Closures and frozen environments

Hello,

I have been reading the book and implementing the Lox language in Java following it. I am really thankful because I've always wanted to see a book like this one.

My background is in statistics so everything I've learned about programming has not been very formal (blog posts, some books I've bought, documentation, a few papers, etc.) and obviously not every piece of knowledge is connected in my mind.

Currently I am on chapter 11: Resolving and binding. I would like to consult something related to the bug seen when showA() is introduced. Specifically, it is related to whether functions should remember the enclosing environment as it was when the function was created or not.

The author says: "The function should capture a frozen snapshot of the environment as it existed at the moment the function was declared."

However, I don't completely understand why this is the behavior you would expect by default.

As a person whose first language has been R, I thought it was normal to have dynamic enclosing environments. Here I share an example of what you have in R.

> # Global. Behaves as expected.
> a <- "Global"
> # Defines a function in the global environment
> f1 <- function() print(a)
> f1()
[1] "Global"
> 
> # Create a new environment whose parent is the global environment
> my_env <- new.env()
> # Bind the value "Internal" to the name "a" in this new environemnt
> my_env$a <- "Internal"
> # Create a function and bind it to `f1` in `my_env`. 
> # Important detail: the function is binded to a name in `my_env`
> # but since the code is executed from the global environment,
> # its enclosing environment is the global environment (as we'll see)
> my_env$f1 <- function() print(a)
> 
> # Result:
> my_env$f1()
[1] "Global"
> 
> # Reason:
> # The enclosing environment is the global environemnt,
> # then the function finds "a" in the global environment.
> environment(my_env$f1)
<environment: R_GlobalEnv>
> 
> # Now if we change the value binded to `a` in the global environment,
> # the output changes (no frozen enclosing environment)
> a <- "Internal"
> f1()
[1] "Internal"
> 
> # Change enclosing environment
> environment(my_env$f1) <- my_env
> environment(my_env$f1)
<environment: 0x00000221003e4bf8>
> 
> # See the effect
> my_env$f1()
[1] "Internal"
> 
> # But it is also dynamic
> my_env$a <- "Second internal"
> my_env$f1()
[1] "Second internal"

Maybe my terminology is not the best, but going straight to the point

What is the advantage of functions remembering the state of their enclosing environment when they were defined?

Thank you, Tomas

munificent / craftinginterpreters

Closures and frozen environments #730