* 2.1:
Restrict Monolog version to be in version <1.3
[Console] Make getTerminalWith & getTerminalHeight public
[DependencyInjection] fixed PhpDumper optimizations when an inlined service depends on the current one indirectly
[DependencyInjection] fixed PhpDumper when an inlined service definition has some properties
[DependencyInjection] added some tests for PhpDumper when the container is compiled
[DependencyInjection] fixed CS
[Process] Do not reset stdout/stderr pipes on Interrupted system call
[Locale] Adjust `StubIntlDateFormatter` to have new methods added in PHP 5.5
use the right RequestMatcherInterface
[Locale] Fix failing `StubIntlDateFormatter` tests in PHP 5.5
[Locale] Fix failing `StubIntlDateFormatter` in PHP 5.5
[Form] Fix failing `MonthChoiceList` in PHP 5.5
Update .travis.yml
Conflicts:
src/Symfony/Bridge/Monolog/composer.json
src/Symfony/Component/DependencyInjection/Tests/Fixtures/php/services9.php
* 2.0:
Restrict Monolog version to be in version <1.3
[Console] Make getTerminalWith & getTerminalHeight public
[DependencyInjection] fixed PhpDumper optimizations when an inlined service depends on the current one indirectly
[DependencyInjection] fixed PhpDumper when an inlined service definition has some properties
[DependencyInjection] added some tests for PhpDumper when the container is compiled
[DependencyInjection] fixed CS
[Locale] Adjust `StubIntlDateFormatter` to have new methods added in PHP 5.5
[Locale] Fix failing `StubIntlDateFormatter` tests in PHP 5.5
[Locale] Fix failing `StubIntlDateFormatter` in PHP 5.5
[Form] Fix failing `MonthChoiceList` in PHP 5.5
Update .travis.yml
Conflicts:
src/Symfony/Bridge/Monolog/composer.json
src/Symfony/Component/DependencyInjection/Tests/Fixtures/php/services9.php
src/Symfony/Component/Form/Extension/Core/ChoiceList/MonthChoiceList.php
tests/Symfony/Tests/Component/DependencyInjection/Fixtures/yaml/services9.yml
Commits
-------
e271b17 Remove the string optimization since it causes no real performance gain but increases generation time of the dumped PHP Container
Discussion
----------
PhpDumper and large strings
When the PhpDumper is dealing with longer strings, the regular expression performed to optimize this can be quite a performance hog.
In our case sometimes the dumper takes more then 30 seconds if leaving this enabled. Disabling it will bring it back to sub-second speed.
This patch makes the optimization optional by passing in an additional container option.
---------------------------------------------------------------------------
by fabpot at 2012-08-25T16:57:29Z
I don't like adding yet another option for something that should "just works". It would be better to find a better way to optimise the strings for all cases.
---------------------------------------------------------------------------
by mazen at 2012-08-25T17:22:07Z
I never really tested how much of a runtime difference it incurs when either using the "optimized" version or the non optimized version, so:
Having an example at hand which generates stable results yields (in a non-debug environment with a booted container using either of the optimization methods):
Without optimized strings:
```
Time taken for tests: 14.865 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Total transferred: 217000 bytes
HTML transferred: 8000 bytes
Requests per second: 67.27 [#/sec] (mean)
Time per request: 14.865 [ms] (mean)
Time per request: 14.865 [ms] (mean, across all concurrent requests)
Transfer rate: 14.26 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.0 0 0
Processing: 14 15 19.7 14 632
Waiting: 14 15 19.7 14 632
Total: 14 15 19.7 14 632
Percentage of the requests served within a certain time (ms)
50% 14
66% 14
75% 14
80% 14
90% 14
95% 14
98% 15
99% 23
100% 632 (longest request)
```
With Optimized Strings enabled
```
Time taken for tests: 14.077 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Total transferred: 217000 bytes
HTML transferred: 8000 bytes
Requests per second: 71.04 [#/sec] (mean)
Time per request: 14.077 [ms] (mean)
Time per request: 14.077 [ms] (mean, across all concurrent requests)
Transfer rate: 15.05 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.0 0 0
Processing: 14 14 1.3 14 48
Waiting: 14 14 1.3 14 48
Total: 14 14 1.3 14 48
Percentage of the requests served within a certain time (ms)
50% 14
66% 14
75% 14
80% 14
90% 14
95% 14
98% 14
99% 15
100% 48 (longest request)
```
So the response times differ by around 0.8ms
Building the non-optimized container takes around 800ms
Building the optimized container takes 43 seconds
From my Point of View it would just be viable to remove the optimization (since it already incurred some issues fixed in 808088a3ca).
I do not see a way how to improve the regexps (but by all means I'm no regular expression guru)
---------------------------------------------------------------------------
by fabpot at 2012-08-30T07:12:55Z
I'm also for removing these optimizations. What others think?
---------------------------------------------------------------------------
by Baachi at 2012-08-30T07:54:53Z
I'm +1 for removing this feature.
The performance boost is to small.
This replaces existing use of core SPL exceptions with the equivalent classes defined within the component. Although method documentation has been changed, this change should be BC since the component-specific SPL exceptions extend their core counterpart.
This commit purposely omits any changes to the PhpDumper, which throws several core SPL exceptions.
Previously, the Definition class was used both for type inference and factory construction (if factoryService was absent). This is fine for cases where classes create instances of themselves (e.g. getInstance() or create()), but leads to ambiguity when we have a separate factory class.
Now that we have a compilation phase for the DIC, using tags after compilation
is not needed anymore.
Tags were introduced to allow several independant bundles to be able to
interact which each others (remember that each extension knows nothing about
the others).
But during the compilation phase, the container has been merged ans so, all
the information from all bundles are available. This is then the right place
to deal with tags. That way, less work is needed at runtime and the DIC class
in the cache is also much smaller.
For simple cases, it means that you need to process the tag in a compiler pass
and store the information you need in a DIC parameter (have a look at the
TranslatorPass for a very simple example).
So, the PHP dumper does not add tags to the dumped PHP class anymore (it does
not implements TaggedContainerInterface anymore). But tags are still available
on ContainerBuilder instances.
- inline private services which are references multiple times, but where all references originate from the same definition
- bug fix for non-shared services which were considered shared within the scope in which they were inlined
* removed the __call() method in Container: it means that now, there is only
one way to get a service: via the get() method;
* removed the $shared variable in the dumped Container classes (we now use
the $services variable from the parent class directly -- this is where we
have a performance improvement);
* optimized the PHP Dumper output.
In the dumped PHP class, we must use get() and not get*Service() methods to get services.
That's because all calls must be managed by get(). From the outside, you can call
get*Service() because as they are protected, they are caught by the __call() method;
which is not the case obviously when it is used internally.
If not, if you override a service with set(), this won't work when a service
depends on this one (the default one will still be used).