dose_unit_conversions missing in eicupreparator.py

[kanghyunyu]

Unlike other datasets, eICU seems to be missing the variable "dose_unit_conversions" in "eicupreparator.py".

Is this meant to be? It ends up raising "ValueError: not enough values to unpack (expected 2, got 0)" in "newmedicationprocessor.py", as the return value for "self.dose_unit_expressions, self.dose_unit_replacements = self._dose_unit_expressions()" is an empty list.

[USM-CHU-FGuyon]

eICU has hundreds of units for medication dosage, which is why I did not provide the dose_unit_conversion dictionnary nor support for drug dosage in eICU.

However it should not raise an Error, I'm fixing this as soon as I can.

As an instant fix, replacing

        exprs = []
        if self.dose_unit_conversion_dic is None:
            return exprs

by

        exprs = []
        if self.dose_unit_conversion_dic is None:
            return exprs, {}

Should be fine.

Please tell me if this fixes the issue, I can't run the codes at the moment.

[kanghyunyu]

I see. This works (although sink_parquet fails for medication and gets diverted to pl.writing. Hope that's normal).

[USM-CHU-FGuyon]

For medications it should not be too problematic to do pl.write_parquet because tables size remain reasonable.

I was a lot more focused on enabling sinking for larger tables to avoid memory error.

thanks for contributing, I'll push the change asap
Matthieu

[kanghyunyu]

Great, thanks for all the work!