# Iterators

## Intro

Iterators allow you to traverse over collections of your resources in an efficient and easy way. Currently there are two Iterators provided by the SDK: 

- **ResourceIterator**. The standard iterator class that implements SPL's standard [Iterator](http://php.net/manual/en/class.iterator.php), [ArrayAccess](http://www.php.net/manual/en/class.arrayaccess.php) and [Countable](http://php.net/manual/en/class.countable.php) interfaces. In short, this allows you to traverse this object (using `foreach`), count its internal elements like an array (using `count` or `sizeof`), and access its internal elements like an array (using `$iterator[1]`).


- **PaginatedIterator**. This is a child of ResourceIterator, and as such inherits all of its functionality. The difference however is that when it reaches the end of the current collection, it attempts to construct a URL to access the API based on predictive paginated collection templates.

## Common behaviour

```php
$iterator = $computeService->flavorList();
```

There are two ways to traverse an iterator. The first is the longer, more traditional way:

```php
while ($iterator->valid()) {
	$flavor = $iterator->current();
    
    // do stuff..
    echo $flavor->id;
    
    $iterator->next();
}
```

There is also a shorter and more intuitive version:

```php
foreach ($iterator as $flavor) {
	// do stuff...
    echo $flavor->id;
}
```

Because the iterator implements PHP's native `Iterator` interface, it can inherit all the native functionality of traversible data structures with `foreach`.

## Very important note

Until now, users have been expected to do this:

```php
while ($flavor = $iterator->next()) {
   // ...
}
```

which is **incorrect**. The single responsibility of `next` is to move the internal pointer forward. It is the job of `current` to retrieve the current element.

For your convenience, these two Iterator classes are fully backward compatible: they exhibit all the functionality you'd expect from a correctly implemented iterator, but they also allow previous behaviour.

## Using paginated collections

For large collections, such as retrieving DataObjects from CloudFiles/Swift, you need to use pagination. Each resource will have a different limit per page; so once that page is traversed, there needs to be another API call to retrieve to *next* page's resources.

There are two key concepts:

- **limit** is the amount of resources returned per page
- **marker** is the way you define a starting point. It is some form of identifier that allows the collection to begin from a specific resource

### Resource classes

When the iterator returns a current element in the internal list, it populates the relevant resource class with all the data returned to the API. In most cases, a `stdClass` object will become an instance of `OpenCloud\Common\PersistentObject`.

In order for this instantiation to happen, the `resourceClass` option must correspond to some method in the parent class that creates the resource. For example, if we specify 'ScalingPolicy' as the `resourceClass`, the parent object (in this case `OpenCloud\Autoscale\Group`, needs to have some method will allows the iterator to instantiate the child resource class. These are all valid:

1. `Group::scalingGroup($data);`

2. `Group::getScalingGroup($data);`

3. `Group::resource('ScalingGroup', $data);`

where `$data` is the standard object. This list runs in order of precedence.

## Setting up a PaginatedIterator

```php
use OpenCloud\Common\Collection\PaginatedIterator;

$service = $client->computeService();

$flavors = PaginatedIterator::factory($service, array(
	'resourceClass'  => 'Flavor',
    'baseUrl'        => $service->getUrl('flavors')
    'limit.total'    => 350,
    'limit.page'     => 100,
    'key.collection' => 'flavors'
));

foreach ($flavors as $flavor) {
	echo $flavor->getId();
}
```

As you can see, there are a lot of configuration parameters to pass in - and getting it right can be quite fiddly, involving a lot of API research. For this reason, using the convenience methods like `flavorList` is recommended because it hides the complexity.

### PaginatedIterator options

There are certain configuration options that the paginated iterator needs to work. These are:

Name|Description|Type|Required|Default|
---|---|---|---|---|
resourceClass|The resource class that is instantiated when the current element is retrieved. This is relative to the parent/service which called the iterator.|string|Yes|-
baseUrl|The base URL that is used for making new calls to the API for new pages|`Guzzle\Http\Url`|Yes|-
limit.total|The total amount of resources you want to traverse in your collection. The iterator will stop as this limit is reached, regardless if there are more items in the list|int|No|10000
limit.page|The amount of resources each page contains|int|No|100
key.links|Often, API responses will contain "links" that allow easy access to the next page of a resource collection. This option specifies what that JSON element is called (its key). For example, for Rackspace Compute images it is `images_links`.|string|No|links
key.collection|The top-level key for the array of resources. For example, servers are returned with this data structure: `{"servers": [...]}`. The **key.collection** value in this case would be `servers`.|string|No|`null`
key.collectionElement|Rarely used. But it indicates the key name for each nested resource element. KeyPairs, for example, are listed like this: `{"keypairs": [ {"keypair": {...}} ] }`. So in this case the collectionElement key would be `keypair`.|string|No|`null`
key.marker|The value used as the marker. It needs to represent a valid property in the JSON resource objects. Often it is `id` or `name`.|string|No|name
request.method|The HTTP method used when making API calls for new pages|string|No|GET
request.headers|The HTTP headers to send when making API calls for new pages|array|No|`array()`
request.body|The HTTP entity body to send when making API calls for new pages|`Guzzle\Http\EntityBody`|No|`null`
request.curlOptions|Additional cURL options to use when making API calls for new pages|array|No|`array()`