|
From: Martin K. <ci...@gm...> - 2004-03-15 18:40:03
|
Am Mon, den 15.03.2004 schrieb Matthias Schmidt um 13:04:
> I've some other ideas what to change in the code, if you think these ideas are
> of use, i can do the coding:
>
> a) the shortcut ALT-M for marking all messages read simply don't work on my
> boxes (Linux|Solaris|Mac OS X). Can I assign a new schortcut (F-Something?)
> AND put the Context-Menu-Entry for "Mark all read" much higher, since this is
> what you do on a day-to-day base rather then reconfiguring your
> subscriptions.
On FreeBSD, I don't know how to use the shortcut either. It would
be a good idea to change it, I think.
> b) We can Highlight ALL Subscribtions by selecting root ( ... or use a button/
> Shortcut) and mark them all read.
I want to see "root" showing all articles and having functions
referring to all articles. This is one of the points why I'm still
waiting with a release.
RssChannelList.java (line: 225):
/*
TODO (by Cinek):
I would like to see certain reactions when selecting GroupNodes.
It would be nice to have a view of all Channels (recent articles)
which are in the currently selected group.
E.a. the group node should show all recent articles. I think many
people depend on this feature.
*/
> c) I played around with the "Remember Items" Option. After having several
> hundred (100-300) articles in the DB, things are getting slow. First i
> suspected the DB to be guilty but it isn't. I've used some profilers on the
> code, just for my own curiosity. I've attached a screenshot which shows that
> the vast majority of the CPU is consumed by
> javax.swing.text.HTMLDocument.insertHTML() operations. To get around this,
> one can implement some caching strategy to buffer the generated HTML and only
> append when something new happens AND ( ... and this is not i minor change,
> but it's worth to do it in the long run ) don't do the work in the
> AWT-Thread. This would make the GUI much more touchy, esp. when someone
> change his/her mind which channel to view, while still doing the rendering.
Yupp. The Java HTML parser has problems with efficiency. That's why I
chose the "more difficult" StyledDocument way.
I have looked at the sources, too, recently and I have discovered other
issues. I've made a note about it in ArticleDatabase (line 75):
/*
TODO (by Cinek):
Should be optimized. It's not a good idea to fetch all
articles while trying to find subscriptions. We have
a tree-based model which can fetch articles while
expanding tree-nodes.
Second things is that the 2nd SELECT should be limited
by offset + length.
What I'm also missing here, is to get all recent
articles at once. This was extremely convenient.
I don't like "SELECT *", we should write all column
names after select which are selected. It WILL make
problems when altering the table structure.
*/
Ummm.. by "fetching", I mean fetching from the database (SELECT).
> d) We can do some experiments with changing the Swing Look&Feel to the
> discovered platform. It's a minor change, and the results are worth it.
Unfortunatelly, I don't know which Look'n'Feel is appropriate for which
platform.
Message me, if you want to access the CVS directly. I will add you to
the developers. (Don't forget: I need to know your SF-nick.)
Martin
|