Is there a "unique" switch?

[fzaninotto]

Is there a "unique" switch?

[fzaninotto]

No, but it's a great idea. Look at #166 for inspiration.

[rogamoore]

I will take a look! :)

[rogamoore]

Ok, so I don't think the chaining approach would work in this case:

$faker->unique()->country - the unique method has no access to the result of country
$faker->country->unique() - doesn't work because country will return only the element

Idea 1

in Base.php

function unique($method, $parameters = array())  {
  // call $method with $parameters
  // keep track of the returned value for this method (where? static array?)
  // return value if not used before
}

Usage:

$faker->unique('country');
$faker->unique('randomElement', array(1, 2, 3));

possible problems:

• $method will be called until there is a new value which will lead to:
  • an infinitive loop if there are less values available than calls to unique()
  • slow performance because the values may be generated randomly

Idea 2

in EntityPopulator.php

• check if field is unique (Doctrine: ->isUniqueField()
• generate entities but track the return values for this field
• try to generate a new value if the value was already used

possible problems:

• same as above

Idea 3

• change all basic methods to accept a unique flag
• create then an array copy of the possible values
• use the copy to track changes (remove value from copy once used)
• throw Exception if there is no more value left
  possible problems:
• a lot of changes

Any thoughts?

[fzaninotto]

I do think a proxy object could do the trick. unique() would return an instance of a proxy object, embedding the generators. Each time a user calls a generator on this proxy, it would call the embedded generator and keep the result in an internal array. Something like:

class UniqueProxy
{
  protected $generator;
  protected $uniques = array();

  public function __construct(Faker\Generator $generator) 
  {
    $this->generator = $generator;
  }

  public function __call($name, $value)
  {
    if(!isset($this->uniques[$name])) {
      $this->uniques[$name] = array();
    }
    do {
      $res = call_user_func(array($this->generator, $name), $value);
    } while (in_array($res, $this->uniques[$name])) ;
    $this->uniques[$name][]= $res;

    return $res;
  }
}

Then $generator->unique() would return a UniqueProxy singleton, and I think it works. There are some limit cases to deal with (like when the provider doesn't exist), but if I write all the code then it's not your pull request anymore ;)

[rogamoore]

Great approach! This solves the ugly syntax from my first idea. :)

There is still the infinite loop problem to solve. If you use this with e.g. safeColorName() which has 15 unique values and call it 20 times, on run 16 it will try forever to get a value that isn't already in the $uniquesarray.
One could set a limit of 1000 loops for each call and throw an exception, but this has a very dirty feeling.

Another thing is using this in combination with the ORM Populator class. As far as I understand the generator instance will be reused for all addEntity() calls. So if you want to populate more than one entity with unique countries, it will probably run out of fresh countries at some point.

I will keep this idea in mind and see if I can come up with an easy way to solve these problems.

[rogamoore]

great!