I(list(1:3, 1:4)))

##   x          y
## 1 1    1, 2, 3
## 2 2 1, 2, 3, 4

Easier list-column creation with tibbles

tibble::tibble(x = 1:2, y = list(1:3, 1:4))

## # A tibble: 2 x 2
##       x y        
##   <int> <list>   
## 1     1 <int [3]>
## 2     2 <int [4]>

16 / 17

Non-your-typical column

Data frame columns can be lists

Need to wrap with

$I()$ here. (

$I$ is for identity.)

data.frame(x = 1:2, y = I(list(1:3, 1:4)))

##   x          y
## 1 1    1, 2, 3
## 2 2 1, 2, 3, 4

Easier list-column creation with tibbles

tibble::tibble(x = 1:2, y = list(1:3, 1:4))

## # A tibble: 2 x 2
##       x y        
##   <int> <list>   
## 1     1 <int [3]>
## 2     2 <int [4]>

Columns can even be matrices and data frames

Notably, don't need to wrap with

$I()$ here.

data.frame(x = 1:2, y = matrix(3:6, nrow = 2))
data.frame(x = 1:2, y = data.frame(a = 3:4, b = 5:6))

16 / 17

In Review17 / 17

In Review

17 / 17

↑, ←, Pg Up, k	Go to previous slide
↓, →, Pg Dn, Space, j	Go to next slide
Home	Go to first slide
End	Go to last slide
Number + Return	Go to specific slide
b / m / f	Toggle blackout / mirrored / fullscreen mode
c	Clone slideshow
p	Toggle presenter mode
t	Restart the presentation timer
?, h	Toggle this help

Advanced R by Hadley Wickham

Chapter 3: Vectors

Tony ElHabr

@TonyElHabr

2020-04-16

What's in Chapter 3

What's in Chapter 3

What's in Chapter 3

Vectors

Vectors

Vectors

Atomic Vectors

Atomic Vectors

Atomic Vectors

Atomic Vectors

Coercion

Coercion

Coercion

Coercion

Coercion

NA and NULL

NA and NULL

NA and NULL

Attributes

Attributes

Attributes

Attributes

Attributes

Attributes

names()

names()

dim()

dim()

dim()

dim()

dim()

S3 atomic vectors

S3 atomic vectors

S3 atomic vectors

Factors

Factors

Factors

Factors

Factors

Date, POSIXct, and duration

Date, POSIXct, and duration

Date, POSIXct, and duration

Date, POSIXct, and duration

Date, POSIXct, and duration

Date, POSIXct, and duration

Date, POSIXct, and duration

Date, POSIXct, and duration

Date, POSIXct, and duration

Lists

Lists

Lists

Lists

Lists

Data frames

Data frames

Data frames

Data frames

Data frames

Data frame vs tibble behavior

Data frame vs tibble behavior

Data frame vs tibble behavior

Data frame vs tibble behavior

Data frame vs tibble behavior

Non-your-typical column

Non-your-typical column

Non-your-typical column

Non-your-typical column

In Review

In Review

What's in Chapter 3

Help

`NA` and `NULL`

`NA` and `NULL`

`NA` and `NULL`

`names()`

`names()`

`dim()`

`dim()`

`dim()`

`dim()`

`dim()`

`Date`, `POSIXct`, and `duration`

`Date`, `POSIXct`, and `duration`

`Date`, `POSIXct`, and `duration`

`Date`, `POSIXct`, and `duration`

`Date`, `POSIXct`, and `duration`

`Date`, `POSIXct`, and `duration`

`Date`, `POSIXct`, and `duration`

`Date`, `POSIXct`, and `duration`

`Date`, `POSIXct`, and `duration`