Value error hyperparam

[karinnasvall]

Dear pyrho developers,

Thanks for creating such a great software.

I have some trouble though. Pyrho was installed as a module last week on our cluster, and everything worked perfectly until I came to the hyperparam module.

This is my code:

pyrho hyperparam --samplesize 36 --tablefile ${OUTPUT_FILE}/lookup_tables/${POP}_lookup.hdf --smcpp_file $INPUT_SMC
--mu 2.9e-9 --ploidy 2 --outfile ${OUTPUT_FILE}/hyperparams/${POP}_hyperparam.out
--logfile ${OUTPUT_FILE}/log/${POP}_hyperparam.log

I get this error:

Traceback (most recent call last):
File "/opt/conda/bin/pyrho", line 8, in
sys.exit(main())
File "/opt/conda/lib/python3.8/site-packages/pyrho/frontend.py", line 68, in main
func(args)
File "/opt/conda/lib/python3.8/site-packages/pyrho/hyperparameter_optimizer.py", line 250, in _main
reco_maps = _load_hapmap()
File "/opt/conda/lib/python3.8/site-packages/pyrho/hyperparameter_optimizer.py", line 128, in _load_hapmap
with importlib.resources.open_binary(
File "/opt/conda/lib/python3.8/importlib/resources.py", line 87, in open_binary
resource = _normalize_path(resource)
File "/opt/conda/lib/python3.8/importlib/resources.py", line 61, in _normalize_path
raise ValueError('{!r} must be only a file name'.format(path))
ValueError: 'data/pyrho_hapmap_maps.txt' must be only a file name

Could this be an error with the versions of importlib.resources and it only takes filenames, so data/pyrho_hapmap_maps.txt should be changed to pyrho_hapmap_maps.txt?
Or is it something I am doing wrong? I could not find any old issue on this.
I tried to use conda and install pyrho with python 3.6 and 3.7 but then I got even worse problems with multiple version incompatibilities.

I would really appreciate help with this.

Many thanks, Karin

[gq1]

Just helped with Karin with the installation a bit.

Drop Python version to 3.7, and force to install all the dependencies:

    mamba create -y -n pyrho \
        python=3.7 \
        numpy=1.19.* \
        scipy=1.5.* \
        pandas=1.1.* \
        msprime=1.0.* \
        numba=0.53.* \
        pytables=3.6.* \
        cyvcf2=0.11.* \
        htslib=1.9.* \
        pytest \
        -c conda-forge -c bioconda

pip install pyrho/ --no-build-isolation

The tests can pass:
python -m pytest pyrho/tests/tests.py

Update the way to load the map and it seems working.

def _load_hapmap():
    """
    Load HapMap recombination maps, faithful to the original pyrho logic,
    without using importlib.resources.
    """
    import os
    import pyrho
    import numpy as np

    reco_maps = []

    data_path = os.path.join(
        os.path.dirname(pyrho.__file__),
        "data",
        "pyrho_hapmap_maps.txt",
    )

    # Open in text mode (binary + decode optional)
    with open(data_path, "r") as mapfile:
        for line in mapfile:
            line = line.strip()
            if not line or line.startswith("#"):
                continue

            # Split each token "pos,rate" into two lists
            try:
                positions, rates = zip(*[p.split(",") for p in line.split()])
            except ValueError:
                # Skip malformed lines
                continue

            # Convert positions: handles scientific notation like "1e6"
            positions = list(map(int, map(float, positions)))

            # Convert rates
            rates = list(map(float, rates))

            # Optional safety check: all rates > 0 except last
            assert np.all(np.array(rates)[:-1] > 0)

            reco_maps.append((positions, rates))

    return reco_maps

Can you check all the changes are right and may put a fix? Thanks.

[jeffspence]

Hi @karinnasvall and @gq1 --- thank you for this. Yes, the latest commit moved to more modern ways of accessing package resources, but I hadn't realized it would break back compatibility.

One option would be to install from the previous commit: 61e7178

Otherwise, the latest commit (fe7d1cf) should be back compatible.