Thanks for that, Jan. You make a good argument fo...

2011-08-12T15:56:59.436+01:00

Thanks for that, Jan. You make a good argument for the advantages of just filtering the list, which I hadn't previously considered - largely because, having thought of the tree (and discovered an easy-to-use implementation of it), I stopped looking for alternative representations. However, I think the final say on utility will come from the internal users at NAG; I'll bear your suggestion in mind if they end up seeking other alternatives to the list representation.

Hello Have you considered even a simpler represent...

2011-08-11T11:01:24.186+01:00

Hello
Have you considered even a simpler representation than trees? For example, just filtering the list?

I think that a tree representation is very useful for "heterogeneous data" where branches of the tree look different, for example, one choice (optimization->local) opens other possibilities then the other (optimization->global). If you have several criteria and (more or less) all combinations make sense it is often useful to offer choices in a different manner.

How would filering look like?
Imagine that, by default, you would see all the implementations on the page (as you can see now) but there would be a pulldown menu above each column which would filter out everything except the chosen item.
The categories could be "platform/architecture" "mark" "compiler" or even more dynamically loaded to refine your search.

The advantage would be:
* no enforced order of choices (you don't need to start with architecture, you can ask "what architectures are available for the latest mark" or "what libraries are available for my compiler")
* no need to load Java applet, which took over 20s with the demo example
* in general faster and more portable (because not everyone allows Java to run)
* you don't need to narrow the search to the last leaf/single item

Cheers,
Jan

Comments on The NAG Blog: Seeing the good for the trees

Thanks for that, Jan. You make a good argument fo...

Hello Have you considered even a simpler represent...