Hello all
I'm still following the chain of dependencies for getting metadata in...
I will need to represent the metadata as a tree. In Geotk, I used the
Swing TreeNode interface. Despite being defined in a javax.swing
sub-package, this was a truly generic interface usable without any
dependency to Swing components. However this have a number of issues:
* TreeNode uses old API (java.util.Enumeration, DefaultTreeNode use
java.util.Vector)
* They forgot the getUserObject() method (there is a
setUserObject(...) in MutableTreeNode interface but no corresponding
getUserObject()) and I don't think that Oracle will fix that for
compatibility reasons. In Geotk, this situation forced us to
redefine all Swing TreeNode interfaces and implementations anyway,
extending the Swing type with the addition of the missing
getUserObject() method.
* Using Swing TreeNode interface is cheap up to JDK8, but may not be
cheap anymore in a modularized JDK9 since it may force the JVM to
load an otherwise unneeded Swing module.
* Swing is being phased out, to be replaced by JavaFX.
* The TreeNode interface alone is not sufficient. We actually need
(most of the time) a TreeTableNode interface similar to the one
defined in SwingLab.
I'm tempted to abandon the usage of Swing interface. This raise the
question about which alternative to use. The JavaFX class (TreeItem) is
probably too dedicated to rendering in a JavaFX scene. Consequently I'm
tempted to define our own interfaces in a org.apache.sis.util.tree
package. The interfaces would be TreeNode and TreeTableNode. The
TreeNode methods would be:
* TreeNode getParent();
* List<TreeNode> children();
* Object getUserObject();
The TreeTableNode interface would have in addition:
* int getColumnCount()
* Class<?> getColumnClass(int)
* Object getValueAt(int)
* void setValueAt(int, Object)
* boolean isEditable(int)
I think this is about everything we need. What do people think?
Martin
