Datasets and Items

MIRIAD uses the term `dataset' to refer to an image, a cube or a set of visibilty data. A MIRIAD dataset is made from a host-system directory, i.e., the host operating system sees a `directory' whenever MIRIAD sees a `dataset'. A MIRIAD dataset contains several kinds of data, called `items',   used by the MIRIAD tasks. Items can be quite small (e.g. a single number), intermediate in size (e.g. the history item, use to store history information), or very large (e.g. the image item in an image dataset, which is used to store the pixel data, or the visdata item in a visibility dataset, which contains most of the data). Large items are stored as a host system file, whereas all small items are stored in a common file which is rather inappropriately called header. Indeed, there are a number of instances where the word ``header'' is used where ``item'' or ``small item'' would be a more appropriate description.

The implementation of datasets as directories does complicate some manipulations of your datasets, since your favourite image, etc, is not just a file anymore. On the other hand, as the host system sees a MIRIAD dataset as a normal directory, all the usual host commands to manipulate directories can be used. On UNIX a -r switch often has to be used with the command, to indicate that the operation is to be applied `recursively' (i.e. to all files in the directory). For example, to delete a dataset, use

% rm -r dataset

  If you have aliased rm to prompt you before deleing a file (as is common in a number of the standard login scripts at Epping), you will be prompted before deleting each individual file within a dataset. This can become somewhat tedious, so you might want to make another alias to delete without prompting. For example, insert

     alias rrm 'rm'

% cp -r dataset1 dataset2

Generally the user is insulated from this internal organization of a dataset and can always think of them as a whole. However there are a few MIRIAD utilities to manipulate at the item level. These MIRIAD tasks do not contain any astronomical knowledge. Consequently they may seem somewhat crude. Some of these commands are:

• itemize: List the items in a dataset.
• delhd: Delete an item from a dataset.
• puthd: Add an item which consists of a single-value to a dataset.
• copyhd: Copy an item from one dataset to another.
• gethd: Print the value of an item. This is most useful in scripts.

As an example, consider the itemize task, which lists the items in a dataset. For the test image dataset, gauss (created with imgen   in Chapter 2) itemize will tell us of the following items:

   % itemize in=gauss
   Itemize: version 1.1 4-mar-91
   naxis    = 2
   naxis1   = 256
   naxis2   = 256
   crpix1   = 129
   crpix2   = 129
   cdelt1   = -4.848137e-06
   cdelt2   = 4.848137e-06
   crval1   = 0
   crval2   = 0
   history    (text data, 38 elements)
   image      (real data, 65536 elements)

Note for images that many items have FITS-like names (although they are lower case, and the units can be different from the FITS standard). A list of the items in an image and visibility dataset are given in Appendices B and C.

The task prthd provides a higher level (astronomical) summary of a dataset. For visibility datasets, it will tell you the number of visibilities, the number of spectral-channels, the number of spectral-windows (IFs), the polarisations present and some information about frequencies. Note that if the visibility dataset contains multiple source or frequencies, it will only give you information about the source and frequency of the first record. For images, it will give dimension and axis information, minimum and maximum, etc. The input to prthd is the dataset name via the keyword in. Task prthd has a brief mode, where it lists a single line description of a dataset. This is useful for generating a summary of directory contents (note that the in keyword supports wildcards).