fix: add support for nested arrays returned by oracles

acvm-repo/brillig_vm/src/lib.rs

@@ -482,60 +482,64 @@ impl<'a, F: AcirField, B: BlackBoxFunctionSolver<F>> VM<'a, F, B> {
             destinations.iter().zip(destination_value_types).zip(&values)
         {
             match (destination, value_type) {
-                (ValueOrArray::MemoryAddress(value_index), HeapValueType::Simple(bit_size)) => {
-                    match output {
-                        ForeignCallParam::Single(value) => {
-                            self.write_value_to_memory(*value_index, value, *bit_size)?;
-                        }
-                        _ => return Err(format!(
-                            "Function result size does not match brillig bytecode. Expected 1 result but got {output:?}")
-                        ),
+            (ValueOrArray::MemoryAddress(value_index), HeapValueType::Simple(bit_size)) => {
+                match output {
+                    ForeignCallParam::Single(value) => {
+                        self.write_value_to_memory(*value_index, value, *bit_size)?;
                     }
+                    _ => return Err(format!(
+                        "Function result size does not match brillig bytecode. Expected 1 result but got {output:?}")
+                    ),
                 }
-                (
-                    ValueOrArray::HeapArray(HeapArray { pointer: pointer_index, size }),
-                    HeapValueType::Array { value_types, size: type_size },
-                ) if size == type_size => {
-                    if HeapValueType::all_simple(value_types) {
-                         match output {
-                            ForeignCallParam::Array(values) => {
-                                if values.len() != *size {
-                                    return Err("Foreign call result array doesn't match expected size".to_string());
-                                }
-                                self.write_values_to_memory_slice(*pointer_index, values, value_types)?;
-                            }
-                            _ => {
-                                return Err("Function result size does not match brillig bytecode size".to_string());
+            }
+            (
+                ValueOrArray::HeapArray(HeapArray { pointer: pointer_index, size }),
+                HeapValueType::Array { value_types, size: type_size },
+            ) if size == type_size => {
+                if HeapValueType::all_simple(value_types) {
+                    match output {
+                        ForeignCallParam::Array(values) => {
+                            if values.len() != *size {
+                                return Err("Foreign call result array doesn't match expected size".to_string());
                             }
+                            self.write_values_to_memory_slice(*pointer_index, values, value_types)?;
+                        }
+                        _ => {
+                            return Err("Function result size does not match brillig bytecode size".to_string());
                         }
-                    } else {
-                        unimplemented!("deflattening heap arrays from foreign calls");
                     }
-                }
-                (
-                    ValueOrArray::HeapVector(HeapVector {pointer: pointer_index, size: size_index }),
-                    HeapValueType::Vector { value_types },
-                ) => {
-                    if HeapValueType::all_simple(value_types) {
-                        match output {
-                            ForeignCallParam::Array(values) => {
-                                // Set our size in the size address
-                                self.memory.write(*size_index, values.len().into());
-
-                                self.write_values_to_memory_slice(*pointer_index, values, value_types)?;
-                            }
-                            _ => {
-                                return Err("Function result size does not match brillig bytecode size".to_string());
-                            }
+                } else {
+                    // foreign call returning flatten values into a nested type, so the sizes do not match
+                    let destination = self.memory.read_ref(*pointer_index);
+                    let return_type = value_type;
+                    let mut flatten_values_idx = 0; //index of values read from flatten_values
+                    self.write_slice_of_values_to_memory(destination, &output.fields(), &mut flatten_values_idx, return_type)?;
+            }
+        }
+            (
+                ValueOrArray::HeapVector(HeapVector {pointer: pointer_index, size: size_index }),
+                HeapValueType::Vector { value_types },
+            ) => {
+                if HeapValueType::all_simple(value_types) {
+                    match output {
+                        ForeignCallParam::Array(values) => {
+                            // Set our size in the size address
+                            self.memory.write(*size_index, values.len().into());
+                            self.write_values_to_memory_slice(*pointer_index, values, value_types)?;
+
+                        }
+                        _ => {
+                            return Err("Function result size does not match brillig bytecode size".to_string());
                         }
-                    } else {
-                        unimplemented!("deflattening heap vectors from foreign calls");
                     }
-                }
-                _ => {
-                    return Err(format!("Unexpected value type {value_type:?} for destination {destination:?}"));
+                } else {
+                    unimplemented!("deflattening heap vectors from foreign calls");
                 }
             }
+            _ => {
+                return Err(format!("Unexpected value type {value_type:?} for destination {destination:?}"));
+            }
+        }
         }
 
         let _ =
@@ -596,6 +600,66 @@ impl<'a, F: AcirField, B: BlackBoxFunctionSolver<F>> VM<'a, F, B> {
         Ok(())
     }
 
+    /// Writes flatten values to memory, using the provided type
+    /// Function calls itself recursively in order to work with recursive types (nested arrays)
+    /// values_idx is the current index in the values vector and is incremented every time
+    /// a value is written to memory
+    /// The function returns the address of the next value to be written
+    fn write_slice_of_values_to_memory(
+        &mut self,
+        destination: MemoryAddress,
+        values: &Vec<F>,
+        values_idx: &mut usize,
+        value_type: &HeapValueType,
+    ) -> Result<MemoryAddress, String> {
+        let mut current_pointer = destination;
+        match value_type {
+            HeapValueType::Simple(bit_size) => {
+                self.write_value_to_memory(destination, &values[*values_idx], *bit_size)?;
+                *values_idx += 1;
+                Ok(MemoryAddress(destination.to_usize() + 1))
+            }
+            HeapValueType::Array { value_types, size } => {
+                for _ in 0..*size {
+                    for typ in value_types {
+                        match typ {
+                            HeapValueType::Simple(len) => {
+                                self.write_value_to_memory(
+                                    current_pointer,
+                                    &values[*values_idx],
+                                    *len,
+                                )?;
+                                *values_idx += 1;
+                                current_pointer = MemoryAddress(current_pointer.to_usize() + 1);
+                            }
+                            HeapValueType::Array { .. } => {
+                                let destination = self.memory.read_ref(current_pointer);
+                                let destination = self.memory.read_ref(destination);
+                                self.write_slice_of_values_to_memory(
+                                    destination,
+                                    values,
+                                    values_idx,
+                                    typ,
+                                )?;
+                                current_pointer = MemoryAddress(current_pointer.to_usize() + 1);
+                            }
+                            HeapValueType::Vector { .. } => {
+                                return Err(format!(
+                                    "Unsupported returned type in foreign calls {:?}",
+                                    typ
+                                ));
+                            }
+                        }
+                    }
+                }
+                Ok(current_pointer)
+            }
+            HeapValueType::Vector { .. } => {
+                Err(format!("Unsupported returned type in foreign calls {:?}", value_type))
+            }
+        }
+    }
+
     /// Process a binary operation.
     /// This method will not modify the program counter.
     fn process_binary_field_op(

compiler/noirc_evaluator/src/brillig/brillig_gen/brillig_block.rs

@@ -1737,8 +1737,7 @@ impl<'block> BrilligBlock<'block> {
                     dfg,
                 );
                 let array = variable.extract_array();
-                self.brillig_context.codegen_allocate_fixed_length_array(array.pointer, array.size);
-                self.brillig_context.usize_const_instruction(array.rc, 1_usize.into());
+                self.allocate_nested_array(typ, Some(array));
 
                 variable
             }
@@ -1765,6 +1764,39 @@ impl<'block> BrilligBlock<'block> {
         }
     }
 
+    fn allocate_nested_array(
+        &mut self,
+        typ: &Type,
+        array: Option<BrilligArray>,
+    ) -> BrilligVariable {
+        match typ {
+            Type::Array(types, size) => {
+                let array = array.unwrap_or(BrilligArray {
+                    pointer: self.brillig_context.allocate_register(),
+                    size: *size,
+                    rc: self.brillig_context.allocate_register(),
+                });
+                self.brillig_context.codegen_allocate_fixed_length_array(array.pointer, array.size);
+                self.brillig_context.usize_const_instruction(array.rc, 1_usize.into());
+
+                let mut index = 0_usize;
+                for _ in 0..*size {
+                    for element_type in types.iter() {
+                        if matches!(element_type, Type::Array(_, _)) {
+                            let inner_array = self.allocate_nested_array(element_type, None);
+                            let idx =
+                                self.brillig_context.make_usize_constant_instruction(index.into());
+                            self.store_variable_in_array(array.pointer, idx, inner_array);
+                        }
+                        index += 1;
+                    }
+                }
+                BrilligVariable::BrilligArray(array)
+            }
+            _ => unreachable!("ICE: allocate_nested_array() expects an array, got {typ:?}"),
+        }
+    }
+
     /// Gets the "user-facing" length of an array.
     /// An array of structs with two fields would be stored as an 2 * array.len() array/vector.
     /// So we divide the length by the number of subitems in an item to get the user-facing length.

docs/docs/how_to/how-to-oracles.md

@@ -166,6 +166,10 @@ interface ForeignCallResult {
 }
 ```
 
+::: Multidimensional Arrays
+
+If the Oracle function is returning an array containing other arrays, such as `[['1','2],['3','4']]`, you need to provide the values in json as flattened values. In the previous example, it would be `['1', '2', '3', '4']`. In the noir program, the Oracle signature can use a nested type, the flattened values will be automatically converted to the nested type.
+
 :::
 
 ## Step 3 - Usage with Nargo

test_programs/noir_test_success/regression_4561/Nargo.toml

@@ -0,0 +1,6 @@
+[package]
+name = "regression_4561"
+type = "bin"
+authors = [""]
+
+[dependencies]

test_programs/noir_test_success/regression_4561/src/main.nr

@@ -0,0 +1,44 @@
+// Regression test for issue #4561
+use dep::std::test::OracleMock;
+
+type TReturnElem = [Field; 3];
+type TReturn = [TReturnElem; 2];
+
+#[oracle(simple_nested_return)]
+unconstrained fn simple_nested_return_oracle() -> TReturn {}
+
+unconstrained fn simple_nested_return_unconstrained() -> TReturn {
+    simple_nested_return_oracle()
+}
+
+#[test]
+fn test_simple_nested_return() {
+    OracleMock::mock("simple_nested_return").returns([1, 2, 3, 4, 5, 6]);
+    assert_eq(simple_nested_return_unconstrained(), [[1, 2, 3], [4, 5, 6]]);
+}
+
+#[oracle(nested_with_fields_return)]
+unconstrained fn nested_with_fields_return_oracle() -> (Field, TReturn, Field) {}
+
+unconstrained fn nested_with_fields_return_unconstrained() -> (Field, TReturn, Field) {
+    nested_with_fields_return_oracle()
+}
+
+#[test]
+fn test_nested_with_fields_return() {
+    OracleMock::mock("nested_with_fields_return").returns((0, [1, 2, 3, 4, 5, 6], 7));
+    assert_eq(nested_with_fields_return_unconstrained(), (0, [[1, 2, 3], [4, 5, 6]], 7));
+}
+
+#[oracle(two_nested_return)]
+unconstrained fn two_nested_return_oracle() -> (Field, TReturn, Field, TReturn) {}
+
+unconstrained fn two_nested_return_unconstrained() -> (Field, TReturn, Field, TReturn) {
+    two_nested_return_oracle()
+}
+
+#[test]
+fn two_nested_return() {
+    OracleMock::mock("two_nested_return").returns((0, [1, 2, 3, 4, 5, 6], 7, [1, 2, 3, 4, 5, 6]));
+    assert_eq(two_nested_return_unconstrained(), (0, [[1, 2, 3], [4, 5, 6]], 7, [[1, 2, 3], [4, 5, 6]]));
+}