Technical Meeting 04/20/2006 1:15PM EDT
Agenda
| Item | People | Description | References |
|---|---|---|---|
| 1 | Shead | PGraph and Prism | |
| 2 | Shead | Selection use cases and interface | Selection Use Cases, Selection |
Action Items
- Berk - Document unchecked property changes on the Wiki.
- Clinton - Complete the accept/reset layer for the April release.
Discussion
The team discussed implementing the Prism SESAME view using PGraph, but there was a strong consensus that the SESAME view should continue to be implemented using a render window - "if it ain't broke don't fix it". At most it seemed that the SESAME view might make use of the PGraph extraction filters in the future.
Discussion continued on how to best handle rendering of selected data. There were two proposals: first, have a singleton component that would gather selected data from the pipeline for rendering. Second (and slightly better received) was a suggestion to add a special "selection" field to data sets that would store the selected/unselected state of cells. A special painter or painters would be responsible for highlighting selected cells based on this field. Thus, there could be many different "styles" for displaying selections, e.g: transparent rendering of unselected cells, increased intensity for selected cells, outline selected cells, etc. This approach also solves the problem of managing a separate "selection" pipeline for drawing selected cells.
An issue was raised as to how to show selected cells for a dataset when it is drawn as a bounding-box or an outline. One suggestion was to use texture-mapping to alter the appearance of the outline; an alternative would be to disallow selection of individual cells when cells aren't being explicitly drawn, or (better) to automatically switch the display style for a dataset when the user switches to cell-selection "mode" (principle of implied consent). There are unanswered questions on how to display selections for some render modes, such as volume rendering.
Berk indicated that the selection set object called-for in the selection design will likely be a type of dataset, probably one that only contains fields. We discussed the need for client-side access to selection sets, which led to a suggestion that we create a general-purpose capability for on-demand retrieval of data to the client, which would be useful for more than just selection.
