I feel this was discussed in NumPy before (including the potential .T limitation!), time to revive that discussion!? I find it mildly annoying that we do these things in the new namespace, but seem avers to discussing it for NumPy proper. If we can get away with doing these things in NumPy it would be the best way to move code towards the better standard!

The new APIs for matrix transpose were chosen so that they can be as backwards compatible as possible with libraries like NumPy that already use .T for reversing the axes. See data-apis/array-api#228 and the corresponding PRs.

I feel this was discussed in NumPy before (including the potential .T limitation!), time to revive that discussion!?

Yes I think so.

I find it mildly annoying that we do these things in the new namespace, but seem avers to discussing it for NumPy proper.

I don't think we are, it's just that no one followed up/through on making matrix_transpose public, and the discussion was framed around a big backcompat break in .T, which did not help. I think no one would object to making the private matrix_transpose public (EDIT: and then add .mT), it just needs doing?

If we can get away with doing these things in NumPy it would be the best way to move code towards the better standard!

agreed

Is there a reason why invalid dtypes sometimes raise a ValueError while in others they raise a TypeError?

Likely I changed it at some point in my previous PR but haven't yet backported the change to this one (I originally started this as part of #19739 but moved it to a separate branch). Thanks for the heads up.

I checked this, there are indeed still 2 places where ValueError is used: in __float__ and in __bool__. Let's fix that.

Actually, @asmeurer you seem to have done this on purpose:

def test_python_scalar_construtors():
    b = asarray(False)
    i = asarray(0)
    f = asarray(0.0)

    assert bool(b) == False
    assert int(i) == 0
    assert float(f) == 0.0
    assert operator.index(i) == 0

    # bool/int/float should only be allowed on 0-D arrays.
    assert_raises(TypeError, lambda: bool(asarray([False])))
    assert_raises(TypeError, lambda: int(asarray([0])))
    assert_raises(TypeError, lambda: float(asarray([0.0])))
    assert_raises(TypeError, lambda: operator.index(asarray([0])))

    # bool/int/float should only be allowed on arrays of the corresponding
    # dtype
    assert_raises(ValueError, lambda: bool(i))
    assert_raises(ValueError, lambda: bool(f))

    assert_raises(ValueError, lambda: int(b))
    assert_raises(ValueError, lambda: int(f))

    assert_raises(ValueError, lambda: float(b))
    assert_raises(ValueError, lambda: float(i))

    assert_raises(TypeError, lambda: operator.index(b))
    assert_raises(TypeError, lambda: operator.index(f))

Can you comment?

EDIT: diff I tried to apply but left out in the end:

$ git diff
diff --git a/numpy/array_api/_array_object.py b/numpy/array_api/_array_object.py
index 8794c5ea57..74e659e81e 100644
--- a/numpy/array_api/_array_object.py
+++ b/numpy/array_api/_array_object.py
@@ -416,7 +416,7 @@ def __bool__(self: Array, /) -> bool:
         if self._array.ndim != 0:
             raise TypeError("bool is only allowed on arrays with 0 dimensions")
         if self.dtype not in _boolean_dtypes:
-            raise ValueError("bool is only allowed on boolean arrays")
+            raise TypeError("bool is only allowed on boolean arrays")
         res = self._array.__bool__()
         return res
 
@@ -454,7 +454,7 @@ def __float__(self: Array, /) -> float:
         if self._array.ndim != 0:
             raise TypeError("float is only allowed on arrays with 0 dimensions")
         if self.dtype not in _floating_dtypes:
-            raise ValueError("float is only allowed on floating-point arrays")
+            raise TypeError("float is only allowed on floating-point arrays")
         res = self._array.__float__()
         return res
 
@@ -515,7 +515,7 @@ def __int__(self: Array, /) -> int:
         if self._array.ndim != 0:
             raise TypeError("int is only allowed on arrays with 0 dimensions")
         if self.dtype not in _integer_dtypes:
-            raise ValueError("int is only allowed on integer arrays")
+            raise TypeError("int is only allowed on integer arrays")
         res = self._array.__int__()
         return res