aquarium 0.4.3 → 0.4.4
Sign up to get free protection for your applications and to get access to all the features.
- data/CHANGES +322 -247
- data/README +11 -9
- data/Rakefile +22 -12
- data/UPGRADE +49 -43
- data/examples/aspect_design_example.rb +17 -14
- data/examples/design_by_contract_example.rb +14 -10
- data/examples/exception_wrapping_example.rb +5 -3
- data/examples/introductions_example.rb +2 -1
- data/examples/method_missing_example.rb +14 -13
- data/examples/method_tracing_example.rb +20 -16
- data/examples/reusable_aspect_hack_example.rb +8 -5
- data/lib/aquarium/aspects/exclusion_handler.rb +3 -3
- data/lib/aquarium/version.rb +1 -1
- data/rake_tasks/verify_rcov.rake +1 -1
- data/spec/aquarium/aspects/concurrent_aspects_spec.rb +41 -41
- data/spec/aquarium/aspects/join_point_spec.rb +5 -5
- data/spec/aquarium/aspects/pointcut_spec.rb +4 -1
- data/spec/aquarium/finders/type_finder_with_descendents_and_ancestors_spec.rb +4 -1
- metadata +3 -9
- data/Aquarium-IDEA.ipr +0 -252
- data/Aquarium-IDEA.iws +0 -493
- data/Aquarium.ipr +0 -253
- data/Aquarium.iws +0 -624
- data/ParseTreePlay.rb +0 -25
- data/TODO.rb +0 -186
data/CHANGES
CHANGED
|
@@ -1,100 +1,134 @@
|
|
|
1
|
+
== Version 0.4.4
|
|
2
|
+
|
|
3
|
+
V0.4.4 changes no APIs. It upgrades compatibility with the latest JRuby (1.3.1), RSpec (1.2.8) and Webgen (0.5.10) for the build process. You'll only need to upgrade if you build Aquarium yourself.
|
|
4
|
+
|
|
5
|
+
Enhancements:
|
|
6
|
+
27235 Support RSpec 1.2.8 (latest) and and JRuby 1.3.1 (latest)
|
|
7
|
+
27236 Upgrade website generation to webgen 0.5.X
|
|
8
|
+
|
|
9
|
+
There are no API changes in this release, with one exception. Most of the code changes are minor fixes to RSpec files. The one exception is an apparent change in JRuby 1.3.1 w.r.t. the access restrictions on Java methods. It now appears that protected methods are public, that is they are in the method list returned by :public_methods, but not :protected_methods. Also, private methods are not returned by any of the :*_methods.
|
|
10
|
+
|
|
11
|
+
I used the upgrade to Webgen 0.5.X as an excuse to refresh the web site's appearance. The content is mostly the same as before.
|
|
12
|
+
|
|
1
13
|
== Version 0.4.3
|
|
2
14
|
|
|
3
|
-
V0.4.3 adds a few bug fixes and enhancements, and several internal
|
|
4
|
-
performance, etc.
|
|
15
|
+
V0.4.3 adds a few bug fixes and enhancements, and several internal
|
|
16
|
+
refactorings to improve performance, etc.
|
|
5
17
|
|
|
6
18
|
Bug fixes:
|
|
7
19
|
[none]
|
|
8
20
|
|
|
9
21
|
Enhancements:
|
|
10
22
|
14165 Support Ruby 1.8.2, 1.8.5, 1.8.6, 1.9.X and 2.0 releases and snapshots
|
|
11
|
-
21283 Allow "_of" suffix at end of "*_and_descendents", "*_and_ancestors",
|
|
12
|
-
|
|
13
|
-
|
|
14
|
-
|
|
15
|
-
|
|
16
|
-
|
|
17
|
-
|
|
18
|
-
|
|
19
|
-
|
|
20
|
-
|
|
21
|
-
|
|
22
|
-
|
|
23
|
-
|
|
24
|
-
|
|
25
|
-
|
|
26
|
-
|
|
23
|
+
21283 Allow "_of" suffix at end of "*_and_descendents", "*_and_ancestors",
|
|
24
|
+
and "*_nested_types"
|
|
25
|
+
21924 Should be able to turn off contract-testing aspects globally in the
|
|
26
|
+
Design by Contract extra module
|
|
27
|
+
21993 By default, don't advise "system" methods that being with "_", but
|
|
28
|
+
provide an option to do so
|
|
29
|
+
|
|
30
|
+
Details:
|
|
31
|
+
14165: I closed this one because it's too broad. Aquarium currently runs with
|
|
32
|
+
1.8.6. I will open a new enhancement specifically for 1.9.
|
|
33
|
+
21283: Sometimes appending "_of" makes the specification read more smoothly,
|
|
34
|
+
so it's now an option.
|
|
35
|
+
21924: To class-level methods in DesignByContract, enable_all and disable_all,
|
|
36
|
+
globally turn contracts on or off. See their documentation for details.
|
|
37
|
+
21993: Now, by default, MethodFinder won't match any method that starts with
|
|
38
|
+
two underscores ("__"), so you don't have to :exclude_ancestors as much just
|
|
39
|
+
to avoid matching methods like "__id__" and "__send__". There is also a new
|
|
40
|
+
:method_options called :include_system_methods that will not suppress
|
|
41
|
+
matching these methods. So, I'm calling these special methods "system
|
|
42
|
+
methods". I don't yet provide a way to configure this list (RSpec methods
|
|
43
|
+
would be good additions...). The list is currently defined as an array of
|
|
44
|
+
regex's, MethodFinder::IGNORED_SYSTEM_METHODS. NOTE: This is effectively a
|
|
45
|
+
behavior change, although it's not likely to affect anyone.
|
|
27
46
|
|
|
28
47
|
== Version 0.4.2
|
|
29
48
|
|
|
30
|
-
V0.4.2 adds a few bug fixes and enhancements, greatly improved RDoc output,
|
|
31
|
-
refactorings to improve performance.
|
|
49
|
+
V0.4.2 adds a few bug fixes and enhancements, greatly improved RDoc output,
|
|
50
|
+
and internal refactorings to improve performance.
|
|
32
51
|
|
|
33
52
|
Bug fixes:
|
|
34
53
|
15202 Intermittent confusion between classes and objects when invoking advice
|
|
35
|
-
19262 Just putting join_point argument in advice block causes mysterious
|
|
54
|
+
19262 Just putting join_point argument in advice block causes mysterious
|
|
55
|
+
method missing error.
|
|
36
56
|
19321 Advice overhead is too high (ongoing improvements)
|
|
37
57
|
|
|
38
58
|
Enhancements:
|
|
39
|
-
13403 Support recursion into nested types without requiring "tricky" regular
|
|
59
|
+
13403 Support recursion into nested types without requiring "tricky" regular
|
|
60
|
+
expressions
|
|
40
61
|
13406 "Sugar" for adding methods and attributes to types
|
|
41
62
|
18537 Provide a search for all pointcut constants
|
|
42
63
|
19666 Improve the RDoc output
|
|
43
|
-
19119 Provide an "after_raising" type of advice that lets you handle the
|
|
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
|
-
|
|
64
|
+
19119 Provide an "after_raising" type of advice that lets you handle the
|
|
65
|
+
exception.
|
|
66
|
+
|
|
67
|
+
#15202: I never figured out the cause of this problem, but it hasn't been seen
|
|
68
|
+
since late last
|
|
69
|
+
year, so I suspect it disappeared as a side effect of on-going refactoring and
|
|
70
|
+
enhancements.
|
|
71
|
+
|
|
72
|
+
#19262: If you just specified "|jp|" for an advice block, you would sometimes
|
|
73
|
+
get a method missing error. I never figured out exactly why, but it was
|
|
74
|
+
somehow related to passing the usual three arguments internally, where the
|
|
75
|
+
last two would be ignored in this case. Now, the code checks the arity and
|
|
76
|
+
only passes the join point in this case.
|
|
77
|
+
|
|
78
|
+
#19321: I removed some of the wasted object creation and initialization in
|
|
79
|
+
advice invocation,
|
|
80
|
+
improving the overhead by about 40%. However, it is still at least 10x higher
|
|
81
|
+
than simple method aliasing, so I want to make more improvements. (I did not
|
|
82
|
+
close this item.)
|
|
83
|
+
|
|
84
|
+
#13403: I added new options :types_and_nested_types and :types_and_nested that
|
|
85
|
+
are analogous to the similar "ancestors" and "descendents" options. The
|
|
86
|
+
nested option will return the specified types and any types they "enclose".
|
|
87
|
+
There are also corresponding "exclude" options.
|
|
88
|
+
|
|
89
|
+
#13406: I've decided not to do this, as it really isn't the "simplest thing
|
|
90
|
+
that could possibly work." It's easy enough for the user to define a module
|
|
91
|
+
of new behavior and just use "obj.extend(module)". However, when the user
|
|
92
|
+
needs to do this over a set of types, Aquarium's TypeFinder can be helpful,
|
|
93
|
+
so I added an example to the Examples code and page.
|
|
94
|
+
|
|
95
|
+
#18537: I've provided an example of the design approach where you define
|
|
96
|
+
pointcuts in a class, as a kind of "aspect interface" and write aspects that
|
|
97
|
+
specify those pointcuts. The problem has been that you had to name every
|
|
98
|
+
single such pointcut explicitly. There was no "finder" option, as for types,
|
|
99
|
+
methods, etc. Now there is a pointcut finder option with a new option
|
|
100
|
+
":named_pointcuts" for Aspect.new to specify a search for pointcuts in a set
|
|
101
|
+
of types matching a name specification. Either constants or class variables
|
|
102
|
+
will be matched (or both).
|
|
103
|
+
|
|
104
|
+
#19666: The rdoc for the key classes was cleaned up so it "renders" better.
|
|
105
|
+
Feedback is welcome.
|
|
106
|
+
|
|
107
|
+
#19119: I finished the previously-incomplete support for allowing advice to
|
|
108
|
+
change the raised exception, in both after and after_raising advice. A
|
|
109
|
+
common scenario is to wrap the thrown exception in another. For example, a
|
|
110
|
+
low-level, service-specific exception (like a database error) in a
|
|
111
|
+
higher-level, more generic application exception.
|
|
112
|
+
|
|
113
|
+
You still can't rescue the exception completely in :after_raising and :after
|
|
114
|
+
advice; the value for the exception in
|
|
82
115
|
joinpoint.context.raised_exception
|
|
83
|
-
when the advice returns will be raised by Aquarium. I think that permitting
|
|
84
|
-
after advice to "eat" the exception could cause subtle
|
|
85
|
-
It would also probably violate the
|
|
86
|
-
|
|
87
|
-
|
|
88
|
-
|
|
89
|
-
|
|
116
|
+
when the advice returns will be raised by Aquarium. I think that permitting
|
|
117
|
+
:after_raising or :after advice to "eat" the exception could cause subtle
|
|
118
|
+
issues with scope and variable binding. It would also probably violate the
|
|
119
|
+
"principle of least surprise"; the advice code that rescues the exception would
|
|
120
|
+
not be as "obvious" to the reader as the familiar idiom of rescue clauses
|
|
121
|
+
that we are accustomed to using. Therefore, if you want to recover completely
|
|
122
|
+
from an exception, use rescue clauses in around advice.
|
|
90
123
|
|
|
91
124
|
== Version 0.4.1
|
|
92
125
|
|
|
93
|
-
V0.4.1 adds a few bug fixes, a few more user examples, internal refactoring and
|
|
94
|
-
improvements.
|
|
126
|
+
V0.4.1 adds a few bug fixes, a few more user examples, internal refactoring and
|
|
127
|
+
some performance improvements.
|
|
95
128
|
|
|
96
129
|
Bug fixes:
|
|
97
|
-
19116 When an exception is thrown during advice execution, the error message
|
|
130
|
+
19116 When an exception is thrown during advice execution, the error message
|
|
131
|
+
always reports the advice type is :before!
|
|
98
132
|
19261 after_raising DSL method provides no way to specify exceptions
|
|
99
133
|
|
|
100
134
|
Enhancements:
|
|
@@ -102,17 +136,20 @@ Enhancements:
|
|
|
102
136
|
19320 Move the Aquarium::...::AspectDSL file to Aquarium::DSL for convenience
|
|
103
137
|
19399 Improve the Design by Contract example
|
|
104
138
|
|
|
105
|
-
I added a new :exceptions argument (synonym :exception) that takes a single
|
|
106
|
-
You can only use this argument with :after_raising.
|
|
107
|
-
|
|
139
|
+
I added a new :exceptions argument (synonym :exception) that takes a single
|
|
140
|
+
exception or list thereof. You can only use this argument with :after_raising.
|
|
141
|
+
If you specify exceptions with the latter and use the :exceptions argument,
|
|
142
|
+
the values will be merged.
|
|
108
143
|
|
|
109
|
-
I thought it was ugly to have to type "include
|
|
110
|
-
|
|
144
|
+
I thought it was ugly to have to type "include
|
|
145
|
+
Aquarium::Aspects::DSL::AspectDSL", so I moved the code so now it's "include
|
|
146
|
+
Aquarium::DSL". However, for backwards compatibility, the old module still
|
|
147
|
+
works.
|
|
111
148
|
|
|
112
149
|
== Version 0.4.0
|
|
113
150
|
|
|
114
|
-
V0.4.0 adds specs to characterize and test advising Java classes when running
|
|
115
|
-
API enhancements.
|
|
151
|
+
V0.4.0 adds specs to characterize and test advising Java classes when running
|
|
152
|
+
on JRuby and adds several API enhancements.
|
|
116
153
|
|
|
117
154
|
Bug fixes:
|
|
118
155
|
17844 JRuby - Advising types, Aquarium thinks the type is a string
|
|
@@ -123,89 +160,101 @@ Enhancements:
|
|
|
123
160
|
17834 Allow :class and :module (and variants) wherever :type is allowed
|
|
124
161
|
17881 Add specs that exercise advising Java types and objects using JRuby
|
|
125
162
|
|
|
126
|
-
#17844 occurred because of the way JRuby encodes Java packages into modules.
|
|
127
|
-
JRuby types.
|
|
163
|
+
#17844 occurred because of the way JRuby encodes Java packages into modules.
|
|
164
|
+
Aquarium now properly handles JRuby types.
|
|
128
165
|
|
|
129
166
|
#17883 reflected a JRuby bug, so a workaround was required.
|
|
130
167
|
|
|
131
|
-
#17834 allows the user to substitute the words "class", "classes", "module" or
|
|
132
|
-
"type" and "types" are used in the API, since some
|
|
168
|
+
#17834 allows the user to substitute the words "class", "classes", "module" or
|
|
169
|
+
"modules" anywhere the words "type" and "types" are used in the API, since some
|
|
170
|
+
users might naturally want to write aspects like this:
|
|
133
171
|
|
|
134
172
|
around :calls_to => :my_method, :in_class => MyClass do ...; end
|
|
135
173
|
|
|
136
|
-
However, there is no enforcement to ensure that "class" is only used for
|
|
137
|
-
for modules, etc. Note: it's possible we'll
|
|
138
|
-
|
|
139
|
-
|
|
140
|
-
|
|
141
|
-
|
|
142
|
-
|
|
143
|
-
|
|
144
|
-
|
|
145
|
-
|
|
146
|
-
|
|
147
|
-
|
|
148
|
-
|
|
149
|
-
|
|
150
|
-
|
|
151
|
-
|
|
152
|
-
|
|
153
|
-
|
|
154
|
-
|
|
155
|
-
|
|
156
|
-
|
|
157
|
-
|
|
174
|
+
However, there is no enforcement to ensure that "class" is only used for
|
|
175
|
+
classes and "module" is only used for modules, etc. Note: it's possible we'll
|
|
176
|
+
enforce this in some future release, as a way of saying things like "only advise
|
|
177
|
+
classes that match ...", etc. Caveat Emptor!
|
|
178
|
+
|
|
179
|
+
For #18090, a bug prevented JoinPoint#invoke_original_join_point (which allows
|
|
180
|
+
you to bypass all advices at the join point) from working except for :around
|
|
181
|
+
advice (and yes, the specs didn't cover this adequately - gasp!). Now fixed.
|
|
182
|
+
|
|
183
|
+
For #17881, I created a separate set of specs for JRuby, so it's easier to run
|
|
184
|
+
the "regular" Aquarium specs using MRI and the JRuby-specific specs separately
|
|
185
|
+
with JRuby. The new "jruby" directory contains a Rakefile, another set of specs,
|
|
186
|
+
and Java example code for the specs to use. The default Rakefile task re-runs
|
|
187
|
+
the main Aquarium spec suite using JRuby, to confirm that the suite passes
|
|
188
|
+
successfully, then it runs a set of different specs that load sample Java
|
|
189
|
+
interfaces and classes into JRuby and then advises them with Aquarium.
|
|
190
|
+
|
|
191
|
+
I found a few JRuby bugs and other behavior differences between MRI during this
|
|
192
|
+
release. I was able to handle them with modifications to the Aquarium code. If
|
|
193
|
+
you "grep" the Aquarium "lib" and "spec" directories for the word "jruby"
|
|
194
|
+
(ignoring case), you'll find notes about these issues and the workarounds I
|
|
195
|
+
implemented.
|
|
196
|
+
|
|
197
|
+
The separate JRuby spec suite shows what you can and cannot do with Java types.
|
|
198
|
+
As a side benefit, it also demonstrates how Java types, objects, and methods
|
|
199
|
+
are mapped to Ruby. There are some important limitations, however. See the
|
|
200
|
+
jruby.html page on the website or the README for more details.
|
|
158
201
|
|
|
159
202
|
Note: JRuby 1.1RC2 was used.
|
|
160
|
-
|
|
161
203
|
|
|
162
204
|
== Version 0.3.1
|
|
163
205
|
|
|
164
|
-
V0.3.1 adds numerous performance improvements, especially in the RSpec suite,
|
|
206
|
+
V0.3.1 adds numerous performance improvements, especially in the RSpec suite,
|
|
207
|
+
and some minor API additions.
|
|
165
208
|
|
|
166
209
|
Bug fixes:
|
|
167
210
|
N.A.
|
|
168
211
|
|
|
169
212
|
Enhancements:
|
|
170
213
|
14447 Unify internal handling of reporting errors to the user
|
|
171
|
-
17514 Provide an Aquarium library-wide logger with configuration parameters and
|
|
214
|
+
17514 Provide an Aquarium library-wide logger with configuration parameters and
|
|
215
|
+
instance overrides
|
|
172
216
|
17515 Add an optional warning if an aspect doesn't match any join points
|
|
173
|
-
17516 Remove unnecessary examples that use :types_and_descendents to shorten
|
|
174
|
-
|
|
175
|
-
|
|
176
|
-
|
|
177
|
-
|
|
178
|
-
|
|
179
|
-
|
|
180
|
-
"
|
|
181
|
-
|
|
182
|
-
|
|
183
|
-
|
|
184
|
-
|
|
185
|
-
|
|
186
|
-
|
|
187
|
-
|
|
188
|
-
|
|
189
|
-
|
|
190
|
-
|
|
191
|
-
|
|
192
|
-
|
|
193
|
-
|
|
194
|
-
|
|
195
|
-
|
|
196
|
-
|
|
197
|
-
|
|
198
|
-
|
|
199
|
-
|
|
217
|
+
17516 Remove unnecessary examples that use :types_and_descendents to shorten
|
|
218
|
+
time to run the RSpec suite
|
|
219
|
+
17565 JoinPoint.new should convert a type name, symbol, or regex to the type
|
|
220
|
+
and only allow one type
|
|
221
|
+
|
|
222
|
+
These first two enhancements are related. There is a now an
|
|
223
|
+
Aquarium::Utils::DefaultLogger module with static accessors for getting and
|
|
224
|
+
setting the "system-wide" logger.
|
|
225
|
+
|
|
226
|
+
When instance-level overrides are necessary, the Aquarium::Utils::OptionsUtils
|
|
227
|
+
provides "universal" options (but currently used only by Aspect and Pointcut)
|
|
228
|
+
for specifying a logger (with the new :logger parameter), or alternatively,
|
|
229
|
+
specifying just the output stream (:logger_stream) and/or the severity
|
|
230
|
+
(:severity, one of the standard library's Logger::Severity-defined constants).
|
|
231
|
+
If either of the latter two options is specified, a separate logger instance is
|
|
232
|
+
created, rather than changing parameters for the global logger.
|
|
233
|
+
|
|
234
|
+
OptionsUtils also supports a :noop parameter, which classes interpret to mean
|
|
235
|
+
do none (or perhaps only some) of the processing. Useful for debugging.
|
|
236
|
+
|
|
237
|
+
#17515 adds a helpful warning to the system (or aspect-instance's) logger if
|
|
238
|
+
an aspect matches no join points. This warning will be suppressed if (i) the
|
|
239
|
+
severity level for the logger is above WARN or (ii) the aspect was created
|
|
240
|
+
with the option :ignore_no_matching_join_points => true.
|
|
241
|
+
|
|
242
|
+
#17516 fixes halved the long execution times for the whole RSpec suite by
|
|
243
|
+
refactoring some examples that used type finders with the :types_and_descendents
|
|
244
|
+
option unnecessarily. It is a very intensive computation! Note that I stubbed
|
|
245
|
+
out these calls using an aspect with around advise, a useful "development-time"
|
|
246
|
+
aspect. See Aquarium::TypeUtilsStub (in spec_example_types.rb) and how it's used
|
|
247
|
+
in pointcut_spec.rb. Using this technique and other optimizations, the time to
|
|
248
|
+
run the suite was reduced from ~5 minutes to about 1 minute.
|
|
200
249
|
|
|
201
|
-
#17565 fixes a "hole" in JoinPoint, where it doesn't confirm that a specified
|
|
202
|
-
or regex matches a class that exists and only one class. Now
|
|
203
|
-
than the original "specification" for it.
|
|
250
|
+
#17565 fixes a "hole" in JoinPoint, where it doesn't confirm that a specified
|
|
251
|
+
type string, symbol or regex matches a class that exists and only one class. Now
|
|
252
|
+
it does and it stores the type, rather than the original "specification" for it.
|
|
204
253
|
|
|
205
254
|
== Version 0.3.0
|
|
206
255
|
|
|
207
|
-
V0.3.0 adds numerous improvements to the DSL, making aspect specification more
|
|
208
|
-
English-like. For example, where you previously wrote, e.g.,
|
|
256
|
+
V0.3.0 adds numerous improvements to the DSL, making aspect specification more
|
|
257
|
+
intuitive and English-like. For example, where you previously wrote, e.g.,
|
|
209
258
|
|
|
210
259
|
around :methods => :all, :types => [Foo, Bar], :advice => advice_proc
|
|
211
260
|
after :attribute => name, :attribute_options => [:readers], :objects => [foo, bar] ...
|
|
@@ -215,8 +264,8 @@ Now you can write the same aspects as follows:
|
|
|
215
264
|
around :calls_to => :all_methods, :within_types => [Foo, Bar], :use_advice => advice_proc
|
|
216
265
|
after :reading => name, :on_objects => [foo, bar] ...
|
|
217
266
|
|
|
218
|
-
Other improvements include performance and robustness enhancements and
|
|
219
|
-
refactoring and DRY improvements.
|
|
267
|
+
Other improvements include performance and robustness enhancements and
|
|
268
|
+
miscellaneous internal refactoring and DRY improvements.
|
|
220
269
|
|
|
221
270
|
Bug fixes:
|
|
222
271
|
16267 gem not updating
|
|
@@ -226,78 +275,89 @@ Enhancements:
|
|
|
226
275
|
|
|
227
276
|
For #17154, the following changes were made:
|
|
228
277
|
Added :all_methods as a synonym for the :all special value.
|
|
229
|
-
Added :reading as a synonym for :attributes => ..., :attribute_options =>
|
|
230
|
-
|
|
278
|
+
Added :reading as a synonym for :attributes => ..., :attribute_options =>
|
|
279
|
+
[:readers]
|
|
280
|
+
Added :writing and :changing as synonyms for :attributes => ...,
|
|
281
|
+
:attribute_options => [:writers]
|
|
231
282
|
Added :accessing as a synonym for :attributes => ...
|
|
232
|
-
Added :calls_to, :calling, :invoking, :sending_message_to as synonyms for
|
|
233
|
-
|
|
234
|
-
|
|
235
|
-
|
|
236
|
-
|
|
237
|
-
The full list of possible synonyms are shown in the spec examples. In particular, see "pointcut_spec.rb".
|
|
283
|
+
Added :calls_to, :calling, :invoking, :sending_message_to as synonyms for
|
|
284
|
+
:methods.
|
|
285
|
+
Added :on_types, :in_types, :within_types and :for_types as synonyms for :types.
|
|
286
|
+
The same set of prefixes is supported for :type, :objects, :object, and the
|
|
287
|
+
various :exclude_*, :*_and_ancestors, and :*_and_descendents.
|
|
238
288
|
|
|
289
|
+
The full list of possible synonyms are shown in the spec examples. In
|
|
290
|
+
particular, see "pointcut_spec.rb".
|
|
239
291
|
|
|
240
292
|
== Version 0.2.0
|
|
241
293
|
|
|
242
|
-
V0.2.0 changes the parameter list used for advice blocks and adds numerous
|
|
243
|
-
improvements and more complete "spec'ing".
|
|
294
|
+
V0.2.0 changes the parameter list used for advice blocks and adds numerous
|
|
295
|
+
enhancements, robustness improvements and more complete "spec'ing".
|
|
244
296
|
|
|
245
297
|
Bug fixes:
|
|
246
298
|
none
|
|
247
299
|
|
|
248
300
|
Enhancements:
|
|
249
|
-
13402 Support a subclass syntax like AspectJ's "Type+"
|
|
250
|
-
13984 More flexible argument list to the advise block
|
|
251
|
-
14053 Remove JoinPoint#type, JoinPoint#type=, JoinPoint#object, and
|
|
301
|
+
13402 Support a subclass syntax like AspectJ's "Type+".
|
|
302
|
+
13984 More flexible argument list to the advise block.
|
|
303
|
+
14053 Remove JoinPoint#type, JoinPoint#type=, JoinPoint#object, and
|
|
304
|
+
JoinPoint#object=
|
|
252
305
|
14061 Add a control flow mechanism to skipping (sic) intermediate advice
|
|
253
306
|
15164 Deprecate ObjectFinder
|
|
254
307
|
15413 Remove ObjectFinder
|
|
255
308
|
15710 Eliminate redundant public methods in various "finders"
|
|
256
309
|
|
|
257
|
-
#13402 adds new invocation options to specify types and their descendents
|
|
258
|
-
that include the specified module(s)) and ancestors.
|
|
259
|
-
|
|
260
|
-
|
|
261
|
-
|
|
262
|
-
the
|
|
263
|
-
|
|
264
|
-
:
|
|
265
|
-
|
|
266
|
-
|
|
267
|
-
|
|
268
|
-
|
|
269
|
-
:
|
|
270
|
-
|
|
271
|
-
|
|
272
|
-
|
|
273
|
-
|
|
274
|
-
|
|
275
|
-
|
|
276
|
-
|
|
277
|
-
the
|
|
278
|
-
|
|
279
|
-
|
|
280
|
-
|
|
281
|
-
|
|
282
|
-
|
|
283
|
-
|
|
284
|
-
|
|
285
|
-
|
|
286
|
-
|
|
287
|
-
|
|
288
|
-
|
|
289
|
-
|
|
290
|
-
|
|
291
|
-
|
|
292
|
-
|
|
293
|
-
|
|
294
|
-
|
|
295
|
-
|
|
310
|
+
#13402 adds new invocation options to specify types and their descendents
|
|
311
|
+
(subclasses or modules that include the specified module(s)) and ancestors.
|
|
312
|
+
The latter should be used cautiously as it will include things like Kernel,
|
|
313
|
+
Object, and Class! I used new command options rather than the AspectJ "+"
|
|
314
|
+
suffix (and the proposed, but never implemented "-" suffix for ancestors),
|
|
315
|
+
because the "+" would be confusing with regular expressions and not in the
|
|
316
|
+
spirit of trying to make the pointcut language "easy to read". So, the
|
|
317
|
+
following are now available:
|
|
318
|
+
:type_and_ancestors
|
|
319
|
+
:types_and_ancestors
|
|
320
|
+
:type_and_descendents
|
|
321
|
+
:types_and_descendents
|
|
322
|
+
And the corresponding:
|
|
323
|
+
:exclude_type_and_ancestors
|
|
324
|
+
:exclude_types_and_ancestors
|
|
325
|
+
:exclude_type_and_descendents
|
|
326
|
+
:exclude_types_and_descendents
|
|
327
|
+
If you want both the ancestors and descendents, just use both options with
|
|
328
|
+
the same value.
|
|
329
|
+
|
|
330
|
+
#13984 adds the object as the second argument to the advice block parameter
|
|
331
|
+
list. This change reflects the fact that the object is often needed, but
|
|
332
|
+
calling jp.context.advised_object is a bit tedious. THIS CHANGE BREAKS
|
|
333
|
+
BACKWARDS COMPATIBILITY!! An exception is raised if advice has the signature
|
|
334
|
+
|jp, *args|.
|
|
335
|
+
|
|
336
|
+
#14061 adds a new method, JoinPoint#invoke_original_join_point, which will
|
|
337
|
+
invoke the original method without intermediate advice. If called within
|
|
338
|
+
around advice, you can write advice that vetoes all subsequent advice, yet
|
|
339
|
+
invokes the original method. Use this technique cautiously, however, since you
|
|
340
|
+
may not always know what other advices are involved and what side effects this
|
|
341
|
+
control-flow change might cause.
|
|
342
|
+
|
|
343
|
+
#15164 and 15413 remove ObjectFinder because it is not used and it requires
|
|
344
|
+
ObjectSpace, which has high overhead and won't be enabled, by default, in
|
|
345
|
+
JRuby (it will be optional).
|
|
346
|
+
|
|
347
|
+
#15710 removes redundant methods that were becoming a maintenance issue, in
|
|
348
|
+
particular, MethodFinder#find_all_by and TypeFinder#find_by_name. This is a
|
|
349
|
+
non-backwards-compatible API change.
|
|
350
|
+
|
|
351
|
+
Finally, note that I have not yet been able to resolve bug #15202, "Intermittent
|
|
352
|
+
confusion between classes and objects when invoking advice." I believe this is a
|
|
353
|
+
very rare occurrence and only likely to ever happen during the "torture-test" of
|
|
354
|
+
running the RSpec suite. Please post a comment to Tracker if you encounter it!
|
|
296
355
|
|
|
297
356
|
== Version 0.1.8
|
|
298
357
|
|
|
299
|
-
V0.1.7 did not successfully "register" at rubyforge. This releases fixes that
|
|
300
|
-
adds several feature enhancements and refactorings. There are
|
|
358
|
+
V0.1.7 did not successfully "register" at rubyforge. This releases fixes that
|
|
359
|
+
problem and also adds several feature enhancements and refactorings. There are
|
|
360
|
+
no known upgrade issues.
|
|
301
361
|
|
|
302
362
|
Bug fixes:
|
|
303
363
|
none
|
|
@@ -306,13 +366,15 @@ Enhancements:
|
|
|
306
366
|
13399 Add :exclusion options for methods and types.
|
|
307
367
|
14707 :exclude_ancestor_methods as synonym for :suppress_ancestor_methods
|
|
308
368
|
|
|
309
|
-
13399 adds new :exclude_(pointcuts|join_points|types|objects|methods|attributes)
|
|
310
|
-
Aspect.new, and Pointcut.new that make it easier to specify a list
|
|
311
|
-
various "items" and then to exclude particular
|
|
312
|
-
|
|
369
|
+
13399 adds new :exclude_(pointcuts|join_points|types|objects|methods|attributes)
|
|
370
|
+
options for Aspect.new, and Pointcut.new that make it easier to specify a list
|
|
371
|
+
or regular expression for various "items" and then to exclude particular
|
|
372
|
+
items, e.g., Aspect.new :around, :types => /nterestingType/, :exclude_types =>
|
|
373
|
+
UninterestingType ...
|
|
313
374
|
|
|
314
|
-
The :exclude_ancestor_methods option is now preferred over
|
|
315
|
-
since the former is more consistent with the new
|
|
375
|
+
The :exclude_ancestor_methods option is now preferred over
|
|
376
|
+
:suppress_ancestor_methods, since the former is more consistent with the new
|
|
377
|
+
:exclude_* options.
|
|
316
378
|
|
|
317
379
|
== Version 0.1.7
|
|
318
380
|
|
|
@@ -320,47 +382,51 @@ Bug fixes:
|
|
|
320
382
|
14946 Advice fails when instrumenting methods containing special characters
|
|
321
383
|
15038 Spec for pointcut example variation #1
|
|
322
384
|
15039 Spec for pointcut example variation #2
|
|
323
|
-
15085 Specifying just :attributes for aspects also matches all methods, as if
|
|
385
|
+
15085 Specifying just :attributes for aspects also matches all methods, as if
|
|
386
|
+
:methods => :all specified
|
|
324
387
|
|
|
325
388
|
Enhancements:
|
|
326
389
|
13396 Unify internal handling of types vs. objects
|
|
327
390
|
|
|
328
|
-
15038 and 15039 were bugs in one of the examples (actually in the comments).
|
|
329
|
-
experimenting with them also revealed the nasty 15085 bug!
|
|
330
|
-
|
|
331
|
-
I previously handled some special characters in method names, but not all the possible
|
|
332
|
-
ones, hence 14946. Aquarium should now properly handle any valid Ruby method name.
|
|
391
|
+
15038 and 15039 were bugs in one of the examples (actually in the comments).
|
|
392
|
+
However, experimenting with them also revealed the nasty 15085 bug!
|
|
333
393
|
|
|
394
|
+
I previously handled some special characters in method names, but not all the
|
|
395
|
+
possible ones, hence 14946. Aquarium should now properly handle any valid Ruby
|
|
396
|
+
method name.
|
|
334
397
|
|
|
335
398
|
== Version 0.1.6
|
|
336
399
|
|
|
337
400
|
Bug fixes:
|
|
338
|
-
14353 Advising subclass method that calls super raises exception when method
|
|
339
|
-
|
|
401
|
+
14353 Advising subclass method that calls super raises exception when method
|
|
402
|
+
executed
|
|
403
|
+
14356 Regexps for types must cover the whole name, which is inconsistent with
|
|
404
|
+
method/attribute regexps
|
|
340
405
|
14384 Design by Contract "extra" does not return correct value "invar" handling
|
|
341
406
|
13410 Fix funky navigation bar on website
|
|
342
407
|
|
|
343
|
-
14353 was kind of bad, but it's actually a Ruby bug with a good workaround. If
|
|
344
|
-
advised a method that called "super", Ruby would use the wrong method name
|
|
345
|
-
class in the parent. See the bug description for the details.
|
|
408
|
+
14353 was kind of bad, but it's actually a Ruby bug with a good workaround. If
|
|
409
|
+
you advised a method that called "super", Ruby would use the wrong method name
|
|
410
|
+
to lookup the class in the parent. See the bug description for the details.
|
|
346
411
|
|
|
347
|
-
For 14356, type regular expressions now match on parts of names; they don't have
|
|
348
|
-
the whole name. The exception is regular expressions with module
|
|
349
|
-
case, it seems to make more sense for the regular
|
|
350
|
-
If the expression is /A::B::C::D/,
|
|
351
|
-
|
|
352
|
-
the
|
|
412
|
+
For 14356, type regular expressions now match on parts of names; they don't have
|
|
413
|
+
to match the whole name. The exception is regular expressions with module
|
|
414
|
+
separators "::". In this case, it seems to make more sense for the regular
|
|
415
|
+
expression to be interpreted as follows: If the expression is /A::B::C::D/,
|
|
416
|
+
then for the the outermost types, the expression behaves as /^.*A/, for the
|
|
417
|
+
types between two "::", the expressions behave as /^B$/ and /^C$/, and the
|
|
418
|
+
trailing expression behaves as /D.*$/.
|
|
353
419
|
|
|
354
|
-
14384 was an easy mistake to make with "around" advice; you have to remember to
|
|
355
|
-
result of the "join_point.proceed" call, unless you specifically
|
|
356
|
-
value! Here are two ways to do it:
|
|
420
|
+
14384 was an easy mistake to make with "around" advice; you have to remember to
|
|
421
|
+
return the result of the "join_point.proceed" call, unless you specifically
|
|
422
|
+
want to change the returned value! Here are two ways to do it:
|
|
357
423
|
|
|
358
424
|
do_something_before(...)
|
|
359
425
|
result = join_point.proceed
|
|
360
426
|
do_something_after(...)
|
|
361
427
|
return result
|
|
362
428
|
|
|
363
|
-
or
|
|
429
|
+
or
|
|
364
430
|
|
|
365
431
|
begin
|
|
366
432
|
do_something_before(...)
|
|
@@ -369,8 +435,9 @@ or
|
|
|
369
435
|
do_something_after(...)
|
|
370
436
|
end
|
|
371
437
|
|
|
372
|
-
The latter approach looks "asymmetrical" and it will behave differently if
|
|
373
|
-
However, it eliminates the temporary, if you find that
|
|
438
|
+
The latter approach looks "asymmetrical" and it will behave differently if
|
|
439
|
+
"proceed" raises! However, it eliminates the temporary, if you find that
|
|
440
|
+
desirable.
|
|
374
441
|
|
|
375
442
|
Enhancements:
|
|
376
443
|
13407 Pick a better method name for JoinPoint#type, which hides the Module#type
|
|
@@ -384,22 +451,25 @@ For 13407, new attribute methods have been added
|
|
|
384
451
|
* JoinPoint#target_object return the object that the join_point matches.
|
|
385
452
|
* JoinPoint#target_object= set the object that the join_point matches.
|
|
386
453
|
|
|
387
|
-
The following, older methods are now deprecated and will be removed in the 0.2.0
|
|
454
|
+
The following, older methods are now deprecated and will be removed in the 0.2.0
|
|
455
|
+
release (#14053):
|
|
388
456
|
* JoinPoint#type
|
|
389
457
|
* JoinPoint#type=
|
|
390
458
|
* JoinPoint#object
|
|
391
459
|
* JoinPoint#object=
|
|
392
460
|
|
|
393
|
-
JoinPoint#type method is deprecated because it hides Module#type, which returns
|
|
394
|
-
the corresponding object. For "symmetry", the other three methods
|
|
395
|
-
they will be removed in a future release. Until
|
|
396
|
-
STDOUT. (If you really want the type
|
|
397
|
-
|
|
461
|
+
JoinPoint#type method is deprecated because it hides Module#type, which returns
|
|
462
|
+
the type of the corresponding object. For "symmetry", the other three methods
|
|
463
|
+
are also now deprecated and they will be removed in a future release. Until
|
|
464
|
+
then, all will print a warning message to STDOUT. (If you really want the type
|
|
465
|
+
of what could be a JoinPoint object, you should use #class anyway, as
|
|
466
|
+
Module#type is also deprecated!)
|
|
398
467
|
|
|
399
468
|
== Version 0.1.5
|
|
400
469
|
|
|
401
470
|
Bug fixes:
|
|
402
|
-
13514 Protected and private methods are made public when advised and left that
|
|
471
|
+
13514 Protected and private methods are made public when advised and left that
|
|
472
|
+
way when unadvised
|
|
403
473
|
13650 Loading Aquarium interferes with Rails filters
|
|
404
474
|
13864 Bug with negative object_id
|
|
405
475
|
|
|
@@ -407,21 +477,24 @@ Enhancements:
|
|
|
407
477
|
13392 Convert examples to specs.
|
|
408
478
|
13463 Support running in JRuby
|
|
409
479
|
|
|
410
|
-
Fixing 13650 required an API change, which is why I've tagged this release
|
|
411
|
-
something like "0.1.1" (and the changes don't seem big enough
|
|
480
|
+
Fixing 13650 required an API change, which is why I've tagged this release
|
|
481
|
+
"0.1.5" instead of something like "0.1.1" (and the changes don't seem big enough
|
|
482
|
+
to warrant "0.2.0"...).
|
|
412
483
|
|
|
413
|
-
Previously, requiring "aquarium.rb" in the top-level "lib" directory would
|
|
414
|
-
lib/aquarium/dsl/aspect_dsl.rb, which
|
|
415
|
-
|
|
416
|
-
Unfortunately, those methods collide with methods of the same name that Rails
|
|
417
|
-
also a bit presumptuous of me to assume that everyone
|
|
484
|
+
Previously, requiring "aquarium.rb" in the top-level "lib" directory would
|
|
485
|
+
implicitly require lib/aquarium/dsl/aspect_dsl.rb, which has Object include the
|
|
486
|
+
AspectDSL module. This module adds methods like :before and :after to Object.
|
|
487
|
+
Unfortunately, those methods collide with methods of the same name that Rails
|
|
488
|
+
adds to Object. It was also a bit presumptuous of me to assume that everyone
|
|
489
|
+
wanted those methods on Object ;)
|
|
418
490
|
|
|
419
|
-
In this release, aspect_dsl.rb is still implicitly included and it still defines
|
|
420
|
-
module. Now, however, it does not include the AspectDSL module in
|
|
421
|
-
behavior for all types, you must require the
|
|
491
|
+
In this release, aspect_dsl.rb is still implicitly included and it still defines
|
|
492
|
+
the AspectDSL module. Now, however, it does not include the AspectDSL module in
|
|
493
|
+
Object. Instead, if you want this behavior for all types, you must require the
|
|
494
|
+
new lib/aquarium/aspects/dsl/object_dsl.rb explicitly.
|
|
422
495
|
|
|
423
|
-
As an alternative, if you just want the AspectDSL module included selectively in
|
|
424
|
-
then do the following:
|
|
496
|
+
As an alternative, if you just want the AspectDSL module included selectively in
|
|
497
|
+
certain types, then do the following:
|
|
425
498
|
|
|
426
499
|
class MyClass # reopen "MyClass"
|
|
427
500
|
# Add the methods as _class_ methods
|
|
@@ -443,11 +516,13 @@ To add the methods as _instance_ methods on individual objects:
|
|
|
443
516
|
object.extend(Aquarium::DSL)
|
|
444
517
|
|
|
445
518
|
|
|
446
|
-
Note: as discussed at
|
|
519
|
+
Note: as discussed at
|
|
520
|
+
http://practicalruby.blogspot.com/2007/02/reopen-with-moduleeval.html,
|
|
447
521
|
using "class_eval" or "module_eval" is safer that just reopening a class if
|
|
448
|
-
you're not sure that "MyClass" has actually been defined yet. However, in our
|
|
449
|
-
probably doesn't matter, as AspectDSL doesn't change
|
|
450
|
-
methods. Still, we can't
|
|
522
|
+
you're not sure that "MyClass" has actually been defined yet. However, in our
|
|
523
|
+
particular case, it probably doesn't matter, as AspectDSL doesn't change
|
|
524
|
+
anything about the type, like aliasing existing methods. Still, we can't
|
|
525
|
+
guarantee that this won't change in the future.
|
|
451
526
|
|
|
452
527
|
== Version 0.1.0
|
|
453
528
|
|