feat: Do not infer dtype when data is None in tensor.Box (#261)

* feat: Do not infer dtype when `data` is `None` in `tensor.Box`

Tests have been added accordingly.

Fixes weird behavior where tensor boxes that should be equal are different.

* Implement changes suggested by @toumix

discopy/tensor.py

@@ -525,7 +525,6 @@ class Box(frobenius.Box, Diagram):
         dom : The domain of the box, i.e. its input dimension.
         cod : The codomain of the box, i.e. its output dimension.
         data : The array inside the tensor box.
-        dtype : The datatype for the entries of the array.
 
     Example
     -------
@@ -541,16 +540,17 @@ def __setstate__(self, state):
             state['data'] = state['_array']
             del state["_array"]
         super().__setstate__(state)
-        if self.dtype is None:
+        if self.dtype is None and self.data is not None:
             self.data, self.dtype = self._get_data_dtype(self.data)
             self.__class__ = self.__class__[self.dtype]
 
-    def __new__(cls, *args, **kwargs):
-        if not args and not kwargs or cls.dtype is not None:
+    def __new__(
+            cls, name=None, dom=None, cod=None, data=None, *args, **kwargs):
+        if cls.dtype is not None or data is None:
             return object.__new__(cls)
-        data, dtype = cls._get_data_dtype(kwargs.get("data", []))
-        kwargs["data"] = data
-        return cls.__new__(cls[dtype],  *args, **kwargs)
+        data, dtype = cls._get_data_dtype(data)
+        return cls.__new__(
+            cls[dtype],  name, dom, cod, data, *args, **kwargs)
 
     @staticmethod
     def _get_data_dtype(data):

test/semantics/tensor.py

@@ -230,7 +230,7 @@ def test_Box():
 
 
 def test_Spider():
-    assert repr(Spider(1, 2, Dim(3))) == "tensor.Spider[float64](1, 2, Dim(3))"
+    assert repr(Spider(1, 2, Dim(3))) == "tensor.Spider(1, 2, Dim(3))"
     assert Spider(1, 2, Dim(2)).dagger() == Spider(2, 1, Dim(2))
     with raises(ValueError):
         Spider(1, 2, Dim(2, 3))
@@ -260,6 +260,13 @@ def test_Tensor_adjoint_eval():
     assert tensor1 == tensor2
 
 
+def test_Tensor_dtype_inference():
+    assert Box("F(A)", Dim(1), Dim(1), data=None).dtype is None
+    assert Box("X", Dim(1), Dim(1), data=[0]) == Box[np.int64]("X", Dim(1), Dim(1), data=[0])
+    assert Box("Y", Dim(1), Dim(1), data=[1.]) == Box[np.float64]("Y", Dim(1), Dim(1), data=[1.])
+    assert Box("Y", Dim(1), Dim(1), data=[1]) != Box("Y", Dim(1), Dim(1), data=[1.])
+
+
 def test_non_numpy_eval():
     with backend('pytorch'):
         with raises(Exception):