CONCEPT
structs
INTRODUCTION
structs are, next to arrays and mappings, a way to group a
collection of value together.
A struct holds a fixed number of values, called 'members', and
allows to access them by their given name. The name is resolved
when the LPC code is compiled, making struct member access as fast
as array member access.
structs are passed by reference.
DEFINITION
A new struct type has to be defined at the top level of an
object. For example
struct Foo {
int one, *two;
struct Bar three;
};
defines the new struct 'Foo' with three members: integer 'one',
integer array 'two', and struct Bar 'three'
It is possible to 'inherit' structs from each other. Given above
definition of struct Foo, the following definition
struct Quux (Foo) {
int four;
};
is equivalent to the definition
struct Quux {
int one, *two;
struct Bar three;
int four;
};
The usual visibility modifiers apply, e.g.
protected struct Bang {...};
struct definitions are promoted through inheritance like functions,
with the difference that all structs live in the same flat namespace.
This means: a struct defined in a program is visible in _all_
inherited programs, regardless of how deep the inheritance is
nested. This also means that in one program there must not be
two structs, inherited or not, with the same name.
To declare a struct without defining it, write:
struct Quux;
This notation is useful if you have two structs referencing
each other:
struct Quux;
struct Bar {
struct Quux quux;
};
struct Quux {
struct Bar bar;
};
USAGE
To use a struct, its definition must be visible - either because it
is defined in the object compiled, or it has been inherited.
(Note: #include'ing structs does not what you think it does: in
LPC it constructs a new struct type whereever it is included).
A variable to hold a struct is defined like this:
struct Foo var;
and similar for function arguments:
void fun (struct Foo arg)
Just writing 'struct Foo var' however does not _create_ a struct,
it just creates a variable capable of holding one. To assign a value
to the variable upon creation, assign it with a struct value, either
from another variable or from a literal struct:
struct Foo var = (<Foo>);
Literal structs are written using (<>) as delimiters:
(<Foo>)
creates an empty instance of struct Foo
(<Foo> 1, ({ 2 }), bar)
creates an instance of struct Foo, and assigns 1 to member
'one', ({ 2 }) to member 'two', and the content of variable
bar to member 'three'.
(<Foo> two: ({ 2 }) )
creates an instance of struct Foo which is all empty except
for member 'two' which is assigned the value ({ 2 }).
It is not possible to use both named and unnamed initializers
in the same literal.
A struct member is accessed using the . and -> operators:
struct Foo var = ...;
var.one = 1;
var->one = 1;
It is possible to compute struct lookups at runtime:
struct Foo bar = ...;
string member = "one";
bar.(member) = 1; // Sets bar.one to 1
bar.(0) = 1; // Sets bar.one to 1
bar->(member) = 1; // sets bar->one to 1
bar->(0) = 1; // sets bar->one to 1
If the given member does not exist, the -> operator will return 0,
but the . operator will throw an error.
USAGE IN CLOSURES
The #'(< operator can be used in lambda closures to create a
struct; the type of the struct is given by the 'template'
struct passed as first argument. The content of the template
struct is irrelevant, so an empty struct suffices. For
example, to create an instance of struct Foo:
({ #'(<, (<Foo>), 1, ({ 2 }), (<Bar>) })
The order of the member values is the order in which they
appear in the struct definition.
To access a struct member in a lambda closure, use the #'->
operator with the name of the member as double-quoted symbol
or literal string:
({ #'->, struct-expression, ''one })
({ #'->, struct-expression, "one" })
MISCELLANEOUS
Internally structs can be identified by the ID string
returned from get_type_info(). This string contains the name
of the struct, the name of the program its type was defined in,
and the ID number of the program. However, do not rely on
a particular format of this string!
Support for structs is signaled by the macro __LPC_STRUCTS__.
Though structs are tied to the program the are defined in,
re-compiling a program doesn't make the struct types
incompatible. Even if the newly compiled struct has a
different structure it will be accepted by routines that
expect the old struct definition. When members disappeared
in the new struct definition, read access to those members
will return 0, write access to vanished members however
will result in a runtime error.
EXAMPLES
Suppose we have two objects: a monster, and a monster
coordinate tracker, and we want to use a struct to store the
coordinate:
-- monster_coordinate.c --
struct Coordinate { int x; int y; };
-- monster_tracker.c --
inherit "monster_coordinate";
void track (struct Coordinate coord) { ... }
-- monster.c --
inherit "monster_coordinate";
int move (..) {
...
"monster_tracker"->track( (<Coordinate> my_x, my_y) );
}
Note that using '#include "monster_coordinate.c"' instead of inherit
won't work. While the objects would compile, the first call to
track() would cause a runtime error of the type
Illegal type to struct->(): struct Coordinate (/monster.c #234),
expected struct Coordinate
(/monster_tracker.c #552)
HISTORY
structs were fully implemented first in LDMud 3.3.246.
The implementation was revised in LDMud 3.3.344.
The reactivation of unchanged structs in object updates was
implemented in LDMud 3.3.417.
In LDMud 3.6.2 the . operator for struct member access was introduced
and the -> operator changed into a relaxed access operation.
SEE ALSO
mappings(LPC), get_type_info(E), structp(E), to_mapping(E),
to_struct(E), struct_info(E), baseof(E)
|
