Refactor the config mechanism to use a single-class based tree model.
The class should allow
- serialization/deserialization of the config tree to/from files
- path based (dot and slash delimited) attribute lookup
- 'duck typing' (non-schema values)
- automagic python attribute lookup
- a 'reconfigure' event, which allows subscribers to 'see' what changed, and perform actions based on the changes (e.g. modify backlog limits on the fly).
- automatically fall back to attribute values in parent configs if a value is undefined at the current path level.
- pretty print function
Perhaps the root of the tree should be global, although it'd probably be best if the root was injected into anything needing access to config data. One root per user, or one root with user children?
Refactor the config mechanism to use a single-class based tree model.
The class should allow
Perhaps the root of the tree should be global, although it'd probably be best if the root was injected into anything needing access to config data. One root per user, or one root with user children?