> On Sept. 22, 2015, 5:26 p.m., Vishesh Handa wrote:
> > Oh wow. This is bad. Thanks for diagnosing it.
> >
> > I'm wondering if this is the best way of solving it. The assumption always
> > is that if a DB does not have any values then it can just return 0. If we
> > change this assumption in PostingDB, then we should change it in every
> > other DB (specially PositionDB). Maybe it would just be easier to update
> > the logic in the SearchStore::constructQuery.
> >
> > I haven't tested this, but this is what I was imagining -
> >
> > diff --git a/src/lib/searchstore.cpp b/src/lib/searchstore.cpp
> > index 806ea5f..26b4430 100644
> > --- a/src/lib/searchstore.cpp
> > +++ b/src/lib/searchstore.cpp
> > @@ -157,6 +157,9 @@ PostingIterator*
> > SearchStore::constructQuery(Transaction* tr, const Term& term)
> > PostingIterator* piter = constructQuery(tr, t);
> > if (piter) {
> > vec << piter;
> > + } else {
> > + qDeleteAll(vec);
> > + return 0;
> > }
> > }
> >
> > What do you think?
>
> Vishesh Handa wrote:
> Also, there is no bug report about this, but someone did mention it on
> reddit - https://www.reddit.com/r/kde/comments/3kucp8/krunner_problems/ So,
> I'm super happy that you have figured it out.
>
> Igor Poboiko wrote:
> That might work (although it should happen only for AND operations, so
> need additional check there).
> But for me that behavior wasn't obvious. Also it seems like same issue
> appears in Transaction::postingIterator(const EngineQuery&) (which, btw,
> duplicates the functionality of SearchStore).
You're right. Maybe then we should just allow the AndPostingIterator to handle
null pointers? This way the code will not be duplicated and it will also be
easily testable.
- Vishesh
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://git.reviewboard.kde.org/r/125351/#review85773
-----------------------------------------------------------
On Sept. 22, 2015, 5:06 p.m., Igor Poboiko wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://git.reviewboard.kde.org/r/125351/
> -----------------------------------------------------------
>
> (Updated Sept. 22, 2015, 5:06 p.m.)
>
>
> Review request for Baloo.
>
>
> Repository: baloo
>
>
> Description
> -------
>
> This is a simple one-line fix for a bug (not sure there is one in bugzilla):
> # balooshow -x baloo-fix-empty.patch
> 7347941719148552 2056 1710826 /home/eol/baloo-fix-empty.patch
>
> Internal Info
> Terms: Mpatch Mtext Mx T8
> File Name Terms: Fbaloo Fempty Ffix Fpatch baloo empty fix patch
> XAttr Terms:
>
> lineCount: 13
>
> You can clearly see that type is 'T8' (which is 'text'). But:
> # baloosearch --type video patch
> /home/eol/baloo-fix-empty.patch
>
> I don't have any 'video' file in baloo index. Because of that corresponding
> PostingDB::iter() returns zero, and in SearchStore::constructQuery() this
> term just gets ignored.
> This leads to bug that same file is displayed in different categories (Audio,
> Video, Documents, Image) in KRunner search.
>
> Instead we should return an empty iterator; this leads to zero results in
> baloosearch (as it should be)
>
>
> Diffs
> -----
>
> src/engine/postingdb.cpp 8ca118c
>
> Diff: https://git.reviewboard.kde.org/r/125351/diff/
>
>
> Testing
> -------
>
> Works as expected:
> # baloosearch --type text patch
> /home/eol/baloo-fix-empty.patch
> Elapsed: 2.12026 msecs
> and
> # baloosearch --type video patch
> Elapsed: 1.71155 msecs
>
>
> Thanks,
>
> Igor Poboiko
>
>
>> Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<
