An in-memory graph data store consisting of maps with links.

## Introduction

This is a very basic database that stores entities. An *entity* is a
Clojure map from *attributes* to values.

An *attribute* is a keyword, usually namespace-qualified.

An entity has exactly one *unique identity* attribute whose value
acts as a globally-unique identifier for that entity. An entity may
not have more than one unique identity attribute; if it does, the
behavior is undefined.

A *lookup ref* is a vector pair of a unique identity attribute and a
value, such as `[:user/id 1234]`. The combination of a unique
identity attribute and its value uniquely identifies a single entity
in the database.

## Usage

Create a new database with `new-db`.

The schema of the database is the set of all unique identity
attributes. `add-id-attr` extends the schema with a new unique
identity attribute.

`add` inserts entities in the database. Entities already in the
database are updated as by `clojure.core/merge`. (To replace an
entity without merging, `dissoc` it first.)

All entites added to a database are automatically *normalized*: all
nested entities are replaced with lookup refs to other entities in
the database.

Entity normalization recognizes the following forms:

  - any non-entity value, including collections
  - a single entity
  - a collection (list, vector, set) of entities
  - a map where keys are any type, vals are single entities

Collection values may not mix entities and non-entities. Collections
are searched only one layer deep: `add` will not recursively walk
arbitrary data structures to search for entities to normalize.

Databases can be manipulated with primitive map operations such as
`get` and `dissoc`. Keep in mind that there are no indexes: If you
`dissoc` an entity from the database you may leave behind broken
links to that entity.

To get nested maps back out, use `pull`, which follows a pull
pattern to recursively expand entity lookup refs.