newRPL: Alpha demo 0.9 released [UPDATED 2017-10-25]

[Claudio L.]

(08-29-2017 04:25 AM)The Shadow Wrote:  Something weird is going on with EVAL. And after a lot of experimentation after getting very strange answers, I think I've figured out what.

If I type in '2/3', then press EVAL, I get something that *looks* exactly like '2/3', but if I do OBJ-> on it, rather than:

2
3
2
/

I get:

2
INV(3)
2
*

I would have expected the latter to look like:

'2*INV(3)', not '2/3'. So my FXND algorithm is getting very confused.

It's not weird at all once you understand some basic concepts of expressions in newRPL:

1. Expressions remain as entered by the user until operated upon.
2. Operations in expressions produce new expressions always in canonical form (more on this below).
   

Canonical form enforces a few internal rules that make manipulating an expression a lot simpler:

• Negative numbers don't exist. They are replaced with unary minus operator on a positive number.
• Subtractions don't exist. They are replaced with additions of negative terms.
• Divisions don't exist. They are replaced with multiplication of INVerted factors.
• Addition trees are flattened: The + operator can take multiple arguments.
• Multiplication trees are flattened: The * operator can take multiple arguments.
  

And a few other rules that I can't recall.
So when you type: '2+3+1'
you generate an object with a tree structure (+ (+ 2 3) 1) because for the parser + is a binary operator, but for the (future) CAS to operate properly, it flattens the tree to (+ 2 3 1), which makes it simpler to simplify terms, search for expressions, cancel out terms, etc.
Your fraction '2/3' is converted after EVAL to '2*INV(3)' per the rules above. What you get from OBJ-> is one level expansion of the main operator. Perhaps OBJ-> should convert to canonical form always before expanding, so you get consistent results every time.

Other than that, everything works as designed.