Re: Generalizing find-definition

[Dmitry Gutov]

Stefan Monnier <address@hidden> writes:

>> When it's t, the function would return an approximate string (basically
>> symbol at point, but maybe with minor syntax adjustments), and that
>> value would be used both in the error message (when we can't find
>> anything), and as initial input for identifier completion, from
>> xref-identifier-completion-table.
>
> I don't care about which values are really allowed as identifiers.
> I just the docstring to say clearly that non-strings can be used.
> And I'd like to get rid of the "identifier->string" method because
> I think it's over-engineering (it's only used to give
> very-slightly-prettier error messages).

The message you're responding to assumes familiarity with the patch,
which, I think, already addresses both your points (and identifiers can
only be strings, nil or t, like you suggested before).