set-word!is - it's just this type of marker.
set-word!points to location in source code, while variables usually point to location in memory. But there's no variables in Redbol. In a vague sense, script's body is it's own memory. Entire script is just one big chunk of self-modifiable state, which can be trivially serialized, saved and restored.
STR: "" ;"" is data structure position 2 insert STR "ah" STR: "" ;"" is data structure position 7 insert STR "bb"
you are stating: SET STR to what you find at position 2 of your data structure ("")
Then: insert on what is rapresented in position 5 (word "ah") on position 2
Then: set STR to what is rapresented in position 7 (second "") and use THIS POSITION in further reference to get/set the contend of STR
then insert on POSITION 7 "bb"
If this code loops, the first STR will take again the value "ah" insted of "BB" as it will use the content of position 2. The second one will use te conent of position 7.
Carl also refers to Rebol scripts as expressions. He says "REBOL Expressions are based on this concept: you combine values and words into blocks".
The chapter Expressions in the Rebol User Guide is pretty much essential reading.
@GiuseppeChillemi, lexical scanning is the first step. Parsing and tokenizing the input to find Red values in the text, and determining if the input is valid. That is, it contains all valid Red values.
At that point you have a block of values. Think about how you, as a human, interpret that. You know some values are direct values, like numbers. But what happens when you find a word? You need to determine what to do. What kind of word is it? What is the behavior of each kind of word?
Glad if it helped a bit @RiVeND.
Parse is deep, and hard to describe in prose/text, because the behavior and data mixture will be verbose. An interactive tool will be much better, so you can see what rule is being evaluated, the current location in the input, and maybe hints about keyword behavior. e.g. if
ahead is evaluated, a "does not advance" indicator could light up.
There was an old parsing tool, VisualParse++, from Sandstone technology, that did something like that. It was a shift-reduce parsing tool, and was pretty cool.