rlang包中sym()与parse_expr()的区别及差异原因咨询

Let's break down why you're seeing different results when using sym() vs parse_expr() with expressions like 'a - b', even though they behave identically for single symbols.

1. Single symbols: They work the same #

When you pass a simple variable name string like 'a':

• sym('a') directly creates a symbol object (a representation of a variable name).
• parse_expr('a') parses the string as an expression, which for a single token also resolves to a symbol object.

This is why identical(sym('a'), parse_expr('a')) returns TRUE—they're the exact same type of object pointing to the same variable. This consistency carries over when unquoting into a quosure with !!, since both resolve to the same symbol inside the quosure environment.

2. Compound expressions: The critical difference #

The split happens when you pass a string that represents an operation (like 'a - b'):

• sym('a - b') treats the entire string as one single, literal symbol. It doesn't interpret the subtraction—it just creates a variable name that includes spaces and a minus sign (R allows these "non-standard" variable names if you wrap them in backticks, e.g., `a - b`).
• parse_expr('a - b') actually parses the string into a call expression: it recognizes the subtraction operator and builds an object representing the operation of subtracting b from a.

Let's use code to see the structural difference clearly:

library(rlang)

# Inspect the raw objects
str(sym('a - b'))
#>  symbol a - b

str(parse_expr('a - b'))
#>  language a - b

# Inspect the quosures
ex5 <- quo(!!sym('a - b'))
ex6 <- quo(!!parse_expr('a - b'))

str(ex5)
#>  quosure
#>  expr: ^`a - b`
#>  env:  global

str(ex6)
#>  quosure
#>  expr: ^a - b
#>  env:  global

You can see ex5 holds a single quoted symbol `a - b` (a weirdly named variable), while ex6 holds the actual arithmetic expression a - b. That's why identical(ex5, ex6) returns FALSE—they're fundamentally different types of expressions.

3. Quick rule of thumb for usage #

• Use sym() when you need to reference a single variable by its name string (e.g., dynamically selecting a data frame column).
• Use parse_expr() when you need to turn a full expression string into executable R code (like arithmetic, function calls, or complex logic).

内容的提问来源于stack exchange，提问作者Rtist