GSoC idea: Create transitive relationship field

The "parent term" property on a taxonomy term is a unique, non-field-api, non-transferrable relation that has not been upgraded to Field API. This GSoC idea does not propose to replace this in core as the short term goal, because a transitive relationship field can exist side-by-side with the current "parent term" field and the "parent term" field can be drupal_alter'ed out of existence. However, if this project succeeds, it would be developed further to include in core.

A transitive relationship is what the parent term property describes. For example, Orlando is in Florida, Florida is in the United States, therefore by the mathematical laws of transitive relations we can know that Orlando is in the United States.

If you check in Taxonomy module, you'll see that there's a lot of "one-off" code for this relationship around this parent term property. There's UI code for generating widget values. There's code for generating SQL that lets us get a term's relationships to nodes, that term's children's relationships to nodes and that term's grandchildren's relationships to nodes. Views also duplicates or repeats much of this in its various handlers.

To express these relations in Field API presents some significant challenges. It requires taking a concrete thing (the parent column in the term_hierarchy table) and making it abstract and reusable. I see this work having three phases. Accomplishing the first is the baseline for success of the GSoC proposal.

1. Let any entities have transitive (hierarchical) relationships--not just terms--using Entity API or Field API to store data and produce widgets.
2. Add (dynamically when needed?) new indexes on the field to support higher performing SELECT queries at the cost of rarer INSERT queries. I'm thinking of the left-and-right "nested set" method, but there may be others.
3. Support non-SQL data storage.

It may be possible to contribute to an existing Drupal project, but I still need to do my research. Relation module is a broader initiative to improve relationship handling in Drupal. I contributed only a seed of what this project has become, and I really don't know much about it at the moment. Please suggest any other Drupal projects I should review before finalizing my proposal.

What does this project lead to? Here are a few ideas:

• Any entity can have a transitive relationship with another entity. One user might supervise another user so that users can have authority over other users. Because nodes have revisions, authorship and status, nodes might be more useful for creating content hierarchies.
• More consistent RDF expressions.
• The notion of controlled vocabularies as a defined set of terms that share a vocabulary ID might be redefined to mean "any entities that have transitive relationships with each other through a particular field."
• Terms whither away as a special class of entities because their character of having relationships is no longer unique.

My qualifications for this project: I'm responsible for taxonomy term reference field, the only relationship-like field in core. As a result of this, I'm a maintainer of taxonomy.module. I wrote a little (tiny) bit of Field API, but I'm competent with Field API in general. I presented some slides and spoke about my scattered thoughts about Drupal relationships and links at the Drupalcon San Francisco core developer summit. I also am working on oEmbed which is also all about links and relationships. I also love taxonomy, vocabularies. I am a big vocabulary nerd.