Overview of NAP (N-Dimensional Array Processor)

NAP is a loadable extension of Tcl. NAP provides a powerful and efficient facility for processing data in the form of n-dimensional arrays. It has been designed to provide a tcl-flavoured array-processing facility with much of the functionality of languages such as APL, Fortran-90, IDL, J, matlab and octave. Three other tcl extensions which provide array-processing facilities are TiM, BLT and Tk3D.

Existing Tcl facilities (e.g. Tcl variables and procedures) are used where appropriate. The new facilities have been designed to match similar existing ones. In particular, NAP expressions use conventions which are essentially a superset of those of the Tcl expr command. Support is provided for data based on n-dimensional grids, where the dimensions correspond to continuous spatial coordinates. There are interfaces to the HDF and netCDF file formats commonly used for such data, especially in Earth sciences such as Oceanography and Meteorology. There is a new photo image format for NAP data.

NAP was developed as part of the CSIRO CAPS project, but can be loaded and used without the (satellite oriented) CAPS extension. However the CAPS extension requires NAP since most CAPS data are stored as NAOs.

Data are stored in memory as n-dimensional array objects (NAOs), which include information such as:

• data-type
• unique ID (handle generated by NAP) called the OOC-name (OOCs are discussed below)
• optional label
• optional C format
• optional unit of measure
• reference count (allowing automatic deletion of the NAO when it is no longer needed)
• optional missing-value (used to indicate undefined data, etc.)
• rank (number of dimensions)
• dimension sizes
• optional dimension names
• optional pointers to coordinate-variable NAOs associated with each dimension

NAOs are created by the Tcl commands nap and nap_get.

The nap command takes arguments specifying an expression in a manner similar to the expr command. However, unlike expr, nap provides:

• assignment (to a Tcl variable whose value is set to the OOC-name of the resultant NAO)
• substitution of Tcl names (obviating the need for "$" prefixes)
• array facilities (constants, operators, functions, indexing)

The nap_get command creates a NAO from data read from a binary, HDF or netCDF file. Some platforms (only Linux on Intel 386 at the time of writing) support reading of remote virtual netCDF files provided by OPeNDAP (a.k.a. DODS) web servers.

Every NAO has an associated Tcl command called an object-oriented command (OOC). This is used to:

• display the data in the NAO
• display other information about the NAO such as its data-type and dimensions
• change data and other details
• write data from the NAO to a binary, HDF or netCDF file

As usual in Tcl, OOC and nap_get command options can be abbreviated provided there is no ambiguity.

NAP provides many operators and built-in functions. One can define new functions by defining Tcl procedures with the proc command. It is also possible to call functions written in C and Fortran.

The caps/nap GUI provides browsers for:

• Tcl variables
• image files (e.g. GIF, JPEG)
• AVHRR satellite files (caps package required)
• ATSR satellite files (caps package required)
• CIF files (Melbourne University format)
• HDF/netCDF files

• displayed as text
• graphed
• shown as various kinds of images and maps
• animated
• used to create a NAO for further processing using NAP

Author: Harvey Davies       © 2002, CSIRO Australia.       Legal Notice and Disclaimer CVS Version Details: $Id: overview.html,v 1.11 2005/01/19 03:31:41 dav480 Exp $