Improved collections performance in common use cases

This improves the performance of most collection methods sometimes
by making it 100% faster compared to the previous implementation.

The improvement is not for free, with this change we are giving up
the lazyness feature (not iterate unless requested). The giving
up of the feature only happens when the collection has been initialized
with an array, as oposed to being initialized with another iterator or
generator; and this is the reson I consider this change a safe one.

For the curious, the improvement comes from the (sad) fact that calling
functions in php is extremenly expensive, specially when using iterators
since each iteration will call at least 4 functions (valid, next, current, key).

This becomes even worse as `IteratorIterator` does not have any optimizations,
so for each wrapped iterator, the number of functions is multiplied by 2
for each iteration.

The change proposed here will unwrap nested iterators as much as possible
to avoid the function call explosion, and in some (safe) cases, will
iterator the collection immediately as an array before wrapping it again
in an iterator.

I was inspired by Haskell when implementing this, as the language is lazy
by default, but the compiler optimizes the cases where code is safe to be
called strictly. Thats is called strictness analysis.

Author: lorenzo

src/Collection/CollectionTrait.php

@@ -47,7 +47,7 @@ trait CollectionTrait
      */
     public function each(callable $c)
     {
-        foreach ($this->unwrap() as $k => $v) {
+        foreach ($this->optimizeUnwrap() as $k => $v) {
             $c($v, $k);
         }
 
@@ -87,7 +87,7 @@ public function reject(callable $c)
      */
     public function every(callable $c)
     {
-        foreach ($this->unwrap() as $key => $value) {
+        foreach ($this->optimizeUnwrap() as $key => $value) {
             if (!$c($value, $key)) {
                 return false;
             }
@@ -101,7 +101,7 @@ public function every(callable $c)
      */
     public function some(callable $c)
     {
-        foreach ($this->unwrap() as $key => $value) {
+        foreach ($this->optimizeUnwrap() as $key => $value) {
             if ($c($value, $key) === true) {
                 return true;
             }
@@ -115,7 +115,7 @@ public function some(callable $c)
      */
     public function contains($value)
     {
-        foreach ($this->unwrap() as $v) {
+        foreach ($this->optimizeUnwrap() as $v) {
             if ($value === $v) {
                 return true;
             }
@@ -145,7 +145,7 @@ public function reduce(callable $c, $zero = null)
         }
 
         $result = $zero;
-        foreach ($this->unwrap() as $k => $value) {
+        foreach ($this->optimizeUnwrap() as $k => $value) {
             if ($isFirst) {
                 $result = $value;
                 $isFirst = false;
@@ -205,7 +205,7 @@ public function groupBy($callback)
     {
         $callback = $this->_propertyExtractor($callback);
         $group = [];
-        foreach ($this as $value) {
+        foreach ($this->optimizeUnwrap() as $value) {
             $group[$callback($value)][] = $value;
         }
 
@@ -219,7 +219,7 @@ public function indexBy($callback)
     {
         $callback = $this->_propertyExtractor($callback);
         $group = [];
-        foreach ($this as $value) {
+        foreach ($this->optimizeUnwrap() as $value) {
             $group[$callback($value)] = $value;
         }
 
@@ -255,7 +255,7 @@ public function sumOf($matcher = null)
 
         $callback = $this->_propertyExtractor($matcher);
         $sum = 0;
-        foreach ($this as $k => $v) {
+        foreach ($this->optimizeUnwrap() as $k => $v) {
             $sum += $callback($v, $k);
         }
 
@@ -500,7 +500,7 @@ public function compile($preserveKeys = true)
      */
     public function buffered()
     {
-        return new BufferedIterator($this);
+        return new BufferedIterator($this->unwrap());
     }
 
     /**
@@ -534,7 +534,7 @@ public function stopWhen($condition)
             $condition = $this->_createMatcherFilter($condition);
         }
 
-        return new StoppableIterator($this, $condition);
+        return new StoppableIterator($this->unwrap(), $condition);
     }
 
     /**
@@ -550,7 +550,7 @@ public function unfold(callable $transformer = null)
 
         return new Collection(
             new RecursiveIteratorIterator(
-                new UnfoldIterator($this, $transformer),
+                new UnfoldIterator($this->unwrap(), $transformer),
                 RecursiveIteratorIterator::LEAVES_ONLY
             )
         );
@@ -571,7 +571,7 @@ public function through(callable $handler)
      */
     public function zip($items)
     {
-        return new ZipIterator(array_merge([$this], func_get_args()));
+        return new ZipIterator(array_merge([$this->unwrap()], func_get_args()));
     }
 
     /**
@@ -586,7 +586,7 @@ public function zipWith($items, $callable)
             $items = [$items];
         }
 
-        return new ZipIterator(array_merge([$this], $items), $callable);
+        return new ZipIterator(array_merge([$this->unwrap()], $items), $callable);
     }
 
     /**
@@ -640,7 +640,7 @@ public function chunkWithKeys($chunkSize, $preserveKeys = true)
      */
     public function isEmpty()
     {
-        foreach ($this->unwrap() as $el) {
+        foreach ($this as $el) {
             return false;
         }
 
@@ -657,6 +657,10 @@ public function unwrap()
             $iterator = $iterator->getInnerIterator();
         }
 
+        if ($iterator !== $this && $iterator instanceof CollectionInterface) {
+            $iterator = $iterator->unwrap();
+        }
+
         return $iterator;
     }
 
@@ -747,4 +751,21 @@ public function transpose()
 
         return new Collection($result);
     }
+
+     /**
+      * Unwraps this iterator and returns the simplest
+      * traversable that can be used for getting the data out
+      *
+      * @return \Traversable|array
+      */
+    protected function optimizeUnwrap()
+    {
+        $iterator = $this->unwrap();
+
+        if ($iterator instanceof ArrayIterator) {
+            $iterator = $iterator->getArrayCopy();
+        }
+
+        return $iterator;
+    }
 }

src/Collection/Iterator/ExtractIterator.php

@@ -14,7 +14,9 @@
  */
 namespace Cake\Collection\Iterator;
 
+use ArrayIterator;
 use Cake\Collection\Collection;
+use Cake\Collection\CollectionInterface;
 
 /**
  * Creates an iterator from another iterator that extract the requested column
@@ -69,4 +71,37 @@ public function current()
 
         return $extractor(parent::current());
     }
+
+    /**
+     * {@inheritDoc}
+     *
+     * We perform here some stricness analysys so that the
+     * iterator logic is bypassed entirely.
+     *
+     * @return \Iterator
+     */
+    public function unwrap()
+    {
+        $iterator = $this->getInnerIterator();
+
+        if ($iterator instanceof CollectionInterface) {
+            $iterator = $iterator->unwrap();
+        }
+
+        if (!$iterator instanceof ArrayIterator) {
+            return $this;
+        }
+
+        // ArrayIterator can be traversed strictly.
+        // Let's do that for performance gains
+
+        $callback = $this->_extractor;
+        $res = [];
+
+        foreach ($iterator->getArrayCopy() as $k => $v) {
+            $res[$k] = $callback($v);
+        }
+
+        return new ArrayIterator($res);
+    }
 }

src/Collection/Iterator/FilterIterator.php

@@ -14,7 +14,9 @@
  */
 namespace Cake\Collection\Iterator;
 
+use ArrayIterator;
 use Cake\Collection\Collection;
+use Cake\Collection\CollectionInterface;
 use CallbackFilterIterator;
 use Iterator;
 
@@ -26,6 +28,13 @@
 class FilterIterator extends Collection
 {
 
+    /**
+     * The callback used to filter the elements in this collection
+     *
+     * @var callable
+     */
+    protected $_callback;
+
     /**
      * Creates a filtered iterator using the callback to determine which items are
      * accepted or rejected.
@@ -37,9 +46,50 @@ class FilterIterator extends Collection
      * @param \Iterator $items The items to be filtered.
      * @param callable $callback Callback.
      */
-    public function __construct(Iterator $items, callable $callback)
+    public function __construct($items, callable $callback)
     {
+        if (!$items instanceof Iterator) {
+            $items = new Collection($items);
+        }
+
+        $this->_callback = $callback;
         $wrapper = new CallbackFilterIterator($items, $callback);
         parent::__construct($wrapper);
     }
+
+    /**
+     * {@inheritDoc}
+     *
+     * We perform here some stricness analysys so that the
+     * iterator logic is bypassed entirely.
+     *
+     * @return \Iterator
+     */
+    public function unwrap()
+    {
+        $filter = $this->getInnerIterator();
+        $iterator = $filter->getInnerIterator();
+
+        if ($iterator instanceof CollectionInterface) {
+            $iterator = $iterator->unwrap();
+        }
+
+        if (!$iterator instanceof ArrayIterator) {
+            return $filter;
+        }
+
+        // ArrayIterator can be traversed strictly.
+        // Let's do that for performance gains
+
+        $callback = $this->_callback;
+        $res = [];
+
+        foreach ($iterator as $k => $v) {
+            if ($callback($v, $k, $iterator)) {
+                $res[$k] = $v;
+            }
+        }
+
+        return new ArrayIterator($res);
+    }
 }

src/Collection/Iterator/ReplaceIterator.php

@@ -14,7 +14,9 @@
  */
 namespace Cake\Collection\Iterator;
 
+use ArrayIterator;
 use Cake\Collection\Collection;
+use Cake\Collection\CollectionInterface;
 
 /**
  * Creates an iterator from another iterator that will modify each of the values
@@ -24,7 +26,7 @@ class ReplaceIterator extends Collection
 {
 
     /**
-     * The callback function to be used to modify each of the values
+     * The callback function to be used to transform values
      *
      * @var callable
      */
@@ -67,4 +69,38 @@ public function current()
 
         return $callback(parent::current(), $this->key(), $this->_innerIterator);
     }
+
+
+    /**
+     * {@inheritDoc}
+     *
+     * We perform here some stricness analysys so that the
+     * iterator logic is bypassed entirely.
+     *
+     * @return \Iterator
+     */
+    public function unwrap()
+    {
+        $iterator = $this->_innerIterator;
+
+        if ($iterator instanceof CollectionInterface) {
+            $iterator = $iterator->unwrap();
+        }
+
+        if (!$iterator instanceof ArrayIterator) {
+            return $this;
+        }
+
+        // ArrayIterator can be traversed strictly.
+        // Let's do that for performance gains
+
+        $callback = $this->_callback;
+        $res = [];
+
+        foreach ($iterator as $k => $v) {
+            $res[$k] = $callback($v, $k, $iterator);
+        }
+
+        return new ArrayIterator($res);
+    }
 }

src/Collection/Iterator/SortIterator.php

@@ -59,11 +59,10 @@ class SortIterator extends Collection
      */
     public function __construct($items, $callback, $dir = SORT_DESC, $type = SORT_NUMERIC)
     {
-        if (is_array($items)) {
-            $items = new Collection($items);
+        if (!is_array($items)) {
+            $items = iterator_to_array((new Collection($items))->unwrap(), false);
         }
 
-        $items = iterator_to_array($items, false);
         $callback = $this->_propertyExtractor($callback);
         $results = [];
         foreach ($items as $key => $value) {
@@ -81,4 +80,14 @@ public function __construct($items, $callback, $dir = SORT_DESC, $type = SORT_NU
         }
         parent::__construct($results);
     }
+
+    /**
+     * {@inheritDoc}
+     *
+     * @return \Iterator
+     */
+    public function unwrap()
+    {
+        return $this->getInnerIterator();
+    }
 }