Allow copy for scalar and nested sequences when converting data to numpy arrays

[loichuder]

Fix #94

Also brings support for Numpy v2. See the discussion at #94 for more info:

As a rule of thumb, using np.asarray should be fine. Also, we have to remove the copy=False arg since it is visibly impossible to avoid copying when the argument is a scalar or a nested sequence (lists): https://numpy.org/devdocs/reference/generated/numpy.asarray.html#numpy.asarray

By using copy=None (the default behaviour), copy is avoided when possible (ex: when the argument is a numpy array) but allowed when there is no other choice (as I said, scalars and nested sequences)

[loichuder]

@t20100 ☝️

[t20100]

🤷 What does it currently returns? It doesn't sounds useful anyway

[loichuder]

The server encounters an error:

• csv:

  File ".../h5grove/h5grove/fastapi_utils.py", line 113, in get_data
    h5grove_response = encode(data, format)
  File ".../h5grove/h5grove/encoders.py", line 131, in encode
    csv_encode(content_array),
  File ".../h5grove/h5grove/encoders.py", line 58, in csv_encode
    np.savetxt(buffer, data, delimiter=",")
  File ".../h5grove/lib/python3.10/site-packages/numpy/lib/_npyio_impl.py", line 1573, in savetxt
    raise ValueError(
ValueError: Expected 1D or 2D array, got 0D array instead

• tiff:

  File ".../h5grove/h5grove/encoders.py", line 149, in encode
    tiff_encode(content_array),
  File ".../h5grove/h5grove/encoders.py", line 78, in tiff_encode
    tifffile.imwrite(buffer, data, photometric="minisblack")
  File ".../h5grove/lib/python3.10/site-packages/tifffile/tifffile.py", line 966, in imwrite
    result = tif.write(data, shape, dtype, **kwargs)
  File ".../h5grove/lib/python3.10/site-packages/tifffile/tifffile.py", line 2020, in write
    raise RuntimeError(
RuntimeError: length of storedshape 4 not in (5, 6)

In both cases, the server returns an HTTP Error 500 to the user. I wondered if we could not make it more explicit.

[t20100]

I wondered if we could not make it more explicit.

+1

What would be the best error code? 415 Unsupported Media Type ?

[axelboc]

Just 400 - Bad Request, I think. Those format values just aren't supported by the /data/ endpoint for scalar datasets. Clients are expected to know about this (ideally from h5grove's documentation) and not try to make the request.

[loichuder]

We already stuff put in place to raise 422 Unprocessable Content for some encodings if the data is not numeric:

h5grove/h5grove/encoders.py

Line 124 in 53d7d0f

if not is_numeric_data(content_array):

and

h5grove/h5grove/content.py

Line 297 in 53d7d0f

except QueryArgumentError as e:

We could reuse it to raise a 422 error in case of unsupported encoding for scalars.

[loichuder]

6028d89

[t20100]

Sounds good!