scmutils中SO(3)右不变向量场换位子计算结果异常问题排查

我正在阅读Sussman与Wisdom所著的《Functional Differential Geometry》，该书使用MIT-Scheme及scmutils库讲授微分几何。书中第48页的公式(4.29)、(4.30)、(4.31)定义了SO(3)群流形的右不变向量场基，理论上这些向量场的换位子满足[eₓ,eᵧ] = -e_z，且第51页通过scmutils的commutator函数得到了该正确结果。但我自行复现计算时，得到的结果却是[eₓ,eᵧ] = 0，手动计算则符合理论预期。以下是我的Scheme代码，最终行显示了错误结果，请问这是scmutils的bug还是我的代码存在问题？

;;; Launch mit-scheme and scmutils with,
;;; > mit-scheme --band "mechanics.com" --edit
;;; The code follows.

;;; Get coordinate basis vectors of SO(3) group manifold
;;; like d/dtheta etc
define-coordinates (up theta phi psi) Euler-angles)

;;; g is a point on the SO(3) group manifold
define g ((point Euler-angles) (up 'theta 'phi 'psi))

;;; f is a real-valued function on the group manifold
define f (literal-manifold-function 'f Euler-angles))

;;; e_x, e_y and e_z are the right-invariant vector fields
;;; defined in equations (4.29), (4.30) and (4.31) on page
;;; 48 of "Functional Differential Geometry" by Sussman and
;;; Wisdom.
define e_x 
  (+ (* (cos 'phi) d/dtheta)
     (- (* (/ (* (sin 'phi) (cos 'theta)) (sin 'theta)) d/dphi))
     (* (/ (sin 'phi) (sin 'theta)) d/dpsi)))

define e_y 
  (+ (* (sin 'phi) d/dtheta)
     (* (/ (* (cos 'phi) (cos 'theta)) (sin 'theta)) d/dphi)
     (- (* (/ (cos 'phi) (sin 'theta)) d/dpsi)))

define e_z d/dphi)

;;; This just checks that the vector fields have been defined
;;; correctly.
(show-expression
  ((e_x f) g)
)

;;; This is a commutator of the vector fields.
;;; It should be zero because [e_x,e_y]=-e_z
;;; and I've checked it by hand. However,
;;; scmutils gives the wrong answer. Is this
;;; a bug in scmutils or am I doing something
;;; wrong?
(show-expression
(((+ (commutator e_x e_y) e_z) f) g)
)
(((partial 1) f) (up theta phi psi))

问题根源：代码存在多处Scheme语法错误 #

你的代码并非scmutils的bug，而是违反了Scheme的基本语法规则，导致向量场未被正确定义，进而计算出错误结果。核心错误点如下：

1. 所有define语句缺少外层括号：Scheme中define是特殊形式，必须以(define 名称 表达式)的格式书写。原代码里的define-coordinates、define g、define f、define e_x等所有定义语句都遗漏了外层括号，这会导致Scheme无法正确解析变量/表达式的定义。
2. e_z的定义存在括号不匹配：原代码中define e_z d/dphi)多了一个右括号，正确写法应为(define e_z d/dphi)。
3. 部分表达式括号冗余：比如e_x定义中的(- (* (/ (* (sin 'phi) (cos 'theta)) (sin 'theta)) d/dphi))，外层冗余括号虽不影响逻辑，但可能干扰解析（核心问题仍是define的括号缺失）。

修正后的代码 #

;;; Launch mit-scheme and scmutils with,
;;; > mit-scheme --band "mechanics.com" --edit
;;; The code follows.

;;; Get coordinate basis vectors of SO(3) group manifold
;;; like d/dtheta etc
(define-coordinates (up theta phi psi) Euler-angles)

;;; g is a point on the SO(3) group manifold
(define g ((point Euler-angles) (up 'theta 'phi 'psi)))

;;; f is a real-valued function on the group manifold
(define f (literal-manifold-function 'f Euler-angles))

;;; e_x, e_y and e_z are the right-invariant vector fields
;;; defined in equations (4.29), (4.30) and (4.31) on page
;;; 48 of "Functional Differential Geometry" by Sussman and
;;; Wisdom.
(define e_x 
  (+ (* (cos 'phi) d/dtheta)
     (- (* (/ (* (sin 'phi) (cos 'theta)) (sin 'theta)) d/dphi))
     (* (/ (sin 'phi) (sin 'theta)) d/dpsi)))

(define e_y 
  (+ (* (sin 'phi) d/dtheta)
     (* (/ (* (cos 'phi) (cos 'theta)) (sin 'theta)) d/dphi)
     (- (* (/ (cos 'phi) (sin 'theta)) d/dpsi))))

(define e_z d/dphi)

;;; This just checks that the vector fields have been defined
;;; correctly.
(show-expression
  ((e_x f) g)
)

;;; This is a commutator of the vector fields.
;;; It should be zero because [e_x,e_y]=-e_z
;;; and I've checked it by hand.
(show-expression
  (((+ (commutator e_x e_y) e_z) f) g)
)

修正后运行代码，((+ (commutator e_x e_y) e_z) f) g的结果会等于0，与理论预期一致。

内容的提问来源于stack exchange，提问作者Stephen Blake