Data Cube Access and Creation

[ricroberts]

To make it easier to maintain the cube code (which is currently distributed across a few components of PublishMyData), I propose adding data cube functionality to grafter (somewhere)... maybe as part of this repo.

Some key functionality to include:

• Get datasets which use a certain dimension, or value(s) for a dimension.
• Get all observations in a data cube (or slice), optionally ordered.
• Get all dimensions in a dataset
• Get all values for a dimension in a cube (or slice)
• Get observations for
• Sorting within groups
• Building a CSVW JSON header for a cube or slice
• Pagination of observations

Unanswered questions:

• Do we need to distinguish between 'locked' dimensions (single value for a dim), and 'filters' (looser constraint).
• How to represent the various elements in code: i.e. (collections of) observations / slice / dataset / filters
• When do we serialise to the different formats? In this library, or the wrapping code?
• Do we want to support 'rolling up' multiple measures for the same set of locked dims into a row (to make ordering make more sense).