Hi Pat,

I have the initial ground work for a patch.  No tests yet.

But I figured I would get your input on to whether you would accept
this type of patch or not as it was a little more invasive than just
modifying the datetime-delta.

The basic approach was to modify the sql generation to add a
sql_pre_query to the core models and then add a sql_query_killlist to
the delta models.  It does require the addition of a new table to
track the index times and only adds the killlist if this table option
is passed to the datetime-delta.

Thinking sphinx changes
http://github.com/adamcooper/thinking-sphinx/commit/8e07d16e7e6194205f161c025bbc7c9c7a94eb16

Ts-datetime-delta changes
http://github.com/adamcooper/ts-datetime-delta/commit/73f6061503f9bfa8f5ba616292805031bc3b9279

We are looking to be using this in production and I would like to
ensure it gets merged back to the mainline so we don't have to
maintain a separate fork.

Thanks,
Adam


On Mar 27, 10:12 pm, Pat Allan <[email protected]> wrote:
> Hi Adam
>
> The kill list conf option is in TS (was added a couple of months ago, I 
> think). However, if you wanted a default option for deltas, then it really 
> only applies to the datetime deltas (in other deltas, you don't get the 
> doubling up, and there's no easy way to figure out which documents to delete 
> either).
>
> So: a patch is definitely welcome, and you'll want to add it to 
> ts-datetime-delta, not the thinking-sphinx gem.
>
> Great detective work by both you and Steven.
>
> Cheers
>
> --
> Pat
>
> On 27/03/2010, at 12:15 PM, adamcooper wrote:
>
>
>
> > We've been looking at this issue fairly in depth and it seems to be
> > due to a datetime-delta implementation issue (not sure if it's only
> > specific to ours...)
>
> > The real crux of the issue is because the main and delta index overlap
> > and therefore double the facet counts.
>
> > After a full index, the delta and main index have an overlap of
> > records that have been updated in last hour (or threshold time).  So
> > any facet counting on those records get doubled.  Normal searches are
> > not affected.  Counts are further messed up as records that were
> > originally only in the main index are updated and appear in the delta
> > index and therefore become double counted.  I can explain this issue
> > further if it's not totally clear.  Index merging further causes
> > issues.
>
> > We have manually tweaked the conf file to get the correct behaviour
> > using sql_query_killlist and sql_query_post and an intermediate
> > table.  Basically, we track the last full index time and add any
> > records that have been updated since then to the delta killlist.
> > We'll probably add the deleted records to there too (based on an
> > acts_as_paranoid type deletion).
>
> > Does anyone know if there is an easier way to accomplish this?  Or is
> > this what others are doing too?
>
> > Pat, we are looking to patch Thinking Sphinx to add killlist support,
> > have you done any initial thinking as to how you would want it added
> > or how it best could be added?
>
> > Thanks,
> > Adam
>
> > --
> > You received this message because you are subscribed to the Google Groups 
> > "Thinking Sphinx" group.
> > To post to this group, send email to [email protected].
> > To unsubscribe from this group, send email to 
> > [email protected].
> > For more options, visit this group 
> > athttp://groups.google.com/group/thinking-sphinx?hl=en.

-- 
You received this message because you are subscribed to the Google Groups 
"Thinking Sphinx" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/thinking-sphinx?hl=en.

Reply via email to