id generation for data/metadata in CSV files

[nkrishnaswami]

We may want to indicate how to generate IDs for the triples corresponding to the rows in CSV files.

This will facilitate having a well defined mapping from DSPL 2 datasets to triples, and may make it feasible to use dimension values and footnotes defined in CSV files across datasets.

Tentative proposal

Attempt to generate easy-to-keep-unique IDs, and make no provisions for ID collisions.

codeList

For each CSV row,

• Start with the containing dimension's ID.
• If there is no fragment, set the fragment to the dimension's name, URL encoded.
• Append an = and the URL-encoded codeValue to the fragment.

For example, if a row's codeValue is us and its containing Dimension has @id of #country, the row's triples should be generated as if from equivalent JSON-LD with "@id": "#country=us".

---

footnote

For each CSV row,

• Start with the containing StatisticalDataset's @id.
• If there is a fragment, append a /
• Append footnote= and the URL-encoded codeValue to the fragment

For example, if the dataset's @id is the empty string, a footnote with codeValue of p would yield an ID of #footnote=p. Similarly, if the dataset @id is #my_dataset, the footnote would have @id of #my_dataset/footnote=p.

---

observation

For each CSV row,

• Start with the slice's @id.
• If there is a fragment, append a / to it.
• Sort the dimension values by dimension name.
• For each dimension value, append the URL-encoded name, = and the URL-encoded codeValue to the fragment, separating the entries with /.
• Sort the measure values by measure name
• For each measure value, append the URL-encoded name to the fragment, separating entries with /.

For example, an observation in a slice with an @id of #europe_unemployment_slice with dimensions

• gender of m,
• country of uk, and
• month of 2010-10

and measures

• unemployment_rate and
• unemployment

would have an @id of #europe_unemployment_slice/country=uk/gender=m/month=2010-10/unemployment/unemployment_rate