Make some progress on ASTs.

offchain/Cargo.lock

@@ -879,6 +879,7 @@ dependencies = [
 name = "offchain"
 version = "0.1.0"
 dependencies = [
+ "rand",
  "revm",
  "serde",
  "serde_json",

offchain/Cargo.toml

@@ -6,4 +6,5 @@ edition = "2021"
 [dependencies]
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
-revm = "18.0.0" 
+revm = "18.0.0"
+rand = "0.8"

offchain/src/compiler/ast.rs

@@ -1,15 +1,114 @@
+pub trait Push3Ast {
+    /// Convert this AST into a bytecode array (Push3 tokens).
+    fn to_bytecode(&self) -> Vec<u8>;
+
+    /// Parse an AST from bytecode (the "decompiler"), if possible.
+    /// Possibly returns a Result in case of partial/invalid parse.
+    fn from_bytecode(data: &[u8]) -> Self
+    where
+        Self: Sized;
+
+    /// Randomly mutate this AST in-place (GP mutation).
+    fn mutate(&mut self, rng: &mut impl rand::Rng);
+
+    /// Possibly a "crossover" method, or we do it externally.
+    // fn crossover(&self, other: &Self, rng: &mut impl rand::Rng) -> Self;
+
+    /// Evaluate the AST with some input(s).
+    /// Optionally do we want this or do we always run in EVM? 
+    /// We can stub it or do an interpret approach for local test.
+    fn interpret_locally(&self, inputs: &[i32]) -> Vec<i32>;
+}
+
+
 #[derive(Debug, Clone)]
-pub enum Expr {
+pub enum TypedAst {
     IntLiteral(i32),
-    Instruction(InstructionKind, Vec<Expr>),
-    // You might have a variant for sublists if needed,
-    // or keep it all in the same structure.
+    Instruction(InstructionKind),
 }
 
 #[derive(Debug, Clone)]
 pub enum InstructionKind {
-    // e.g. Plus, Minus, Mult, etc.
+    Plus(Box<TypedAst>, Box<TypedAst>),
+    Minus(Box<TypedAst>, Box<TypedAst>),
+    Mult(Box<TypedAst>, Box<TypedAst>),
+    Dup,  // 0 children
+    Pop,  // 0 children
+}
+
+impl Push3Ast for TypedAst {
+    fn to_bytecode(&self) -> Vec<u8> {
+        // do typed encoding: 
+        // - for Plus(e1, e2): e1 -> bytecode, e2 -> bytecode, then PLUS opcode
+        unimplemented!()
+    }
+
+    fn from_bytecode(data: &[u8]) -> Self {
+        // parse in a strictly typed manner; might fail or panic 
+        // if the code doesn't line up properly
+        unimplemented!()
+    }
+
+    fn mutate(&mut self, rng: &mut impl rand::Rng) {
+        // e.g. randomly pick a node, replace it with something 
+        // that respects the typed structure
+        unimplemented!()
+    }
+
+    fn interpret_locally(&self, inputs: &[i32]) -> Vec<i32> {
+        // do a normal expression evaluation. If a node is "Plus(a,b)",
+        // interpret (a) + interpret(b).
+        unimplemented!()
+    }
+}
+
+#[derive(Debug, Clone)]
+pub enum UntypedAst {
+    /// A single token in a list-based approach, e.g. "Integer literal"
+    IntLiteral(i32),
+
+    /// "Instruction" might be a simple variant with no children, 
+    /// because the stack-based evaluation decides how many arguments it consumes.
+    Instruction(OpCode),  
+
+    /// A sublist of tokens, e.g. "( 2 3 DUP + )"
+    Sublist(Vec<UntypedAst>),
+}
+
+/// Possibly an enum for OpCode:
+#[derive(Debug, Clone)]
+pub enum OpCode {
+    Noop,
     Plus,
     Minus,
     Mult,
+    Dup,
+    Pop,
+    // maybe Unknown => treat as no-op
+}
+
+impl Push3Ast for UntypedAst {
+    fn to_bytecode(&self) -> Vec<u8> {
+        // flatten out sublists vs instructions
+        unimplemented!()
+    }
+
+    fn from_bytecode(data: &[u8]) -> Self {
+        // parse "0x02 => INT_LITERAL => read next 4 bytes" 
+        // instructions might have no subchildren
+        unimplemented!()
+    }
+
+    fn mutate(&mut self, rng: &mut impl rand::Rng) {
+        // can randomly insert unknown tokens, reorder sublists, etc.
+        // possibly produce "invalid ops => noops"
+        unimplemented!()
+    }
+
+    fn interpret_locally(&self, inputs: &[i32]) -> Vec<i32> {
+        // push-based interpretation:
+        // sublist => interpret each item in sequence
+        // if an instruction sees insufficient args => no-op
+        unimplemented!()
+    }
 }

offchain/src/compiler/bytecode_generator.rs

@@ -1,7 +0,0 @@
-use super::ast::{Expr, InstructionKind};
-
-pub fn ast_to_bytecode(expr: &Expr) -> Vec<u8> {
-    // for now, just a stub
-    // eventually we produce push3 tokens
-    vec![]
-}

offchain/src/compiler/bytecode_parser.rs

@@ -1,7 +0,0 @@
-use super::ast::{Expr};
-
-pub fn bytecode_to_ast(data: &[u8]) -> Expr {
-    // parse tokens, produce expr
-    // stub for now
-    Expr::IntLiteral(0)
-}

offchain/src/compiler/mod.rs

@@ -1,5 +1 @@
 pub mod ast;
-pub mod bytecode_generator;
-pub mod bytecode_parser;
-
-// In the future, we'll define or `pub use` them.

offchain/src/gp/genetic_ops.rs

@@ -1,10 +1,10 @@
-use crate::compiler::ast::{Expr};
+// use crate::compiler::ast::{Expr};
 
-pub fn crossover(a: &Expr, b: &Expr) -> (Expr, Expr) {
-    // stub
-    (a.clone(), b.clone())
-}
+// pub fn crossover(a: &Expr, b: &Expr) -> (Expr, Expr) {
+//     // stub
+//     (a.clone(), b.clone())
+// }
 
-pub fn mutate(expr: &mut Expr) {
-    // stub
-}
+// pub fn mutate(expr: &mut Expr) {
+//     // stub
+// }