catseye / flobnar Goto Github PK

This should permit us to add I/O while only making a handful of functions monadic. (Basically the "trampoline" would be monadic and would do I/O on behalf of the continuations it executes.) This seems marginally cleaner than making the entire evaluator monadic. But experiment and see.

The README states without explanation that

|    0
|  5 p  <
|    0  +@
|    g  <
|    0
= Result: 5

and also that

8
*,<  5
9 +@>*
  >,*7
    3

is supposed to print Hi, not iH. In both examples, it seems like the north branch is getting evaluated "before" the south branch, but I don't know if that's meant to be a general rule. (Maybe the right-hand branch is always evaluated before the left-hand branch!)

A more "functional-spirited" rule would be that each branch spawns a completely new copy of funge-space, so that side-effects happening in one branch can never be observed by the other branch, and thus the interpreter could memoize and/or multi-thread any way it wanted. However, this rule clearly conflicts with existing examples such as the p/g one above.

If we are indeed guaranteed that the north operand is always evaluated before the south operand, then we can write a "non-explicitly recursive" factorial program like this:

|  0  0
| >g >p:
| *  /0#  <<<  \@
| >    v       6
| #0201
|  g` |<
|  0>g
|  #00
|  -p<
|  10+<
| ^  <
= Result: 720

Notice the application of / as a short-circuiting conditional operator: "evaluate to the north, then, if the condition to the south is true, return the result of the north evaluation; else return the result of evaluating to the east." Here the condition to the south is always 0, so I'm just using / as a more compact version of + that evaluates north-then-east instead of north-then-south.

If you go into this project's settings and mark master as your GitHub Pages branch, then navigating to https://catseye.github.io/Flobnar/demo/flobnar.html should Just Work, eliminating the need for people to download the HTML demo in order to see it.

Maybe these rules are intentional, but without explicit calling-out in the README, it feels like they might have been oversights (of the kind that reduce fun more than increase it).

The p term evaluates to the north to get an x coordinate, then to the south to get a y coordinate. It then evaluates what is on the other side of it to get a value.

Does this imply that

|   v<
|   #+@
|   1#
|   p1
|   6g
|    6
= Result: 6

because the 6 south of the p is evaluated twice — once as the y coordinate and then again as the "what is on the other side of" the p?

Similarly:

The \ term takes what to the south of it evaluates to, and uses that as the argument as it "applies" the "one-argument" "function" on the other side of it.

|   v\@
|   \2
|   >*
|    :
= Result: 8

because the second slash first evaluates "to the south" (2*:) to get the value 4, and then uses 4 as the argument to the function "on the other side of it" (identically 2*:) to get 2*4=8.

I'm generally surprised by the fixed cardinal directions in Flobnar, btw. It would feel much more "Befungey" if you had made

|  1+2
|   ^@
= Hypothetical Result: 3

(whereas with the fixed north/south rule, my understanding is that this program should produce infinite-recursion-and-crash while trying to evaluate the north operand of +).