What is metaprogramming?

What is metaprogramming?

What is metaprogramming?

What is metaprogramming?

What is metaprogramming?

Writing programs (or code) that can manipulate other programs (or code).

Does metaprogamming == Non Standard Evaluation (NSE)?

[Metaprogramming] is the idea that code is data that can be inspected and modified programmatically... Closely related to metaprogramming is non-standard evaluation, NSE for short. This term, which is commonly used to describe the behaviour of R functions, is problematic.. NSE is actually a property of the argument (or arguments) of a function, so talking about NSE functions is a little sloppy.

Big Ideas

Big Ideas

Big Ideas

Big Ideas

1. R code is a tree

1. R code is a tree

But what does that mean???

"R code can be described as expressions, which can be drawn as trees."

An expression is anything that has a value. The simplest expressions are literal values like the number 1, the string "stuff", and the Boolean TRUE. A variable like least is also an expression: its value is whatever the variable currently refers to. Complex expressions are built out of simpler expressions: 1 + 2 is an expression that uses + to combine 1 and 2, while the expression c(10, 20, 30) uses the function c to create a vector out of the values 10, 20, 30. Expressions are often drawn as trees.

lobstr::ast(f(x, 'y', 1))

## o-f 
## +-x 
## +-"y" 
## \-1

Colours will be shown when you call ast(), but do not appear in the book for complicated technical reasons.

1. R code is a tree

But what does that mean???

"R code can be described as expressions, which can be drawn as trees."

An expression is anything that has a value. The simplest expressions are literal values like the number 1, the string "stuff", and the Boolean TRUE. A variable like least is also an expression: its value is whatever the variable currently refers to. Complex expressions are built out of simpler expressions: 1 + 2 is an expression that uses + to combine 1 and 2, while the expression c(10, 20, 30) uses the function c to create a vector out of the values 10, 20, 30. Expressions are often drawn as trees.

lobstr::ast(f(x, 'y', 1))

## o-f 
## +-x 
## +-"y" 
## \-1

Colours will be shown when you call ast(), but do not appear in the book for complicated technical reasons.

1. R code is a tree

Everything is a tree!

x <- 1
lobstr::ast(y <- 2 * x)

## o-`<-` 
## +-y 
## \-o-`*` 
##   +-2 
##   \-x

lobstr::ast(if(x > 1) y else x)

## o-`if` 
## +-o-`>` 
## | +-x 
## | \-1 
## +-y 
## \-x

lobstr::ast(function(x, y) x + y)

## o-`function` 
## +-o-x = `` 
## | \-y = `` 
## +-o-`+` 
## | +-x 
## | \-y 
## \-<inline srcref>

lobstr::ast(lobstr::ast(x + y))

## o-o-`::` 
## | +-lobstr 
## | \-ast 
## \-o-`+` 
##   +-x 
##   \-y

2. Capture trees by quoting

ex1 <- rlang::expr(x + y)
ex1

## x + y

x <- 1
y <- 2
eval(ex1)

## [1] 3

lobstr::ast(1 + 2)

## o-`+` 
## +-1 
## \-2

2. Capture trees by quoting

rlang::expr vs. rlang::enexpr

rlang::expr quotes your expression

f1 <- function(z) expr(z)
f1(a + b)

## z

enexpr quotes user's expression

f2 <- function(z) enexpr(z)
f2(a + b)

## a + b

en = "enriched"

2. Capture trees by quoting

2. Capture trees by quoting

mean(x + y)

library(ggplot2)  

# ggplot(mtcars, aes(disp, mpg)) + geom_point()

mtcars$disp

z <- x + 1

data.frame(z = 3)

Blue: Evaluated using usual R rules

Red: Quoted and evaluated with special rules

3. Build trees by unquoting

lobstr::ast(eval(!!ex1))

## o-eval 
## \-o-`+` 
##   +-x 
##   \-y

ex2 <- rlang::expr(x / !!ex1)
ex2

## x/(x + y)

eval(ex2)

## [1] 0.3333333

lobstr::ast(eval(!!ex2))

## o-eval 
## \-o-`/` 
##   +-x 
##   \-o-`+` 
##     +-x 
##     \-y

3. Build trees by unquoting

Understanding how to build code trees == success in online arguments.

3. Build trees by unquoting

Understanding how to build code trees == success in online arguments.

3. Build trees by unquoting

Understanding how to build code trees == success in online arguments.

3. Build trees by unquoting

Understanding how to build code trees == success in online arguments.

x <- 8 / 2 * (2 + 2)

lobstr::ast(x <- 8 / 2 * (2 + 2))

## o-`<-` 
## +-x 
## \-o-`*` 
##   +-o-`/` 
##   | +-8 
##   | \-2 
##   \-o-`(` 
##     \-o-`+` 
##       +-2 
##       \-2

x

## [1] 16

4. Capture trees AND
environments

quosure == closure + quote

Quosures aremportant for disambiguating the context in which expressions are evaluated (e.g. a column in a data frame or a variable in the parent environment).

4. Capture trees AND
environments

Result is tibble(x = 0, y = 1).

update <- function(df, col) {
  n <- 1
  col <- rlang::enexpr(col)
  res <- dplyr::mutate(df, y = !!col)
  res
}

df <- tibble::tibble(x = 0)
n <- 2
update(df, x + n)

4. Capture trees AND
environments

4. Capture trees AND
environments

Result is tibble(x = 0, y = 2).

update <- function(df, col) {
  n <- 1
  col <- rlang::enquo(col)
  res <- dplyr::mutate(df, y = !!col)
  res
}

df <- tibble::tibble(x = 0)
n <- 2
update(df, x + n)

4. Capture trees AND
environments

The Power of Metaprogramming

Metaprogramming is awesome. R is great because of how much control it allows user to have.

What's next

• Chapter 18: More about R code as a tree

• Chapter 19: More about evaluating (quoting) unevaluated code

• Chapter 20: More about evaluating (unquoting) captured code

Aside about trees

Understanding how "code is a tree" would have helped me with my final project in my Intro to Programming class.

Aside about trees