Grids define the spatial regions over which the grid functions will be defined as well as their storage. A grid declaration can take one of several forms, the longest of which would be something like:
uniform rect[x,z] grid g1 [1:Nx][1:Nz] {xmin:xmax}{zmin:zmax}
The first word can be uniform or nonuniform,
though only the former is
currently defined. Next comes the name of the coordinate system followed
by a list of coordinates on which the grid is defined. The above grid is
two dimensional with coordinates x and z.
After the coordinate
system comes the reserved word grid followed by the grid name. Next
comes the index region. In this example, the first index starts at 1 and
goes to Nx,
while the second starts at 1 and goes to Nz. Nx
and Nz must be defined elsewhere. The index regions can contain
arbitrary expressions such as [A*B+C-2:4*Nx-5/a], however, as discussed
in section 2.2, it is best to keep to forms like [1:Nx]
and [0:Nx-1], where Nx has been declared as a parameter. Finally,
comes the coordinate region which gives the actual spatial ranges of the
coordinates. In the example, we have 
and
. Coordinate regions must be of the form
{name1:name2}, where name1 and name2
have been declared as parameters.
Other forms of the grid declaration leave out one or more of the above parts. The minimum allowable declaration is:
uniform rect grid g2
This declaration (along with the example coordinate declaration in section 1.3) declares g2 to be a three dimensional grid with coordinates x, y and z. The index region will be [0:Nx-1][0:Ny-1][0:Nz-1] for C output and [1:Nx][1:Ny][1:Nz] for FORTRAN output. The coordinate region will be {xmin:xmax}{ymin:ymax}{zmin:zmax}