Suggested plan for nested set prototype

ns-plan.txt

* Create working NS implementation
  - The basic algorithm just requires 'left' and 'right' nodes, but it is a good idea to add in a
    'parent' node, as this simplifies some of the SQL
  - There are plenty of implementations on the web already, check GitHub and do a search
* Add the per-node columns as discussed on Stack Overflow
  - 'table_name' for table, 'primary_key' for ID
* Add some data tables
  - 'customer', 'address', 'book', etc.
* Write some code to handle a hierarchical string expression
  - Convert an expression like `customer.address.postcode` to three levels of hierarchy plus a value
    (or this could be two levels and the last one is just detected as a column)
  - the postcode would go in the address table
* Add some test data
  - customer.name
  - customer.address.line1
  - customer.address.postcode
* Write some code to read a hierarchical string expression
  - From 'customer.address.postcode' you'd expect a string
  - From 'customer.address' that's a row (i.e. an array/object)
  - From 'customer' that's a nested array/object
  - Your nested set implementation will help here (getChildren/getDescendants etc)
* Work on how to share nodes
  - Create two customers and give them different favouriteBooks with the same author
  - Create two customers and get them to share the same favourite book
* Decide how your 'persistence' flag is going to work
  - Is this per node?
  - Do you walk down a hierarchy path and, when encountering this flag = true, share the remainder of the
    descendent path?
  - This bit is probably dependent on your requirements, but I'll try to advise if I can
* Shared node amendment
  - If you delete any element from a shared node (except a column) you'll need to delete those references
    from the tree for all nodes that refer to this
  - If you add an element into a shared node, you'll need to add those references to to each tree that refers
    to it
* Load testing
  - Add in test data of the sizes you expect (1000 rows? 1M rows?)
  - What are the response times? Are they acceptable? Will they work on your hardware?

The trick with prototypes is:

- Do them quickly
- Reuse as much code as you can
- Don't reinvent the wheel
- Be willing to throw them away if they don't work