summaryrefslogtreecommitdiff
path: root/pod/perlmroapi.pod
blob: 54da7dded254e77a7f7536197984154ae40cbbe5 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
=head1 NAME

perlmroapi - Perl method resolution plugin interface

=head1 DESCRIPTION

As of Perl 5.10.1 there is a new interface for plugging and using method
resolution orders other than the default (linear depth first search).
The C3 method resolution order added in 5.10.0 has been re-implemented as
a plugin, without changing its Perl-space interface.

Each plugin should register itself by providing
the following structure

    struct mro_alg {
        AV *(*resolve)(pTHX_ HV *stash, U32 level);
        const char *name;
        U16 length;
        U16 kflags;
        U32 hash;
    };

and calling C<Perl_mro_register>:

    Perl_mro_register(aTHX_ &my_mro_alg);

=over 4

=item resolve

Pointer to the linearisation function, described below.

=item name

Name of the MRO, either in ISO-8859-1 or UTF-8.

=item length

Length of the name.

=item kflags

If the name is given in UTF-8, set this to C<HVhek_UTF8>. The value is passed
direct as the parameter I<kflags> to C<hv_common()>.

=item hash

A precomputed hash value for the MRO's name, or 0.

=back

=head1 Callbacks

The C<resolve> function is called to generate a linearised ISA for the
given stash, using this MRO. It is called with a pointer to the stash, and
a I<level> of 0. The core always sets I<level> to 0 when it calls your
function - the parameter is provided to allow your implementation to track
depth if it needs to recurse.

The function should return a reference to an array containing the parent
classes in order. The names of the classes should be the result of calling
C<HvENAME()> on the stash. In those cases where C<HvENAME()> returns null,
C<HvNAME()> should be used instead.

The caller is responsible for incrementing the reference count of the array
returned if it wants to keep the structure. Hence, if you have created a
temporary value that you keep no pointer to, C<sv_2mortal()> to ensure that
it is disposed of correctly. If you have cached your return value, then
return a pointer to it without changing the reference count.

=head1 Caching

Computing MROs can be expensive. The implementation provides a cache, in
which you can store a single C<SV *>, or anything that can be cast to
C<SV *>, such as C<AV *>. To read your private value, use the macro
C<MRO_GET_PRIVATE_DATA()>, passing it the C<mro_meta> structure from the
stash, and a pointer to your C<mro_alg> structure:

    meta = HvMROMETA(stash);
    private_sv = MRO_GET_PRIVATE_DATA(meta, &my_mro_alg);

To set your private value, call C<Perl_mro_set_private_data()>:

    Perl_mro_set_private_data(aTHX_ meta, &c3_alg, private_sv);

The private data cache will take ownership of a reference to private_sv,
much the same way that C<hv_store()> takes ownership of a reference to the
value that you pass it.

=head1 Examples

For examples of MRO implementations, see C<S_mro_get_linear_isa_c3()>
and the C<BOOT:> section of F<mro/mro.xs>, and C<S_mro_get_linear_isa_dfs()>
in F<mro.c>

=head1 AUTHORS

The implementation of the C3 MRO and switchable MROs within the perl core was
written by Brandon L Black. Nicholas Clark created the pluggable interface, 
refactored Brandon's implementation to work with it, and wrote this document.

=cut