Issue running example from README.md with decode_batch

[cicero-a]

While the example with decode works as expected, using the example with decode_batch returns the following error:

Traceback (most recent call last):
  File "/home/accel-team/ciceroa/MWPM-pyMatching/test-pymatching2.py", line 12, in <module>
    predicted_observables = matching.decode_batch(syndrome)
  File "/home/accel-team/ciceroa/MWPM-pyMatching/.venv-pymatch/lib/python3.10/site-packages/pymatching/matching.py", line 419, in decode_batch
    predictions, weights = self._matching_graph.decode_batch(
ValueError: resize only works on single-segment arrays

Is this due to a not specified library dependency?

[oscarhiggott]

Thanks for flagging this. This issue has been caused by the recent release of numpy 2.0. You might consider downgrading to pip install numpy==1.26.4 while I get this fixed.

[jeromejavelle-seeqc]

Same issue experienced on my end on a simple decoding using a check matrix and an 8-bit trivial syndrome. Maybe a first fix could to explicitly restrict the numpy dependency to major version 1.x.x until support for numpy 2.x.x is complete, in the pymatching toml file?

[oscarhiggott]

I've pinned to numpy==1.* in #100, which is now released in pymatching v2.2.1 on pypi.

Ideally pymatching would support numpy v2, however that will require fixing a bug introduced when updating the the required version of pybind11. Since this problem is described in the open issue #102, I'll close this issue since the pin to numpy v1 fixes it in the meantime.