Performance merged

[patrickallaert]

Pull request merging the changes made by Kore in performance branch.

[ezrobot]

This Pull Request does not respect our Coding Standards, please, see the report below:

FILE: .../Publish/Core/Persistence/Legacy/Content/Search/Gateway/EzcDatabase.php
--------------------------------------------------------------------------------
FOUND 1 ERROR(S) AFFECTING 1 LINE(S)
--------------------------------------------------------------------------------
 208 | ERROR | A cast statement must not be followed by a space
--------------------------------------------------------------------------------


FILE: ...Z/Publish/Core/Persistence/Legacy/Content/Type/MemoryCachingHandler.php
--------------------------------------------------------------------------------
FOUND 3 ERROR(S) AFFECTING 2 LINE(S)
--------------------------------------------------------------------------------
  11 | ERROR | The "use" keyword must be used to import one class/namespace at
     |       | a time.
  11 | ERROR | The "use" keyword must not be spread over more than one line.
 317 | ERROR | File must not contain multiple empty lines in a row; found 2
     |       | empty lines
--------------------------------------------------------------------------------

[bdunogier]

Wrong :-)

[patrickallaert]

Damn, I forgot creating a Sniff for checking that ;-)

[bdunogier]

Shouldn't we, for all of those, use {@inheritdoc} instead ?

[andrerom]

we probably should, would save us lots of unaligned doc.

[patrickallaert]

The thing is that documentation tools behaves differently. Some supports inheritance with a dedicated flag, some inherit by default,... We should probably decide what to do based on some research in this regard.

I am personally in favour of not documenting anything if there is no doc change, and use {@inheritdoc} in the case we have something that is changed and use it for copy/paste parent doc. Like in:

/**
 * Wrapper for in-memory cache. {@inheritdoc}
 */

The following shouldn't be used IMHO as it should be documentation tool's default behaviour:

/**
 * {@inheritdoc}
 */

[gggeek]

-1 for "standalone" {@inheritdoc} tags. Line noise (as in: makes reading the actual code harder, and for the generated doc, well, let's improve doc generators instead)

[bdunogier]

The fact that @inheritdoc reduces readability is an argument. But honestly,
I'd rather have to jump to ther parent method, rather than having outdated
documentation due to copy/paste + entropy.

I think that we can require that our users, at least those who have
interest in reading our code will rely either on the generated
documentation OR on a feature-rich IDE rather than on reading code only.
This assumption makes it really easier to provide, accurate, up-to-date
documentation.

In any case, I think we at least all agree that if we choose not to
copy/paste parent documentation when it is identical, it is useless to add
@inheritdoc. This should be automatically handled by IDEs and doc
generators. And actually, it is.

I have checked support on sami (http://ezsystems.github.com/ezp-next/sami),
and it indeed isn't perfect. I have also checked PhpStorm 5. There are
the results with a validate function, in the Image\Type class, inheriting
from FieldType.

No phpdoc at all:

• sami: documentation from parent is re-used
• phpstorm: documentation from parent is re-used

Inline {@inheritdoc} with extra documentation
/**
* Overridden documentation.
* {@inheritdoc}
*/

• sami: Prints out {@inheritdoc}
• phpstorm: Doesn't take the local changes into account

Override short description in the child method

• sami: replaces the parent's short description with the child's. Parameter
  & return & @throws docs are maintained.
• phpdoc: replaces the parent's short description with the child's.
  Parameter & return & @throws docs are maintained.

[ezrobot]

This Pull Request does not respect our Coding Standards, please, see the report below:

FILE: ...workspace/eZ/Publish/API/Repository/Tests/Stubs/URLAliasServiceStub.php
--------------------------------------------------------------------------------
FOUND 1 ERROR(S) AFFECTING 1 LINE(S)
--------------------------------------------------------------------------------
 368 | ERROR | Public method name
     |       | "URLAliasServiceStub::_removeAliasesForLocation" is not in camel
     |       | caps format
--------------------------------------------------------------------------------


FILE: .../eZ/Publish/Core/Persistence/Legacy/Tests/Content/SearchHandlerTest.php
--------------------------------------------------------------------------------
FOUND 1 ERROR(S) AFFECTING 1 LINE(S)
--------------------------------------------------------------------------------
 23 | ERROR | File must not contain multiple empty lines in a row; found 2
    |       | empty lines
--------------------------------------------------------------------------------

[ezrobot]

This Pull Request does not respect our Coding Standards, please, see the report below:

FILE: .../eZ/Publish/Core/Persistence/Legacy/Tests/Content/SearchHandlerTest.php
--------------------------------------------------------------------------------
FOUND 1 ERROR(S) AFFECTING 1 LINE(S)
--------------------------------------------------------------------------------
 23 | ERROR | File must not contain multiple empty lines in a row; found 2
    |       | empty lines
--------------------------------------------------------------------------------

[dpobel]

Tested on a real life website and those patches make a big difference :-) Still not perfect, but way better.

[patrickallaert]

All unit and integration tests are now passing.
All CS issues are solved as well.

[andrerom]

Only thing would be that I think "eZ/Publish/Core/Persistence/Legacy/Content/Type/MemoryCachingHandler.php" is probably unneeded when SPI cache is introduced. But given this is pr instance it does not hurt, and refactoring needed to de couple handlers to be able to inject outer spi persistence cache handler is to high. So: +1 as is

[patrickallaert]

Thanks for your comment @andrerom, I second your opinion.

[gggeek]

BUG - you need an array_unique here, or different SQL, as you can get back duplicate object ids

[gggeek]

OPTIMIZATION: if there is no offset and limit set, just execute 1 query instead of 2, and do a php count() on the resulting array. It makes db happier

[andrerom]

Use issue tracker please.

[lolautruche]

Ref EZP-21906