Re: [dtp-dev] Fine-grained refresh control on ICatalogObject

[rcernich@xxxxxxxxxx]

Thanks Larry,

I have a couple of follow ups.

> Here is the explanation for the object re-use I got from our catalog
> loading expert:
>
> "It is used to keep the same reference when the target (like PK
> table) is re-loaded.   This is needed when the table references are
> cross-schema or the target is filtered out at that time of the refresh."

What happens if the PK column or table is removed?  It seems like the
references will still exist.  I could be wrong, but it looks like the code
doesn't do a very good job cleaning up these dangling references (this
includes the refactored, reusable catalog loading code).

Also, since refresh() is invoked on the child references, won't this cause
the columns and PK's to be reloaded anyway?  I guess that's the thing I
find most confusing.  Although, I guess if those loaders are doing the same
thing, the instances will be reused if they remain in existence, even if
their underlying meta-data changes (e.g. the PK's column).  If that's the
case, is there also a need to update the elements referencing the PK?
(Note, I'm just using PK as an example, I assume this generally
applicable.)

I'd like to understand this better so I can improve the reusable catalog
loaders.

Thanks again for the explanation.
Rob