1
|
# Prophecy
|
2
|
|
3
|
[](https://packagist.org/packages/phpspec/prophecy)
|
4
|
[](https://travis-ci.org/phpspec/prophecy)
|
5
|
|
6
|
Prophecy is a highly opinionated yet very powerful and flexible PHP object mocking
|
7
|
framework. Though initially it was created to fulfil phpspec2 needs, it is flexible
|
8
|
enough to be used inside any testing framework out there with minimal effort.
|
9
|
|
10
|
## A simple example
|
11
|
|
12
|
```php
|
13
|
<?php
|
14
|
|
15
|
class UserTest extends PHPUnit_Framework_TestCase
|
16
|
{
|
17
|
private $prophet;
|
18
|
|
19
|
public function testPasswordHashing()
|
20
|
{
|
21
|
$hasher = $this->prophet->prophesize('App\Security\Hasher');
|
22
|
$user = new App\Entity\User($hasher->reveal());
|
23
|
|
24
|
$hasher->generateHash($user, 'qwerty')->willReturn('hashed_pass');
|
25
|
|
26
|
$user->setPassword('qwerty');
|
27
|
|
28
|
$this->assertEquals('hashed_pass', $user->getPassword());
|
29
|
}
|
30
|
|
31
|
protected function setup()
|
32
|
{
|
33
|
$this->prophet = new \Prophecy\Prophet;
|
34
|
}
|
35
|
|
36
|
protected function tearDown()
|
37
|
{
|
38
|
$this->prophet->checkPredictions();
|
39
|
}
|
40
|
}
|
41
|
```
|
42
|
|
43
|
## Installation
|
44
|
|
45
|
### Prerequisites
|
46
|
|
47
|
Prophecy requires PHP 5.3.3 or greater.
|
48
|
|
49
|
### Setup through composer
|
50
|
|
51
|
First, add Prophecy to the list of dependencies inside your `composer.json`:
|
52
|
|
53
|
```json
|
54
|
{
|
55
|
"require-dev": {
|
56
|
"phpspec/prophecy": "~1.0"
|
57
|
}
|
58
|
}
|
59
|
```
|
60
|
|
61
|
Then simply install it with composer:
|
62
|
|
63
|
```bash
|
64
|
$> composer install --prefer-dist
|
65
|
```
|
66
|
|
67
|
You can read more about Composer on its [official webpage](http://getcomposer.org).
|
68
|
|
69
|
## How to use it
|
70
|
|
71
|
First of all, in Prophecy every word has a logical meaning, even the name of the library
|
72
|
itself (Prophecy). When you start feeling that, you'll become very fluid with this
|
73
|
tool.
|
74
|
|
75
|
For example, Prophecy has been named that way because it concentrates on describing the future
|
76
|
behavior of objects with very limited knowledge about them. But as with any other prophecy,
|
77
|
those object prophecies can't create themselves - there should be a Prophet:
|
78
|
|
79
|
```php
|
80
|
$prophet = new Prophecy\Prophet;
|
81
|
```
|
82
|
|
83
|
The Prophet creates prophecies by *prophesizing* them:
|
84
|
|
85
|
```php
|
86
|
$prophecy = $prophet->prophesize();
|
87
|
```
|
88
|
|
89
|
The result of the `prophesize()` method call is a new object of class `ObjectProphecy`. Yes,
|
90
|
that's your specific object prophecy, which describes how your object would behave
|
91
|
in the near future. But first, you need to specify which object you're talking about,
|
92
|
right?
|
93
|
|
94
|
```php
|
95
|
$prophecy->willExtend('stdClass');
|
96
|
$prophecy->willImplement('SessionHandlerInterface');
|
97
|
```
|
98
|
|
99
|
There are 2 interesting calls - `willExtend` and `willImplement`. The first one tells
|
100
|
object prophecy that our object should extend specific class, the second one says that
|
101
|
it should implement some interface. Obviously, objects in PHP can implement multiple
|
102
|
interfaces, but extend only one parent class.
|
103
|
|
104
|
### Dummies
|
105
|
|
106
|
Ok, now we have our object prophecy. What can we do with it? First of all, we can get
|
107
|
our object *dummy* by revealing its prophecy:
|
108
|
|
109
|
```php
|
110
|
$dummy = $prophecy->reveal();
|
111
|
```
|
112
|
|
113
|
The `$dummy` variable now holds a special dummy object. Dummy objects are objects that extend
|
114
|
and/or implement preset classes/interfaces by overriding all their public methods. The key
|
115
|
point about dummies is that they do not hold any logic - they just do nothing. Any method
|
116
|
of the dummy will always return `null` and the dummy will never throw any exceptions.
|
117
|
Dummy is your friend if you don't care about the actual behavior of this double and just need
|
118
|
a token object to satisfy a method typehint.
|
119
|
|
120
|
You need to understand one thing - a dummy is not a prophecy. Your object prophecy is still
|
121
|
assigned to `$prophecy` variable and in order to manipulate with your expectations, you
|
122
|
should work with it. `$dummy` is a dummy - a simple php object that tries to fulfil your
|
123
|
prophecy.
|
124
|
|
125
|
### Stubs
|
126
|
|
127
|
Ok, now we know how to create basic prophecies and reveal dummies from them. That's
|
128
|
awesome if we don't care about our _doubles_ (objects that reflect originals)
|
129
|
interactions. If we do, we need to use *stubs* or *mocks*.
|
130
|
|
131
|
A stub is an object double, which doesn't have any expectations about the object behavior,
|
132
|
but when put in specific environment, behaves in specific way. Ok, I know, it's cryptic,
|
133
|
but bear with me for a minute. Simply put, a stub is a dummy, which depending on the called
|
134
|
method signature does different things (has logic). To create stubs in Prophecy:
|
135
|
|
136
|
```php
|
137
|
$prophecy->read('123')->willReturn('value');
|
138
|
```
|
139
|
|
140
|
Oh wow. We've just made an arbitrary call on the object prophecy? Yes, we did. And this
|
141
|
call returned us a new object instance of class `MethodProphecy`. Yep, that's a specific
|
142
|
method with arguments prophecy. Method prophecies give you the ability to create method
|
143
|
promises or predictions. We'll talk about method predictions later in the _Mocks_ section.
|
144
|
|
145
|
#### Promises
|
146
|
|
147
|
Promises are logical blocks, that represent your fictional methods in prophecy terms
|
148
|
and they are handled by the `MethodProphecy::will(PromiseInterface $promise)` method.
|
149
|
As a matter of fact, the call that we made earlier (`willReturn('value')`) is a simple
|
150
|
shortcut to:
|
151
|
|
152
|
```php
|
153
|
$prophecy->read('123')->will(new Prophecy\Promise\ReturnPromise(array('value')));
|
154
|
```
|
155
|
|
156
|
This promise will cause any call to our double's `read()` method with exactly one
|
157
|
argument - `'123'` to always return `'value'`. But that's only for this
|
158
|
promise, there's plenty others you can use:
|
159
|
|
160
|
- `ReturnPromise` or `->willReturn(1)` - returns a value from a method call
|
161
|
- `ReturnArgumentPromise` or `->willReturnArgument($index)` - returns the nth method argument from call
|
162
|
- `ThrowPromise` or `->willThrow($exception)` - causes the method to throw specific exception
|
163
|
- `CallbackPromise` or `->will($callback)` - gives you a quick way to define your own custom logic
|
164
|
|
165
|
Keep in mind, that you can always add even more promises by implementing
|
166
|
`Prophecy\Promise\PromiseInterface`.
|
167
|
|
168
|
#### Method prophecies idempotency
|
169
|
|
170
|
Prophecy enforces same method prophecies and, as a consequence, same promises and
|
171
|
predictions for the same method calls with the same arguments. This means:
|
172
|
|
173
|
```php
|
174
|
$methodProphecy1 = $prophecy->read('123');
|
175
|
$methodProphecy2 = $prophecy->read('123');
|
176
|
$methodProphecy3 = $prophecy->read('321');
|
177
|
|
178
|
$methodProphecy1 === $methodProphecy2;
|
179
|
$methodProphecy1 !== $methodProphecy3;
|
180
|
```
|
181
|
|
182
|
That's interesting, right? Now you might ask me how would you define more complex
|
183
|
behaviors where some method call changes behavior of others. In PHPUnit or Mockery
|
184
|
you do that by predicting how many times your method will be called. In Prophecy,
|
185
|
you'll use promises for that:
|
186
|
|
187
|
```php
|
188
|
$user->getName()->willReturn(null);
|
189
|
|
190
|
// For PHP 5.4
|
191
|
$user->setName('everzet')->will(function () {
|
192
|
$this->getName()->willReturn('everzet');
|
193
|
});
|
194
|
|
195
|
// For PHP 5.3
|
196
|
$user->setName('everzet')->will(function ($args, $user) {
|
197
|
$user->getName()->willReturn('everzet');
|
198
|
});
|
199
|
|
200
|
// Or
|
201
|
$user->setName('everzet')->will(function ($args) use ($user) {
|
202
|
$user->getName()->willReturn('everzet');
|
203
|
});
|
204
|
```
|
205
|
|
206
|
And now it doesn't matter how many times or in which order your methods are called.
|
207
|
What matters is their behaviors and how well you faked it.
|
208
|
|
209
|
#### Arguments wildcarding
|
210
|
|
211
|
The previous example is awesome (at least I hope it is for you), but that's not
|
212
|
optimal enough. We hardcoded `'everzet'` in our expectation. Isn't there a better
|
213
|
way? In fact there is, but it involves understanding what this `'everzet'`
|
214
|
actually is.
|
215
|
|
216
|
You see, even if method arguments used during method prophecy creation look
|
217
|
like simple method arguments, in reality they are not. They are argument token
|
218
|
wildcards. As a matter of fact, `->setName('everzet')` looks like a simple call just
|
219
|
because Prophecy automatically transforms it under the hood into:
|
220
|
|
221
|
```php
|
222
|
$user->setName(new Prophecy\Argument\Token\ExactValueToken('everzet'));
|
223
|
```
|
224
|
|
225
|
Those argument tokens are simple PHP classes, that implement
|
226
|
`Prophecy\Argument\Token\TokenInterface` and tell Prophecy how to compare real arguments
|
227
|
with your expectations. And yes, those classnames are damn big. That's why there's a
|
228
|
shortcut class `Prophecy\Argument`, which you can use to create tokens like that:
|
229
|
|
230
|
```php
|
231
|
use Prophecy\Argument;
|
232
|
|
233
|
$user->setName(Argument::exact('everzet'));
|
234
|
```
|
235
|
|
236
|
`ExactValueToken` is not very useful in our case as it forced us to hardcode the username.
|
237
|
That's why Prophecy comes bundled with a bunch of other tokens:
|
238
|
|
239
|
- `IdenticalValueToken` or `Argument::is($value)` - checks that the argument is identical to a specific value
|
240
|
- `ExactValueToken` or `Argument::exact($value)` - checks that the argument matches a specific value
|
241
|
- `TypeToken` or `Argument::type($typeOrClass)` - checks that the argument matches a specific type or
|
242
|
classname
|
243
|
- `ObjectStateToken` or `Argument::which($method, $value)` - checks that the argument method returns
|
244
|
a specific value
|
245
|
- `CallbackToken` or `Argument::that(callback)` - checks that the argument matches a custom callback
|
246
|
- `AnyValueToken` or `Argument::any()` - matches any argument
|
247
|
- `AnyValuesToken` or `Argument::cetera()` - matches any arguments to the rest of the signature
|
248
|
- `StringContainsToken` or `Argument::containingString($value)` - checks that the argument contains a specific string value
|
249
|
|
250
|
And you can add even more by implementing `TokenInterface` with your own custom classes.
|
251
|
|
252
|
So, let's refactor our initial `{set,get}Name()` logic with argument tokens:
|
253
|
|
254
|
```php
|
255
|
use Prophecy\Argument;
|
256
|
|
257
|
$user->getName()->willReturn(null);
|
258
|
|
259
|
// For PHP 5.4
|
260
|
$user->setName(Argument::type('string'))->will(function ($args) {
|
261
|
$this->getName()->willReturn($args[0]);
|
262
|
});
|
263
|
|
264
|
// For PHP 5.3
|
265
|
$user->setName(Argument::type('string'))->will(function ($args, $user) {
|
266
|
$user->getName()->willReturn($args[0]);
|
267
|
});
|
268
|
|
269
|
// Or
|
270
|
$user->setName(Argument::type('string'))->will(function ($args) use ($user) {
|
271
|
$user->getName()->willReturn($args[0]);
|
272
|
});
|
273
|
```
|
274
|
|
275
|
That's it. Now our `{set,get}Name()` prophecy will work with any string argument provided to it.
|
276
|
We've just described how our stub object should behave, even though the original object could have
|
277
|
no behavior whatsoever.
|
278
|
|
279
|
One last bit about arguments now. You might ask, what happens in case of:
|
280
|
|
281
|
```php
|
282
|
use Prophecy\Argument;
|
283
|
|
284
|
$user->getName()->willReturn(null);
|
285
|
|
286
|
// For PHP 5.4
|
287
|
$user->setName(Argument::type('string'))->will(function ($args) {
|
288
|
$this->getName()->willReturn($args[0]);
|
289
|
});
|
290
|
|
291
|
// For PHP 5.3
|
292
|
$user->setName(Argument::type('string'))->will(function ($args, $user) {
|
293
|
$user->getName()->willReturn($args[0]);
|
294
|
});
|
295
|
|
296
|
// Or
|
297
|
$user->setName(Argument::type('string'))->will(function ($args) use ($user) {
|
298
|
$user->getName()->willReturn($args[0]);
|
299
|
});
|
300
|
|
301
|
$user->setName(Argument::any())->will(function () {
|
302
|
});
|
303
|
```
|
304
|
|
305
|
Nothing. Your stub will continue behaving the way it did before. That's because of how
|
306
|
arguments wildcarding works. Every argument token type has a different score level, which
|
307
|
wildcard then uses to calculate the final arguments match score and use the method prophecy
|
308
|
promise that has the highest score. In this case, `Argument::type()` in case of success
|
309
|
scores `5` and `Argument::any()` scores `3`. So the type token wins, as does the first
|
310
|
`setName()` method prophecy and its promise. The simple rule of thumb - more precise token
|
311
|
always wins.
|
312
|
|
313
|
#### Getting stub objects
|
314
|
|
315
|
Ok, now we know how to define our prophecy method promises, let's get our stub from
|
316
|
it:
|
317
|
|
318
|
```php
|
319
|
$stub = $prophecy->reveal();
|
320
|
```
|
321
|
|
322
|
As you might see, the only difference between how we get dummies and stubs is that with
|
323
|
stubs we describe every object conversation instead of just agreeing with `null` returns
|
324
|
(object being *dummy*). As a matter of fact, after you define your first promise
|
325
|
(method call), Prophecy will force you to define all the communications - it throws
|
326
|
the `UnexpectedCallException` for any call you didn't describe with object prophecy before
|
327
|
calling it on a stub.
|
328
|
|
329
|
### Mocks
|
330
|
|
331
|
Now we know how to define doubles without behavior (dummies) and doubles with behavior, but
|
332
|
no expectations (stubs). What's left is doubles for which we have some expectations. These
|
333
|
are called mocks and in Prophecy they look almost exactly the same as stubs, except that
|
334
|
they define *predictions* instead of *promises* on method prophecies:
|
335
|
|
336
|
```php
|
337
|
$entityManager->flush()->shouldBeCalled();
|
338
|
```
|
339
|
|
340
|
#### Predictions
|
341
|
|
342
|
The `shouldBeCalled()` method here assigns `CallPrediction` to our method prophecy.
|
343
|
Predictions are a delayed behavior check for your prophecies. You see, during the entire lifetime
|
344
|
of your doubles, Prophecy records every single call you're making against it inside your
|
345
|
code. After that, Prophecy can use this collected information to check if it matches defined
|
346
|
predictions. You can assign predictions to method prophecies using the
|
347
|
`MethodProphecy::should(PredictionInterface $prediction)` method. As a matter of fact,
|
348
|
the `shouldBeCalled()` method we used earlier is just a shortcut to:
|
349
|
|
350
|
```php
|
351
|
$entityManager->flush()->should(new Prophecy\Prediction\CallPrediction());
|
352
|
```
|
353
|
|
354
|
It checks if your method of interest (that matches both the method name and the arguments wildcard)
|
355
|
was called 1 or more times. If the prediction failed then it throws an exception. When does this
|
356
|
check happen? Whenever you call `checkPredictions()` on the main Prophet object:
|
357
|
|
358
|
```php
|
359
|
$prophet->checkPredictions();
|
360
|
```
|
361
|
|
362
|
In PHPUnit, you would want to put this call into the `tearDown()` method. If no predictions
|
363
|
are defined, it would do nothing. So it won't harm to call it after every test.
|
364
|
|
365
|
There are plenty more predictions you can play with:
|
366
|
|
367
|
- `CallPrediction` or `shouldBeCalled()` - checks that the method has been called 1 or more times
|
368
|
- `NoCallsPrediction` or `shouldNotBeCalled()` - checks that the method has not been called
|
369
|
- `CallTimesPrediction` or `shouldBeCalledTimes($count)` - checks that the method has been called
|
370
|
`$count` times
|
371
|
- `CallbackPrediction` or `should($callback)` - checks the method against your own custom callback
|
372
|
|
373
|
Of course, you can always create your own custom prediction any time by implementing
|
374
|
`PredictionInterface`.
|
375
|
|
376
|
### Spies
|
377
|
|
378
|
The last bit of awesomeness in Prophecy is out-of-the-box spies support. As I said in the previous
|
379
|
section, Prophecy records every call made during the double's entire lifetime. This means
|
380
|
you don't need to record predictions in order to check them. You can also do it
|
381
|
manually by using the `MethodProphecy::shouldHave(PredictionInterface $prediction)` method:
|
382
|
|
383
|
```php
|
384
|
$em = $prophet->prophesize('Doctrine\ORM\EntityManager');
|
385
|
|
386
|
$controller->createUser($em->reveal());
|
387
|
|
388
|
$em->flush()->shouldHaveBeenCalled();
|
389
|
```
|
390
|
|
391
|
Such manipulation with doubles is called spying. And with Prophecy it just works.
|