Why does loop variable in repeat not work - red

Question

Loop variable of repeat can be used in its block like this:
>> b: func [x] [x + i]
== func [x][x + i]
>> repeat i 2 [print reduce append copy [b] [3]]
4
5
You can see that the variable "i" is used by the function "b".
However,in the following example, the loop variable "idx" is not seen by the function. Error message is given:
*** Script Error: idx has no value
Red []
map: function [a-func a-block][
result: [] args: [] clear result
either parse a-block [some block!][
repeat idx length? a-block/1 [
clear args
foreach x a-block [append args to-block x/:idx]
append result reduce append copy [a-func] args
]
return result
]
[
repeat idx length? a-block [
append result to-block a-func a-block/:idx
]
]
]
map func [x y] [x + y - idx] [[1 2] [3 4]]
Why is the second code wrong? And how to make it working?

Answer 1

Loop variable of repeat can be used in its block like this
Unfortunately, this is conceptually wrong. Loop "variable" is not used in the block, it is "globally visible" both to function's body block and to repeat.
>> repeat index 5 []
>> index
== 5
Diagrammatically, that's lambda → <idx in global context> ← repeat, not lambda → repeat → <idx in repeat's "scope"> as you probably think.
However,in the following example, the loop variable "idx" is not seen by the function.
This is because you are using function constructor, which makes idx word local to its context, like so:
>> function [][repeat index 3 []]
== func [/local index][repeat index 3 []]
By contrast, func (used in your first example) doesn't do that.
>> func [][repeat index 3 []]
== func [][repeat index 3 []]
That is, in this code:
map func [x y] [x + y - idx] [[1 2] [3 4]]
idx within the anonymous function that you map over the block and idx in map implementation are two completely different "variables", bound to different contexts: one to the global (in which it has no value, hence the error message), and another to a local (where it is set to none by default).
The mechanics of Red's "scoping" model (or rather complete absence thereof) is a bit of an advanced topic, though I can elaborate on it if needed.
Suffice to say that it does not rely on traditional lexical scoping (like in most of the Lisp dialects), nor does it have variables in the strict sense of the term. Rather, it relies on symbolic values (aka words) that carry with them a binding to namespaces (aka contexts), which can be changed at will during the runtime (see bind in the example below) — a bit like f-expressions in Kernel and older Lisps, or perhaps anaphoric macro, of which collect (see below also) is a prime example: note that it "captures" keep word, which from then on refers to its internal context, where it is defined as an ad-hoc alias for append. Check out source collect output to see what I mean.
Here's a sketch of map (yours technically is more like zip, but anyway) to give you a hint at the potential solution.
map: function [
function [function!]
series [series!]
][
spec: spec-of :function ; ideally needs to be cleaned up
step: length? spec
index: 1
bind body-of :function 'index
collect [
foreach :spec series [
keep/only do compose [(:function) (:spec)]
index: index + step
]
]
]
Example:
>> map func [x] [reduce [index x]][a b c d]
== [[1 a] [2 b] [3 c] [4 d]]
>> map func [x y] [reduce [index x + y - index]][9 2 1 4]
== [[1 10] [3 2]]