Re: How many use the diagram tree?



On Sun, 10 Aug 2003 <horkana maths tcd ie> wrote:
On Sun, 10 Aug 2003, James K. Lowden wrote:

Date: Sun, 10 Aug 2003 06:12:01 -0400
From: James K. Lowden <jklowden schemamania org>

"Tabbed diagrams" would seem to me to be the natural way to present
named subsets to the UI, FWIW.

If by Tabbed Diagrams you mean representing multiple pages in single
workbook then that sounds reasonable and consistant with other
diagramming software.  I would really hate to see someone trying to
reinvent the window manager and turn Dia in to a tabbed interface, which
is quite differnt from the idea of a workbook with multiple pages/sheets
which we will eventually want to support anyway for compatibility
reasons.

I generally avoid pie-in-the-sky suggestions because I'm not knee deep in
code.  It's just that someone sometime (possibly me, but not in C) is
going create useful database shapes, whether based on the UML shapes or
not.  When that happens, it will be only a short step to lots of people
using Dia for database design, and they won't be interested in 10-table
models.  You can keep a 10-table model in your head.  It's 100-table
models that require diagrams.  And, as you say, that's a problem of a
different magnitude.  ;-)

And I've never thought Visio was a paradigm worthy of emulation.  Maybe
there are Visio diagrams out there with hundreds of objects on them, but
to me that would be the work of a masochist.  Visio is a drawing tool; it
demands much attention to visual aspects (and too many clicks for most of
them).  Tools like ERwin are data tools with a visual representation
capability.  I've always thought that Dia's XML foundation is a terrific
asset in creating data-rich diagrams, something Visio really doesn't even
try to do.  And, yes, by the way, your guess is correct (sorry not to have
been clear): I was talking about ERwin.  I've modelled thousands of tables
in ERwin, I'm sure.  

<rant> One thing I've never understood is Visio's facination with handles.
 Why?  Any 14-year-old knows a line consists of an infinite set of points.
 What's the purpose of picking out some few as more significant than the
rest?  Why isn't every point on the line a handle?  Draw a line, drag it
to another line.  Of course they connect.  What in world else would anyone
want?  But, no, isn't it much more fun, more like a video game, to play
"Hit the Handle" and see it go from red to green or whatever?  Why do real
work when you could be improving your eye-hand coördination?  </rant>

I think many too-many-objects-sire problems could be solved/mitigated with
linked diagrams, especially if the jump-to-parent-diagram shape were
automatically inserted in dependendent diagrams.  Andy reminded me that
it's often useful to have several views of the same objects, too.  

To clarify what I meant by "tabbed diagrams".  I've found that even for
very complex diagrams, only a few named subsets were worth maintaining. 
Each one is a kind of sub-diagram, really another perspective on some of
the objects.  Suppose each named subset would appear as a tab in the
diagram's window, similar to the way tabs work in Mozilla.  I suggest this
approach because I've often wanted to switch quickly between views, and
seldom had any reason to compare them side-by-side (they have identical
objects, after all).  Erwin's UI is rather clumsy on that score; it has
tabs, but they're not very useful IMO.  

When suggesting features it always helps if you can identify a program
that already succesfully implements the behaviour 
...
I dont particularly want to register just to try it out for a few weeks
but as you already seem to be using it extensively screenshots and
descriptions of things about it you like would be helpful.

I'll be more explicit in the future about my references, and I'll make you
some screen shots so you can see what I mean.  You're right not to bother
registering, btw.  ERwin is a little like single malt scotch or Mahler. 
Something of an acquired taste, and something that grows on you.  

This might be an interesting (potentially better) way of looking at the
behaviours that are currently dealt with in the Selection menu, we might
be able to seperate out the concept of going to and actually selecting
these objects.

If I can further the discussion, I'm only too happy to.  

--jkl



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]