[Nitro] Concerns over Og mandating the form of initialize
james_b at neurogami.com
Thu Apr 14 12:28:23 EDT 2005
Anastasios Koutoumanos wrote:
> i think you're missing the fact that tstamp is just a class variable,
> not a prop_* variable!
No, it's an instance variable. It's read-only, and not marked for
persistence, as the whole point is that it gets instantiated by process,
not client assignment.
> james is right in the fact that we should allow for internal class
> variables to be persisted some way.
I do not want certain instance values saved, as they may be derived data
or related to specific circumstances (such as the time a particular
instance is created).
> but you're right saying that initialised should _not_ be used when
> restoring an object from the backend.
I think there is confusion over the use of the word 'restore'. If the
idea is to revive a previously-existing object (similar in spirit to
YAML.load( yml ), then do not call initialize.
If the the idea is use data (which may or may not have come from a
previously existing object) to drive the creation of objects, then
initialize may be needed because that is where internal state may set up
and validity checks performed.
If I understand George, any such internal initialization and state
validation logic performed at object creation would have to be moved to
another method that gets called by Og during og_read.
More information about the Nitro-general