Writing `MuData` objects to .h5mu after using `mudata.concat()`

[IsaacUtah1379]

Describe the bug After combining one or more MuData objects with the mudata.concat(), the resulting MuData object throws the following error when MuData.write() is called:

The .h5mu file is still written despite the error. If the .h5mu file is read with mudata.read(), the resulting object appears to be the same as the original object, but I have not verified whether any data is changed or lost.

To Reproduce Example using simulated data:

import mudata
import anndata
import numpy as np

mudata.set_options(pull_on_update=False)

np.random.seed(1)

n, d, k = 1000, 100, 10

z = np.random.normal(loc=np.arange(k), scale=np.arange(k)*2, size=(n,k))
w = np.random.normal(size=(d,k))
y = np.dot(z, w.T)

adata = anndata.AnnData(y)
adata.obs_names = [f"obs_{i+1}" for i in range(n)]
adata.var_names = [f"var_{j+1}" for j in range(d)]

d2 = 50
w2 = np.random.normal(size=(d2,k))
y2 = np.dot(z, w2.T)

adata2 = anndata.AnnData(y2)
adata2.obs_names = [f"obs_{i+1}" for i in range(n)]
adata2.var_names = [f"var2_{j+1}" for j in range(d2)]

mdata = mudata.MuData({"A": adata, "B": adata2})

# This works without an error:
mdata.write('test_1.h5mu')

# Copy the two halves of mdata to separate variables, then recombine them
half_1 = mdata[:500,:].copy()
half_2 = mdata[500:,:].copy()
mdata2 = mudata.concat([half_1, half_2])

# This throws an error, but still writes the file:
mdata2.write('test_2.h5mu')

Expected behaviour The .h5mu file is written without any errors.

System

• OS: Windows 11
• Python version: 3.12.4
• Versions of libraries involved:
  • MuData 0.3.0
  • AnnData 0.10.8
  • h5py 3.11.0
  • hdf5 1.14.3

[gtca]

Thanks for spotting that, @IsaacUtah1379! Seems like np.str_ is not appreciated by h5py.