Marvin - MarvinView table JTable cell

Hello,

I'm implementing support for multiple molecules in one record. It is usefull for example to show results of few chemical terms (e.g. conformormers). I use this code to show table:

 

            TableSupport tableSupport = mvp.getTableSupport();
            MDocument  mdoc[]  = new MDocument[mols.length];
            for (int i = 0 ; i < mols.length ; i++) {
                mdoc = new MDocument(mols);
            }
            tableSupport.start(new ArrayMDocSource(mdoc), ""); // NOI18N

 

It doesn't work if the component is visualized in JTable to show multiple records. Probably the main problem is that  molecules are processed from MDocSource asynchronously. Is there any way how to process the molecules synchronously in AWT-EVENT-QEUE thread? The TableSupport.startNow method causes deadlock.

Thanks

Petr

Hi,

No, there is no way. Since the startNow method starts and waits for new thread(s) in which calls like EventQueue.invokeAndWait can occur, it cannot be used from the AWT thread. Generally, it is not even a good idea to make GUI update wait for molecule loading. (even though it may be fast in your special case)

Peter

Here is a simple example exemplifying the problem (code attached):

Aim: to use MViewPane as TableCellRenderer and to have MViewPane displaying multiple cells withing the column of the JTable using the MViewPan.setM(index, mol) functionality.

Problem: structures are set to MViewPane component within the getTableCellRendererComponent() method, but the display of the structures is erratic.

 

Is there a solution to this?

Thanks

Tim

Another proof that creating convenience methods is a mistake. Users tend to misuse them. To solve the current problem, I deprecated all asynchronous molecule setting methods in MViewPane. As of Marvin 5.2.5, setM(int, String), setM(int, String, String) and setM(int, File, String) are not recommended to use.

Attached a slightly modified version of your test application. Molecules are loaded using MolImporter, then set using a synchronous method, setM(int, Molecule).

Thanks for the example. It works fine.

So if I understand this correctly then setM(index, Molecule) is synchronous, while setM(index, String) is asynchronous?

If so, then I just see this as a problem in the javadocs that does not suggest this.  I see no reason to remove (deprecate) the methods, just make it clear what they do and how to use them correctly.

Thanks

Tim

 

I think there is no problem with the javadocs, please compare them for the two methods:

http://chemaxon.com/marvin/help/developer/beans/api/chemaxon/marvin/beans/MViewPane.html#setM(int,%20chemaxon.struc.Molecule)

"Sets the nth molecule."

http://chemaxon.com/marvin/help/developer/beans/api/chemaxon/marvin/beans/MViewPane.html#setM(int,%20java.lang.String)

"Starts loading a molecule into the nth cell."

I think it is obvious from the doc that the second method is asynchronous: it does not "load" but "starts loading". It is no use writing a clear documentation if users misunderstand it anyway (if they read it at all). I think the only working solution is the deprecation of such "convenience" methods. Then the user has to write a bit longer code but at least it is more likely that he will know what he is doing.

Peter

Well, I've been programming with Marvin for about 8 years, and I didn't know the difference!

Tim

tdudgeon wrote:

Well, I've been programming with Marvin for about 8 years, and I didn't know the difference! Tim

Usually there is no noticable difference.
The name of the asynchronous methods is also misleading, "startMolLoading" were better than setM. But the removal is even better.