Q: How to bind positional arguments in HatchDict?

[npow]

I'm trying to use functools.partial in HatchDict, but it seems to always require one argument which is not named. Is there any way to work around this?

(Pdb) config.get("metrics")
*** TypeError: type 'partial' takes at least one argument

config:

metrics:
  - =: functools.partial
    func:
      =: sklearn.metrics.roc_auc_score
    multiclass: ovr

[Minyus]

HI @npow ,

You can use _ key to specify positional (not named) arguments if any.

from pipelinex import HatchDict
import yaml
from pprint import pprint  # pretty-print for clearer look

params_yaml = """
metrics:
  - =: functools.partial
    _:
      =: sklearn.metrics.roc_auc_score
    multiclass: ovr
"""
parameters = yaml.safe_load(params_yaml)

metrics_dict = parameters.get("metrics")

print("### Before ###")
pprint(metrics_dict)

metrics = HatchDict(parameters).get("metrics")

print("\n### After ###")
print(metrics)

### Before ###
[{'=': 'functools.partial',
  '_': {'=': 'sklearn.metrics.roc_auc_score'},
  'multiclass': 'ovr'}]

### After ###
[functools.partial(<function roc_auc_score at 0x16bcf19d0>, multiclass='ovr')]

[npow]

Awesome, that did it!