switch db

[Doozers]

relate on #649

It isn't a better idea to abstract the db in a first place instead of replacing it ?

advantage:

• It will decrease the complexity of use.
• smaller access will act like documentation
• It will keep the optimized and fast advantage of the cayleygraphDB
• If after abstracting, it still to hard to use, it will be easy to implement the db with a gorm/sqlite for example

// database.go
type Database interface {
    GetTasks(...)
    PushTasks(...)
    ...
}

// cayley.go
type cayleyDB {
}

func (c *cayleyDB) GetTasks(...) {
[...]
}

and if needed

// gorm.go
type gormDB {
}

func (c *gormDB) GetTasks(...) {
[...]
}

wdy think about it @moul ?

┆Issue is synchronized with this Trello card

[moul]

If the target is to have multiple databases, the this separation can make the hybrid approach very difficult to implement and maintain, and make transaction (and rollback) harder to manage too.

What about keeping your idea but without the need of an interface and multiple implementations; one implementation that can have one or multiple underlying databases; and clear function names and more isolated logic as you suggested?

Cayley is great but it's hard; I've strong believes that the best target is the hybrid approach; probably sql as the source; and Cayley as a smart cache synchronized when we need its extra features.