N-gram

gistfile1.aw

<?php
class Text_Ngram implements Countable, SeekableIterator, ArrayAccess {
    private
        $text       = '',
        $chunk_size = 0,
        $max_count  = 0,
        $charset    = 'UTF-8',
        $current    = 0;    // for iterator

    public function __construct($text, $n, $charset = 'AUTO') {
        if($charset === 'AUTO') {
            $charset = mb_internal_encoding();
        }
        $this->text         = (string)$text;
        $this->charset      = (string)$charset;
        $this->chunk_size   = (int)$n;
        $this->max_count    = ($text === '') ? 0 : max(1, mb_strlen($this->text, $this->charset) - ($this->chunk_size - 1));
    }

    // Countable
    public function count() {
        return $this->max_count;
    }

    // Iterator
    public function current() {
        return $this->get($this->current);
    }

    // Iterator
    public function key() {
        return $this->current;
    }

    // Iterator
    public function next() {
        ++$this->current;
    }

    // Iterator
    public function rewind() {
        $this->seek(0);
    }

    // Iterator
    public function valid() {
        return $this->offsetExists($this->current);
    }

    // SeekableIterator
    public function seek($pos) {
        $this->current = (int)$pos;
    }

    // ArrayAccess
    public function offsetExists($offset) {
        return is_int($offset) && (0 <= $offset) && ($offset < $this->count());
    }

    // ArrayAccess
    public function offsetGet($offset) {
        return $this->get($offset);
    }

    // ArrayAccess
    public function offsetSet($offset, $value) {
    }

    // ArrayAccess
    public function offsetUnset($offset) {
    }

    public function toArray() {
        $retval = array();
        foreach($this as $value) {
            $retval[] = $value;
        }
        return $retval;
    }

    public function toString($glue = ' ') {
        return implode($glue, $this->toArray());
    }

    public function __toString() {
        return $this->toString();
    }

    private function get($offset) {
        if(!$this->offsetExists($offset)) {
            return false;
        }
        return mb_substr($this->text, $offset, $this->chunk_size, $this->charset);
    }
}