10.1 Referencing Column Values

To create a useful expression which can be evaluated for each row in a table, you will have to refer to cells in different columns of that row. You can do this in three ways:

By Name

The Name of the column may be used if it is unique (no other column in the table has the same name) and if it has a suitable form. This means that it must have the form of a Java variable - basically starting with a letter and continuing with letters, numbers, underscores and currency symbols. In particular it cannot contain spaces, commas, parentheses etc.

As a special case, if an expression contains just a single column name, rather than some more complicated expression, then any column name may be used, even one containing non-alphanumeric characters.

Column names are treated case-insensitively.

By $ID

The "$ID" identifier of the column may always be used to refer to it; this is a useful fallback if the column name isn't suitable for some reason (for instance it contains spaces or is not unique). This is just a "$" sign followed by the column index - the first column is $1.

By ucd$ specifier

If the column has a Unified Content Descriptor (this will usually only be the case for VOTable or possibly FITS format tables) you can refer to it using an identifier of the form "ucd$<ucd-spec>". Depending on the version of UCD scheme used, UCDs can contain various punctuation marks such as underscores, semicolons and dots; for the purpose of this syntax these should all be represented as underscores ("_"). So to identify a column which has the UCD "phot.mag;em.opt.R", you should use the identifier "ucd$phot_mag_em_opt_r". Matching is not case-sensitive. Futhermore, a trailing underscore acts as a wildcard, so that the above column could also be referenced using the identifier "ucd$phot_mag_". If multiple columns have UCDs which match the given identifer, the first one will be used.

Note that the same syntax can be used for referencing table parameters (see the next section); columns take preference so if a column and a parameter both match the requested UCD, the column value will be used.

By utype$ specifier

If the column has a Utype (this will usually only be the case for VOTable or possibly FITS format tables) you can refer to it using an identifier of the form "utype$<utype-spec>". Utypes can contain various punctuation marks such as colons and dots; for the purpose of this syntax these should all be represented as underscores ("_"). So to identify a column which has the Utype "ssa:Access.Format", you should use the identifier "utype$ssa_Access_Format". Matching is not case-sensitive. If multiple columns have Utypes which match the given identifier, the first one will be used.

Note that the same syntax can be used for referencing table parameters (see the next section); columns take preference so if a column and a parameter both match the requested Utype, the column value will be used.

With the Object$ prefix

If a column is referenced with the prefix "Object$" before its identifier (e.g. "Object$BMAG" for a column named BMAG) the result will be the column value as a java Object. Without that prefix, numeric columns are evaluated as java primitives. In most cases, you don't want to do this, since it means that you can't use the value in arithmetic expressions. However, if you need the value to be passed to a (possibly user-defined) method, and you need that method to be invoked even when the value is null, you have to do it like this. Null-valued primitives otherwise cause expression evaluation to abort.

There is also a special column:

$index

The value of this is the current row number (the first row is 1). You can alternatively use the form $0. (The form index is also permitted, but deprecated). Note that this value is a long (8-byte integer); when using it in certain expressions you may find it necessary to convert it to an int (4-byte integer) using the toInteger() function.

The value of the variables so referenced will be a primitive (boolean, byte, short, char, int, long, float, double) if the column contains one of the corresponding types. Otherwise it will be an Object of the type held by the column, for instance a String. In practice this means: you can write the name of a column, and it will evaluate to the numeric (or string) value that that column contains in each row. You can then use this in normal algebraic expressions such as "B_MAG-U_MAG" as you'd expect.