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
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
|
Smart pointers in aptitude
aptitude uses smart pointers in a number of ways internally. These
techniques generally simplify the code and make it easy to program
safely, but C++ being what it is, you should be aware of how they work
and what the caveats are.
* threading
As of this writing, none of the smart pointers lock their reference
counts (implementing locks in the imm::wtree class caused the
problem resolver's running time to increase by 50%!). Some of the
less-commonly-used ones may get locking in the future, but for the
time being you should handle objects that are given to another
thread carefully -- do deep copies of anything that's
reference-counted. (this is only the imm::* stuff at the moment, so
the problem is managable) For more information on threading, see
README.THREADS.
* reference-counted immutable values
aptitude employs reference-counting of immutable objects in several
places. For the most part this is invisible to clients; the typical
implementation looks like this:
class foo
{
class foo_impl
{
...
public:
void m1();
void m2(int);
...
};
public:
void m1();
void m2(int);
...
};
When you create a new "foo" object, a corresponding foo_impl is
created with reference-count 1. Reference counts are managed
in all the ways you expect, and the object is deleted when you're
done with it.
This use of reference-counting can be viewed as a performance hack:
since the objects are immutable, pass-by-value is indistinguishable
from pass-by-reference; if we can prove that no strong cycles will
be created, it's always safe to reference-count objects like this.
* auto_ptr for temporary return values
When building up a temporary return value, several routines use
auto_ptrs to store intermediates:
auto_ptr<T> v1 = parse_T_from_string(str1);
auto_ptr<T> v2 = parse_T_from_string(str2);
The advantage here is that you can freely throw exceptions and/or
break out of the function with an early "return" without worrying
about which of {v1,v2} have to be deleted. To actually return them,
you'd do something like this:
return pair<T*,T*>(v1.release(), v2.release());
The "release" calls tell v1 and v2 that they no longer "own" their
respective pointers.
* reference-counting of display widgets
Display widgets are also reference-counted. This is a
generalization of the idea that "the enclosing widget owns this
one"; it provides more support for extending the lifetime of a
widget if, for instance, you want to extract values from it after
its parent is done with it. Unfortunately, this reference counting
is NOT fully transparent, and you should be aware of some basic
principles:
- As with any reference counting scheme, it is up to you to
avoid creating strong cycles. By default, the only
non-transient strong references are references from a parent to
a child; your best bet is to ensure that all additional strong
references either have stack lifetime or do not close cycles.
If you must create a cycle, make sure it gets explicitly broken
before all external references to it are lost.
- Reference-counting is done via the generic ref_ptr class
(ref_ptr.h). This is a templated class, like auto_ptr, and
generally works as you expect. To enforce the use of ref_ptr,
all widget constructors are protected; static ::create methods
are provided to actually allocate a new widget. Because the
vscreen_widget class initializes its reference count to 1, the
::create method should explicitly decref() its return value; see
the existing ::create routines for examples.
To make code a bit more readable, adopt the convention of
creating typedefs for ref_ptr wrappers around new classes. For
instance, if you have just created the class vs_moo, add the
following line to its header file:
typedef ref_ptr<vs_moo> vs_moo_ref;
- Watch out for deletion of "this". If you aren't sure whether
"this" will be deleted in a method, I recommend creating a
method-scoped strong reference to it:
ref_ptr<this_type> thisref = this;
Doing so will prevent "this" from being deleted until the
current method terminates and is probably good practice in
general.
- Beware sigc::bind. sigc::bind is an easy way to create bad
circularities; moreover, it's actually unsafe to bind a ref_ptr
as a slot argument. The solution adopted in vscreen/ is to
exploit sigc++ weak references. If w is a ref_ptr, then rather
than closing w, you should close over w.weak_ref().
Unfortunately, w.weak_ref() will appear to the callee as a C++
reference, not a ref_ptr: if w has type T, you need a slot that
accepts a T&, not a ref_ptr<T>. To solve this problem, it is
conventional for widgets defining public interfaces that accept
a ref_ptr<T> to define a corresponding _bare method that accepts
a T&; the _bare method should simply instantiate a ref_ptr and
call the main interface. For instance,
void add_widget(const ref_ptr<vscreen_widget> &w);
void add_widget_bare(vscreen_widget &w)
{
add_widget(ref_ptr<vscreen_widget>(&w));
}
Obviously this is less than ideal, but it will work. Be aware,
though, that the bound argument is a *weak* reference: if there
are no strong references to a bound widget, the signal
connection will simply disappear.
|
