/ 
parrot-0.1.1
/ docs/pdds/pdd04_datatypes.pod

NAME

docs/pdds/pdd04_datatypes.pod - Parrot's internal data types

ABSTRACT

This PDD describes Parrot's internal data types.

DESCRIPTION

This PDD details the basic datatypes that the Parrot core knows how to deal with. Three of these (the integer, floating point and string datatypes) have no additional semantics. The fourth datatype, the Parrot Magic Cookie (PMC) acts as the basis for all high level languages running on top of Parrot; only the most basic aspects are described here.

Note that PMC and string internals are volatile and may be changed in the future (although this will become increasingly unlikely as we near v1.0). Access from external code to the internals of particular datatypes should be via the extension mechanism (see docs/pdds/pdd11_extending.pod, which has more explicit guarantees of stability.

IMPLEMENTATION

Integer data types

Integer data types are generically referred to as INTs. INTs are conceptual things and there is no data structure that corresponds to them.

Platform-native integer

These are whatever size native integer was chosen at Parrot configuration time. The C-level typedefs INTVAL and UINTVAL get you a platform-native signed and unsigned integer respectively.

Arbitrary precision integers

Big integers, or bigints, are arbitrary-length integer numbers. The only limit to the number of digits in a bigint is the lesser of the amount of memory available or the maximum value that can be represented by a UINTVAL. This will generally allow at least 4 billion digits, which ought to be far more than enough for anyone.

The internal representation of a bigint has not yet been determined. One possible implementation is described below; for an alternative implementation, see docs/pdds/pdd14_bignum.pod.

A possible bigint implementation

We can represent a bigint with:

struct bigint {
     void *buffer;
     UINTVAL length;
     INTVAL exponent;
     UINTVAL flags;
}

The buffer pointer points to the buffer holding the actual number, length is the length of the buffer, exponent is the base 10 exponent for the number (so 2e4532 doesn't take up much space), and flags are some flags for the bigint.

Floating point data types

Floating point data types are generically referred to as NUMs. Like INTs, NUMs are conceptual things, not real data structures.

Platform native float

These are whatever size float was chosen when parrot was configured. The C level typedef FLOATVAL will get you one of these.

Arbitrary precision decimal numbers

Arbitrary precision decimal numbers, or bignums, can have any number of digits before and after the decimal point.

The internal representation of a bigint has not yet been determined. One possible implementation is described below; for an alternative implementation, see docs/pdds/pdd14_bignum.pod.

A possible bignum implementation

Bignums could be represented by the structure:

struct bignum {
    void *buffer;
    UINTVAL length;
    INTVAL exponent;
    UINTVAL flags;
}

The similarity to the suggested bigint structure is not accidental; we want upgrading from bigint to bignum to be quick.

String data types

Parrot has a single internal string form:

struct parrot_string_t {
    pobj_t obj;
    UINTVAL bufused;
    void *strstart;
    UINTVAL strlen;
    const ENCODING *encoding;
    const CHARTYPE *type;
    INTVAL language;
}

The fields are:

obj

A pointer to a Parrot object, Parrot's most general internal data type. In this case, it holds the buffer for the string data, the size of the buffer in bytes, and any applicable flags.

bufused

The amount of the buffer currently in use, in bytes.

strstart

A pointer to the beginning of the actual string (which may not be positioned at the start of the buffer).

strlen

The length of the string, in characters.

encoding

How the data is encoded (e.g. fixed 8-bit characters, UTF-8, or UTF-32). Note that this specifies encoding only -- it's valid to encode EBCDIC characters with the UTF-8 algorithm. Silly, but valid.

The ENCODING structure specifies the encoding (by index number and by name, for ease of lookup), the maximum number of bytes that a single character will occupy in that encoding, as well as functions for manipulating strings with that encoding.

type

What sort of string data is in the buffer, for example ASCII, EBCDIC, or Unicode.

The CHARTYPE structure specifies the character type (by index number and by name) and provides functions for transcoding to and from that character type.

language

This specifies the language corresponding to the string. This is to allow for locale-based data to be attached to strings. To give an example of the use of this: strings in German may not sort in the same way as strings in French, even when both types use the Latin-1 charset and are encoded in UTF-8.

Note that language-agnostic utilities are at liberty to ignore this entry.

Parrot Magic Cookies (PMCs)

Parrot Magic Cookies, or PMCs, are the last of Parrot's basic datatypes, but are also potentially the most important. Their basic structure is as follows. All PMCs have the form:

   struct PMC {
       pobj_t obj;
       VTABLE *vtable;
#if ! PMC_DATA_IN_EXT
       DPOINTER *data;
#endif
       struct PMC_EXT *pmc_ext;
   };

where obj is a pointer to an pobj_t structure:

   typedef struct pobj_t {
       UnionVal u;
       Parrot_UInt flags;
#if ! DISABLE_GC_DEBUG
       UINTVAL _pobj_version;
#endif
   } pobj_t;

and where:

typedef union UnionVal {
    struct {
        void * _bufstart;
        size_t _buflen;
    } _b;
    struct {
        DPOINTER* _struct_val;
        PMC* _pmc_val;
    } _ptrs;
    INTVAL _int_val;
    FLOATVAL _num_val;
    struct parrot_string_t * _string_val;
} UnionVal;

u holds data associated with the PMC. This can be in the form of an integer value, a floating point value, a string value, or a pointer to other data. u may be empty, since the PMC structure also provides a more general data pointer, but is useful for PMCs which hold only a single piece of data (e.g. PerlInts).

flags holds a set of flags associated with the PMC; these are documented in include/parrot/pobj.h, and are generally only used within the Parrot internals.

_pobj_version is only used for debugging Parrot's garbage collector. It is documented elsewhere (well, it will be once we get around to doing that...).

vtable holds a pointer to the vtable associated with the PMC. This points to a set of functions, with interfaces described in docs/pdds/pdd02_vtables.pod that implement the basic behaviour of the PMC (i.e. how it behaves under addition, subtraction, cloning etc.)

data (if present) holds a pointer to any additional data associated with the PMC. This may be NULL.

pmc_ext points to an extended PMC structure. This has the form:

   struct PMC_EXT {
#if PMC_DATA_IN_EXT
       DPOINTER *data;
#endif
       PMC *_metadata;
       struct _Sync *_synchronize;
       PMC *_next_for_GC;
   };

data is a generic data pointer, as described above.

_metadata holds internal PMC metadata. The specification for this has not yet been finalized.

XXX: what does _synchronize do?

XXX: ditto _next_for_GC...

PMCs are not required to have a PMC_EXT structure (i.e. pmc_ext can be null).

PMCs are used to implement the basic data types of the high level languages running on top of Parrot. For instance, a Perl 5 SV will map onto one (or more) types of PMC, while particular Python datatypes will map onto different types of PMC.

ATTACHMENTS

None.

REFERENCES

The perl modules Math::BigInt and Math::BigFloat. Alex Gough's suggestions for bigint/bignum implementation, outlined in docs/pdds/pdd14_bignum.pod. The Unicode standard at http://www.unicode.org.

GLOSSARY

Type

Type refers to a basic Parrot data type. There are four such: integers, floating point numbers (often just numbers), strings and Parrot Magic Cookies (PMCs).

VERSION

1.4

CURRENT

Maintainer: Dan Sugalski <dan@sidhe.org>
Class: Internals
PDD Number: 4
Version: 1.4
Status: Developing
Last Modified: 20 February 2004
PDD Format: 1
Language: English

HISTORY

Version 1.4, 20 February 2004
Version 1.3, 2 July 2001
Version 1.2, 2 July 2001
Version 1.1, 2 March 2001
Version 1, 1 March 2001

CHANGES

Version 1.4

Document basic PMC internals. Make clear the fact that the bigint/bignum description is still provisional. Other minor fixups to make the documentation match reality.

Version 1.3

Fixed some silly typos and dropped phrases.

Took all the underscores out of the field names.

Version 1.2

The string header format has changed some to allow for type tagging. The flags information for strings has changed as well.

Version 1.1

INT and NUM are now concepts rather than data structures, as making them data structures was a Bad Idea.

Version 1

None. First version