wiki:Meetings/Meet-2011-11-02
Last modified 5 months ago

  1. KIELER Meeting 2011/11/02
    1. Agenda
    2. KLePto Demo (cmot)
    3. SyncCharts? Demo (chsch)
    4. Menges KLighD Demo (chsch)
    5. Ptolemy Demo (msp)
    6. Release

KIELER Meeting 2011/11/02

  • Attendees:
    • msp
    • cmot
    • cds
    • rvh
    • chsch
    • ckru
    • ima
  • Protocol: cds
  • Begin: 10:05h
  • End: 11:12h

Agenda

  • KLePto Demo (cmot)
  • SyncCharts? Demo (chsch)
  • Menges KLighD Demo (chsch)
  • Ptolemy Demo (msp)
  • Release (msp)

KLePto Demo (cmot)

  • Support for immediate transitions and conditional states in KlePto? simulations.
  • Also works perfectly for parallel states. (which are simulated using different modal models)
  • Also works across different levels of hierarchy.
  • No support for immediate transitions originating at initial states. (Chris Shaver is working on providing Ptolemy support for that)
  • Is able to detect immediate transition cycles and throws up an error message for the user.

SyncCharts? Demo (chsch)

  • KITS allows comments in the code.
  • Which element a comment belongs to may be ambiguous when the parser doesn't know whether the comment is followed by a state or by a transition.
  • Solved by adding a semicolon to a stat's last transition to resolve the ambiguity.
  • Discussion about why semicolons where left out of the textual syntax to begin with.
    • chsch: Because it was tempting to remove them when they are not absolutely necessary.
    • rvh: Doesn't see why semicolons shouldn't be included in the textual syntax.
    • chsch: This would have to be done properly throughout the syntax, such that multiple signals or transitions could be separated e.g. by commas.
  • GraphViz? segmentation fault during a model demo.
    • #1799: Collect examples for which GraphViz? fails and file a bug report.
  • Since the new Xtext version, certain characters cannot be entered into the text view anymore.
  • The serializer takes quite a long time to turn the model into a textual representation. Xtext 2 provides a new implementation for the serializer which should take much less time.

Menges KLighD Demo (chsch)

  • Semantic zooming: the level of detail is changed dynamically. This is for instance done by replacing multiple edges by a single edge.
  • rvh: The basic problem to be solved here is whether to apply a completely new layout when the level of detail is reduced. If so, the model doesn't retain its basic layout, interfering with the user's mental map of the model. If not, it may be difficult for a layout algorithm to only route edges. This would be an interesting research topic, and a tradeoff between keeping the basic layout of the diagram the same while still providing a good layout for the reduced versions.

Ptolemy Demo (msp)

  • Comments can now be associated with nodes by dragging the comments over the nodes.
  • KLay Layered recognices comments and places them either above or below a node, depending on a node's labels and ports.
  • Comments are currently not imported into KIELER. The importer rewrite scheduled by cds should fix that. (#1785)
  • rvh: It would be nice for Ptolemy to have node labels centered on their respective nodes instead of them being left-aligned.
  • Relation node placing was improved by introducing a post processing step.
  • msp: Many Ptolemy models have too many relation nodes that throw off the layout algorithm. It would be better for the layout algorithm to throw them away prior to running the layout algorithm.
    • rvh: Relation nodes may not be discarded in every case because they may carry important semantics.
  • There's a new layout configuration dialog that seemlessly integrates into Ptolemy. It provides the most important options.
    • Aspect ratio only has an influence on the box layout algorithm.
    • rvh: #1800 The aspect ratio slider is not intuitive; perhaps add some hints as to what the layout result might be.
  • The new interaction mode allows the user to influence cycle breaking, crossing minimization and layering. This can greatly improve the layout of certain diagrams, and is a solution for a comment by Edward Lee.
    • The interactive column mode sometimes doesn't work too well for Ptolemy, with ndoes being placed in the wrong layer.
  • Relation node placement is not ideal for Edward's Brewery example.

Release

  • Once the ticket bashing sessions are over with, everyone should go through their tickets and fix them.
  • We would like to target a release date two weeks from now.