We are (finally!) getting around to moving some of our stuff to the
new ugrid standard, and we've come up with some questions.
At this point, we are only interested in triangular mesh grids, and
while I like that the standard is very flexible, it's also nice that
the simple stuff is simple, for the most part.
(thanks Bert for making that available again...)
On to the questions:
1) Where are we at with the clockwise / counterclockwise issue? I
personally prefer counter-clockwise always, but if others want it to
be a specify-able attribute, that's OK too. But I'd like to nail that
down so we know what to put in our client software.
2) I see that "edge_node_connectivity" is listed as a "Optionally
required attribute" -- I"m not sure what the difference is from an
"Optional attribute", but anyway -- we often need to specify data on
some edges, but not all -- can we put only those edges that we care
about in the edge_node_connectivity array? Or do we need to put them
all in, then use a Location index set to specify which ones we have
data on -- it seems somewhat redundant to do so.
If we can specify only those edges that we care about, there should be
a standard way to specify whether the edge_node_connectivity is a
complete set or not. Maybe a different name?
3) I see" face_face_connectivity" as an optional variable. While we
often do want to know what the neighbors of a given triangle are,
specifying it in this way requires one to search through the whole
list. On the other hand, we (and some of the models we've looked at)
often work with a neighbors set -- i.e for each triangle, there are
three neighbors (though some could be outside the domain, and thus get
an invalid flag or something). So we'd like a
face_neighbor_connectivity array that was (in the case of triangles)
(nMesh2_face X 3). Not sure about the name, though. GRanted, this is
re-constructable, but often the data is there in model output file,s
so why not use it?
4) Not a standard issue, per se, but we often need to know what the
boundary types are for a model -- i.e. is it an open or closed
boundary. There are ways in the standard to specify data on the edges,
so the mechanism is there, but it would be nice to officially make
this an optional (and encouraged) data -- give it a standard name,
maybe even a standardized set of flags, or something (that may no be
possible, each model has its own boundary types(conditions) ).
I noticed that the data sets put together for the Modeling Test Bed
project don't have this information, but we'd really like it!
Thanks for your comments,
Christopher Barker, PhD
Python Language Consulting
- Scientific Software Development
- Desktop GUI and Web Development
- wxPython, numpy, scipy, Cython