<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Aaron J. Seigo a écrit :
<blockquote cite="mid:200809121834.26450.aseigo@kde.org" type="cite">
<pre wrap="">On Friday 12 September 2008, Adrien BUSTANY wrote:
</pre>
<blockquote type="cite">
<pre wrap="">Sebastian Kügler a écrit :
</pre>
<blockquote type="cite">
<pre wrap="">On Friday 12 September 2008 16:57:51 Alexander Dymo wrote:
</pre>
<blockquote type="cite">
<blockquote type="cite">
<pre wrap="">i actully added the search widget yesterday, at least on the file
dialog side. now i need to hook it up to .. nepomuk?
</pre>
</blockquote>
<pre wrap="">Actually I think filtering files by name in the directory would be ok as
an intitial implementation.
</pre>
</blockquote>
<pre wrap="">filename and tag would be good. From there we can see how this 'feels'
and improve ...
</pre>
</blockquote>
<pre wrap="">That's possible, as SPARQL (the query language used for Nepomuk) allows
that quite easily. I also started to work on the concept of "Project" in
KDE-Nepomuk, although for now it's mostly similar to tags, so not of a
great use. So let's restrict to filename and tags for now.
</pre>
</blockquote>
<pre wrap=""><!---->
if we're going to do filename and tags, i may as well just use a regular
indexer directly via e.g. xesam.
the reason Alexander suggested name was in case connecting to nepomuk was
harder; obviously that was a premature suggestion. =)
there's no point, at least imho, in hiding the power of nepomuk by only
searching tags and filenames. i should be able to search by any meta data i so
desire in the file dialog that nepomuk understands.
</pre>
</blockquote>
<br>
The two points are not exclusive :<br>
When you ask Nepomuk to return all the files related to keyword foo,
what does the runner doo ? Only filter on filename, or use tags too,
etc...<br>
Searching in any metadata is of course possible, but obviously we need
some "weighting" of the properties here (a simple answer could be :
weight = # of properties matching keyword, most "heavy" answers first).<br>
<br>
</body>
</html>