Next Previous Up Contents
Next: Column-oriented FITS
Up: Supplied Output Handlers
Previous: Supplied Output Handlers

3.7.1 FITS

FITS is a very well-established format for storage of astronomical table or image data (see This writer stores tables in a FITS file consisting of two HDUs (Header+Data Units): a Primary HDU as required by the FITS standard, and a single extension of type BINTABLE containing the table data.

There are a few variants of this format:

The primary HDU contains an array of bytes which stores the full table metadata as the text of a VOTable document, along with headers that mark this has been done. Most FITS table readers will ignore this altogether and treat the file just as if it contained only the table. When it is re-read by this or compatible applications however, they can read out the metadata and make it available for use. In this way you can store your data in the efficient and widely portable FITS format without losing the additional metadata such as table parameters, column UCDs, lengthy column descriptions etc that may be attached to the table.
The primary HDU contains only very minimal headers and no data.
Behaves like fits-basic, but columns containing variable-length numeric array data are stored using the P and Q formats where appropriate, rather than padding smaller arrays to the size of the largest. This can make for more compact storage of variable-length array-valued column data but may also result in tables less suitable for streaming.
Used for storing HEALPix pixel data in a way that conforms to the HEALPix-FITS serialization convention. In most ways it behaves the same as fits-basic, but it will rearrange and rename columns as required to follow the convention, and it will fail if the table does not contain the required HEALPix metadata (STIL_HPX_* parameters).
The default output format is fits-plus; in general you don't need to worry about this, it just gives you some hidden benefits over fits-basic.

A private convention is used where required to support encoding of tables with more than 999 columns (not possible in standard FITS); this was discussed on the FITSBITS mailing list in July 2017 in the thread BINTABLE convention for >999 columns. If software unaware of this convention (e.g. CFITSIO) is used to read such tables, it will only see the first 998 columns written as intended, plus a column 999 containing an undescribed byte buffer.

The handler behaviour may be modified by specifying one or more comma-separated name=value configuration options in parentheses after the handler name, e.g. "fits-plus(date=true)". The following options are available:

date = true|false
If true, the DATE-HDU header is filled in with the current date; otherwise it is not included.

Multiple tables may be written to a single output file using this format.

If no output format is explicitly chosen, writing to a filename with the extension ".fit", ".fits" or ".fts" (case insensitive) will select fits-plus format for output.

The handler classes for these formats are FitsTableWriter, FitsPlusTableWriter, HealpixFitsTableWriter and VariableFitsTableWriter.

To write the FITS header for the table extension, certain things need to be known which may not be available from the StarTable object being written; in particular the number of rows and the size of any variable-sized arrays (including variable-length strings) in the table. This may necessitate two passes through the data to do the write.

To fix the value of the TNULLn magic value for a given column on output, set the value of the Tables.NULL_VALUE_INFO auxiliary metadata value, e.g.:

    colInfo.setAuxDatum(new DescribedValue(Tables.NULL_VALUE_INFO, new Integer(-99)));

Writing columns containing (scalar or array) unsigned byte values (TFORMnn = 'B') cannot be done simply by providing byte data to write, since the java byte type is signed. To do it, you must provide a column of java short (16-bit) integers, and set the value of the Tables.UBYTE_FLAG_INFO auxiliary metadata item to Boolean.TRUE, e.g.:

    colInfo.setAuxDatum(new DescribedValue(Tables.UBYTE_FLAG_INFO, Boolean.TRUE));

Writing columns containing (scalar or array) unsigned long values cannot be done straightforwardly, since, like FITS, the Java long type is signed. Instead, you can provide a column of java String values giving the integer representation of the numbers required, and set the value of the BintableStarTable.LONGOFF_INFO auxiliary metadata item to the string representation of the offset 2**63, e.g.:

   colInfo.setAuxDatum(new DescribedValue(BintableStarTable.LONGOFF_INFO, "9223372036854775808263"));

This will result in the values being written, where in range, with FITS headers TFORMnn = 'K', TZEROnn = '9223372036854775808263'. The same mechanism can be used for other long offsets if required (though not for other integer types).

See the "Binary Table Extension" section of the FITS standard for more details of the FITS BINTABLE format. These handler can write tables with more than the BINTABLE limit of 999 columns, as discussed in Section 3.8.2.

There is some support for the semi-standard HEALPix-FITS serialization convention. If some of the HPX_*_INFO metadata items defined by the class HealpixTableInfo are present in the output table's parameter list, the corresponding HEALPix-specific FITS headers will be written on a best-efforts basis into the output BINTABLE HDU. This may or may not be good enough to make that FITS file readable by external HEALPix-FITS-aware applications; one of the requirements of the convention is that the HEALPix pixel index, if present, appears in the first column of the table under the name "PIXEL". An alternative is to use the handler HealpixFitsTableWriter, which tries harder to write tables using the HEALPix convention. This will fail unless the required HPX_*_INFO metadata items mentioned above are present, and will reorder and rename columns as required for maximum compatibility.

For column-oriented FITS output, see Section 3.7.2.

Next Previous Up Contents
Next: Column-oriented FITS
Up: Supplied Output Handlers
Previous: Supplied Output Handlers

STIL - Starlink Tables Infrastructure Library
Starlink User Note252
STIL web page:
Author email: