[GH-606] Don't add empty Expr to another one

This issue is created automatically through a Github pull request on behalf of jean-gui:

Url: https://github.com/doctrine/doctrine2/pull/606

Message:

Doctrine should not allow to add an empty Expr to another one. Current code allows that, which can lead to wrong DQL.
Example 1:
```php
$andExpr = $this->_expr->andx();
$andExpr->add($this->_expr->andx());
$andExpr->add($this->_expr->eq(1, 1));
echo $andExpr;
```
will output:
```sql
AND 1 = 1
```
instead of:
```sql
1 = 1
```

Example 2:
```php
echo $andExpr;
```
will output:
```sql
AND AND
```
instead of nothing.

IRC log of the discusion on #doctrine:
```irc
[jeudi 7 mars 2013] [14:36:18] <Jean-Gui> hi
[jeudi 7 mars 2013] [14:36:35] <Jean-Gui> I have a question about Doctrine/ORM/Query/Expr/Base.php
[jeudi 7 mars 2013] [14:37:17] <Jean-Gui> looking at the add function (https://github.com/doctrine/doctrine2/blob/master/lib/Doctrine/ORM/Query/Expr/Base.php#L89), the first if seems wrong
[jeudi 7 mars 2013] [14:37:39] <Jean-Gui> [[ if ( $arg !== null || ($arg instanceof self && $arg->count() > 0) ) ]]
[jeudi 7 mars 2013] [14:39:49] <Jean-Gui> if $arg is not null, then it will get added even if $arg->count === 0
[jeudi 7 mars 2013] [14:41:03] <ocramius> yes
[jeudi 7 mars 2013] [14:41:42] <Jean-Gui> that doesn't seem right, does it?
[jeudi 7 mars 2013] [14:43:29] <ocramius> why not?
[jeudi 7 mars 2013] [14:43:34] <ocramius> can you elaborate?
[jeudi 7 mars 2013] [14:45:37] <Jean-Gui> that "if" seems to be meaning that the function shouldn't add $arg if $arg->count() equals 0
[jeudi 7 mars 2013] [14:45:45] <Ninj> ocramius: i think Jean-Gui means that the right side of the OR will be called only if the left side is false, which means $arg is null. And if $arg is null it cannot be an object
[jeudi 7 mars 2013] [14:46:07] <Jean-Gui> right
[jeudi 7 mars 2013] [14:46:30] <Ninj> this condition is indeed bugged
[jeudi 7 mars 2013] [14:47:12] <alcuadradoatwork> I see Jean-Gui's point too
[jeudi 7 mars 2013] [14:47:18] <ocramius> write a test case then
[jeudi 7 mars 2013] [14:47:24] <alcuadradoatwork> maybe it works right, but it's confusing
[jeudi 7 mars 2013] [14:48:07] <Ninj> this condition will not throw any error because the "instanceof" will prevent the $arg::count function to be called on a null object, but it is still useless
[jeudi 7 mars 2013] [14:48:20] <Jean-Gui> I think [[ if ( $arg Unable to render embedded object: File (== null && () not found.($arg instanceof self) || ($arg instanceof self && $arg->count() > 0)) ) ]] would work better
[jeudi 7 mars 2013] [14:48:31] * Jean-Gui will look into how to submit bug reports
[jeudi 7 mars 2013] [14:48:49] <alcuadradoatwork> isn't this enough? if ($arg instanceof self && $arg->count() > 0)
[jeudi 7 mars 2013] [14:49:18] <Ninj> i think it is, alcuadradoatwork
[jeudi 7 mars 2013] [14:49:28] <Jean-Gui> no, because if $arg is not an instance of self, I think we still want to add it
[jeudi 7 mars 2013] [14:49:45] <ocramius> alcuadradoatwork: again... if it is a buggy condition, write a small test and open a PR
[jeudi 7 mars 2013] [14:50:03] <alcuadradoatwork> ocramius, it depends on your definition of "buggy"
[jeudi 7 mars 2013] [14:50:32] <alcuadradoatwork> it won't damage the behavior of the software, but it's quality its kind of pour
[jeudi 7 mars 2013] [14:51:19] <ocramius> alcuadradoatwork: yes, still it needs coverage. I didn't say it needs a FAILING test
[jeudi 7 mars 2013] [14:51:38] <alcuadradoatwork> I see your point
[jeudi 7 mars 2013] [14:51:47] <Ninj> if the logic is : "add any non-null object, but if it's self so add it only if count>0" then it must be rewriten
[jeudi 7 mars 2013] [14:52:46] <Jean-Gui> $args should be added if it's an instance of self or $_allowedClasses
[jeudi 7 mars 2013] [14:53:16] <Ninj> hum
[jeudi 7 mars 2013] [14:53:37] <Ninj> $args should be added if it's an instance of self with count > 0 or any other $_allowedClasse
[jeudi 7 mars 2013] [14:53:55] <Jean-Gui> yes, Ninj
[jeudi 7 mars 2013] [14:55:01] <Jean-Gui> thanks for your help guys, I'll submit a bug report with some test cases
[jeudi 7 mars 2013] [14:55:43] <Ninj> your welcome Jean-Gui
```