Lloyd Treinish:
 |Randall Hopper:
 |> Is there a way to adjoin grids in DX without interpolation?
 |...
 |>         0 -----> (m-1)       m -----> (2m-1)      0 ---------------> (2m-1)
 |>          ----------           ----------           --------------------
 |>         |          |         |          |         |                    |
 |>         |          |    +    |          |    =    |                    |
 |>         |          |         |          |         |                    |
 |>         |          |         |          |         |                    |
 |>          ----------           ----------           --------------------
 |
 |Well, you can create a composite field (i.e., what Partition does on an SMP
 |machine for modules to run in parallel) or a multigrid.  The composite
 |field would be appropriate for tiled images, DEMs, etc.  The latter
 |supports overlap and you can indicate what part of the other grid(s) to
 |ignore.  It can be used for data from simulations on multizone or nested
 |grids.

Thanks for the suggestion.  I hadn't worked with composite fields before,
and it's good to know they're there.

After reading up on them however, it appears I can't use them to solve the
problem (unless I regenerate the datasets, or write a custom loader -- what
I'm trying to avoid).  There are two reasons for this:

   1) Composite fields presume that the edge positions and data are repeated
      in adjoining subfields, and

   2) There's no way (that I could see) to generate a correct composite
      field from pre-partitioned individual fields with the standard DX
      modules.

To the 1st point, this topo dataset and other USGS topo I've worked with
doesn't replicate data across tile boundaries (i.e. it's 0..(m-1),
m..(2m-1); versus 0..m, m..2m).  The application operates on a scale where
this can make a difference.

To the 2nd, I don't see how to set meshoffsets with a DX module.  With
Collect and ChangeGroupType("compositefield"), I can get close to
constructing a composite field given data that repeats boundaries, but I
don't see how to set the mesh offsets.  I noticed that "compositefield" is
an undocumented "newtype" value for the ChangeGroupType module (though it's
valid in the module source code), and maybe this is why.

Does this make sense?  It's possible I've missed something here.

Thanks,

Randall

-- 
Randall Hopper (mailto:[EMAIL PROTECTED])
EPA Scientific Visualization Center
US EPA MD/24 ERC-1A; RTP, NC 27711




------------------------------------------------------------------------------
object 2 class gridconnections counts 5 3
object 3 class gridpositions counts 5 3 
object 4 class array type float rank 1 shape 1 items 15 data follows
object 1 class field
  component "connections" value 2
  component "positions" value 3
  component "data" value 4

object 6 class gridconnections counts 5 3
  meshoffsets 0 2                                 <----------------------------
object 7 class gridpositions counts 5 3 
object 8 class array type float rank 1 shape 1 items 15 data follows
object 5 class field
  component "connections" value 6
  component "positions" value 7
  component "data" value 8

object "default" class compositefield
  member 0 value 1
  member 1 value 5

Reply via email to