Top-level variables, local variables, variable scoping and the difference between set and setq

Question

I've been reading about emacs dynamic and lexical bindings. While I generally get the difference between the two types, there is one example that is not clear to me. I've checked this question and I think I understand why the second example prints nil instead of t (I've read about set and setq when lexical-binding: t

;; -*- lexical-binding: t -*-

(let ((a nil))
  (setq a t)
  (print a))


(let ((a nil))
  (set 'a t)
  (print a))

From the elisp manual I've read that when lexical-binding is in effect set affects the dynamic value of a variable where as setq affects its current lexical value. However from the following example it looks like (setq a 5) sets the dynamic value of a since later (set 'a t) changes top level a.

;;; -*- lexical-binding: t; -*-
(setq a 5)

(let ((a nil))
  (setq a t)
  (print a)) ;; prints t 


(let ((a nil))
  (set 'a t)
  (print a)) ;; prints nil

a ;; prints t (I've expected this to be unchanged i.e. 5)

My question is when lexical-binding is on, what is the binding for the top level variables ?

Answer 1

Looking at Using Lexical Binding it appears that a needs to be defined special which in using setq it is not. Modifying your example to use defvar gives the answers expected.

;;; -*- lexical-binding: t; -*-
(defvar b 5)

(let ((b nil))
  (setq b t)
  (print b)) ;; prints t 

(print b) ;; prints 5

(let ((b nil))
  (set 'b t)
  (print b)) ;; prints t

(print b) ;; prints 5