Creating mappings for RDF/JSON based on values of multiple columns

Hi - 

Based on the [historical weather observations example][1] I am working on, I've come across a difficulty.

The CSV has two columns `yyyy` and `mm` representing year and month relating to a given monthly observation record; e.g. 

```
yyyy,mm,tmax,,tmin,,af,,rain,,sun,,
1978,9,17.5,,11.3,,0,,26.7,,---,,
1978,10,15.6,,10.7,,0,,20.4,,---,,
```

I want to map the values of `yyyy` and `mm` into a single target; in this case an object representing the target month (expressed using the scheme from reference.gov.uk), e.g.

```
_:b1 sdmx:refPeriod <http://reference.data.gov.uk/id/gregorian-interval/1978-09-01T00:00:00/P1M> .
```

I don't want to create triples for the `yyyy` and `mm` properties at all - only the aggregated property for `sdmx:refPeriod`.

(this overlaps a little with [Dan's email about RML capabilities][2] ... in my case I am trying to steer clear of external templating).

Any ideas?

Thanks, Jeremy



[1]: https://github.com/w3c/csvw/tree/gh-pages/experiments/historical-weather-observation-dataset 
[2]: http://lists.w3.org/Archives/Public/public-csv-wg/2014Oct/0015.html