@spec build_slug(Keyword.t(), Ecto.Changeset.t() | nil) :: String.t() | nil

This function is used to build the slug itself.

This function is a place to modify the result slug. For convenience you can call super(sources, changeset) which will return the slug binary. super(sources) uses slugify, but you can completely change slug-engine to your own.

Note: this function will only be called if sources is not empty. Also important this function will be called only once for the normal workflow. And every time for :always_change. So you can do some heavy computations.

If for some reason slug should not be set - just return nil or empty binary.

It should return a binary or nil.

@spec get_sources(Ecto.Changeset.t(), Keyword.t()) :: [atom() | binary()]

This function is used to get sources for the slug.

There can be many use cases when this behavior is required, here are the brief examples:

1. Conditional slug sources
2. Add any data from different sources

This function should return list of atoms or binaries, or nil.

When processing the returned list:

1. atom-key is supposed to identify the model field
2. binary-key is treated as a data itself, it won't be changed

@spec maybe_generate_slug(Ecto.Changeset.t(), atom() | list() | none(), Keyword.t()) ::
  Ecto.Changeset.t()

This function is used to generate slug.

It is called 'maybe' since it may not generate slug for several reasons:

1. It was already created and :always_change option was not set
2. The source fields for the slug were empty

This function takes changeset as an input and returns changeset.

@spec unique_constraint(Ecto.Changeset.t(), atom(), Keyword.t()) :: Ecto.Changeset.t()

This is just a helper function to check for uniqueness.

It basically just wraps Ecto.Changeset and set a proper field name.