[Dbix-class] dirty-flag on inflated columns

[Marc Mims]

* Matt S Trout <dbix-class at trout.me.uk> [090513 09:29]:
> On Tue, May 12, 2009 at 01:51:00PM -0700, Marc Mims wrote:
> > * Matt S Trout <dbix-class at trout.me.uk> [090512 11:34]:
> > > On Mon, May 11, 2009 at 12:14:07PM -0700, Marc Mims wrote:
> > > > * Matt S Trout <dbix-class at trout.me.uk> [090511 10:29]:
> > > > > Thoughts on whether make_column_dirty should clear the deflated value if an
> > > > > inflated one is present, guys?
> > > > 
> > > > That would break DBIx::Class::InflateColumn::FS, causing it to leave an
> > > > orphaned file in the file system.
> > > 
> > > Why?
> > 
> > FS uses the deflated value to determine whether it needs to overwrite an
> > existing file or create a new file.  If the deflated value is cleared by
> > make_column_dirty, FS will always make a copy, potentially orphaning a
> > file in fs_column_path.
> 
> That's a bug. Deflation should not be dependent on the deflated value even
> being there - store_inflated_column for e.g. deletes the deflated value.

Ok.  So, now how to fix it?

I need a unique filename to deflate to, but I only want to create it
once.  Rather than using the deflated value {_column_data}{$column}, I
can create {_fs_column_filename}{$column} and use it, instead.  I need
to populate {_fs_column_filename} when a row is read from the db.  Looks
like I can do that by extending inflate_result.

Does that sound like the correct way for FS to deal with it?

> So your code was already broken, you just hadn't noticed yet - doc patches
> to ::InflateColumn on writing safe inflators very welcome :)

Obviously, I don't know how to write a safe inflator, yet. :)

	-Marc