Interpolate a quoted expression.

f_quote() and expr_quote() return their arguments as an unevaluated expression. The former returns a formula, which bundles an expression and a scope, while the latter returns a raw expression. Like all capturing functions in the tidy evaluation framework, the expression is interpolated on capture (see arg_capture()) and vignette("tidy-eval"). Alternatively, interp() allows you to interpolate manually when you have constructed a raw expression or formula by yourself.

An expression.
A function, raw expression, or formula to interpolate. When interpolating a formula, other formulas are treated as promises (see section on tidy evaluation).

When an expression is interpolated, all sub-expressions within unquoting operators (like UQ(x) and UQS(x)) are evaluated and inlined. This provides a powerful mechanism for manipulating expressions. Since unquoting is such an important operation, !! and !!! are provided as syntactic shortcuts for unquoting and unquote-splicing (see examples).


For f_quote(), a formula (which contains information about where to find the objects mentioned in the formula). For expr_quote(), a raw quoted expression.

Tidy manipulation of expressions

Interpolating an expression allows you to inline any value within the expression. In particular, you can transform a quoted expression by unquoting another quoted expression into it. The latter expression gets inlined within the former. This mechanism allows you to easily program with NSE functions. E.g. var <- ~baz; f_quote(foo(bar, !! var)) produces the formula-quote ~foo(bar, baz).

Tidy evaluation of expressions

The purpose of a quoted expression is to be eventually evaluated in a modified scope (e.g. a scope where the columns of a data frame are directly accessible). While interpolation allows you to easily manipulate expressions, you need to be a bit careful with the scope of transplanted subexpressions. If they refer to variables and functions that are only available in a given environment (in a given scope), that environment should be bundled with the expression. That is exactly the purpose of a one-sided formula: bundling an expression and a scope. In the general case, i.e. interpolating a function with interp() or capturing a raw expression with expr_quote(), interpolation evaluates unquoted expressions in the calling environment. While useful, raw interpolation does not allow you to keep track of the scope of an expression. Interpolating a formula with interp() (or during capture by NSE functions or f_quote()) provides a much more powerful mechanism. The key is to unquote formulas rather than raw expressions. The formulas are inlined in the expression and carry their scope. While such formulas would confuse the base R function eval(), they become self-evaluating expressions when passed to f_eval(). To sum up, one-sided formulas are very much similar to promises in the tidy evaluation framework. Promises are R objects bound to function arguments. They make lazy evaluation in R possible: they bundle an expression (the argument supplied by the caller of the function) and an environment (the original call site of the function), and they self-evaluate to return a value as soon as you touch them. Similarly, formulas self-evaluate when called within f_eval(). However, unlike promises, they are first-class objects: you can pass a formula around and use it to transform another formula or expression. Formulas are thus treated as reified promises. Being able to manipulate a formula has important practical purposes: you can create them, inspect them interactively, and modify them (see previous section). Taken together, tidy modification and tidy evaluation of formulas provide a powerful mechanism for metaprogramming and programming with DSLs.


Formally, f_quote() and expr_quote() are quasiquote functions, UQ() is the unquote operator, and UQS() is the unquote splice operator. These terms have a rich history in LISP, and live on in modern languages like http://docs.julialang.org/en/release-0.1/manual/metaprogramming/ and https://docs.racket-lang.org/reference/quasiquote.html.

  • UQ
  • UQE
  • UQF
  • UQS
  • expr_quote
  • f_quote
  • interp
# When a tidyeval function captures an argument, it is wrapped in a
# formula and interpolated. f_quote() is a simple wrapper around
# arg_capture() and as such is the fundamental tidyeval
# function. It allows you to quote an expression and interpolate
# unquoted parts:
f_quote(!! 1 + 2)
f_quote(paste0(!! letters[1:2], "foo"))

# Alternatively you can interpolate a formula that is already
# constructed:
interp(~!! 1 + 2)
f <- ~paste0(!! letters[1:2], "foo")

# The !! operator is a syntactic shortcut for unquoting. However
# you need to be a bit careful with operator precedence. All
# arithmetic and comparison operators bind more tightly than `!`:
interp(x ~ 1 +  !! (1 + 2 + 3) + 10)

# For this reason you should always wrap the unquoted expression
# with parentheses when operators are involved:
interp(x ~ 1 + (!! 1 + 2 + 3) + 10)

# Or you can use the explicit unquote function:
interp(x ~ 1 + UQ(1 + 2 + 3) + 10)

# Use !!! or UQS() if you want to add multiple arguments to a
# function It must evaluate to a list
args <- list(1:10, na.rm = TRUE)
interp(~mean(!!! args))
f_quote(mean( UQS(args) ))

# You can combine the two
var <- quote(xyz)
extra_args <- list(trim = 0.9, na.rm = TRUE)
f_quote(mean(UQ(var) , UQS(extra_args)))
interp(~mean(!!var , !!!extra_args))

# Unquoting is especially useful for transforming a captured
# expression:
expr <- quote(foo(bar))
expr <- expr_quote(inner(!! expr, arg1))
expr <- expr_quote(outer(!! expr, !!! lapply(letters[1:3], as.symbol)))

# Quasiquotation of formulas is much more powerful. Formulas carry
# scope information about the inner expression inlined in the outer
# expression upon unquoting. Let's create a formula that quotes a
# symbol that only exists in a local scope (a child environment of
# the current environment):
f1 <- local({ foo <- "foo"; ~foo })

# You can evaluate that expression with f_eval():

# But you can also inline it in another expression before
# evaluation:
f2 <- local({ bar <- "bar"; ~toupper(bar)})
f3 <- f_quote(paste(!!f1, !!f2, "!"))

# f_eval() treats one-sided formulas like promises to be evaluated:

# The formula-promise representation is necessary to preserve scope
# information and make sure objects are looked up in the right
# place. However, there are situations where it can get in the way.
# This is the case when you deal with non-tidy NSE functions that do
# not understand formulas. You can inline the RHS of a formula in a
# call thanks to the UQE() operator:
nse_function <- function(arg) substitute(arg)
var <- local(~foo(bar))

# This is equivalent to unquoting and taking the RHS:
f_quote(nse_function(!! f_rhs(var)))

# One of the most important old-style NSE function is the dollar
# operator. You need to use UQE() for subsetting with dollar:
var <- ~cyl

# `!!`() is also treated as a shortcut. It is meant for situations
# where the bang operator would not parse, such as subsetting with
# $. Since that's its main purpose, we've made it a shortcut for
# UQE() rather than UQ():
var <- ~cyl

# Sometimes you would like to unquote an object containing a
# formula but include it as is rather than treating it as a
# promise. You can use UQF() for this purpose:
var <- ~letters[1:2]
f <- f_quote(list(!!var, UQF(var)))

# Note that two-sided formulas are never treated as fpromises:
f_eval(f_quote(a ~ b))

# Finally, you can also interpolate a closure's body. This is
# useful to inline a function within another. The important
# limitation is that all formal arguments of the inlined function
# should be defined in the receiving function:
other_fn <- function(x) toupper(x)

fn <- interp(function(x) {
  x <- paste0(x, "_suffix")
  !!! body(other_fn)
Documentation reproduced from package rlang, version, License: GPL-3

Community examples

Looks like there are no examples yet.