[rekonq] Review Request: Activities support for Rekonq

Sat Oct 20 06:12:06 UTC 2012

> On Oct. 17, 2012, 12:07 p.m., David Narváez wrote:
> > I'm not sure why am I under the impression that adjam does not use activities. Anyhow, I'll pitch in with a couple observations:
> > 
> > First of all, it is cool to see activities integration in rekonq (I'm sure at least mathieson has been asking for this) and I hope we can get more of this.
> > 
> > On the other hand, the target audience of this review is probably not related to the SLC concept so besides mentioning that you can use the SLC applet, it would be useful to give an overview of what does that do and explain how does that make Rekonq a better browser etc.
> > 
> > Another comment, and this applies to all reviews in the platform, not this one in particular, is the Testing Done. People seem to think it is a yes/no question or answer "Works" or "Compiles". According to ReviewBoard documentation "The Testing Done field describes how this change has been tested [...] This should cover any and all testing scenarios that have been done, in order to help reviewers feel more confident about the stability and design of the change."
> > 
> > Now, my review:
> > 
> > I patched my Rekonq master (btw, you probably want to specify that in the branch field) and was unable to configure Rekonq with my current KActivities version (6.0.0) so I updated KActivities and Nepomuk Core to latest master and was able to compile everything. Is that intended? I see the _OPTIONAL cmake macro used there, but I'm not sure why did it fail - I could test that again if you need me to. After installing the patched Rekonq, I can only report it doesn't break, but I'm not sure how to test if it is doing anything with the resources, is there a sparql query I can throw into, e.g., NepSaK to see this working in the backend?
> > 
> > Thanks.
> 
> Ivan Čukić wrote:
>     Kinda expected that people follow planetkde - seemed like other applications' reviewers (okular, gwenview, kate ...) knew about SLC.
>     
>     Will update the description above.
>     
>     > I could test that again if you need me to
>     
>     Can you give me the output?
> 
> David Narváez wrote:
>     Minutes after writing the review I went over my steps and noticed the configuration error I received was because of a mistake in my initial attempt to patch the sources and was not related to the KActivities version, so nvm that. I'm not at home right now to test the sqlite3 thing described above but I'd bet it is working :)
> 
> Andrea Diamantini wrote:
>     Yes, I just read about this "Share-Like-Connect" thing. And yes, it's true I'm not using activities. I just don't feel the need to switch wallpaper or plasmoids. But that's just me, I know.
>     In general, I'm just a bit worried about all the things you are tracing about users "activities". And where you are storing it (are we using another database somewhere?). 
>     Anyway, talking about the patch. First, it needs to be changed to NOT notify anything if rekonq is in "private browsing" mode. Second, I'd like to understand how the user has been made aware another software but rekonq is storing his browsing data.
> 
> Ivan Čukić wrote:
>     >  And yes, it's true I'm not using activities. I just don't feel the need to switch wallpaper or plasmoids.
>     Not really the point of activities, but not important :)
>     
>     > it needs to be changed to NOT notify anything if rekonq
>     Good point - completely slipped my mind
>     
>     > I'd like to understand how the user has been made aware another software
>     
>     It is a KDE-wide thing. There's a systemsettings module to make the user aware of it.
>     http://ivan.fomentgroup.org/blog/wp-content/uploads/2012/10/kamdsettings.png
> 
> Andrea Diamantini wrote:
>     Ivan, ok for the settings module and for the kde-wide thing. I just was aware of it, reading the planet :)
>     My second question wasn't really about your patch for rekonq, but about your job on Activities/SLC in general: how are all these information stored?
> 
> Ivan Čukić wrote:
>     That's a long tale of woe, treachery and deceit :)
>     
>     Joking aside, kamd has a plugin system behind it, and plugins are responsible for the stats storage.
>     
>     At first, everything was in Nepomuk (the obvious choice), but that turned out not to be performant enough, so the data corpus now resides in a sqlite database mentioned above (.kde/share/apps/activitymanager/resources/database).
>     
>     That file contains all events and scores (might contain more in the future). For the sake of easier integration with other KDE applications, only the score objects are pushed to Nepomuk (kao:ResourceScoreCache instances). This way, we don't get a performance hit (there are *a lot* of events that need storing) and don't loose anything significant for most applications (those usually need top-rated stuff, not exact timestamps of when a file was accessed).

One of the last questions, just to understand things :)
You trace everytime a link is opening and a tab is switching, but never took care about when a tab is closed. Can this storing QHash going to explode? This because we have lots of reporters used to never close their browser in weeks (browse, suspend, browser... crash! oops... and so again..)

- Andrea

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://git.reviewboard.kde.org/r/106912/#review20495
-----------------------------------------------------------

On Oct. 18, 2012, 8:48 p.m., Ivan Čukić wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://git.reviewboard.kde.org/r/106912/
> -----------------------------------------------------------
> 
> (Updated Oct. 18, 2012, 8:48 p.m.)
> 
> 
> Review request for rekonq and Andrea Diamantini.
> 
> 
> Description
> -------
> 
> Rekonq reports the open/close document events to activity manager daemon.
> 
> By knowing which window contains which documents and which one is in focus, we can do the following:
> 
> - collect the statistics about visited pages. Further, this provides a score for each document visited, that depends on the number of times it was open, the time the user spent on that location, and the time passed since the last visit.
> - availability of a global/workspace applet that allows sharing the current document via e-mail, social networks; bookmarking and rating the link, or connecting it to the current activity. (advantage of this is a unified UI for sharing/rating/linking that works with any application)
> - jump-lists (not impl. yet in plasma) to list top rated documents on a launcher icon or in the task manager applet
> - krunner can sort the documents based on the score
> - more things that I haven't thought of yet
> 
> There is no need to *use* ativities to have these benefits. Activities just serve as manual data clustering to provide more useful scores compared to the one-activity approach.
> 
> 
> Diffs
> -----
> 
>   CMakeLists.txt 030f0fb 
>   src/CMakeLists.txt 6c27376 
>   src/mainview.h 89ee36e 
>   src/mainview.cpp b71a81e 
> 
> Diff: http://git.reviewboard.kde.org/r/106912/diff/
> 
> 
> Testing
> -------
> 
> The event reporting was watched in two ways:
> - by using the SLC applet (planned to be included in 4.10, currently plasma-mobile) - when rekonq opens a url, the icons light-up and when clicking on an icon, it show whether it reports the document properly
> - by using sqlite3 to browse the .kde/share/apps/activitymanager/resources/database - 'select * from nuao_DesktopEvent;'
> 
> The tests consisted of using rekonq in two modes - the single-window and multiple-window mode. They consisted of switching between different webpages back-and-forth.
> 
> 
> Thanks,
> 
> Ivan Čukić
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/rekonq/attachments/20121020/2afddba5/attachment.html>