Merge pull request #6 from NillionNetwork/feature/alphas-array

Feature/alphas array

nada_algebra/array.py

@@ -4,7 +4,7 @@
 """
 
 from dataclasses import dataclass
-from typing import Callable, Union
+from typing import Any, Callable, Union
 
 import numpy as np
 from nada_dsl import (
@@ -31,6 +31,34 @@ class NadaArray:
 
     inner: np.ndarray
 
+    SUPPORTED_OPERATIONS = {
+        "compress",
+        "copy",
+        "cumprod",
+        "cumsum",
+        "diagonal",
+        "fill",
+        "flatten",
+        "item",
+        "itemset",
+        "prod",
+        "put",
+        "ravel",
+        "repeat",
+        "reshape",
+        "resize",
+        "shape",
+        "size",
+        "squeeze",
+        "sum",
+        "swapaxes",
+        "T",
+        "take",
+        "tolist",
+        "trace",
+        "transpose",
+    }
+
     def __getitem__(self, item):
         """
         Get an item from the array.
@@ -54,23 +82,10 @@ def __setitem__(self, key, value):
             value: The value to set.
         """
         if isinstance(value, NadaArray):
-            # print("NadaArray")
             self.inner[key] = value.inner
         else:
             self.inner[key] = value
 
-    def __getattr__(self, name: str):
-        """
-        Get an attribute from the array.
-
-        Args:
-            name (str): The attribute name.
-
-        Returns:
-            NadaArray: A new NadaArray representing the retrieved attribute.
-        """
-        return getattr(self.inner, name)
-
     def __add__(
         self,
         other: Union[
@@ -220,15 +235,6 @@ def dot(self, other: "NadaArray") -> "NadaArray":
         """
         return NadaArray(self.inner.dot(other.inner))
 
-    def sum(self) -> Union[SecretInteger, SecretUnsignedInteger]:
-        """
-        Compute the sum of the elements in the array.
-
-        Returns:
-            Union[SecretInteger, SecretUnsignedInteger]: The sum of the array elements.
-        """
-        return NadaArray(self.inner.sum())
-
     def hstack(self, other: "NadaArray") -> "NadaArray":
         """
         Horizontally stack two NadaArray objects.
@@ -424,3 +430,39 @@ def random(
                 )
             )
         )
+
+    def __getattr__(self, name: str) -> Any:
+        """Routes other attributes to the inner NumPy array.
+
+        Args:
+            name (str): Attribute name.
+
+        Raises:
+            AttributeError: Raised if attribute not supported.
+
+        Returns:
+            Any: Result of attribute.
+        """
+        if name not in self.SUPPORTED_OPERATIONS:
+            raise AttributeError("NumPy method `%s` is not (currently) supported by NadaArrays." % name)
+
+        attr = getattr(self.inner, name)
+
+        if callable(attr):
+            def wrapper(*args, **kwargs):
+                result = attr(*args, **kwargs)
+                if isinstance(result, np.ndarray):
+                    return NadaArray(result)
+                return result
+            return wrapper
+
+        if isinstance(attr, np.ndarray):
+            attr = NadaArray(attr)
+
+        return attr
+
+    def __setattr__(self, name, value):
+        if name == 'inner':
+            super().__setattr__(name, value)
+        else:
+            setattr(self.inner, name, value)

nada_algebra/funcs.py

@@ -3,6 +3,7 @@
 and manipulation of arrays and party objects.
 """
 
+from typing import Any, Iterable
 from nada_dsl import (
     Party,
     SecretInteger,
@@ -30,20 +31,20 @@ def parties(num: int, prefix: str = "Party") -> list:
     return [Party(name=f"{prefix}{i}") for i in range(num)]
 
 
-def __from_list(lst: list, nada_type: Integer | UnsignedInteger) -> list:
+def __from_numpy(arr: np.ndarray, nada_type: Integer | UnsignedInteger) -> list:
     """
-    Recursively convert a nested list to a list of NadaInteger objects.
+    Recursively convert a n-dimensional NumPy array to a nested list of NadaInteger objects.
 
     Args:
-        lst (list): A nested list of integers.
+        arr (np.ndarray): A NumPy array of integers.
         nada_type (type): The type of NadaInteger objects to create.
 
     Returns:
         list: A nested list of NadaInteger objects.
     """
-    if len(lst.shape) == 1:
-        return [nada_type(int(elem)) for elem in lst]
-    return [__from_list(lst[i], nada_type) for i in range(len(lst))]
+    if len(arr.shape) == 1:
+        return [nada_type(int(elem)) for elem in arr]
+    return [__from_numpy(arr[i], nada_type) for i in range(arr.shape[0])]
 
 
 def from_list(lst: list, nada_type: Integer | UnsignedInteger = Integer) -> NadaArray:
@@ -59,15 +60,15 @@ def from_list(lst: list, nada_type: Integer | UnsignedInteger = Integer) -> Nada
     """
     if not isinstance(lst, np.ndarray):
         lst = np.array(lst)
-    return NadaArray(np.array(__from_list(lst, nada_type)))
+    return NadaArray(np.array(__from_numpy(lst, nada_type)))
 
 
-def ones(dims: list, nada_type: Integer | UnsignedInteger = Integer) -> NadaArray:
+def ones(dims: Iterable[int], nada_type: Integer | UnsignedInteger = Integer) -> NadaArray:
     """
     Create a cleartext NadaArray filled with ones.
 
     Args:
-        dims (list): A list of integers representing the dimensions of the array.
+        dims (Iterable[int]): A list of integers representing the dimensions of the array.
         nada_type (type, optional): The type of NadaInteger objects to create. Defaults to Integer.
 
     Returns:
@@ -76,12 +77,28 @@ def ones(dims: list, nada_type: Integer | UnsignedInteger = Integer) -> NadaArra
     return from_list(np.ones(dims), nada_type)
 
 
-def zeros(dims: list, nada_type: Integer | UnsignedInteger = Integer) -> NadaArray:
+def ones_like(a: np.ndarray | NadaArray, nada_type: Integer | UnsignedInteger = Integer) -> NadaArray:
+    """
+    Create a cleartext NadaArray filled with one with the same shape and type as a given array.
+
+    Args:
+        a (np.ndarray | NadaArray): A reference array.
+        nada_type (type, optional): The type of NadaInteger objects to create. Defaults to Integer.
+
+    Returns:
+        NadaArray: The created NadaArray filled with ones.
+    """
+    if isinstance(a, NadaArray):
+        a = a.inner
+    return from_list(np.ones_like(a), nada_type)
+
+
+def zeros(dims: Iterable[int], nada_type: Integer | UnsignedInteger = Integer) -> NadaArray:
     """
     Create a cleartext NadaArray filled with zeros.
 
     Args:
-        dims (list): A list of integers representing the dimensions of the array.
+        dims (Iterable[int]): A list of integers representing the dimensions of the array.
         nada_type (type, optional): The type of NadaInteger objects to create. Defaults to Integer.
 
     Returns:
@@ -90,8 +107,56 @@ def zeros(dims: list, nada_type: Integer | UnsignedInteger = Integer) -> NadaArr
     return from_list(np.zeros(dims), nada_type)
 
 
+def zeros_like(a: np.ndarray | NadaArray, nada_type: Integer | UnsignedInteger = Integer) -> NadaArray:
+    """
+    Create a cleartext NadaArray filled with zeros with the same shape and type as a given array.
+
+    Args:
+        a (np.ndarray | NadaArray): A reference array.
+        nada_type (type, optional): The type of NadaInteger objects to create. Defaults to Integer.
+
+    Returns:
+        NadaArray: The created NadaArray filled with zeros.
+    """
+    if isinstance(a, NadaArray):
+        a = a.inner
+    return from_list(np.zeros_like(a), nada_type)
+
+
+def alphas(dims: Iterable[int], alpha: Any) -> NadaArray:
+    """
+    Create a NadaArray filled with a certain constant value.
+
+    Args:
+        dims (Iterable[int]): A list of integers representing the dimensions of the array.
+        alpha (Any): Some constant value.
+
+    Returns:
+        NadaArray: NadaArray filled with constant value.
+    """
+    ones_array = np.ones(dims)
+    return NadaArray(np.frompyfunc(lambda _: alpha, 1, 1)(ones_array))
+
+
+def alphas_like(a: np.ndarray | NadaArray, alpha: Any) -> NadaArray:
+    """
+    Create a NadaArray filled with a certain constant value with the same shape and type as a given array.
+
+    Args:
+        a (np.ndarray | NadaArray): Reference array.
+        alpha (Any): Some constant value.
+
+    Returns:
+        NadaArray: NadaArray filled with constant value.
+    """
+    if isinstance(a, NadaArray):
+        a = a.inner
+    ones_array = np.ones_like(a)
+    return NadaArray(np.frompyfunc(lambda _: alpha, 1, 1)(ones_array))
+
+
 def array(
-    dims: list,
+    dims: Iterable[int],
     party: Party,
     prefix: str,
     nada_type: (
@@ -102,7 +167,7 @@ def array(
     Create a NadaArray with the specified dimensions and elements of the given type.
 
     Args:
-        dims (list): A list of integers representing the dimensions of the array.
+        dims (Iterable[int]): A list of integers representing the dimensions of the array.
         party (Party): The party object.
         prefix (str): A prefix for naming the array elements.
         nada_type (type, optional): The type of elements to create. Defaults to SecretInteger.
@@ -114,13 +179,13 @@ def array(
 
 
 def random(
-    dims: list, nada_type: SecretInteger | SecretUnsignedInteger = SecretInteger
+    dims: Iterable[int], nada_type: SecretInteger | SecretUnsignedInteger = SecretInteger
 ) -> NadaArray:
     """
     Create a random NadaArray with the specified dimensions.
 
     Args:
-        dims (list): A list of integers representing the dimensions of the array.
+        dims (Iterable[int]): A list of integers representing the dimensions of the array.
         nada_type (type, optional): The type of elements to create. Defaults to SecretInteger.
 
     Returns:

tests/generate_array/nada-project.toml

@@ -0,0 +1,7 @@
+name = "generate_array"
+version = "0.1.0"
+authors = [""]
+
+[[programs]]
+path = "src/main.py"
+prime_size = 128

tests/generate_array/src/main.py

@@ -0,0 +1,23 @@
+from nada_dsl import *
+import nada_algebra as na
+
+
+def nada_main():
+    party = Party("party_0")
+
+    a = SecretInteger(Input("a", party))
+
+    ones1 = na.ones([2, 3])
+    ones2 = na.ones_like(ones1)
+
+    zeros1 = na.zeros([2, 3])
+    zeros2 = na.zeros_like(zeros1)
+
+    alphas1 = na.alphas([2, 3], alpha=a)
+    alphas2 = na.alphas_like(alphas1, alpha=a)
+
+    two_a = alphas1 + alphas2
+
+    out = two_a + zeros1 + zeros2 + ones1 + ones2
+
+    return out.output(party, "my_output")

tests/generate_array/target/.gitignore

@@ -0,0 +1,5 @@
+# This directory is kept purposely, so that no compilation errors arise.
+# Ignore everything in this directory
+*
+# Except this file
+!.gitignore

tests/generate_array/tests/base.yaml

@@ -0,0 +1,20 @@
+---
+program: main
+inputs:
+  secrets:
+    a:
+      SecretInteger: "3"
+  public_variables: {}
+expected_outputs:
+  my_output_0_0:
+    SecretInteger: "8"
+  my_output_0_1:
+    SecretInteger: "8"
+  my_output_0_2:
+    SecretInteger: "8"
+  my_output_1_0:
+    SecretInteger: "8"
+  my_output_1_1:
+    SecretInteger: "8"
+  my_output_1_2:
+    SecretInteger: "8"

tests/sum/src/main.py

@@ -9,4 +9,4 @@ def nada_main():
 
     result = a.sum()
 
-    return result.output(parties[1], "my_output")
+    return [Output(result, "my_output", parties[1])]

tests/supported_operations/nada-project.toml

@@ -0,0 +1,7 @@
+name = "supported_operations"
+version = "0.1.0"
+authors = [""]
+
+[[programs]]
+path = "src/main.py"
+prime_size = 128