Consider introducing a ORM

[lottspot]

Given that building some of our API objects requires recursive queries, it might be beneficial to hand this complexity off to a ORM. Or it might not be. We should consider the possibility, look into what options might be available to us, and weigh the benefits and drawbacks of our available options.

[lottspot]

Revisiting this codebase a couple months later confirms without a doubt in my mind the need for this change; the long chain of Promise pipelines are difficult to read and understand, and are too easy to make mistakes while implementing. The querying as it's written now is a major inhibition on the maintainability of this codebase.

• Sequelize certainly seems to be the popular choice, but I would very much like to explore other options as well.
• Bookshelf.js looks like another fully-featured option, and provides both Promise and callback based interfaces. Neat.
• CaminteJS is one more possible option.

The issue of primary concern I want to examine with each of these ORMs is how effectively we can build the ORM into our abstracted data access methods so that implementing new storage backends remains as simple as possible

[lottspot]

After browsing documentation, Bookshelf.js is the clear leader here. It appears to be the least intrusive ORM of the three, requiring almost no schema definition at all. Declaring relations is also simple and javascript-y (and can easily be done in a nested way). Bookshelf should be the first ORM we attempt to integrate.