Merge pull request #4 from gizatechxyz/refactor-signed-int

Refactor signed int

osiris/cairo/data_converter/data_statement_generator.py

@@ -40,12 +40,8 @@ def get_data_statement(data: np.ndarray, dtype: Dtype) -> list[str]:
         list[str]: The generated data statements.
     """
     match dtype:
-        case Dtype.U32:
+        case Dtype.U32 | Dtype.I32 | Dtype.I8:
             return [f"{int(x)}" for x in data.flatten()]
-        case Dtype.I32:
-            return ["i32 { "+f"mag: {abs(int(x))}, sign: {str(x < 0).lower()} "+"}" for x in data.flatten()]
-        case Dtype.I8:
-            return ["i8 { "+f"mag: {abs(int(x))}, sign: {str(x < 0).lower()} "+"}" for x in data.flatten()]
         case Dtype.FP8x23:
             return ["FP8x23 { "+f"mag: {abs(int(x))}, sign: {str(x < 0).lower()} "+"}" for x in data.flatten()]
         case Dtype.FP16x16:
@@ -99,12 +95,8 @@ def get_data_statement_for_sequences(data: Sequence, dtype: Dtype) -> list[list[
 
 dtype_to_numbers = {
     Dtype.U32: [],
-    Dtype.I32: [
-        "orion::numbers::{IntegerTrait, i32}",
-    ],
-    Dtype.I8: [
-        "orion::numbers::{IntegerTrait, i8}",
-    ],
+    Dtype.I32: [],
+    Dtype.I8: [],
     Dtype.FP8x23: [
         "orion::numbers::{FixedTrait, FP8x23}",
     ],

osiris/cairo/serde/data_structures.py

@@ -8,15 +8,9 @@ def __init__(self, shape: tuple, data):
         self.shape = shape
         self.data = data
 
-class UnsignedInt:
-    def __init__(self, mag):
-        self.mag = mag
-
-class SignedInt:
-    def __init__(self, mag, sign):
-        self.mag = mag
-        self.sign = sign
-
+class Int:
+    def __init__(self, val):
+        self.val = val
 
 class FixedPoint:
     def __init__(self, mag, sign):
@@ -32,11 +26,8 @@ def create_tensor_from_array(arr: np.ndarray, fp_impl='FP16x16'):
     tensor_data = []
 
     for value in flat_array:
-        if isinstance(value, (np.unsignedinteger)):
-            tensor_data.append(UnsignedInt(value))
-        elif isinstance(value, (int, np.integer, np.signedinteger)):
-            sign = 0 if value >= 0 else 1
-            tensor_data.append(SignedInt(abs(value), sign))
+        if isinstance(value, (np.integer)):
+            tensor_data.append(Int(value))
         elif isinstance(value, (float, np.floating)):
             (mag, sign) = to_fp(value, fp_impl)
             tensor_data.append(FixedPoint(mag, sign))

osiris/cairo/serde/deserialize.py

@@ -2,7 +2,7 @@
 
 import numpy as np
 
-from .utils import from_fp
+from .utils import felt_to_int, from_fp
 
 
 def deserializer(serialized: str, data_type: str, fp_impl='FP16x16'):
@@ -17,22 +17,16 @@ def deserializer(serialized: str, data_type: str, fp_impl='FP16x16'):
 
     serialized = convert_data(serialized)
 
-    if data_type == 'unsigned_int':
-        return deserialize_unsigned_int(serialized)
-    elif data_type == 'signed_int':
-        return deserialize_signed_int(serialized)
+    if data_type == 'int':
+        return deserialize_int(serialized)
     elif data_type == 'fixed_point':
         return deserialize_fixed_point(serialized, fp_impl)
-    elif data_type == 'arr_uint':
-        return deserialize_arr_uint(serialized)
-    elif data_type == 'arr_signed_int':
-        return deserialize_arr_signed_int(serialized)
+    elif data_type == 'arr_int':
+        return deserialize_arr_int(serialized)
     elif data_type == 'arr_fixed_point':
         return deserialize_arr_fixed_point(serialized, fp_impl)
-    elif data_type == 'tensor_uint':
-        return deserialize_tensor_uint(serialized)
-    elif data_type == 'tensor_signed_int':
-        return deserialize_tensor_signed_int(serialized)
+    elif data_type == 'tensor_int':
+        return deserialize_tensor_int(serialized)
     elif data_type == 'tensor_fixed_point':
         return deserialize_tensor_fixed_point(serialized)
     # TODO: Support Tuples
@@ -87,23 +81,13 @@ def convert_data(data):
     return result
 
 
-# ================= UNSIGNED INT =================
+# ================= INT =================
 
 
-def deserialize_unsigned_int(serialized: list) -> np.int64:
-    return np.int64(serialized[0])
+def deserialize_int(serialized: list) -> np.int64:
+    return np.int64(felt_to_int(serialized[0]))
 
 
-# ================= SIGNED INT =================
-
-
-def deserialize_signed_int(serialized: list) -> np.int64:
-    serialized_mag = serialized[0]
-    serialized_sign = serialized[1]
-
-    deserialized = serialized_mag if serialized_sign == 0 else -serialized_mag
-    return np.int64(deserialized)
-
 # ================= FIXED POINT =================
 
 
@@ -114,31 +98,17 @@ def deserialize_fixed_point(serialized: list, impl='FP16x16') -> np.float64:
     deserialized = serialized_mag if serialized_sign == 0 else -serialized_mag
     return np.float64(deserialized)
 
-# ================= ARRAY UINT =================
 
+# ================= ARRAY INT =================
 
-def deserialize_arr_uint(serialized: list) -> np.array:
-    return np.array(serialized[0], dtype=np.int64)
 
-# ================= ARRAY SIGNED INT =================
-
-
-def deserialize_arr_signed_int(serialized):
+def deserialize_arr_int(serialized):
 
     serialized = serialized[0]
 
-    if len(serialized) % 2 != 0:
-        raise ValueError("Array length must be even")
-
     deserialized = []
-    for i in range(0, len(serialized), 2):
-        mag = serialized[i]
-        sign = serialized[i + 1]
-
-        if sign == 1:
-            mag = -mag
-
-        deserialized.append(mag)
+    for ele in serialized:
+        deserialized.append(felt_to_int(ele))
 
     return np.array(deserialized)
 
@@ -162,21 +132,12 @@ def deserialize_arr_fixed_point(serialized: list, impl='FP16x16'):
     return np.array(deserialized)
 
 
-# ================= TENSOR UINT =================
-
-
-def deserialize_tensor_uint(serialized: list) -> np.array:
-    shape = serialized[0]
-    data = serialized[1]
-
-    return np.array(data, dtype=np.int64).reshape(shape)
-
-# ================= TENSOR SIGNED INT =================
+# ================= TENSOR INT =================
 
 
-def deserialize_tensor_signed_int(serialized: list) -> np.array:
+def deserialize_tensor_int(serialized: list) -> np.array:
     shape = serialized[0]
-    data = deserialize_arr_signed_int([serialized[1]])
+    data = deserialize_arr_int([serialized[1]])
 
     return np.array(data, dtype=np.int64).reshape(shape)
 

osiris/cairo/serde/serialize.py

@@ -1,29 +1,25 @@
 from osiris.cairo.serde.data_structures import (
     FixedPoint,
-    SignedInt,
+    Int,
     Tensor,
-    UnsignedInt,
 )
+from osiris.cairo.serde.utils import int_to_felt
 
 
 def serializer(data):
     if isinstance(data, bool):
         return "1" if data else "0"
     elif isinstance(data, int):
-        if data >= 0:
-            return f"{data}"
-        else:
-            raise ValueError("Native signed integers are not supported yet")
-            # TODO: Support native singned-int
+            return f"{int_to_felt(data)}"
+    elif isinstance(data, Int):
+            return f"{int_to_felt(data.val)}"
     elif isinstance(data, (list, tuple)):
         joined_elements = ' '.join(serializer(e) for e in data)
         return f"[{joined_elements}]"
     elif isinstance(data, Tensor):
         return f"{serializer(data.shape)} {serializer(data.data)}"
-    elif isinstance(data, (SignedInt, FixedPoint)):
+    elif isinstance(data, FixedPoint):
         return f"{serializer(data.mag)} {serializer(data.sign)}"
-    elif isinstance(data, UnsignedInt):
-        return f"{data.mag}"
 
     else:
         raise ValueError("Unsupported data type for serialization")

osiris/cairo/serde/utils.py

@@ -22,3 +22,19 @@ def from_fp(value, fp_impl='FP16x16'):
             return value / 2**32
         case 'FP64x64':
             return value / 2**64
+
+def int_to_felt(n):
+    PRIME_FIELD = 2**251 + 17 * 2**192 + 1
+    if n < 0:
+        return (PRIME_FIELD + n) % PRIME_FIELD
+    else:
+        return n % PRIME_FIELD
+
+def felt_to_int(felt):
+    PRIME_FIELD = 2**251 + 17 * 2**192 + 1
+    HALF_FIELD = PRIME_FIELD // 2
+
+    if felt > HALF_FIELD:
+        return felt - PRIME_FIELD
+    else:
+        return felt

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "giza-osiris"
-version = "0.1.8"
+version = "0.1.9"
 description = "Osiris is a Python library designed for efficient data conversion and management, primarily transforming data into Cairo programs"
 authors = ["Fran Algaba <[email protected]>"]
 readme = "README.md"

tests/test_deserialize.py

@@ -5,13 +5,13 @@
 from osiris.cairo.serde.deserialize import *
 
 
-def test_deserialize_signed_int():
-    serialized = '[{"Int":"2A"}, {"Int":"0"}]'
-    deserialized = deserializer(serialized, 'signed_int')
+def test_deserialize_int():
+    serialized = '[{"Int":"2A"}]'
+    deserialized = deserializer(serialized, 'int')
     assert deserialized == 42
 
-    serialized = '[{"Int":"2A"}, {"Int":"0x1"}]'
-    deserialized = deserializer(serialized, 'signed_int')
+    serialized = '[{"Int":"800000000000010FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFD7"}]'
+    deserialized = deserializer(serialized, 'int')
     assert deserialized == -42
 
 
@@ -25,15 +25,13 @@ def test_deserialize_fp():
     assert isclose(deserialized, -42.42, rel_tol=1e-7)
 
 
-def test_deserialize_array_uint():
+def test_deserialize_array_int():
     serialized = '[{"Array": [{"Int": "0x1"}, {"Int": "0x2"}]}]'
-    deserialized = deserializer(serialized, 'arr_uint')
+    deserialized = deserializer(serialized, 'arr_int')
     assert np.array_equal(deserialized, np.array([1, 2], dtype=np.int64))
 
-
-def test_deserialize_array_signed_int():
-    serialized = '[{"Array": [{"Int": "2A"}, {"Int": "0"}, {"Int": "2A"}, {"Int": "0x1"}]}]'
-    deserialized = deserializer(serialized, 'arr_signed_int')
+    serialized = '[{"Array": [{"Int": "2A"}, {"Int": "800000000000010FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFD7"}]}]'
+    deserialized = deserializer(serialized, 'arr_int')
     assert np.array_equal(deserialized, np.array([42, -42], dtype=np.int64))
 
 
@@ -44,16 +42,14 @@ def test_deserialize_arr_fixed_point():
     assert np.all(np.isclose(deserialized, expected, atol=1e-7))
 
 
-def test_deserialize_tensor_uint():
+def test_deserialize_tensor_int():
     serialized = '[{"Array": [{"Int": "0x2"}, {"Int": "0x2"}]}, {"Array": [{"Int": "0x1"}, {"Int": "0x2"}, {"Int": "0x3"}, {"Int": "0x4"}]}]'
-    deserialized = deserializer(serialized, 'tensor_uint')
+    deserialized = deserializer(serialized, 'tensor_int')
     assert np.array_equal(deserialized, np.array(
         ([1, 2], [3, 4]), dtype=np.int64))
 
-
-def test_deserialize_tensor_signed_int():
-    serialized = '[{"Array": [{"Int": "0x2"}, {"Int": "0x2"}]}, {"Array": [{"Int": "2A"}, {"Int": "0x0"}, {"Int": "2A"}, {"Int": "0x0"}, {"Int": "2A"}, {"Int": "0x1"}, {"Int": "2A"}, {"Int": "0x1"}]}]'
-    deserialized = deserializer(serialized, 'tensor_signed_int')
+    serialized = '[{"Array": [{"Int": "0x2"}, {"Int": "0x2"}]}, {"Array": [{"Int": "2A"}, {"Int": "2A"},{"Int": "800000000000010FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFD7"}, {"Int": "800000000000010FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFD7"}]}]'
+    deserialized = deserializer(serialized, 'tensor_int')
     assert np.array_equal(deserialized, np.array([[42, 42], [-42, -42]]))
 
 

tests/test_serialize.py

@@ -1,6 +1,6 @@
 import numpy as np
 import pytest
-from osiris.cairo.serde.data_structures import create_tensor_from_array, Tensor, SignedInt, FixedPoint
+from osiris.cairo.serde.data_structures import create_tensor_from_array, Tensor, FixedPoint, Int
 from osiris.cairo.serde.serialize import serializer
 
 
@@ -9,7 +9,7 @@ def test_create_tensor_from_array_with_integers():
     tensor = create_tensor_from_array(arr)
     assert isinstance(tensor, Tensor)
     assert tensor.shape == arr.shape
-    assert all(isinstance(x, SignedInt) for x in tensor.data)
+    assert all(isinstance(x, Int) for x in tensor.data)
 
 
 def test_create_tensor_from_array_with_floats():
@@ -48,19 +48,22 @@ def test_serializer_for_tuple():
     serialized_data = serializer(data)
     assert serialized_data == "[1 2 3]"
 
+
 def test_serializer_for_fixedpoint():
     data = FixedPoint(42, True)
     serialized_data = serializer(data)
-    assert serialized_data == "42 1" 
+    assert serialized_data == "42 1"
+
 
-def test_serializer_for_tensor_uint():
+def test_serializer_for_tensor_int():
     arr = np.array([[1, 2], [3, 4]], dtype=np.uint64)
     tensor = create_tensor_from_array(arr)
     serialized_data = serializer(tensor)
-    assert serialized_data == "[2 2] [1 2 3 4]" 
+    assert serialized_data == "[2 2] [1 2 3 4]"
+
 
 def test_serializer_for_tensor_fixedpoint():
     arr = np.array([[1, 2], [3, 4]], dtype=np.float32)
     tensor = create_tensor_from_array(arr)
     serialized_data = serializer(tensor)
-    assert serialized_data == "[2 2] [65536 0 131072 0 196608 0 262144 0]" 
+    assert serialized_data == "[2 2] [65536 0 131072 0 196608 0 262144 0]"