[MKDoc-dev] Handling Component Deletion

[Bruno Postle]

On Sun 24-Oct-2004 at 12:53 -0400, Sam Tregar wrote:

> Then when the user clicks save the objects that are present get
> saved to the DB.  This makes it hard to call methods on the
> missing objects since the data needed to instantiate them is gone.

My idea was to delete all entries in the Document_TimeRange table
matching the current document ID as the first step after someone
clicks 'save document' (I'm hand-waving here as I don't know that
bit of code so well).

> > We could actually run it every minute with a parameter of 90
> > seconds, but it then causes high server-loads when somebody does a
> > mass import of ten thousand documents and you end up with multiple
> > indexer instances running concurrently.
> 
> It would be possible to solve this using a lock-file to inhibit
> concurrent runs.

Sorry I realise I was describing a non-problem, subsequent indexer
instances will be looking at a different (or empty) list of
documents - The only reason we run it every hour instead of every
minute is because that's how we've always done it.

-- 
Bruno