Explore how to adapt the csv schema so it is more robust to term additions

[ColmMassey]

Oxford and Newbridge now have a different schema from Liverpool.

But the surveys are only different in that some of the multi answer questions have more options in Liverpool.
We would want the schema to proscribe that the question is present, but not that all possible answers are present.
The problem stems from how the Limesurvey describes the answers, by creating a column for every possible answer and having a binary answer for each.

[ColmMassey]

See #54 (comment)

(I checked, and these new terms are not options in their secondary activities
multiple-choice). So the list of required fields for Mersey
includes fields Oxford and Newbridge don't. And this breaks their
conversion, which is designed to catch mismatches like this and
throw an error. )

I guess this is because each answer option in those questions requires
a new column, and then the schema defines these new columns as required?
Yes, each new column has to be added to the open-data incoming schema
(schema.yml) for it to be noticed. Anything not in there is ignored by
design. This is because all the New/Ox/Mer schemas are slightly
different, but include a common core (or used to). At one point any
differences wouldn't be tolerated, but I loosened it to allow
superfluous fields.