Some feedback

Joachim Breitner mail at joachim-breitner.de
Sun Jun 16 22:08:41 CEST 2013 

Hi,

Am Sonntag, den 16.06.2013, 17:49 +0100 schrieb Waldir Pimenta:


> Sure, two questions about that: First, do you have any software
> project page which you think works well and we could emulate (not
> necessarily in appearance, but in content and organization)?

Sadly, no. There is http://sm.nomeata.de/ which is the web-version of
one of my programs, so quite different, all others have just blog posts.

> Second, I was actually thinking of a single page, i.e. an index.html
> in the current repo, nothing too fancy. The url could change (I have
> no strong opinions on that), but regarding the content, I don't see
> the need for anything more elaborated than a simple page, maybe with
> some tabs for different sections but that's all.

Agreed, lets keep it simple. Put it in the main repo and I’ll make it
available from http://arbtt.nomeata.de.

>         So maybe arbtt-dump should get new options: The same sample-selecting
>         options that arbtt-stats supports (--filter, --exclude, --only), and in
>         addition a --format parameter that allows to switch from the „internal“
>         format to some human-readable format or some programming-accessible
>         format (e.g. JSON).
> 
> 
> Yes, all that sounds great, and would be very useful to me (see
> below). 
>         
>         An option to more easily select the last n samples or the samples of the
>         last n minutes/hours/days/week/months/years would probably be handy as
>         well.
>         
>         Do you have a concrete usecase in mind? I prefer to develop features
>         when there is a need for it, rather than adding features that are only
>         potentially useful (and otherwise just add bloat).
> 
> 
> Yes, I actually wrote little shell aliases called arbtt-log-hour,
> arbtt-log-day, etc., which get arbtt-dump's output and apply some sed
> commands to clean it up for consumption, resulting in entries in the
> format:
> YYYY-MM-DD HH:mm  [program-name]  "Top window title"
> This basically gives me a readable log of which windows were the top
> ones at every minute for the past, say, hour. I use this to quickly
> summarize my recent activities in my personal log. I would love if
> arbtt would provide this functionality out of the box, so I wouldn't
> have to get the full arbtt-dump output and then remove entries I don't
> need (which is obviously slow).

Sounds useful.

> Additionally, it would be great to have it merge entries where the top
> window remains the same through a continuous set of samples (or the
> list of windows doesn't change, if one chooses to view also non-top
> windows).

That sounds a lot like
        arbtt-stats --intervals=Program: --filter '$sampleage <= 1:00'
with
        -- Simple rule that just tags the current program
        tag Program:$current.program,
in categorize.cfg:

$ arbtt-stats --intervals=Program: --filter '$sampleage <= 1:00'
Intervals for category "Program"                                                                                                                                              
================================
__________________Tag_|______________From_|_____________Until_|_Duration_
            Navigator | 06/16/13 19:01:33 | 06/16/13 19:01:33 |    1m00s
                gnote | 06/16/13 19:02:33 | 06/16/13 19:03:33 |    2m00s
            Navigator | 06/16/13 19:04:33 | 06/16/13 19:06:33 |    3m00s
                gnote | 06/16/13 19:07:33 | 06/16/13 19:07:33 |    1m00s
            Navigator | 06/16/13 19:08:33 | 06/16/13 19:17:33 |   10m00s
                gnote | 06/16/13 19:18:34 | 06/16/13 19:19:34 |    2m00s
            Navigator | 06/16/13 19:20:34 | 06/16/13 19:23:34 |    4m00s
                gnote | 06/16/13 19:24:34 | 06/16/13 19:24:34 |    1m00s
                gnote | 06/16/13 19:26:34 | 06/16/13 19:26:34 |    1m00s
            Navigator | 06/16/13 19:27:34 | 06/16/13 19:28:34 |    2m00s
            Navigator | 06/16/13 19:30:34 | 06/16/13 19:38:34 |    9m00s
            evolution | 06/16/13 19:39:34 | 06/16/13 19:39:34 |    1m00s
            Navigator | 06/16/13 19:40:34 | 06/16/13 19:41:34 |    2m00s
gnome-terminal-server | 06/16/13 19:42:34 | 06/16/13 19:42:34 |    1m00s
            evolution | 06/16/13 19:43:34 | 06/16/13 19:45:35 |    3m00s
            Navigator | 06/16/13 19:46:35 | 06/16/13 19:46:35 |    1m00s
gnome-terminal-server | 06/16/13 19:47:35 | 06/16/13 19:47:35 |    1m00s
            evolution | 06/16/13 19:48:35 | 06/16/13 19:48:35 |    1m00s
gnome-terminal-server | 06/16/13 19:49:35 | 06/16/13 19:49:35 |    1m00s
            Navigator | 06/16/13 19:50:35 | 06/16/13 19:50:35 |    1m00s
            evolution | 06/16/13 19:51:35 | 06/16/13 19:54:35 |    4m00s
              gnucash | 06/16/13 19:55:35 | 06/16/13 19:55:35 |    1m00s
            evolution | 06/16/13 19:56:35 | 06/16/13 19:57:35 |    2m00s
gnome-terminal-server | 06/16/13 19:58:35 | 06/16/13 19:58:35 |    1m00s
            Navigator | 06/16/13 19:59:35 | 06/16/13 19:59:35 |    1m00s
gnome-terminal-server | 06/16/13 20:00:35 | 06/16/13 20:01:35 |    2m00s

Is that not sufficient for your use case?

> This sort of access would also make it possible to use arbtt to
> generate personal statistics and visualizations (Stephen Wolfram has
> done some interesting stuff in this vein, see
> here: http://blog.stephenwolfram.com/2012/03/the-personal-analytics-of-my-life/). An open-source program similar to RescueTime would also be possible (to some degree - actual urls aren't captured by arbtt and that's important data since a large part of today's digital activity occurs online). By the way, have you ever thought of using, say, sqlite instead of a Haskell-specific (I assume) binary format? I think that would open up the door to even more possibilities :)

Of course graphs were on my mind as well, but never got around to
implement them. Also capturing URLs (or at least domain names) would be
nice; are there no firefox plugins that set an X property with the URL?

I like to know that the background program uses minimal resources;
appending a few bytes to a file is good there. If the possibilities you
talk about are about other people writing code to use the data, then a
nicer dumping format can serve the same purpose.

As more more analysis in arbtt-stats (or arbtt-graph or whatever): There
are endless possibilities, and so little time :-). I’d prefer to
implement stuff with a concrete usecase, preferably also with a little
outline (e.g. a sketch of an graph or of intended output) and some
discussion if it is not possibly covered by exiting features, or can
maybe simplified or generalized.

Greetings,
Joachim

-- 
Joachim “nomeata” Breitner
  mail at joachim-breitner.de • http://www.joachim-breitner.de/
  Jabber: nomeata at joachim-breitner.de  • GPG-Key: 0x4743206C
  Debian Developer: nomeata at debian.org
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <https://lists.nomeata.de/pipermail/arbtt/attachments/20130616/d721d109/attachment.asc>

More information about the arbtt mailing list