AVS Field Data


This is an extremely general file format, with a wide variety of techniques for data input. Here we shall only consider a subset, please refer to the program documentation for further information.

Files of this format can be read using the read field module. There are two variants: we shall consider first data-parsing mode, which is more general, and then look briefly at native mode.

Data-parsing input

To input a dataset in this mode, you must first create a description file. This is a text file whose name must end in .fld, and which contains a description of the layout of the accompanying file (or files) which contain the actual data. The dataset is an n-dimensional array, each element of which may itself be an array; this data structure is known as a field. There are three types of field.

  • Uniform

    The dataset contains a value for every element in the array, supplied in order with the leftmost subscript varying most rapidly.

  • Irregular

    Each value supplied in the dataset it accompanied by a full set of coordinates, indicating where it belongs in the array. There does not have to be a value for each element.

  • Rectilinear

    On each axis, there are certain indexes for which values are supplied. The values themselves are supplied in order, with the leftmost subscript varying most rapidly. One sequence of coordinates for each axis is supplied.

Description file

Suppose that the dataset we have includes data for a number of points scattered over a 100x100x50 3-dimensional space. For each point, we have recorded two floating-point values, followed by three coordinates. The data are held as ASCII floating-point values in a file mydata.dat. The description file might look like this:

# AVS 
ndim=3
dim1=100
dim2=100
dim3=50
nspace=3
veclen=2
data=float
field=irregular
variable 1 file=mydata.dat filetype=ascii offset=0 stride=5
variable 2 file=mydata.dat filetype=ascii offset=1 stride=5
coord 1 file=mydata.dat filetype=ascii offset=2 stride=5
coord 2 file=mydata.dat filetype=ascii offset=3 stride=5
coord 3 file=mydata.dat filetype=ascii offset=4 stride=5

The first five characters of the description file must be # AVS. A # character can be used to introduce a comment on any line.

The ndim line specifies the number of dimensions in the array, and the dim1, dim2 and dim3 lines specify the size of the dimensions. The nspace line defines the number of coordinates for each point - often this will be the same as ndim.

veclen=2 indicates that there are 2 values for each data point, and data=float specifies that these are binary values. The field=irregular line specifies what kind of field this is (see the last section).

The variable and coord lines have the same format, and explain how these values are to be found. In our case, they are all in the same file, but this is not compulsory. The filetype is given as ascii; it could also be unformatted, corresponding to unformatted Fortran data, or binary for pure binary data.

The offset specifies how many data items to skip before reading the first value for that variable. Note how it varies for the two variables and three coordinates. The stride specifies how far apart the values for this variable are. So, to read the values for the second variable, AVS will first skip one value (offset=1) and thereafter read every fifth value (stride=5).

Using these directives, you can describe a wide variety of file formats. For example, the values for the two variables might be in separate files; or we might have all the values for one variable, followed by all the values for the other.

Native-mode input

This time we have only one file, instead of separate description and data files. The file starts with an text header, similar in format to a description file, but omitting the variable and coord specifiers. This header ends with two form-feed characters; the idea is that you should be able to look at the file with more, for example.

After the two form-feeds, the data follow in binary; first the avlues for the elements, element by element (ie, with the values for each element together). The coordinates (if any) follow the data; first all the x-coordinates, then all the y-coordinates, and so on.

Native-mode field files can also be read by read field. Their names should end with .fld.