Re: [Gramps-users] Database Verification Warnings

[Michael L. <mic...@ca...>]

On Tuesday 28 February 2006 08:03, Alex Roitman wrote:
> On Mon, 2006-02-27 at 23:04 +0200, Eero Tamminen wrote:
> > Hm.  Why not make this automated?  Flagging the entries sounds
> > complicated and not necessarily what user wanted, ignoring just the
> > output lines sounds more robust. :-)
> >
> > Add option for "Ignore all current warnings", which would just save all
> > the output lines, and on next rounds those lines would be removed from
> > the output:
> > 	if warning not in ignored_warnings:
> > 		output(warning)
> >
> > User would use this option after he had fixed all the real warnings.
>
> This is what I was thinking. I just said that one can do it with grep
> and other tools manually, but if this seems reasonable then we may
> automate it.
>
The only potential problem with this is that as I work thru the report fixing 
what are usually typos, they will still be saved in the ignore file.  This 
will potentially lead to a very large file of warnings that no longer exist.

> > It would then require also option for "Clear ignored warnings" which
> > would just delete the saved warnings.
> >
This would lead us back to square one, not the most desirable.  The "best" 
solution would be to generate a new warnings file each time which did not 
include the fixed ones from the previous run.  A Clear warnings option would 
be useful, or we could just delete the warnings file.

> > The downside of this is that if user changes the criterias for warnings,
> > the ignore list of warnings might not be valid anymore.  Either this
> > issue could be ignored, or the ignore list might be hashed by all the
> > check
> > parameters.
>
> I think a healthy conservatism will do good here: we should
> only "ignore" things if they've already been approved as harmless.
> If you changed verification criteria and get new warnings, you'd
> have to declare them harmless at least once.
>
Yes, if you change anything that is used to generate the warnings, then you 
should have to return to square one.  There is no simple way out of that.

I think this is beginning to look like a solid proposal.

I would like to raise one other thing about the warnings, but perhaps in 
another thread.

-- 
==================================
Michael Lightfoot

mic...@ca...
http://www.tip.net.au/~michal
==================================