unity-design team mailing list archive

[Matthew Paul Thomas <mpt@xxxxxxxxxxxxx>]

Anup Verma wrote on 03/10/11 12:26:
>
> Friends, don't you find any problem with the algorithm used to search
> in the software center. Giving you an example:
>
> Let us search for MultiGet. When I write "Mult" in the search MultiGet
> appears at the 6th position. As soon as I add 'i', I see that
> surprisingly, MultiGet now appears at 10th position with some rather
> irrelevant options before it.
>
> Not understandable...
> ...

These are the items that appear for "multi" before MultiGet:
*   Qtractor, "a multi-track sequencer"
*   qutIM, "a multi-protocol IM client"
*   Teeworlds, "an online multi-player platform 2D shooter"
*   BasKet, "a multi-purpose note-taking application for KDE"
*   ROXTerm, "Multi-tabbed GTK/VTE terminal emulator"
*   Babiloo, "dictionary viewer with multi-languages support"
*   Jokosher, "Simply and easily create multi-track audio"
*   Empathy, "GNOME multi-protocol chat and call client"
*   ACM, "a multi-player aerial combat simulation".

Though USC does take into account the possibility that the last word you 
typed is only part of a word, it gives more weight to results where that 
last word is a complete word. And it can't tell the difference between a 
complete word and a hyphenated prefix like "multi-".

All of those higher results use "multi-" in their summary or synopsis. 
MultiGet uses "multi-" only in its description.

So that's why more items appear above MultiGet for "multi" than for 
"mult": because as soon as you type the "i", the packages that use 
"multi-anything" start being more important than those which use 
"multianything". And the packages that use "multi-anything" in their 
title or summary are treated as more important than those that use 
"multi" in their name *or* "multi-" in their description.

If there is a bug here, it is that BasKet, ROXTerm, and Empathy use 
"multi-" words only in their package synopsis, which USC doesn't show at 
all -- so it's not obvious why they're being weighted as they are. I've 
reported that now. <http://launchpad.net/bugs/865294>

Thanks
--
mpt