RE: [mylar-dev] DOI fisheye view

["Mik Kersten" <beatmik@xxxxxxx>]

I only skimmed it too, but found two notable things about the approach: the
DOI function and the display method.  The DOI function is very similar to
other Focus+Context approaches, where it uses the distance from the focus
point and some structural properties (e.g. they rank types higher than
fields).  Mylar's DOI function is different because it's purely determined
by usage and is a measure of the frequency and recency of access to the
element, where access can include predicted or propagated interest.  The
structure-based approaches are useful for a lot of things, but what we have
found is that ranking based on interaction is a good way to solve the
information overload problem when working on tasks that involve a small
subset of a large system.  In a couple of weeks I'll be presenting our
results at the Foundations of Software Engineering (FSE) conference in
Portland, and will post the corresponding paper before then. 

The other interesting thing about their approach is the display mechanism.
But I wonder how useful text that's scaled down below an easily readable
font size really is.  That's why for focusing the editor on the task context
we automatically drive the editor folding, since folding provides a kind of
semantic zooming (you see the declaration of the method, not the body).
While I currently think that this approach provides better integration and
simpler interaction than scaling, you can also combine the two.  For
example, at PARC I experimented with using Fluid Documents
(http://www2.parc.com/istl/projects/fluid/) to provide Focus+Context views
that could do things like dynamically inline AspectJ advice blocks.
Something like that could provide the desired context while retaining the
readability of the text.  But it would be challenging to layer over the
Eclipse text editors, ant the resulting interaction with the document can
end up quite complex.

Mik

> -----Original Message-----
> From: mylar-dev-bounces@xxxxxxxxxxx [mailto:mylar-dev-bounces@xxxxxxxxxxx]
> On Behalf Of Raphael Ackermann
> Sent: Thursday, October 19, 2006 8:00 AM
> To: Mylar developer discussions
> Subject: [mylar-dev] DOI fisheye view
> 
> I just came across this paper "Evaluating a fisheye view of source
> code" by Jakobsen
> http://portal.acm.org/citation.cfm?id=1124830&dl=ACM&coll=&CFID=15151515&C
> FTOKEN=6184618
> available from: http://www.diku.dk/~kash/papers/CHI2006_jakobsen.pdf
> 
> Disclaimer: I've only skimmed the paper.
> and thought it might be quite interesting to have such a fisheye view
> combined with Mylar.  They use DOI  on a single java source file,
> resulting in a compact fisheye representation of the file in the
> editor.
> 
> Mylar lets you use active folding, which results in all the methods of
> no interest being folded. Active folding however doesn't get rid of
> white space for example and only hides elements in the views and not
> in the editor. The fisheye view seems to show the class declaration,
> relevant fields and methods which seems to be quite useful.
> 
> Raphael
> 
> 
> --
> Raphael Ackermann
> Glasgow, G20 8QT, United Kingdom
> E-mail: raphael.ackermann@xxxxxxxxx
> _______________________________________________
> mylar-dev mailing list
> mylar-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/mylar-dev