Difficult to override lexeme `ID`

[SimonGreenhill]

I'm creating a dataset at the moment from a relational database and I'd like the lexeme IDs to be the primary keys from the database. I assumed that if I specified ID=x in the add_forms* call, then the ID would be kept, however, the code always calls self.lexeme_id:

pylexibank/src/pylexibank/cldf.py

Lines 115 to 120 in 8ae170c

	def lexeme_id(self, kw):
	self._count[(kw['Language_ID'], kw['Parameter_ID'])] += 1
	return '{0}-{1}-{2}'.format(
	kw['Language_ID'],
	kw['Parameter_ID'],
	self._count[(kw['Language_ID'], kw['Parameter_ID'])])

Can we make lexeme_id be ignored if ID is given?

[LinguList]

I remember we decided to leave Local_ID for that purpose of the original ID, as done in the dataset shared (this can also be in conflict with lexibank ID format, if I understand properly and is not checked). Would local_id be enough for you?

[LinguList]

I just checked here: https://github.com/intercontinental-dictionary-series/keypano/blob/main/lexibank_keypano.py If ID is always created, we need indeed discuss that, as it means we would also have to revise datasets like the keypano, where we want to inherit IDS ids.

[SimonGreenhill]

Yes -- I'm using Local_ID but this feels like a workaround. If someone is explicitly specifying an ID then we should respect that.

At the very least silently replacing given ID's with a generated one is surprising and unexpected which could lead to glitches (I only spotted this when I was analysing the CLDF data sometime down the track and could not figure out why the list of lexemes I needed was wrong). It's worse because it's only the lexemes (add_form*) that do this and languages, parameters, etc are quite happy to have my manually curated IDs.

[xrotwang]

I think I agree that the call to lexeme_id should only happen in add_forms_from_value, because it's the potential splitting of values into several forms that made this functionality necessary in the first place. (Maybe add_form_* should have something like kw.setdefault('ID', self.lexeme_id(...)) for backwards compat.)