rice2 2.2.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +7 -0
- data/COPYING +23 -0
- data/Doxyfile +2268 -0
- data/Makefile.am +26 -0
- data/Makefile.in +929 -0
- data/README.md +1054 -0
- data/README.mingw +8 -0
- data/Rakefile +24 -0
- data/aclocal.m4 +1090 -0
- data/bootstrap +8 -0
- data/config.guess +1667 -0
- data/config.sub +1793 -0
- data/configure +8209 -0
- data/configure.ac +55 -0
- data/depcomp +791 -0
- data/doxygen.ac +314 -0
- data/doxygen.am +186 -0
- data/extconf.rb +66 -0
- data/install-sh +529 -0
- data/missing +215 -0
- data/post-autoconf.rb +22 -0
- data/post-automake.rb +28 -0
- data/rice/Address_Registration_Guard.cpp +22 -0
- data/rice/Address_Registration_Guard.hpp +7 -0
- data/rice/Address_Registration_Guard.ipp +37 -0
- data/rice/Address_Registration_Guard_defn.hpp +75 -0
- data/rice/Arg.hpp +8 -0
- data/rice/Arg_impl.hpp +129 -0
- data/rice/Arg_operators.cpp +21 -0
- data/rice/Arg_operators.hpp +19 -0
- data/rice/Array.hpp +214 -0
- data/rice/Array.ipp +256 -0
- data/rice/Builtin_Object.hpp +8 -0
- data/rice/Builtin_Object.ipp +50 -0
- data/rice/Builtin_Object_defn.hpp +50 -0
- data/rice/Class.cpp +57 -0
- data/rice/Class.hpp +8 -0
- data/rice/Class.ipp +6 -0
- data/rice/Class_defn.hpp +83 -0
- data/rice/Constructor.hpp +47 -0
- data/rice/Data_Object.hpp +8 -0
- data/rice/Data_Object.ipp +133 -0
- data/rice/Data_Object_defn.hpp +138 -0
- data/rice/Data_Type.cpp +54 -0
- data/rice/Data_Type.hpp +8 -0
- data/rice/Data_Type.ipp +365 -0
- data/rice/Data_Type_defn.hpp +261 -0
- data/rice/Data_Type_fwd.hpp +12 -0
- data/rice/Director.cpp +13 -0
- data/rice/Director.hpp +39 -0
- data/rice/Enum.hpp +117 -0
- data/rice/Enum.ipp +246 -0
- data/rice/Exception.cpp +59 -0
- data/rice/Exception.hpp +13 -0
- data/rice/Exception_Base.hpp +8 -0
- data/rice/Exception_Base.ipp +13 -0
- data/rice/Exception_Base_defn.hpp +27 -0
- data/rice/Exception_defn.hpp +69 -0
- data/rice/Hash.hpp +227 -0
- data/rice/Hash.ipp +329 -0
- data/rice/Identifier.cpp +8 -0
- data/rice/Identifier.hpp +50 -0
- data/rice/Identifier.ipp +33 -0
- data/rice/Jump_Tag.hpp +24 -0
- data/rice/Makefile.am +122 -0
- data/rice/Makefile.in +885 -0
- data/rice/Module.cpp +84 -0
- data/rice/Module.hpp +8 -0
- data/rice/Module.ipp +6 -0
- data/rice/Module_defn.hpp +88 -0
- data/rice/Module_impl.hpp +281 -0
- data/rice/Module_impl.ipp +345 -0
- data/rice/Object.cpp +169 -0
- data/rice/Object.hpp +8 -0
- data/rice/Object.ipp +33 -0
- data/rice/Object_defn.hpp +214 -0
- data/rice/Require_Guard.hpp +21 -0
- data/rice/String.cpp +94 -0
- data/rice/String.hpp +91 -0
- data/rice/Struct.cpp +117 -0
- data/rice/Struct.hpp +162 -0
- data/rice/Struct.ipp +26 -0
- data/rice/Symbol.cpp +25 -0
- data/rice/Symbol.hpp +66 -0
- data/rice/Symbol.ipp +44 -0
- data/rice/config.hpp +47 -0
- data/rice/config.hpp.in +46 -0
- data/rice/detail/Arguments.hpp +118 -0
- data/rice/detail/Auto_Function_Wrapper.hpp +898 -0
- data/rice/detail/Auto_Function_Wrapper.ipp +3694 -0
- data/rice/detail/Auto_Member_Function_Wrapper.hpp +897 -0
- data/rice/detail/Auto_Member_Function_Wrapper.ipp +2774 -0
- data/rice/detail/Caster.hpp +103 -0
- data/rice/detail/Exception_Handler.hpp +8 -0
- data/rice/detail/Exception_Handler.ipp +68 -0
- data/rice/detail/Exception_Handler_defn.hpp +96 -0
- data/rice/detail/Iterator.hpp +93 -0
- data/rice/detail/Not_Copyable.hpp +25 -0
- data/rice/detail/Wrapped_Function.hpp +33 -0
- data/rice/detail/cfp.hpp +24 -0
- data/rice/detail/cfp.ipp +51 -0
- data/rice/detail/check_ruby_type.cpp +27 -0
- data/rice/detail/check_ruby_type.hpp +23 -0
- data/rice/detail/creation_funcs.hpp +37 -0
- data/rice/detail/creation_funcs.ipp +36 -0
- data/rice/detail/default_allocation_func.hpp +23 -0
- data/rice/detail/default_allocation_func.ipp +11 -0
- data/rice/detail/define_method_and_auto_wrap.hpp +31 -0
- data/rice/detail/define_method_and_auto_wrap.ipp +30 -0
- data/rice/detail/demangle.cpp +56 -0
- data/rice/detail/demangle.hpp +19 -0
- data/rice/detail/env.hpp +11 -0
- data/rice/detail/from_ruby.hpp +43 -0
- data/rice/detail/from_ruby.ipp +60 -0
- data/rice/detail/method_data.cpp +92 -0
- data/rice/detail/method_data.hpp +21 -0
- data/rice/detail/node.hpp +13 -0
- data/rice/detail/protect.cpp +29 -0
- data/rice/detail/protect.hpp +34 -0
- data/rice/detail/ruby.hpp +74 -0
- data/rice/detail/ruby_version_code.hpp +6 -0
- data/rice/detail/ruby_version_code.hpp.in +6 -0
- data/rice/detail/st.hpp +22 -0
- data/rice/detail/to_ruby.hpp +22 -0
- data/rice/detail/to_ruby.ipp +36 -0
- data/rice/detail/traits.hpp +43 -0
- data/rice/detail/win32.hpp +16 -0
- data/rice/detail/wrap_function.hpp +66 -0
- data/rice/global_function.hpp +33 -0
- data/rice/global_function.ipp +22 -0
- data/rice/protect.hpp +38 -0
- data/rice/protect.ipp +1134 -0
- data/rice/ruby_mark.hpp +13 -0
- data/rice/ruby_try_catch.hpp +86 -0
- data/rice/rubypp.rb +97 -0
- data/rice/to_from_ruby.hpp +8 -0
- data/rice/to_from_ruby.ipp +418 -0
- data/rice/to_from_ruby_defn.hpp +70 -0
- data/ruby.ac +135 -0
- data/ruby/Makefile.am +1 -0
- data/ruby/Makefile.in +628 -0
- data/ruby/lib/Makefile.am +3 -0
- data/ruby/lib/Makefile.in +506 -0
- data/ruby/lib/mkmf-rice.rb.in +217 -0
- data/ruby/lib/version.rb +3 -0
- data/sample/Makefile.am +47 -0
- data/sample/Makefile.in +489 -0
- data/sample/enum/extconf.rb +3 -0
- data/sample/enum/sample_enum.cpp +54 -0
- data/sample/enum/test.rb +8 -0
- data/sample/inheritance/animals.cpp +98 -0
- data/sample/inheritance/extconf.rb +3 -0
- data/sample/inheritance/test.rb +7 -0
- data/sample/map/extconf.rb +3 -0
- data/sample/map/map.cpp +81 -0
- data/sample/map/test.rb +7 -0
- data/test/Makefile.am +72 -0
- data/test/Makefile.in +1213 -0
- data/test/ext/Makefile.am +41 -0
- data/test/ext/Makefile.in +483 -0
- data/test/ext/t1/Foo.hpp +10 -0
- data/test/ext/t1/extconf.rb +2 -0
- data/test/ext/t1/t1.cpp +15 -0
- data/test/ext/t2/extconf.rb +2 -0
- data/test/ext/t2/t2.cpp +11 -0
- data/test/test_Address_Registration_Guard.cpp +43 -0
- data/test/test_Array.cpp +248 -0
- data/test/test_Builtin_Object.cpp +71 -0
- data/test/test_Class.cpp +496 -0
- data/test/test_Constructor.cpp +128 -0
- data/test/test_Data_Object.cpp +275 -0
- data/test/test_Data_Type.cpp +348 -0
- data/test/test_Director.cpp +308 -0
- data/test/test_Enum.cpp +215 -0
- data/test/test_Exception.cpp +47 -0
- data/test/test_Hash.cpp +212 -0
- data/test/test_Identifier.cpp +70 -0
- data/test/test_Jump_Tag.cpp +17 -0
- data/test/test_Memory_Management.cpp +50 -0
- data/test/test_Module.cpp +497 -0
- data/test/test_Object.cpp +159 -0
- data/test/test_String.cpp +107 -0
- data/test/test_Struct.cpp +205 -0
- data/test/test_Symbol.cpp +63 -0
- data/test/test_To_From_Ruby.cpp +428 -0
- data/test/test_global_functions.cpp +114 -0
- data/test/test_rice.rb +41 -0
- data/test/unittest.cpp +136 -0
- data/test/unittest.hpp +294 -0
- metadata +297 -0
data/README.md
ADDED
@@ -0,0 +1,1054 @@
|
|
1
|
+
# Rice - Ruby Interface for C++ Extensions {#mainpage}
|
2
|
+
|
3
|
+
# Introduction {#intro}
|
4
|
+
|
5
|
+
Rice is a C++ interface to Ruby's C API. It provides a type-safe and
|
6
|
+
exception-safe interface in order to make embedding Ruby and writing
|
7
|
+
Ruby extensions with C++ easier. It is similar to Boost.Python in many
|
8
|
+
ways, but also attempts to provide an object-oriented interface to all
|
9
|
+
of the Ruby C API.
|
10
|
+
|
11
|
+
What Rice gives you:
|
12
|
+
- A simple C++-based syntax for wrapping and defining classes
|
13
|
+
- Automatic conversion of exceptions between C++ and Ruby
|
14
|
+
- Smart pointers for handling garbage collection
|
15
|
+
- Wrappers for most builtin types to simplify calling code
|
16
|
+
|
17
|
+
# Project Details {#project}
|
18
|
+
|
19
|
+
The source is hosted on github: http://github.com/jasonroelofs/rice
|
20
|
+
|
21
|
+
Bug tracking: http://github.com/jasonroelofs/rice/issues
|
22
|
+
|
23
|
+
API documentation: http://jasonroelofs.github.io/rice
|
24
|
+
|
25
|
+
# Installation {#installation}
|
26
|
+
|
27
|
+
~~~
|
28
|
+
gem install rice
|
29
|
+
~~~
|
30
|
+
|
31
|
+
Building it locally from a clone of the repository is as follows:
|
32
|
+
|
33
|
+
~~~
|
34
|
+
./bootstrap
|
35
|
+
ruby extconf.rb
|
36
|
+
make
|
37
|
+
~~~
|
38
|
+
|
39
|
+
Rice is known to work on *nix and OSX. Windows is not currently
|
40
|
+
supported.
|
41
|
+
|
42
|
+
Rice requires a C++ compiler with support for C++14 or later.
|
43
|
+
|
44
|
+
# Tutorial {#tutorial}
|
45
|
+
|
46
|
+
## Getting started {#getting_started}
|
47
|
+
|
48
|
+
Writing an extension with Rice is very similar to writing an extension
|
49
|
+
with the C API.
|
50
|
+
|
51
|
+
The first step is to create an extconf.rb file:
|
52
|
+
|
53
|
+
~~~{.cpp}
|
54
|
+
require 'mkmf-rice'
|
55
|
+
create_makefile('test')
|
56
|
+
~~~
|
57
|
+
|
58
|
+
Note that we use `mkmf-rice` instead of `mkmf`. This will ensure that the
|
59
|
+
extension will be linked with standard C++ library along with the Rice
|
60
|
+
library, and allow access to the Rice header files.
|
61
|
+
|
62
|
+
Next we create our extension and save it to test.cpp:
|
63
|
+
|
64
|
+
~~~{.cpp}
|
65
|
+
extern "C"
|
66
|
+
void Init_test()
|
67
|
+
{
|
68
|
+
}
|
69
|
+
~~~
|
70
|
+
|
71
|
+
Note the extern "C" line above. This tells the compiler that the
|
72
|
+
function `Init_test` should have C linkage and calling convention. This
|
73
|
+
turns off name mangling so that the Ruby interpreter will be able to
|
74
|
+
find the function (remember that Ruby is written in C, not C++).
|
75
|
+
|
76
|
+
So far we haven't put anything into the extension, so it isn't
|
77
|
+
particularly useful. The next step is to define a class so we can add
|
78
|
+
methods to it.
|
79
|
+
|
80
|
+
|
81
|
+
## Defining clases {#classes}
|
82
|
+
|
83
|
+
Defining a class in Rice is easy:
|
84
|
+
|
85
|
+
~~~{.cpp}
|
86
|
+
#include "rice/Class.hpp"
|
87
|
+
|
88
|
+
using namespace Rice;
|
89
|
+
|
90
|
+
extern "C"
|
91
|
+
void Init_test()
|
92
|
+
{
|
93
|
+
Class rb_cTest = define_class("Test");
|
94
|
+
}
|
95
|
+
~~~
|
96
|
+
|
97
|
+
This will create a class called `Test` that inherits from `Object`. If we
|
98
|
+
wanted to inherit from a different class, we could easily do so:
|
99
|
+
|
100
|
+
~~~{.cpp}
|
101
|
+
#include "rice/Class.hpp"
|
102
|
+
|
103
|
+
using namespace Rice;
|
104
|
+
|
105
|
+
extern "C"
|
106
|
+
void Init_test()
|
107
|
+
{
|
108
|
+
Class rb_cMySocket = define_class("MySocket", rb_cIO);
|
109
|
+
}
|
110
|
+
~~~
|
111
|
+
|
112
|
+
Note the prefix rb_c on the name of the class. This is a convention
|
113
|
+
that the Ruby interpreter and many extensions tend to use. It signifies
|
114
|
+
that this is a class and not some other type of object. Some other
|
115
|
+
naming conventions that are commonly used:
|
116
|
+
|
117
|
+
- rb_c variable name prefix for a Class
|
118
|
+
- rb_m variable name prefix for a Module
|
119
|
+
- rb_e variable name prefix for an Exception type
|
120
|
+
- rb_ function prefix for a function in the Ruby C API
|
121
|
+
- rb_f_ function prefix to differentiate between an API function that
|
122
|
+
takes Ruby objects as arguments and one that takes C argument types
|
123
|
+
- rb_*_s_ indicates the function is a singleton function
|
124
|
+
- *_m suffix to indicate the function takes variable number of
|
125
|
+
arguments
|
126
|
+
|
127
|
+
|
128
|
+
Also note that we don't include "ruby.h" directly. Rice has a wrapper
|
129
|
+
for ruby.h that handles some compatibility issues across platforms and
|
130
|
+
Ruby versions. Always include Rice headers before including anything
|
131
|
+
that might include "ruby.h".
|
132
|
+
|
133
|
+
## Defining methods {#methods}
|
134
|
+
|
135
|
+
Now let's add a method to our class:
|
136
|
+
|
137
|
+
~~~{.cpp}
|
138
|
+
#include "rice/Class.hpp"
|
139
|
+
#include "rice/String.hpp"
|
140
|
+
|
141
|
+
using namespace Rice;
|
142
|
+
|
143
|
+
Object test_hello(Object /* self */)
|
144
|
+
{
|
145
|
+
String str("hello, world");
|
146
|
+
return str;
|
147
|
+
}
|
148
|
+
|
149
|
+
extern "C"
|
150
|
+
void Init_test()
|
151
|
+
{
|
152
|
+
Class rb_cTest =
|
153
|
+
define_class("Test")
|
154
|
+
.define_method("hello", &test_hello);
|
155
|
+
}
|
156
|
+
~~~
|
157
|
+
|
158
|
+
Here we add a method `%Test#hello` that simply returns the string
|
159
|
+
"Hello, World". The method takes self as an implicit parameter, but
|
160
|
+
isn't used, so we comment it out to prevent a compiler warning.
|
161
|
+
|
162
|
+
We could also add an `#initialize` method to our class:
|
163
|
+
|
164
|
+
~~~{.cpp}
|
165
|
+
#include "rice/Class.hpp"
|
166
|
+
#include "rice/String.hpp"
|
167
|
+
|
168
|
+
using namespace Rice;
|
169
|
+
|
170
|
+
Object test_initialize(Object self)
|
171
|
+
{
|
172
|
+
self.iv_set("@foo", 42);
|
173
|
+
}
|
174
|
+
|
175
|
+
Object test_hello(Object /* self */)
|
176
|
+
{
|
177
|
+
String str("hello, world");
|
178
|
+
return str;
|
179
|
+
}
|
180
|
+
|
181
|
+
extern "C"
|
182
|
+
void Init_test()
|
183
|
+
{
|
184
|
+
Class rb_cTest =
|
185
|
+
define_class("Test")
|
186
|
+
.define_method("initialize", &test_initialize)
|
187
|
+
.define_method("hello", &test_hello);
|
188
|
+
}
|
189
|
+
~~~
|
190
|
+
|
191
|
+
The `initialize` method sets an instance variable `@foo` to the value 42.
|
192
|
+
The number is automatically converted to a `Fixnum` before doing the
|
193
|
+
assignment.
|
194
|
+
|
195
|
+
Note that we're chaining calls on the `Class` object. Most member
|
196
|
+
functions in `Module` and `Class` return a reference to self, so we can
|
197
|
+
chain as many calls as we want to define as many methods as we want.
|
198
|
+
|
199
|
+
|
200
|
+
## Wrapping C++ Types {#data_types}
|
201
|
+
|
202
|
+
It's useful to be able to define Ruby classes in a C++ style rather than
|
203
|
+
using the Ruby API directly, but the real power Rice is in wrapping
|
204
|
+
already-defined C++ types.
|
205
|
+
|
206
|
+
Let's assume we have the following C++ class that we want to wrap:
|
207
|
+
|
208
|
+
~~~{.cpp}
|
209
|
+
class Test
|
210
|
+
{
|
211
|
+
public:
|
212
|
+
Test();
|
213
|
+
std::string hello();
|
214
|
+
};
|
215
|
+
~~~
|
216
|
+
|
217
|
+
This is a C++ version of the Ruby class we just created in the previous
|
218
|
+
section. To wrap it:
|
219
|
+
|
220
|
+
~~~{.cpp}
|
221
|
+
#include "rice/Data_Type.hpp"
|
222
|
+
#include "rice/Constructor.hpp"
|
223
|
+
|
224
|
+
using namespace Rice;
|
225
|
+
|
226
|
+
extern "C"
|
227
|
+
void Init_test()
|
228
|
+
{
|
229
|
+
Data_Type<Test> rb_cTest =
|
230
|
+
define_class<Test>("Test")
|
231
|
+
.define_constructor(Constructor<Test>())
|
232
|
+
.define_method("hello", &Test::hello);
|
233
|
+
}
|
234
|
+
~~~
|
235
|
+
|
236
|
+
This example is similar to the one before, but we use `Data_Type<>`
|
237
|
+
instead of `Class` and the template version of define_class() instead of
|
238
|
+
the non-template version. This creates a binding in the Rice library
|
239
|
+
between the Ruby class `Test` and the C++ class Test such that Rice passes
|
240
|
+
member function pointers to `define_method()`.
|
241
|
+
|
242
|
+
It is possible to write the conversion functions ourself (as we'll see
|
243
|
+
below), but Rice does all the dirty work for us.
|
244
|
+
|
245
|
+
|
246
|
+
## Type conversions {#conversions}
|
247
|
+
|
248
|
+
Let's look again at our example class:
|
249
|
+
|
250
|
+
~~~{.cpp}
|
251
|
+
class Test
|
252
|
+
{
|
253
|
+
public:
|
254
|
+
Test();
|
255
|
+
std::string hello();
|
256
|
+
};
|
257
|
+
~~~
|
258
|
+
|
259
|
+
When we wrote our class, we never wrote a single line of code to convert
|
260
|
+
the `std::string` returned by `hello()` into a Ruby type. Neverthless, the
|
261
|
+
conversion works, and when we write:
|
262
|
+
|
263
|
+
~~~{.cpp}
|
264
|
+
test = Test.new
|
265
|
+
puts test.hello
|
266
|
+
~~~
|
267
|
+
|
268
|
+
We get the expected result.
|
269
|
+
|
270
|
+
Rice has two template conversion functions to convert between C++ and
|
271
|
+
Ruby types:
|
272
|
+
|
273
|
+
~~~{.cpp}
|
274
|
+
template<typename T>
|
275
|
+
T from_ruby(Object x);
|
276
|
+
|
277
|
+
template<typename T>
|
278
|
+
Object to_ruby(T const & x);
|
279
|
+
~~~
|
280
|
+
|
281
|
+
Rice has included by default specializations for many of the builtin
|
282
|
+
types. To define your own conversion, you can write a specialization:
|
283
|
+
|
284
|
+
~~~{.cpp}
|
285
|
+
template<>
|
286
|
+
Foo from_ruby<Foo>(Object x)
|
287
|
+
{
|
288
|
+
// ...
|
289
|
+
}
|
290
|
+
|
291
|
+
template<>
|
292
|
+
Object to_ruby<Foo>(Foo const & x)
|
293
|
+
{
|
294
|
+
// ...
|
295
|
+
}
|
296
|
+
~~~
|
297
|
+
|
298
|
+
The implementation of these functions would, of course, depend on the
|
299
|
+
implementation of `Foo`.
|
300
|
+
|
301
|
+
|
302
|
+
## Conversions for wrapped C++ types {#data_conversions}
|
303
|
+
|
304
|
+
Take another look at the wrapper we wrote for the `Test` class:
|
305
|
+
|
306
|
+
~~~{.cpp}
|
307
|
+
extern "C"
|
308
|
+
void Init_test()
|
309
|
+
{
|
310
|
+
Data_Type<Test> rb_cTest =
|
311
|
+
define_class<Test>("Test")
|
312
|
+
.define_constructor(Constructor<Test>())
|
313
|
+
.define_method("hello", &Test::hello);
|
314
|
+
}
|
315
|
+
~~~
|
316
|
+
|
317
|
+
When we called `define_class<Test>`, it created a Class for us and
|
318
|
+
automatically registered the new Class with the type system, so that the
|
319
|
+
calls:
|
320
|
+
|
321
|
+
~~~{.cpp}
|
322
|
+
Data_Object<Foo> obj(new Foo);
|
323
|
+
Foo * f = from_ruby<Foo *>(obj);
|
324
|
+
Foo const * f = from_ruby<Foo const *>(obj);
|
325
|
+
~~~
|
326
|
+
|
327
|
+
works as expected.
|
328
|
+
|
329
|
+
The `Data_Object` class is a wrapper for the `Data_Wrap_Struct` and the
|
330
|
+
`Data_Get_Struct` macros in C extensions. It can be used to wrap or
|
331
|
+
unwrap any class that has been assigned to a `Data_Type`. It inherits
|
332
|
+
from `Object`, so any member functions we can call on an `Object` we can
|
333
|
+
also call on a `Data_Object`:
|
334
|
+
|
335
|
+
~~~{.cpp}
|
336
|
+
Object object_id = obj.call("object_id");
|
337
|
+
std::cout << object_id << std::endl;
|
338
|
+
~~~
|
339
|
+
|
340
|
+
The `Data_Object` class can be used to wrap a newly-created object:
|
341
|
+
|
342
|
+
~~~{.cpp}
|
343
|
+
Data_Object<Foo> foo(new Foo);
|
344
|
+
~~~
|
345
|
+
|
346
|
+
or to unwrap an already-created object:
|
347
|
+
|
348
|
+
~~~{.cpp}
|
349
|
+
VALUE obj = ...;
|
350
|
+
Data_Object<Foo> foo(obj);
|
351
|
+
~~~
|
352
|
+
|
353
|
+
A `Data_Object` functions like a smart pointer:
|
354
|
+
|
355
|
+
~~~{.cpp}
|
356
|
+
Data_Object<Foo> foo(obj);
|
357
|
+
foo->foo();
|
358
|
+
std::cout << *foo << std::endl;
|
359
|
+
~~~
|
360
|
+
|
361
|
+
Like a `VALUE` or an `Object`, data stored in a `Data_Object` will be marked
|
362
|
+
by the garbage collector as long as the `Data_Object` is on the stack.
|
363
|
+
|
364
|
+
|
365
|
+
## Exceptions {#exception}
|
366
|
+
|
367
|
+
Suppose we added a member function to our example class that throws an
|
368
|
+
exception:
|
369
|
+
|
370
|
+
~~~{.cpp}
|
371
|
+
class MyException
|
372
|
+
: public std::exception
|
373
|
+
{
|
374
|
+
};
|
375
|
+
|
376
|
+
class Test
|
377
|
+
{
|
378
|
+
public:
|
379
|
+
Test();
|
380
|
+
std::string hello();
|
381
|
+
void error();
|
382
|
+
};
|
383
|
+
~~~
|
384
|
+
|
385
|
+
If we were to wrap this function:
|
386
|
+
|
387
|
+
~~~{.cpp}
|
388
|
+
extern "C"
|
389
|
+
void Init_test()
|
390
|
+
{
|
391
|
+
Data_Type<Test> rb_cTest =
|
392
|
+
define_class<Test>("Test")
|
393
|
+
.define_constructor(Constructor<Test>())
|
394
|
+
.define_method("hello", &Test::hello)
|
395
|
+
.define_method("error", &Test::error);
|
396
|
+
}
|
397
|
+
~~~
|
398
|
+
|
399
|
+
and call it from inside Ruby:
|
400
|
+
|
401
|
+
~~~{.cpp}
|
402
|
+
test = Test.new
|
403
|
+
test.error()
|
404
|
+
~~~
|
405
|
+
|
406
|
+
we would get an exception. Rice will automatically convert any
|
407
|
+
C++ exception it catches into a Ruby exception. But what if we wanted
|
408
|
+
to use a custom error message when we convert the exception, or what if
|
409
|
+
we wanted to convert to a different type of exception? We can write
|
410
|
+
this:
|
411
|
+
|
412
|
+
~~~{.cpp}
|
413
|
+
extern "C"
|
414
|
+
void Init_test()
|
415
|
+
{
|
416
|
+
Data_Type<Test> rb_cTest =
|
417
|
+
define_class<Test>("Test")
|
418
|
+
.add_handler<MyException>(handle_my_exception)
|
419
|
+
.define_constructor(Constructor<Test>())
|
420
|
+
.define_method("hello", &Test::hello)
|
421
|
+
.define_method("error", &Test::error);
|
422
|
+
}
|
423
|
+
~~~
|
424
|
+
|
425
|
+
The `handle_my_exception` function need only rethrow the exception as a
|
426
|
+
`Rice::Exception`:
|
427
|
+
|
428
|
+
~~~{.cpp}
|
429
|
+
void handle_my_exception(MyException const & ex)
|
430
|
+
{
|
431
|
+
throw Exception(rb_eRuntimeError, "Goodnight, moon");
|
432
|
+
}
|
433
|
+
~~~
|
434
|
+
|
435
|
+
And what if we want to call Ruby code from C++? These exceptions are
|
436
|
+
also converted:
|
437
|
+
|
438
|
+
~~~{.cpp}
|
439
|
+
Object o;
|
440
|
+
o.call("some_function_that_raises", 42);
|
441
|
+
|
442
|
+
protect(rb_raise, rb_eRuntimeError, "some exception msg");
|
443
|
+
~~~
|
444
|
+
|
445
|
+
Internally whenever Rice catches a C++ or a Ruby exception, it converts
|
446
|
+
it to an `Exception` object. This object will later be re-raised as a
|
447
|
+
Ruby exception when control is returned to the Ruby VM.
|
448
|
+
|
449
|
+
Rice uses a similar class called `Jump_Tag` to handle symbols thrown by
|
450
|
+
Ruby's `throw`/`catch` or other non-local jumps from inside the Ruby VM.
|
451
|
+
|
452
|
+
|
453
|
+
## Builtin Types {#builtin}
|
454
|
+
|
455
|
+
You've seen this example:
|
456
|
+
|
457
|
+
~~~{.cpp}
|
458
|
+
Object object_id = obj.call("object_id");
|
459
|
+
std::cout << object_id << std::endl;
|
460
|
+
~~~
|
461
|
+
|
462
|
+
Rice mimics the Ruby class hierarchy as closely as it can.
|
463
|
+
In fact, the above code also works for Classes:
|
464
|
+
|
465
|
+
~~~{.cpp}
|
466
|
+
Class rb_cTest = define_class<Test>("Test");
|
467
|
+
Object object_id = rb_cTest.call("object_id");
|
468
|
+
std::cout << object_id << std::endl;
|
469
|
+
~~~
|
470
|
+
|
471
|
+
Rice provides builtin wrappers for many builtin Ruby types, including:
|
472
|
+
|
473
|
+
- Object
|
474
|
+
- Module
|
475
|
+
- Class
|
476
|
+
- String
|
477
|
+
- Array
|
478
|
+
- Hash
|
479
|
+
- Struct
|
480
|
+
- Symbol
|
481
|
+
- Exception
|
482
|
+
|
483
|
+
The `Array` and `Hash` types can even be iterated over the same way one
|
484
|
+
would iterate over an STL container:
|
485
|
+
|
486
|
+
~~~{.cpp}
|
487
|
+
Array a;
|
488
|
+
a.push(to_ruby(42));
|
489
|
+
a.push(to_ruby(43));
|
490
|
+
a.push(to_ruby(44));
|
491
|
+
Array::iterator it = a.begin();
|
492
|
+
Array::iterator end = a.end();
|
493
|
+
for(; it != end; ++it)
|
494
|
+
{
|
495
|
+
std::cout << *it << std::endl;
|
496
|
+
}
|
497
|
+
~~~
|
498
|
+
|
499
|
+
STL algorithms should also work as expected on `Array` and `Hash` containers.
|
500
|
+
|
501
|
+
|
502
|
+
## Inheritance {#inheritance}
|
503
|
+
|
504
|
+
Inheritance is a tricky problem to solve in extensions. This is because
|
505
|
+
wrapper functions for base classes typically don't know how to accept
|
506
|
+
pointers to derived classes. It is possible to write this logic, but
|
507
|
+
the code is nontrivial.
|
508
|
+
|
509
|
+
Forunately Rice handles this gracefully:
|
510
|
+
|
511
|
+
~~~{.cpp}
|
512
|
+
class Base
|
513
|
+
{
|
514
|
+
public:
|
515
|
+
virtual void foo();
|
516
|
+
};
|
517
|
+
|
518
|
+
class Derived
|
519
|
+
: public Base
|
520
|
+
{
|
521
|
+
};
|
522
|
+
|
523
|
+
extern "C"
|
524
|
+
void Init_test()
|
525
|
+
{
|
526
|
+
Data_Type<Base> rb_cBase =
|
527
|
+
define_class<Base>("Base")
|
528
|
+
.define_method("foo", &Base::foo);
|
529
|
+
Data_Type<Derived> rb_cDerived =
|
530
|
+
define_class<Derived, Base>("Derived");
|
531
|
+
}
|
532
|
+
~~~
|
533
|
+
|
534
|
+
The second template parameter to define_class indicates that `Derived`
|
535
|
+
inherits from `Base`.
|
536
|
+
|
537
|
+
Rice does not support multiple inheritance.
|
538
|
+
|
539
|
+
|
540
|
+
## Overloaded functions {#overloading}
|
541
|
+
|
542
|
+
If you try to create a member function pointer to an overloaded
|
543
|
+
function, you will get an error. So how do we wrap classes that have
|
544
|
+
overloaded functions?
|
545
|
+
|
546
|
+
Consider a class that uses this idiom for accessors:
|
547
|
+
|
548
|
+
~~~{.cpp}
|
549
|
+
class Container
|
550
|
+
{
|
551
|
+
size_t capacity(); // Get the capacity
|
552
|
+
void capacity(size_t cap); // Set the capacity
|
553
|
+
};
|
554
|
+
~~~
|
555
|
+
|
556
|
+
We can wrap this class by using `typedef`s:
|
557
|
+
|
558
|
+
~~~{.cpp}
|
559
|
+
extern "C"
|
560
|
+
void Init_Container()
|
561
|
+
{
|
562
|
+
typedef size_t (Container::*get_capacity)();
|
563
|
+
typedef void (Container::*set_capacity)(size_t);
|
564
|
+
|
565
|
+
Data_Type<Container> rb_cContainer =
|
566
|
+
define_class<Container>("Container")
|
567
|
+
.define_method("capacity", get_capacity(&Container::capacity))
|
568
|
+
.define_method("capacity=", set_capacity(&Container::capacity))
|
569
|
+
}
|
570
|
+
~~~
|
571
|
+
|
572
|
+
|
573
|
+
## User-defined type conversions {#user_defined_conversions}
|
574
|
+
|
575
|
+
Rice provides default conversions for many built-in types. Sometimes,
|
576
|
+
however, the default conversion is not what is expected. For
|
577
|
+
example, consider a function:
|
578
|
+
|
579
|
+
~~~{.cpp}
|
580
|
+
void foo(char * x);
|
581
|
+
~~~
|
582
|
+
|
583
|
+
Is `x` a pointer to a single character or a pointer to the first character
|
584
|
+
of a null-terminated string or a pointer to the first character of an
|
585
|
+
array of char?
|
586
|
+
|
587
|
+
Because the second case is the most common use case (a pointer to the
|
588
|
+
first character of a C string), Rice provides a default conversion that
|
589
|
+
treats a `char *` as a C string. But suppose the above function takes a
|
590
|
+
pointer to a char instead?
|
591
|
+
|
592
|
+
If we write this:
|
593
|
+
|
594
|
+
~~~{.cpp}
|
595
|
+
extern "C"
|
596
|
+
void Init_test()
|
597
|
+
{
|
598
|
+
define_global_function("foo", &foo);
|
599
|
+
}
|
600
|
+
~~~
|
601
|
+
|
602
|
+
It will likely have the wrong behavior.
|
603
|
+
|
604
|
+
To avoid this problem, it is necessary to write a wrapper function:
|
605
|
+
|
606
|
+
~~~{.cpp}
|
607
|
+
Object wrap_foo(Object o)
|
608
|
+
{
|
609
|
+
char c = from_ruby<char>(o);
|
610
|
+
foo(&c);
|
611
|
+
return to_ruby(c);
|
612
|
+
}
|
613
|
+
|
614
|
+
extern "C"
|
615
|
+
void Init_test()
|
616
|
+
{
|
617
|
+
define_global_function("foo", &wrap_foo);
|
618
|
+
}
|
619
|
+
~~~
|
620
|
+
|
621
|
+
Note that the out parameter is returned from `wrap_foo`, as Ruby does not
|
622
|
+
have pass-by-variable-reference (it uses pass-by-object-reference).
|
623
|
+
|
624
|
+
|
625
|
+
## Default Arguments {#default_arguments}
|
626
|
+
|
627
|
+
Going back to our initial C++ class example, lets say that `hello()` now
|
628
|
+
takes more arguments, one of which has a default value:
|
629
|
+
|
630
|
+
~~~{.cpp}
|
631
|
+
class Test
|
632
|
+
{
|
633
|
+
public:
|
634
|
+
Test();
|
635
|
+
std::string hello(std::string first, std::string second = "world");
|
636
|
+
};
|
637
|
+
~~~
|
638
|
+
|
639
|
+
As default parameter information is not available through templates,
|
640
|
+
it is necessary to define this in Rice explicitly using `Rice::Arg`:
|
641
|
+
|
642
|
+
~~~{.cpp}
|
643
|
+
#include "rice/Data_Type.hpp"
|
644
|
+
#include "rice/Constructor.hpp"
|
645
|
+
|
646
|
+
using namespace Rice;
|
647
|
+
|
648
|
+
extern "C"
|
649
|
+
void Init_test()
|
650
|
+
{
|
651
|
+
Data_Type<Test> rb_cTest =
|
652
|
+
define_class<Test>("Test")
|
653
|
+
.define_constructor(Constructor<Test>())
|
654
|
+
.define_method("hello",
|
655
|
+
&Test::hello,
|
656
|
+
(Arg("hello"), Arg("second") = "world")
|
657
|
+
);
|
658
|
+
}
|
659
|
+
~~~
|
660
|
+
|
661
|
+
The syntax here is simply `Arg(nameOfParameter)[ = defaultValue]`. The name of the
|
662
|
+
parameter is not important here (a readability tool), but the value set via `operator=`
|
663
|
+
must match the type of the parameter. As such it may be necessary to
|
664
|
+
explicitly cast the default value.
|
665
|
+
|
666
|
+
~~~{.cpp}
|
667
|
+
.define_method("hello",
|
668
|
+
&Test::hello,
|
669
|
+
(Arg("hello"), Arg("second") = (std::string)"world")
|
670
|
+
);
|
671
|
+
~~~
|
672
|
+
|
673
|
+
These `Rice::Arg` objects must be in the correct order and must be
|
674
|
+
surrounded with parentheses if more than one exists.
|
675
|
+
|
676
|
+
Now, Ruby will now know about the default arguments, and this wrapper
|
677
|
+
can be used as expected:
|
678
|
+
|
679
|
+
~~~{.cpp}
|
680
|
+
t = Test.new
|
681
|
+
t.hello("hello")
|
682
|
+
t.hello("goodnight", "moon")
|
683
|
+
~~~
|
684
|
+
|
685
|
+
This also works with Constructors:
|
686
|
+
|
687
|
+
~~~{.cpp}
|
688
|
+
.define_constructor(Constructor<SomeClass, int, int>(),
|
689
|
+
( Arg("arg1") = 1, Arg("otherArg") = 12 );
|
690
|
+
~~~
|
691
|
+
|
692
|
+
## Director {#director}
|
693
|
+
|
694
|
+
As polymorphism is the most important tennant of Object Oriented Programming,
|
695
|
+
it is important that Rice supports polymorphic calls travelling between C++
|
696
|
+
and Ruby seemlessly. Super calls from Ruby subclasses back into C++ already work,
|
697
|
+
but enabling the other direction requires some extra effort. Rice
|
698
|
+
supplies the the `Rice::Director` class and
|
699
|
+
`Rice::Data_Type::define_director` to expose this functionality.
|
700
|
+
|
701
|
+
Like `SWIG_Director`, `Rice::Director` is a class that is used to build a proxy class
|
702
|
+
to properly send execution up or down the object hierarchy for that class. Take
|
703
|
+
the following class:
|
704
|
+
|
705
|
+
~~~{.cpp}
|
706
|
+
class VirtualBase {
|
707
|
+
public:
|
708
|
+
VirtualBase();
|
709
|
+
virtual int doWork();
|
710
|
+
virtual int processWorker() = 0;
|
711
|
+
};
|
712
|
+
~~~
|
713
|
+
|
714
|
+
Due to the abstract nature of this class, it will not work at all with Rice
|
715
|
+
in its current form. Any attempt to do so will cause a compilation error due to
|
716
|
+
this class not being constructable. Even without the pure virtual function, any
|
717
|
+
call to `VirtualBase::doWork` will stop at the C++ level and will not pass down into
|
718
|
+
any Ruby subclasses.
|
719
|
+
|
720
|
+
To properly wrap both of these methods, use a `Rice::Director` subclass as a proxy:
|
721
|
+
|
722
|
+
~~~{.cpp}
|
723
|
+
#include "rice/Director.hpp"
|
724
|
+
|
725
|
+
class VirtualBaseProxy : public VirtualBase, public Rice::Director {
|
726
|
+
public:
|
727
|
+
VirtualBaseProxy(Object self) : Rice::Director(self) { }
|
728
|
+
|
729
|
+
virtual int doWork() {
|
730
|
+
return from_ruby<int>( getSelf().call("do_work") );
|
731
|
+
}
|
732
|
+
|
733
|
+
int default_doWork() {
|
734
|
+
return VirtualBase::doWork();
|
735
|
+
}
|
736
|
+
|
737
|
+
virtual int processWorker() {
|
738
|
+
return from_ruby<int>( getSelf().call("process_worker") );
|
739
|
+
}
|
740
|
+
|
741
|
+
int default_processWorker() {
|
742
|
+
raisePureVirtual();
|
743
|
+
}
|
744
|
+
};
|
745
|
+
~~~
|
746
|
+
|
747
|
+
There is a lot going on here, so we'll go through each part.
|
748
|
+
|
749
|
+
~~~{.cpp}
|
750
|
+
class VirtualBaseProxy : public Virtualbase, public Rice::Director {
|
751
|
+
~~~
|
752
|
+
|
753
|
+
First, the class needs to subclass both the virtual class in question and `Rice::Director`.
|
754
|
+
|
755
|
+
~~~{.cpp}
|
756
|
+
public:
|
757
|
+
VirtualBaseProxy(Object self) : Rice::Director(self) { }
|
758
|
+
~~~
|
759
|
+
|
760
|
+
For `Rice::Director` to work its magic, every instance of this class needs to
|
761
|
+
have a handle to the Ruby instance. The constructor
|
762
|
+
must take a `Rice::Object` as the first argument and pass it up into
|
763
|
+
`Rice::Director`. The code here is the minimum required for a `Rice::Director` proxy.
|
764
|
+
|
765
|
+
~~~{.cpp}
|
766
|
+
virtual int doWork() {
|
767
|
+
return from_ruby<int>( getSelf().call("do_work") );
|
768
|
+
}
|
769
|
+
|
770
|
+
int default_doWork() {
|
771
|
+
return VirtualBase::doWork();
|
772
|
+
}
|
773
|
+
~~~
|
774
|
+
|
775
|
+
Here the directory proxy overrides the methods for Ruby exposure and
|
776
|
+
implements the required actions to pass flow around the hierarchy
|
777
|
+
appropriately. The pattern shown here is that the actual override will
|
778
|
+
call down into Ruby, handling any type conversions, while a
|
779
|
+
`default_methodName` method handles calling up into C++ and will be the
|
780
|
+
method wrapped into Rice.
|
781
|
+
|
782
|
+
The `default_doWork` method will be used as Rice's hookup of calling back up the
|
783
|
+
hierarchy (wrapping is below). This method needs to do one of two things: call
|
784
|
+
up the class hierarchy, as seen here, or call `raisePureVirtual()` as seen in the
|
785
|
+
processWorker example:
|
786
|
+
|
787
|
+
~~~{.cpp}
|
788
|
+
int default_processWorker() {
|
789
|
+
raisePureVirtual();
|
790
|
+
}
|
791
|
+
~~~
|
792
|
+
|
793
|
+
The method `raisePureVirtual()` exists to allow wrapping a pure virtual method into Ruby
|
794
|
+
(and ensuring compliation is possible) but making sure any users of this extension are
|
795
|
+
informed quickly that there's nothing callable in the C++ side of the library.
|
796
|
+
|
797
|
+
Once the proxy class is built, it's time to wrap it into Ruby:
|
798
|
+
|
799
|
+
~~~{.cpp}
|
800
|
+
extern "C"
|
801
|
+
void Init_virtual() {
|
802
|
+
define_class<VirtualBase>("VirtualBase")
|
803
|
+
.define_director<VirtualBaseProxy>()
|
804
|
+
.define_constructor(Constructor<VirtualBaseProxy, Rice::Object>())
|
805
|
+
.define_method("do_work", &VirtualBaseProxy::default_doWork)
|
806
|
+
.define_method("process_worker", &VirtualBaseProxy::default_processWorker);
|
807
|
+
}
|
808
|
+
~~~
|
809
|
+
|
810
|
+
The wrapping is the same as is described earlier in this document. Expose the class
|
811
|
+
`VirtualBase`, and register `VirtualBaseProxy` as a director proxy of `VirtualBase` with
|
812
|
+
`Rice::Data_Type::define_director`, then `define methods` pointing to the proxy object as necessary.
|
813
|
+
|
814
|
+
You must use the `Rice::Director` proxy class in the Constructor line, this allows proper
|
815
|
+
object construction / destruction of the types in question.
|
816
|
+
|
817
|
+
## Implicit Casting {#implicit_cast}
|
818
|
+
|
819
|
+
There are times when a library exposes classes that, while unrelated, are
|
820
|
+
built to be interchangeable across the library. One example of this is found in
|
821
|
+
the Open Source 3d rendering engine <a
|
822
|
+
href="http://www.ogre3d.org/">OGRE</a>: Ogre::Degree and Ogre::Radian.
|
823
|
+
When a given method takes a Radian, you're free to pass in a Degree, and vice versa.
|
824
|
+
|
825
|
+
Rice cannot automatically figure out if this kind of functionality is
|
826
|
+
possible in a given library but it does provide an API for defining
|
827
|
+
these relationships: `Rice::define_implicit_cast<From, To>()`.
|
828
|
+
|
829
|
+
~~~{.cpp}
|
830
|
+
class Degree { ... };
|
831
|
+
class Radian { ... };
|
832
|
+
|
833
|
+
extern "C"
|
834
|
+
void Init_implicit() {
|
835
|
+
define_class<Degree>()
|
836
|
+
...;
|
837
|
+
define_class<Radian>()
|
838
|
+
...;
|
839
|
+
|
840
|
+
define_implicit_cast<Degree, Radian>();
|
841
|
+
define_implicit_cast<Radian, Degree>();
|
842
|
+
}
|
843
|
+
~~~
|
844
|
+
|
845
|
+
Using `Rice::define_implicit_cast` has the following requirements:
|
846
|
+
|
847
|
+
- The two types must be bound in Rice before defining the cast.
|
848
|
+
- The classes must have constructors that take the other type.
|
849
|
+
- This feature cannot be used with fundamental types.
|
850
|
+
|
851
|
+
To see a full example of this feature, please check out
|
852
|
+
test/test_Data_Type.cpp.
|
853
|
+
|
854
|
+
# Motivation {#motivation}
|
855
|
+
|
856
|
+
There are a number of common problems when writing C or C++ extensions
|
857
|
+
for Ruby:
|
858
|
+
|
859
|
+
- Type safety. It is easy to mix-up integral types such as ID and
|
860
|
+
VALUE. Some of the functions in the Ruby API are not consistent with
|
861
|
+
which types they take (e.g. rb_const_defined takes an ID and
|
862
|
+
rb_mod_remove_const takes a Symbol).
|
863
|
+
|
864
|
+
- DRY principle. Specifying the number of arguments that each wrapped
|
865
|
+
function takes is easy to get wrong. Adding a new argument to the
|
866
|
+
function means that the number of arguments passed to rb_define_method
|
867
|
+
must also be updated.
|
868
|
+
|
869
|
+
- Type conversion. There are many different functions to convert data
|
870
|
+
to and from ruby types. Many of them have different semantics or
|
871
|
+
different forms. For example, to convert a string, one might use the
|
872
|
+
StringValue macro, but to convert a fixnum, one might use FIX2INT.
|
873
|
+
Unwrapping previously wrapped C data uses yet another form.
|
874
|
+
|
875
|
+
- Exception safety. It is imperative that C++ exceptions never make
|
876
|
+
their way into C code, and it is also imperative that a Ruby exception
|
877
|
+
never escape while there are objects on the stack with nontrivial
|
878
|
+
destructors. Rules for when it is okay to use which exceptions are
|
879
|
+
difficult to get right, especially as code is maintained through time.
|
880
|
+
|
881
|
+
- Thread safety. Because the Ruby interpreter is not threads-safe,
|
882
|
+
the Ruby interpreter must not be run from more than one thread.
|
883
|
+
Because of tricks the GC and scheduler play with the C stack, it's not
|
884
|
+
enough to ensure that only one thread runs the interpreter at any
|
885
|
+
given time; once the interpreter has been run from one thread, it must
|
886
|
+
only ever be run from that thread in the future. Additionally,
|
887
|
+
because Ruby copies the stack when it switches threads, C++ code must
|
888
|
+
be careful not to access objects in one Ruby thread that were created
|
889
|
+
on the stack in another Ruby thread.
|
890
|
+
|
891
|
+
- C-based API. The Ruby API is not always convenient for accessing
|
892
|
+
Ruby data structurs such as Hash and Array, especially when writing C++
|
893
|
+
code, as the interface for these containers is not consistent with
|
894
|
+
standard containers.
|
895
|
+
|
896
|
+
- Calling convention. Function pointers passed into the Ruby API must
|
897
|
+
follow the C calling convention. This means that it is not possible to
|
898
|
+
pass a pointer to a template function or static member function (that
|
899
|
+
is, it will work on some platforms, but isn't portable).
|
900
|
+
|
901
|
+
- Inheritance. When wrapping C++ objects, it is easy to store a
|
902
|
+
pointer to a derived class, but then methods in the base class must have
|
903
|
+
knowledge of the derived class in order to unwrap the object. It is
|
904
|
+
possible to always store a pointer to the base class and then
|
905
|
+
dynamic_cast the pointer to the derived type when necessary, but this
|
906
|
+
can be slow and cumbersome, and it isn't likely to work with multiple
|
907
|
+
inheritance. A system that properly handles inheritance for all corner
|
908
|
+
cases is nontrivial.
|
909
|
+
|
910
|
+
- Multiple inheritance. C++ supports true multiple inheritance, but
|
911
|
+
the Ruby object model uses single inheritance with mixins. When
|
912
|
+
wrapping a library whose public interface uses multiple inheritance,
|
913
|
+
care must be taken in constructing the mapping.
|
914
|
+
|
915
|
+
- GC safety. All live Ruby objects must be marked during the garbage
|
916
|
+
collector's mark phase, otherwise they will be prematurely destroyed.
|
917
|
+
The general rule is that object references stored on the heap should be
|
918
|
+
either registered with rb_gc_register_address or marked by a data
|
919
|
+
object's mark function; object references stored on the stack will be
|
920
|
+
automatically marked, provided the Ruby interpreter was properly
|
921
|
+
initialized at startup.
|
922
|
+
|
923
|
+
- Callbacks. C implements callbacks via function pointers, while ruby
|
924
|
+
typically implements callbacks via procs. Writing an adapter function
|
925
|
+
to call the proc is not difficult, but there is much opportunity for
|
926
|
+
error (particularly with exception-safety).
|
927
|
+
|
928
|
+
- Data serialization. By default data objects defined at the C layer
|
929
|
+
are not marshalable. The user must explicitly define functions to
|
930
|
+
marshal the data member-by-member.
|
931
|
+
|
932
|
+
Rice addresses these issues in many ways:
|
933
|
+
|
934
|
+
- Type safety. Rice provides encapsulation for all builtin types,
|
935
|
+
such as Object, Identifier, Class, Module, and String. It
|
936
|
+
automatically checks the dynamic type of an object before constructing
|
937
|
+
an instance of a wrapper.
|
938
|
+
|
939
|
+
- DRY principle. Rice uses introspection through the use of templates
|
940
|
+
and function overloading to automatically determine the number and types
|
941
|
+
of arguments to functions. Default arguments must still be handled
|
942
|
+
explicitly, however.
|
943
|
+
|
944
|
+
- Type conversions. Rice provides cast-style to_ruby<> and
|
945
|
+
from_ruby<> template functions to simplify explicit type conversions.
|
946
|
+
Automatic type conversions for parameters and return values are
|
947
|
+
generated for all wrapped functions.
|
948
|
+
|
949
|
+
- Exception safety. Rice automatically converts common exceptions and
|
950
|
+
provides a mechanism for converting user-defined exception types. Rice
|
951
|
+
also provides convenience functions for converting exceptions when
|
952
|
+
calling back into ruby code.
|
953
|
+
|
954
|
+
- Thread safety. Rice provides no mechanisms for dealing with thread
|
955
|
+
safety. Many common thread safety issues should be alleviated by YARV,
|
956
|
+
which supports POSIX threads.
|
957
|
+
|
958
|
+
- C++-based API. Rice provides an object-oriented C++-style API to
|
959
|
+
most common functions in the Ruby C API.
|
960
|
+
|
961
|
+
- Calling convention. Rice automatically uses C calling convention
|
962
|
+
for all function pointers passed into the Ruby API.
|
963
|
+
|
964
|
+
- Inheritance. Rice provides automatic conversion to the base class
|
965
|
+
type when a wrapped member function is called on the base class.
|
966
|
+
|
967
|
+
- Multiple inheritance. Rice provides no mechanism for multiple
|
968
|
+
inheritance. Multiple inheritance can be simulated via mixins, though
|
969
|
+
this is not yet as easy as it could be.
|
970
|
+
|
971
|
+
- GC safety. Rice provides a handful of convenience classes for
|
972
|
+
interacting with the garbage collector. There are still basic rules
|
973
|
+
which must be followed to ensure that objects get properly destroyed.
|
974
|
+
|
975
|
+
- Callbacks. Rice provides a handful of convenience classes for
|
976
|
+
dealing with callbacks.
|
977
|
+
|
978
|
+
- Data serialization. Rice provides no mechanism for data
|
979
|
+
serialization, but it is likely this may be added in a future release.
|
980
|
+
|
981
|
+
|
982
|
+
# What Rice is Not {#what_not}
|
983
|
+
|
984
|
+
There are a number projects which server similar functions to Rice. Two
|
985
|
+
such popular projects are SWIG and Boost.Python. Rice has some
|
986
|
+
distinct features which set it apart from both of these projects.
|
987
|
+
|
988
|
+
Rice is not trying to replace SWIG. Rice is not a generic wrapper
|
989
|
+
interface generator. Rice is a C++ library for interfacing with the
|
990
|
+
Ruby C API. This provides a very natural way for C++ programmers to
|
991
|
+
wrap their C++ code, without having to learn a new domain-specific
|
992
|
+
language. However, there is no reason why SWIG and Rice could not work
|
993
|
+
together; a SWIG module could be written to generate Rice code. Such a
|
994
|
+
module would combine the portability of SWIG with the maintainability of
|
995
|
+
Rice (I have written extensions using both, and I have found Rice
|
996
|
+
extensions to be more maintainable when the interface is constantly
|
997
|
+
changing. Your mileage may vary).
|
998
|
+
|
999
|
+
Rice is also not trying to simply be a Ruby version of Boost.Python.
|
1000
|
+
Rice does use some of the same template tricks that Boost.Python uses,
|
1001
|
+
however there are some important distinctions. First of all,
|
1002
|
+
Boost.Python attempts to create a declarative DSL in C++ using
|
1003
|
+
templates. Rice is a wrapper around the Ruby C API and attempts to make
|
1004
|
+
its interface look like an OO version of the API; this means that class
|
1005
|
+
declarations look procedural rather than declarative. Secondly, the
|
1006
|
+
Ruby object model is different from the python object model. This is
|
1007
|
+
reflected in the interface to Rice; it mimics the Ruby object model at
|
1008
|
+
the C++ level. Thirdly, Rice uses Ruby as a code generator; I find this
|
1009
|
+
to be much more readable than using the Boost preprocessor library.
|
1010
|
+
|
1011
|
+
|
1012
|
+
# History {#history}
|
1013
|
+
|
1014
|
+
Rice originated as Excruby, a project to interface with C++-based trading
|
1015
|
+
software at Automated Trading Desk in Mount Pleasant, South Carolina.
|
1016
|
+
The Ruby bindings for Swig were at the time less mature than they are
|
1017
|
+
today, and did not suit the needs of the project.
|
1018
|
+
|
1019
|
+
Excruby was written not as a wrapper for the Ruby API, but rather as a
|
1020
|
+
set of helper functions and classes for interfacing with the Ruby
|
1021
|
+
interpreter in an exception-safe manner. Over the course of five years,
|
1022
|
+
the project grew into wrappers for pieces of the API, but the original
|
1023
|
+
helper functions remained as part of the public interface.
|
1024
|
+
|
1025
|
+
This created confusion for the users of the library, because there were
|
1026
|
+
multiple ways of accomplishing most tasks -- directly through the C API,
|
1027
|
+
through a low-level wrapper around the C API, and through a high-level
|
1028
|
+
abstraction of the lower-level interfaces.
|
1029
|
+
|
1030
|
+
Rice was then born in an attempt to clean up the interface. Rice keeps
|
1031
|
+
the lower-level wrappers, but as an implementation detail; the public
|
1032
|
+
interface is truly a high-level abstraction around the Ruby C API.
|
1033
|
+
|
1034
|
+
|
1035
|
+
# The GC {#gc}
|
1036
|
+
|
1037
|
+
- Objects are not automatically registered with the garbage collector.
|
1038
|
+
|
1039
|
+
- If an Object is on the stack, it does not need to be registered with
|
1040
|
+
the garbage collector.
|
1041
|
+
|
1042
|
+
- If an Object is allocated on the heap or if it is a member of an
|
1043
|
+
object that might be allocated on the heap, use an
|
1044
|
+
Rice::Address_Registration_Guard to register the object with the garbage
|
1045
|
+
collector.
|
1046
|
+
|
1047
|
+
- If a reference counted object is being wrapped, or if another type
|
1048
|
+
of smart pointer is wrapped, ensure that only one mechanism is used to
|
1049
|
+
destroy the object. In general, the smart pointer manages the
|
1050
|
+
allocation of the object, and Ruby should hold only a reference to the
|
1051
|
+
smart pointer. When the garbage collector determines that it is time to
|
1052
|
+
clean up the object, the smart pointer will be destroyed, decrementing
|
1053
|
+
the reference count; when the reference count drops to 0, underlying
|
1054
|
+
object will be destroyed.
|