GNU Astronomy Utilities Processing options

Some processing steps are common to several programs, so they are defined as common options to all programs. Note that this class of common options is thus necessarily less common between all the programs than those described in Input/Output options, or Operating mode options options. Also, if they are irrelevant for a program, these options will not display in the --help output of the program.


The minimum size (in bytes) to memory-map a processing/internal array as a file (on the non-volatile HDD/SSD), and not use the system’s RAM. Before using this option, please read Memory management. By default processing arrays will only be memory-mapped to a file when the RAM is full. With this option, you can force the memory-mapping, even when there is enough RAM. To ensure this default behavior, the pre-defined value to this option is an extremely large value (larger than any existing RAM).

Please note that using a non-volatile file (in the HDD/SDD) instead of RAM can significantly increase the program’s running time, especially on HDDs (where read/write is slower). Also, note that the number of memory-mapped files that your kernel can support is limited. So when this option is necessary, it is best to give it values larger than 1 megabyte (--minmapsize=1000000). You can then decrease it for a specific program’s invocation on a large input after you see memory issues arise (for example, an error, or the program not aborting and fully consuming your memory). If you see randomly named files remaining in this directory when the program finishes normally, please send us a bug report so we address the problem, see Report a bug.

Limited number of memory-mapped files: The operating system kernels usually support a limited number of memory-mapped files. Therefore never set --minmapsize to zero or a small number of bytes (so too many files are created). If the kernel capacity is exceeded, the program will crash.


Do not print any message when an array is stored in non-volatile memory (HDD/SSD) and not RAM, see the description of --minmapsize (above) for more.

-Z INT[,INT[,...]]

The size of regular tiles for tessellation, see Tessellation. For each dimension an integer length (in units of data-elements or pixels) is necessary. If the number of input dimensions is different from the number of values given to this option, the program will stop with an error. Values must be separated by commas (,) and can also be fractions (for example, 4/2). If they are fractions, the result must be an integer, otherwise an error will be printed.

-M INT[,INT[,...]]

The number of channels for larger input tessellation, see Tessellation. The number and types of acceptable values are similar to --tilesize. The only difference is that instead of length, the integers values given to this option represent the number of channels, not their size.


The fraction of remainder size along all dimensions to add to the first tile. See Tessellation for a complete description. This option is only relevant if --tilesize is not exactly divisible by the input dataset’s size in a dimension. If the remainder size is larger than this fraction (compared to --tilesize), then the remainder size will be added with one regular tile size and divided between two tiles at the start and end of the given dimension.


Ignore the channel borders for the high-level job of the given application. As a result, while the channel borders are respected in defining the small tiles (such that no tile will cross a channel border), the higher-level program operation will ignore them, see Tessellation.


Make a FITS file with the same dimensions as the input but each pixel is replaced with the ID of the tile that it is associated with. Note that the tile IDs start from 0. See Tessellation for more on Tiling an image in Gnuastro.


When showing the tile values (for example, with --checktiles, or when the program’s output is tessellated) only use one element for each tile. This can be useful when only the relative values given to each tile compared to the rest are important or need to be checked. Since the tiles usually have a large number of pixels within them the output will be much smaller, and so easier to read, write, store, or send.

Note that when the full input size in any dimension is not exactly divisible by the given --tilesize in that dimension, the edge tile(s) will have different sizes (in units of the input’s size), see --remainderfrac. But with this option, all displayed values are going to have the (same) size of one data-element. Hence, in such cases, the image proportions are going to be slightly different with this option.

If your input image is not exactly divisible by the tile size and you want one value per tile for some higher-level processing, all is not lost though. You can see how many pixels were within each tile (for example, to weight the values or discard some for later processing) with Gnuastro’s Statistics (see Statistics) as shown below. The output FITS file is going to have two extensions, one with the median calculated on each tile and one with the number of elements that each tile covers. You can then use the where operator in Arithmetic to set the values of all tiles that do not have the regular area to a blank value.

$ aststatistics --median --number --ontile input.fits    \
                --oneelempertile --output=o.fits
$ REGULAR_AREA=1600    # Check second extension of `o.fits'.
$ astarithmetic o.fits o.fits $REGULAR_AREA ne nan where \
                -h1 -h2

Note that if input.fits also has blank values, then the median on tiles with blank values will also be ignored with the command above (which is desirable).


When values are to be interpolated, only change the values of the blank elements, keep the non-blank elements untouched.


The metric to use for finding nearest neighbors. Currently it only accepts the Manhattan (or taxicab) metric with manhattan, or the radial metric with radial.

The Manhattan distance between two points is defined with \(|\Delta{x}|+|\Delta{y}|\). Thus the Manhattan metric has the advantage of being fast, but at the expense of being less accurate. The radial distance is the standard definition of distance in a Euclidean space: \(\sqrt{\Delta{x}^2+\Delta{y}^2}\). It is accurate, but the multiplication and square root can slow down the processing.


The number of nearby non-blank neighbors to use for interpolation.