New document class

src/Camspiers/StatisticalClassifier/Document/Document.php

@@ -0,0 +1,84 @@
+<?php
+namespace Camspiers\StatisticalClassifier\Document;
+
+use Camspiers\StatisticalClassifier\Normalizer\Token;
+use Camspiers\StatisticalClassifier\Tokenizer\Word;
+
+class Document
+    extends \ArrayObject
+    implements DocumentInterface
+{
+    /**
+     * @var callable $normalizer
+     */
+    public static $documentNormalizer;
+
+    /**
+     * @var callable $tokenizer
+     */
+    public static $tokenizer;
+
+    /**
+     * @var callable $normalizer
+     */
+    public static $tokenNormalizer;
+
+    /**
+     * @param  string|array|Traversable  $data
+     * @throws \InvalidArgumentException
+     */
+    public function __construct($data = null)
+    {
+        parent::setFlags(parent::ARRAY_AS_PROPS);
+
+        if (! empty($data)) {
+            if (is_scalar($data)) {
+                if (is_callable(static::$documentNormalizer)) {
+                    $data = call_user_func(static::$documentNormalizer, $data);
+                }
+
+                $data = call_user_func(static::$tokenizer ?: new Word(), $data);
+            } elseif (! is_array($data) and ! $data instanceof \Traversable) {
+                $type = gettype($data);
+                throw new \InvalidArgumentException("Data must be either a string or an array or a traversable, $type given.");
+            }
+
+            $data = array_count_values((array) $data);
+
+            $this->exchangeArray($data);
+        }
+    }
+
+    public function offsetSet($token, $count)
+    {
+        if (is_callable(static::$tokenNormalizer)) {
+            $token = call_user_func(static::$tokenNormalizer, $token);
+        }
+
+        parent::offsetSet($token, $count);
+    }
+
+    /**
+     * @param callable $normalizer
+     */
+    public static function setDocumentNormalizer(callable $normalizer)
+    {
+        static::$documentNormalizer = $normalizer;
+    }
+
+    /**
+     * @param callable $tokenizer
+     */
+    public static function setTokenizer(callable $tokenizer)
+    {
+        static::$tokenizer = $tokenizer;
+    }
+
+    /**
+     * @param callable $normalizer
+     */
+    public static function setTokenNormalizer(callable $normalizer)
+    {
+        static::$tokenNormalizer = $normalizer;
+    }
+}

src/Camspiers/StatisticalClassifier/Document/DocumentInterface.php

@@ -0,0 +1,20 @@
+<?php
+namespace Camspiers\StatisticalClassifier\Document;
+
+interface DocumentInterface
+{
+    /**
+     * @param callable $normalizer
+     */
+    public static function setDocumentNormalizer(callable $normalizer);
+
+    /**
+     * @param callable $tokenizer
+     */
+    public static function setTokenizer(callable $tokenizer);
+
+    /**
+     * @param callable $normalizer
+     */
+    public static function setTokenNormalizer(callable $normalizer);
+}

src/Camspiers/StatisticalClassifier/Tokenizer/Word.php

@@ -17,6 +17,11 @@
  */
 class Word implements TokenizerInterface
 {
+    public function __invoke($document)
+    {
+        return $this->tokenize($document);
+    }
+
     /**
      * @{inheritdoc}
      */

tests/Camspiers/StatisticalClassifier/Document/DocumentTest.php

@@ -0,0 +1,73 @@
+<?php
+namespace Camspiers\StatisticalClassifier\Document;
+
+/**
+ * @backupStaticAttributes disabled
+ */
+class DocumentTest extends \PHPUnit_Framework_TestCase
+{
+    public function constructDataProvider()
+    {
+        return [
+            [
+                'Some document data in a document',
+                ['document' => 2, 'data' => 1, 'in' => 1, 'a' => 1, 'Some' => 1],
+            ],
+            [
+                ['Some', 'document', 'data', 'in', 'a', 'document'],
+                ['document' => 2, 'data' => 1, 'in' => 1, 'a' => 1, 'Some' => 1],
+            ],
+            [
+                new \ArrayObject(['Some', 'document', 'data', 'in', 'a', 'document']),
+                ['document' => 2, 'data' => 1, 'in' => 1, 'a' => 1, 'Some' => 1],
+            ],
+            [
+                new \stdClass(),
+                'InvalidArgumentException',
+            ],
+        ];
+    }
+
+    /**
+     * @covers \Camspiers\StatisticalClassifier\Document\Document::__construct
+     * @dataProvider constructDataProvider
+     * @param mixed        $data
+     * @param string|array $expected
+     */
+    public function testConstruct($data, $expected)
+    {
+        if (is_string($expected)) {
+            $this->setExpectedException($expected);
+        }
+
+        $document = new Document($data);
+
+        $this->assertEquals($document->getArrayCopy(), (array) $document);
+        $this->assertEquals($expected, $document->getArrayCopy());
+    }
+
+    public function testExchangeArrayUsesOffsetSet()
+    {
+        Document::setTokenNormalizer('strtolower');
+
+        $document           = new Document();
+        $document->Some     = 1;
+        $document->document = 2;
+        $document->data     = 1;
+        $document->in       = 1;
+        $document->a        = 1;
+
+        $this->assertSame(
+            ['some' => 1, 'document' => 2, 'data' => 1, 'in' => 1, 'a' => 1],
+            (array) $document
+        );
+        $this->assertCount(5, $document); // Token frequency, thus, document is counted only once
+    }
+
+    public function testIsCountable()
+    {
+        $document = new Document(['Some', 'document', 'data', 'in', 'a', 'document']);
+
+        $this->assertCount(5, $document); // Token frequency, thus, document is counted only once
+    }
+}