Error in read_bb() in cluster_bins.py due to change in pandas df.groupby()

[BramLimSJ]

When running the demo script (demo-complete.sh), I encountered the following error:

File "/software/team274/bl10/miniconda3/envs/hatchet/lib/python3.9/site-packages/hatchet/utils/cluster_bins.py", line 71, in main assert str(start_row[0]) == my_chr, (start_row[0], my_chr) AssertionError: ('chr22', "('chr22',)")

The error appears to be due to line 144 in cluster_bins.py:

for ch, df0 in bb.groupby(['#CHR']):

The newer version of pandas outputs ch = ('chr22',) instead of ch = 'chr22'. This occurs when a length-1 list is supplied to df.groupby(). Refer to this issue here.

The error can be overcome by changing line 144 to :

for ch, df0 in bb.groupby('#CHR'):

In my environment, pandas (v2.1.2) was installed as a dependency when I installed hatchet (v1.1.1) using:

conda install hatchet=1.1.1

[BramLimSJ]

In addition, there is an extra comma in line 190 after chr_labels, which could potentially return a tuple. However, this does not seem to be the problem here.

[ronkesm]

An issue I've come up against as well. Can also be fixed by installing pandas==1.5.0 - there are no conflicts with other dependencies on install.

[vineetbansal]

Thanks @ronkesm, @BramLimSJ - I'm inclined to pin the version of pandas in our pyproject.toml so that this error does not happen if you do a pip install of our package. Does this seem like an acceptable solution at this time, or are you using the latest features of pandas elsewhere in your code such that we need to fix this issue?

[BramLimSJ]

Pinning an older version of pandas could work. However, I personally feel the newer version of pandas is more stable as df.groupby() with a list argument always returns a tuple when iterated (including a length-1 tuple) and df.groupby() with a string argument always returns a string.