PackageKit need to be more intelligent!

[Seth Vidal]

On Wed, 12 May 2010, David Malcolm wrote:

> I don't know if that's the issue, but this reminded me of this:
> http://www.secretgeek.net/ms_search.asp
>
> Implementing "search" is difficult.  In a good implementation, the user
> shouldn't need to understand the schema of the underlying database
> ("name" vs "description"): you should be able to type text and get good
> results, in a meaningful order.  You should be able to misspell things
> and have the program cope.  You should be able to type synonyms of words
> and get good suggestions.  It should cope with non-English queries.
>
> I don't know how to solve this well for Fedora; FWIW, I saw that the
> package database recently gained an Applications search feature.  I'd
> love it if software search results within Fedora could direct the user
> to a pending package review, or to a relevant SIG within Fedora,
> perhaps.
>

1. try running: yum search python rsync ssh | less

look at the results. - we sort them based on the number of items matched 
and where the items were found (pkgnames, summary, descriptions)

2. we implemented support for a tags db/keywords db lookup in yum. If a 
repo has a tags metadata defined in it - yum will download it and search 
it to match it to a pkg.

The format of this db is the one that the pkgdb can generate. We just need 
that added to the fedora repos when they are made.

Seriously folks- try doing a 'yum search foo' before saying we cannot 
provide good search results.


-sv