|Contents | Language Implementations | .Net|
This document discusses .NET arrays and how they can be supported in Parrot.
At the runtime level only one-dimensional arrays of fixed size, a single type and with first element at index zero are supported. Anything beyond this is supported through the System.Array class.
Arrays are implemented using PMCs. PMCs have keyed v-table operations (that is, v-table functions that as one of their parameters take a key). For arrays the key is the array index and is expected to be an integer. A number of PMCs exist ready to use in the Parrot source tree, however languages are free to provide their own. The v-table mechanism to access elements means that the interface to the array is always the same, so translated .NET code will be able to handle being passed a Perl array even though their behaviours are very different (Perl arrays are resizable, for example, where as .NET arrays are of fixed length).
Arrays are created using a "newarr" instruction which takes the type that the array will contain. As the notion of .NET value types isn't really handled so gracefully in the translation, there is a need to special-case them, mapping any type of array that stores integers to be a FixedIntegerArray and any array that stores floating point numbers to be a FixedFloatArray. All other types get stored in a FixedPMCArray.
Array loads and stores can be translated trivially to the PIR keyed syntax.
$I1 = $P1[$I0] # load $P1[$I0] = $I1 # store
There $P1 is the array PMC, $I0 is the array index and $I1 is the register the value is loaded in to or the value to be stored.
The only remaining array instruction is ldlen, which loads the element count of the array onto the stack. This maps to the elements v-table call in Parrot.
$I0 = elements $P0
Where $P0 is the array PMC and $I0 will hold the number of elements.