Project manifest migration strategy update

[pileks]

We should probably rethink our manifest migration strategy. While we can have all of these options set as optional manifest properties, I'd say that we'll likely want to create a proper migration strategy sooner than later, as it will reduce future work.

Current migration strategy:

Auto-generated code requires us to implement certain, fixed migrators.

available manifests:
0.1.0
0.2.0
0.3.0
0.4.0

migrators/migration paths:
0.1.0 -> 0.4.0
0.2.0 -> 0.4.0
0.3.0 -> 0.4.0

After update to 0.5.0:
We need to update/create n+1 migrators if we want full migration support.

available manifests:
0.1.0
0.2.0
0.3.0
0.4.0
0.5.0

migrators/migration paths:
0.1.0 -> 0.5.0 #added/modified
0.2.0 -> 0.5.0 #added/modified
0.3.0 -> 0.5.0 #added/modified
0.4.0 -> 0.5.0 #added

Proposed migration strategy:

Allow developers to declare their own migrators, from whichever version to whichever other version.
A simple search algorithm would then find the appropriate migration path, and could even allow for downgrades if the correct migrator exists.

available manifests:
0.1.0
0.2.0
0.3.0
0.4.0

migrators:
0.1.0 -> 0.2.0
0.2.0 -> 0.3.0
0.3.0 -> 0.4.0
0.1.0 -> 0.3.0

migration paths to 0.4.0:
0.1.0 -> 0.2.0 -> 0.3.0 -> 0.4.0
0.1.0 -> 0.3.0 -> 0.4.0

After update to 0.5.0:

available manifests:
0.1.0
0.2.0
0.3.0
0.4.0
0.5.0

migrators:
0.1.0 -> 0.2.0
0.2.0 -> 0.3.0
0.3.0 -> 0.4.0
0.1.0 -> 0.3.0
0.4.0 -> 0.5.0 #added

migration paths to 0.4.0:
0.1.0 -> 0.2.0 -> 0.3.0 -> 0.4.0 -> 0.5.0
0.1.0 -> 0.3.0 -> 0.4.0 -> 0.5.0

[pileks]

Niraj: Look into Django migrations.

[krisbitney]

I think this is a good idea, but I think we should require the simplest complete path:

migrators:
0.1.0 -> 0.2.0
0.2.0 -> 0.3.0
0.3.0 -> 0.4.0

This ensures any migration can always happen, even when nobody adds a shorter path. Even though this is the longest path, the migrators are inexpensive and it won't be a practical issue.

I think you already had this in mind, but I thought I'd make it explicit just in case.