[Kst] Time index lookup

[Barth Netterfield]

On Tue, Jul 24, 2012 at 5:34 PM, Nicolas Brisset <nicolas.brisset at free.fr>wrote:

> I have some ideas I will detail later, but some quick remarks:
> 1) please make the name of the time vector configurable. Not everybody
> speaks English, remember? I have bunches of files where time is called
> Temps :-)
>

I agree.  But:
A) Should the list of index-able fields be attached
     -to the current session (ie, the kst file)?
     -to a particular data file?
     -to the data source type (eg, all dirfiles),
  or should it be global?

B) Where should the list be managed?  This will depend on the answers to
(A).  If global, then the settings dialog.  Otherwise, either in the data
source configuration, or in the range dialog itself.


> 2) if binary search is not fast enough, maybe we could store an index of
> 100 of 1000 intermediate time values to initialize the search (not sure
> it'S worth it, though, just an idea like that. Nad maybe I should read up
> on what binary search exactly is...)
>

Binary search is a super simple log(N) search for ordered lists.  The big
issue, however, is that we may have telemetry spikes in our index field -
so we have to cope with that.  :-(


> 3) I hope we will quickly have a convenient way to input date and time
> because a float value is not going to be fun to work with!
>

CTime isn't that much different than frame numbers once you have a few
million frames.  I think we should try to eventually implement string based
time, but I suspect we will mainly use ctime indexes in practice.



> ----- Mail original -----
> > De: "Barth Netterfield" <netterfield at astro.utoronto.ca>
> > À: "kst" <kst at kde.org>
> > Envoyé: Mardi 24 Juillet 2012 17:00:32
> > Objet: [Kst] Time index lookup
> >
> >
> > For BLASTpol, I need to implement the ability to index samples by
> > time, rather than by frame number. So I'm going to implement it.
> > Here is my initial plan:
> >
> >
> > In the field range widget, there is a combo box that currently only
> > includes 'frame'.
> > Kst will add to the combo box any vector fields provided by the data
> > source which happen to be called "time", "Time", or "TIME". kst will
> > assume that these fields will be monotonically increasing (though it
> > will be able to look for and ignore short telemetry glitches).
> > The user can then do a lookup with these fields instead of frames.
> > Kst will do a binary search to find the right frames, and then read
> > them in.
> >
> >
> > In the first instance, you will enter the times as numbers in
> > whatever format the fields store them.
> >
> >
> > Eventually, you will be able to tell kst (how?) what to interpret the
> > fields as, and then use text type time entry (eg, yyyy/mm/dd
> > hh:mm:ss). But not in the first instance.
> >
> >
> >
> > --
> > C. Barth Netterfield
> > University of Toronto
> > 416-845-0946
> >
> >
> >
> > _______________________________________________
> > Kst mailing list
> > Kst at kde.org
> > https://mail.kde.org/mailman/listinfo/kst
> >
> _______________________________________________
> Kst mailing list
> Kst at kde.org
> https://mail.kde.org/mailman/listinfo/kst
>



-- 
C. Barth Netterfield
University of Toronto
416-845-0946
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kst/attachments/20120811/1e5879bc/attachment.html>