Use current environment for user-defined functions

[sharkdp]

This is the current behavior:

>>> a=2
>>> f(x)=a*x  # global environment is closed over here
>>> f(1)
   = 2
>>> a=3  # has no effect, environment was closed over at function definition
>>> f(1)
   = 2

However, it would probably be more useful if the current environment would be used at call-time:

>>> a=3
>>> f(1)
   = 3

[fattredd]

I disagree. If you'd like a function to be able to change in the future, use two parameters like this:

>>> f(x, y) = x * y
>>> f(1, 2)
   = 2
>>> f(1, 3)
   = 3

Though it can be tedious to type two parameters every time. Perhaps a better solution would be the ability to set function defaults like this:

>>> f(x, y=2) = x * y
>>> f(1)
   = 2  
>>> f(1, 3)  
   = 3

[sharkdp]

I initially implemented it like this, because I thought that it might be a good idea (and "mutability is bad"). However, most scripting languages (and I tend to think of Insect as a simple scripting language) would imitate the mutable behavior:

Python:

>>> a = 2
>>> def f(x): return a * x
... 
>>> a = 3
>>> f(1)
3

JavaScript:

> var a = 2;
> function f(x) { return a*x; }
> a = 3;
> f(1)
3

It's even possible to define the function f before a is defined. That's currently not possible in Insect:

>>> f(x) = a * x
>>> a = 2
>>> f(1)

  Unknown identifier: a

Things get even more weird (in my opinion) if two functions are involved. At function definition, Insect captures the entire environment - including all function definitions:

>>> f1(x) = x²
>>> f2(x) = 2 f1(3 x)    # environment (including f1) is stored here
>>> f2(4)
   = 288
>>> f1(x) = x³    # this change does not have any effect
>>> f2(4)
   = 288

Coming back to your point: A way to secure a function definition from being changed in the future would be to only use constants, not global variables.

(JavaScript):

> const a = 2;
> function f(x) { return a*x; }
> a = 3;
TypeError: Assignment to constant variable.

Currently, Insect only supports built-in constants (like pi, speedOfLight, etc.). But user-defined constants could be a useful addition.

[sharkdp]

@fattredd Any thoughts on this?

[triallax]

This change makes sense to me. @sharkdp has your opinion on this issue changed? If not, let's add it in v5.9.0.

[sharkdp]

Sounds good. My opinion has not changed.