`BaseFloat._check_scalar` does not properly check string values

[sehoffmann]

Zarr version

commit edd47db

Numcodecs version

Python Version

3.13.3

Operating System

Linux

Installation

uv dev

Description

uv run python
Python 3.13.3 (main, Apr  9 2025, 04:03:52) [Clang 20.1.0 ] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from zarr.core.dtype.npy.float import Float32
>>> Float32().cast_scalar('NaN')
np.float32(nan)
>>> Float32().cast_scalar('not valid')
Traceback (most recent call last):
  File "<python-input-2>", line 1, in <module>
    Float32().cast_scalar('not valid')
    ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
  File "/zarr-python/src/zarr/core/dtype/npy/float.py", line 237, in cast_scalar
    return self._cast_scalar_unchecked(data)
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^
  File "/zarr-python/src/zarr/core/dtype/npy/float.py", line 220, in _cast_scalar_unchecked
    return self.to_native_dtype().type(data)  # type: ignore[return-value]
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^
ValueError: could not convert string to float: 'not valid'

Notice how the above does not produce a TypeError but a ValueError. I don't think this is intended. This actually lets my tests for a "Subarray" dtype fail (via the invalid_scalar_params field), because I pass the check through to the sub-dtype, i.e. return self.subdtype._check_scalar(data).

Steps to reproduce

See above

Additional output

No response

[d-v-b]

thanks for this report, I agree that we should be raising a TypeError here.