PDFTextExtractor

Complexity

Total Complexity

18

Size/Duplication

Total Lines	133
Duplicated Lines	0 %

Coupling/Cohesion

Components	1
Dependencies	2

Importance

Changes	8
Bugs	1	Features	2

7 Methods

Rating	Name	Duplication	Size	Complexity
A	isAvailable()	0	5	3
A	supportsExtension()	0	4	1
A	supportsMime()	0	12	1
B	bin()	0	23	5
A	getContent()	0	8	2
B	getRawOutput()	0	19	5
A	cleanupLigatures()	0	13	1

<?php

/**
 * Text extractor that calls pdftotext to do the conversion.
 * @author mstephens
 *
 */
class PDFTextExtractor extends FileTextExtractor

PSR1 recommends that each class must be in a namespace of at least one level to avoid collisions.

{
    /**
     * Set to bin path this extractor can execute
     *
     * @var string
     */
    private static $binary_location = null;

The property $binary_location is not used and could be removed.

    /**
     * Used if binary_location isn't set.
     * List of locations to search for a given binary in
     *
     * @config
     * @var array
     */
    private static $search_binary_locations = array(

The property $search_binary_locations is not used and could be removed.

        '/usr/bin',
        '/usr/local/bin',
    );

    public function isAvailable()
    {
        $bin = $this->bin('pdftotext');
        return $bin && file_exists($bin) && is_executable($bin);
    }

    public function supportsExtension($extension)
    {
        return strtolower($extension) === 'pdf';
    }

    public function supportsMime($mime)
    {
        return in_array(
            strtolower($mime),
            array(
                'application/pdf',
                'application/x-pdf',
                'application/x-bzpdf',
                'application/x-gzpdf'
            )
        );
    }

    /**
     * Accessor to get the location of the binary
     *
     * @param string $program Name of binary
     * @return string
     */
    protected function bin($program = '')
    {
        // Get list of allowed search paths
        if ($location = $this->config()->binary_location) {
            $locations = array($location);
        } else {
            $locations = $this->config()->search_binary_locations;
        }

        // Find program in each path
        foreach($locations as $location) {
            $path = "{$location}/{$program}";
            if(file_exists($path)) {
                return $path;
            }
            if (file_exists($path.'.exe')) {
                return $path.'.exe';
            }
        }
        
        // Not found
        return null;
    }

    public function getContent($path)
    {
        if (!$path) {
            return "";
        } // no file
        $content = $this->getRawOutput($path);
        return $this->cleanupLigatures($content);
    }

    /**
     * Invoke pdftotext with the given path
     *
     * @param string $path
     * @return string Output
     * @throws FileTextExtractor_Exception
     */
    protected function getRawOutput($path)
    {
        if(!$this->isAvailable()) {
            throw new FileTextExtractor_Exception("getRawOutput called on unavailable extractor");
        }
        exec(sprintf('%s %s - 2>&1', $this->bin('pdftotext'), escapeshellarg($path)), $content, $err);
        if ($err) {

The expression $err of type integer|null is loosely compared to true; this is ambiguous if the integer can be zero. You might want to explicitly use !== null instead.

            if (!is_array($err) && $err == 1) {
                // For Windows compatibility
                $err = $content;
            }
            throw new FileTextExtractor_Exception(sprintf(
                'PDFTextExtractor->getContent() failed for %s: %s',
                $path,
                implode('', $err)
            ));
        }
        return implode('', $content);
    }

    /**
     * Removes utf-8 ligatures.
     *
     * @link http://en.wikipedia.org/wiki/Typographic_ligature#Computer_typesetting
     *
     * @param string $input
     * @return string
     */
    protected function cleanupLigatures($input)
    {
        $mapping = array(
            'ff' => 'ff',
            'fi' => 'fi',
            'fl' => 'fl',
            'ffi' => 'ffi',
            'ffl' => 'ffl',
            'ſt' => 'ft',
            'st' => 'st'
        );
        return str_replace(array_keys($mapping), array_values($mapping), $input);
    }
}