Re: [Tracker] search match on substrings

[Martyn Russell <martyn lanedo com>]

On 03/11/13 15:02, Brian J. Murrell wrote:
> So, I have a PDF indexed which has the following nie:plainTextContent
> from the extractor in it:
>
> $ /usr/libexec/tracker-extract -f tmp/Fence\ Bylaw.pdf
> ...
>   nie:plainTextContent "...\nBY-LAW NO. 2003-405\n..."
>
> Should any subtring return a match hit?  i.e.:
>
> $ tracker-search -l 1000 BY-LAW --disable-color
> Results:

BY and LAW would be separate words, "-" is considered a word gap/space.

> Although even matching the exact string doesn't hit:
>
> $ tracker-search -l 1000 "BY-LAW NO. 2003-405" --disable-color
> Results:

For numbers, you need to make sure you're indexing numbers. Did you 
check that after following Ivan's email?

A reindex is needed after that.

> I am able to match at least one hit in that file, so it must be indexed:

It may well be that tracker-search is just not powerful enough for what 
you want.

The backend FTS functionality offers more than we do with 
tracker-search. To see what you can do, take a look at:

  https://wiki.gnome.org/Tracker/Documentation/Examples/SPARQL/FTS

Also, to see what tracker-search is *doing*, you can use:

  $ export TRACKER_VERBOSITY=3
  $ tracker-search ...

That way you will see what SPARQL it uses.

> P.S. Sorry to be such a pain on all of this, but I'm trying to evaluate
> tracker for suitibuility here and being accurate and complete is a
> requirement.

No problem at all. You're not being a pain, some of these questions are 
useful for others too I would imagine.

You also have to realise, the command line tools are making very simple 
use of the Tracker APIs in many cases and there are always corner cases 
we won't catch. We try to improve the tools for those where possible of 
course :)

--
Regards,
Martyn

Founder & Director @ Lanedo GmbH.
http://www.linkedin.com/in/martynrussell