New addon "tesseract" for OCR

tesseract/vendor/thiagoalessio/tesseract_ocr/src/Command.php

@@ -0,0 +1,80 @@
+<?php namespace thiagoalessio\TesseractOCR;
+
+class Command
+{
+	public $executable = 'tesseract';
+	public $useFileAsInput = true;
+	public $useFileAsOutput = true;
+	public $options = array();
+	public $configFile;
+	public $tempDir;
+	public $threadLimit;
+	public $image;
+	public $imageSize;
+	private $outputFile;
+
+	public function __construct($image=null, $outputFile=null)
+	{
+		$this->image = $image;
+		$this->outputFile = $outputFile;
+	}
+
+	public function build() { return "$this"; }
+
+	public function __toString()
+	{
+		$cmd = array();
+		if ($this->threadLimit) $cmd[] = "OMP_THREAD_LIMIT={$this->threadLimit}";
+		$cmd[] = self::escape($this->executable);
+		$cmd[] = $this->useFileAsInput ? self::escape($this->image) : "-";
+		$cmd[] = $this->useFileAsOutput ? self::escape($this->getOutputFile(false)) : "-";
+
+		$version = $this->getTesseractVersion();
+
+		foreach ($this->options as $option) {
+			$cmd[] = is_callable($option) ? $option($version) : "$option";
+		}
+		if ($this->configFile) $cmd[] = $this->configFile;
+
+		return join(' ', $cmd);
+	}
+
+	public function getOutputFile($withExt=true)
+	{
+		if (!$this->outputFile)
+			$this->outputFile = $this->getTempDir()
+				.DIRECTORY_SEPARATOR
+				.basename(tempnam($this->getTempDir(), 'ocr'));
+		if (!$withExt) return $this->outputFile;
+
+		$hasCustomExt = array('hocr', 'tsv', 'pdf');
+		$ext = in_array($this->configFile, $hasCustomExt) ? $this->configFile : 'txt';
+		return "{$this->outputFile}.{$ext}";
+	}
+
+	public function getTempDir()
+	{
+		return $this->tempDir ?: sys_get_temp_dir();
+	}
+
+	public function getTesseractVersion()
+	{
+		exec(self::escape($this->executable).' --version 2>&1', $output);
+		$outputParts = explode(' ', $output[0]);
+		return $outputParts[1];
+	}
+
+	public function getAvailableLanguages()
+	{
+		exec(self::escape($this->executable) . ' --list-langs 2>&1', $output);
+		array_shift($output);
+		sort($output);
+		return $output;
+	}
+
+	public static function escape($str)
+	{
+		$charlist = strtoupper(substr(PHP_OS, 0, 3)) == 'WIN' ? '$"`' : '$"\\`';
+		return '"'.addcslashes($str, $charlist).'"';
+	}
+}

tesseract/vendor/thiagoalessio/tesseract_ocr/src/FeatureNotAvailableException.php

@@ -0,0 +1,7 @@
+<?php
+
+namespace thiagoalessio\TesseractOCR;
+
+class FeatureNotAvailableException extends TesseractOcrException
+{
+}

tesseract/vendor/thiagoalessio/tesseract_ocr/src/FriendlyErrors.php

@@ -0,0 +1,120 @@
+<?php namespace thiagoalessio\TesseractOCR;
+
+class FriendlyErrors
+{
+	public static function checkImagePath($image)
+	{
+		if (file_exists($image)) return;
+
+		$currentDir = __DIR__;
+		$msg = array();
+		$msg[] = "Error! The image \"$image\" was not found.";
+		$msg[] = '';
+		$msg[] = "The current __DIR__ is $currentDir";
+		$msg = join(PHP_EOL, $msg);
+
+		throw new ImageNotFoundException($msg);
+	}
+
+	public static function checkTesseractPresence($executable)
+	{
+		if (file_exists($executable)) return;
+
+		$cmd = stripos(PHP_OS, 'win') === 0
+			? 'where.exe '.Command::escape($executable).' > NUL 2>&1'
+			: 'type '.Command::escape($executable).' > /dev/null 2>&1';
+		system($cmd, $exitCode);
+
+		if ($exitCode == 0) return;
+
+		$currentPath = getenv('PATH');
+		$msg = array();
+		$msg[] = "Error! The command \"$executable\" was not found.";
+		$msg[] = '';
+		$msg[] = 'Make sure you have Tesseract OCR installed on your system:';
+		$msg[] = 'https://github.com/tesseract-ocr/tesseract';
+		$msg[] = '';
+		$msg[] = "The current \$PATH is $currentPath";
+		$msg = join(PHP_EOL, $msg);
+
+		throw new TesseractNotFoundException($msg);
+	}
+
+	public static function checkCommandExecution($command, $stdout, $stderr)
+	{
+		if ($command->useFileAsOutput) {
+		    $file = $command->getOutputFile();
+		    if (file_exists($file) && filesize($file) > 0)  return;
+		}
+
+		if (!$command->useFileAsOutput && $stdout) {
+			return;
+		}
+
+		$msg = array();
+		$msg[] = 'Error! The command did not produce any output.';
+		$msg[] = '';
+		$msg[] = 'Generated command:';
+		$msg[] = "$command";
+		$msg[] = '';
+		$msg[] = 'Returned message:';
+		$arrayStderr = explode(PHP_EOL, $stderr);
+		array_pop($arrayStderr);
+		$msg = array_merge($msg, $arrayStderr);
+		$msg = join(PHP_EOL, $msg);
+
+		throw new UnsuccessfulCommandException($msg);
+	}
+
+	public static function checkProcessCreation($processHandle, $command)
+	{
+		if ($processHandle !== FALSE) return;
+
+		$msg = array();
+		$msg[] = 'Error! The command could not be launched.';
+		$msg[] = '';
+		$msg[] = 'Generated command:';
+		$msg[] = "$command";
+		$msg = join(PHP_EOL, $msg);
+
+		throw new UnsuccessfulCommandException($msg);
+	}
+
+	public static function checkTesseractVersion($expected, $action, $command)
+	{
+		$actual = $command->getTesseractVersion();
+
+		if ($actual[0] === 'v')
+			$actual = substr($actual, 1);
+
+		if (version_compare($actual, $expected, ">=")) return;
+
+		$msg = array();
+		$msg[] = "Error! $action is not available this tesseract version";
+		$msg[] = "Required version is $expected, actual version is $actual";
+		$msg[] = '';
+		$msg[] = 'Generated command:';
+		$msg[] = "$command";
+		$msg = join(PHP_EOL, $msg);
+
+		throw new FeatureNotAvailableException($msg);
+	}
+
+	public static function checkWritePermissions($path)
+	{
+		if (!is_dir(dirname($path))) mkdir(dirname($path));
+		$writableDirectory = is_writable(dirname($path));
+		$writableFile = true;
+		if (file_exists($path)) $writableFile = is_writable($path);
+		if ($writableFile && $writableDirectory) return;
+
+		$msg = array();
+		$msg[] = "Error! No permission to write to $path";
+		$msg[] = "Make sure you have the right outputFile and permissions "
+			."to write to the folder";
+		$msg[] = '';
+		$msg = join(PHP_EOL, $msg);
+
+		throw new NoWritePermissionsForOutputFile($msg);
+	}
+}

tesseract/vendor/thiagoalessio/tesseract_ocr/src/ImageNotFoundException.php

@@ -0,0 +1,7 @@
+<?php
+
+namespace thiagoalessio\TesseractOCR;
+
+class ImageNotFoundException extends TesseractOcrException
+{
+}

tesseract/vendor/thiagoalessio/tesseract_ocr/src/NoWritePermissionsForOutputFile.php

@@ -0,0 +1,7 @@
+<?php
+
+namespace thiagoalessio\TesseractOCR;
+
+class NoWritePermissionsForOutputFile extends TesseractOcrException
+{
+}

tesseract/vendor/thiagoalessio/tesseract_ocr/src/Option.php

@@ -0,0 +1,79 @@
+<?php namespace thiagoalessio\TesseractOCR;
+
+class Option
+{
+	public static function psm($psm)
+	{
+		return function($version) use ($psm) {
+			$version = preg_replace('/^v/', '', $version);
+			return (version_compare($version, 4, '>=') ? '-' : '')."-psm $psm";
+		};
+	}
+
+	public static function oem($oem)
+	{
+		return function($version) use ($oem) {
+			Option::checkMinVersion('3.05', $version, 'oem');
+			return "--oem $oem";
+		};
+	}
+
+	public static function dpi($dpi)
+	{
+		return function() use ($dpi) {
+			return "--dpi $dpi";
+		};
+	}
+
+	public static function userWords($path)
+	{
+		return function($version) use ($path) {
+			Option::checkMinVersion('3.04', $version, 'user-words');
+			return '--user-words "'.addcslashes($path, '\\"').'"';
+		};
+	}
+
+	public static function userPatterns($path)
+	{
+		return function($version) use ($path) {
+			Option::checkMinVersion('3.04', $version, 'user-patterns');
+			return '--user-patterns "'.addcslashes($path, '\\"').'"';
+		};
+	}
+
+	public static function tessdataDir($path)
+	{
+		return function() use ($path) {
+			return '--tessdata-dir "'.addcslashes($path, '\\"').'"';
+		};
+	}
+
+	public static function lang()
+	{
+		$languages = func_get_args();
+		return function() use ($languages) {
+			return '-l '.join('+', $languages);
+		};
+	}
+
+	public static function config($var, $value)
+	{
+		return function() use($var, $value) {
+			$snakeCase = function($str) {
+				return strtolower(preg_replace('/([A-Z])+/', '_$1', $str));
+			};
+			$pair = $snakeCase($var).'='.$value;
+			return '-c "'.addcslashes($pair, '\\"').'"';
+		};
+	}
+
+	public static function checkMinVersion($minVersion, $currVersion, $option)
+	{
+		$minVersion = preg_replace('/^v/', '', $minVersion);
+		$currVersion = preg_replace('/^v/', '', $currVersion);
+		if (!version_compare($currVersion, $minVersion, '<')) return;
+		$msg = "$option option is only available on Tesseract $minVersion or later.";
+		$msg.= PHP_EOL."Your version of Tesseract is $currVersion";
+		throw new \Exception($msg);
+	}
+}

tesseract/vendor/thiagoalessio/tesseract_ocr/src/Process.php

@@ -0,0 +1,83 @@
+<?php namespace thiagoalessio\TesseractOCR;
+
+class Process {
+
+    private $stdin;
+    private $stdout;
+    private $stderr;
+    private $handle;
+    private $startTime;
+
+    public function __construct($command)
+    {
+        $this->startTime = microtime(true);
+        $streamDescriptors = [
+            array("pipe", "r"),
+            array("pipe", "w"),
+            array("pipe", "w")
+        ];
+        $this->handle = proc_open($command, $streamDescriptors, $pipes, NULL, NULL, ["bypass_shell" => true]);
+        list($this->stdin, $this->stdout, $this->stderr) = $pipes;
+
+        FriendlyErrors::checkProcessCreation($this->handle, $command);
+
+        //This is can avoid deadlock on some cases (when stderr buffer is filled up before writing to stdout and vice-versa)
+        stream_set_blocking($this->stdout, 0);
+        stream_set_blocking($this->stderr, 0);
+    }
+
+    public function write($data, $len)
+    {
+        $total = 0;
+        do
+        {
+            $res = fwrite($this->stdin, substr($data, $total));
+        } while($res && $total += $res < $len);
+        return $total === $len;
+    }
+
+
+    public function wait($timeout = 0)
+    {
+        $running = true;
+        $data = ["out" => "", "err" => ""];
+        while (($running === true) && !$this->hasTimedOut($timeout))
+        {
+            $data["out"] .= fread($this->stdout, 8192);
+            $data["err"] .= fread($this->stderr, 8192);
+            $procInfo = proc_get_status($this->handle);
+            $running = $procInfo["running"];
+            if ($running) {
+                usleep(1000); // Sleep 1ms to yield CPU time
+            }
+        }
+        return $data;
+    }
+
+    public function close()
+    {
+        $this->closeStream($this->stdin);
+        $this->closeStream($this->stdout);
+        $this->closeStream($this->stderr);
+        return proc_close($this->handle);
+    }
+
+    public function closeStdin()
+    {
+        $this->closeStream($this->stdin);
+    }
+
+    private function hasTimedOut($timeout)
+    {
+        return (($timeout > 0) &&  ($this->startTime + $timeout < microtime(true)));    
+    }
+    
+    private function closeStream(&$stream)
+    {
+        if ($stream !== NULL)
+        {
+            fclose($stream);
+            $stream = NULL;
+        }
+    }
+}

tesseract/vendor/thiagoalessio/tesseract_ocr/src/TesseractNotFoundException.php

@@ -0,0 +1,7 @@
+<?php
+
+namespace thiagoalessio\TesseractOCR;
+
+class TesseractNotFoundException extends TesseractOcrException
+{
+}

tesseract/vendor/thiagoalessio/tesseract_ocr/src/TesseractOCR.php

@@ -0,0 +1,181 @@
+<?php namespace thiagoalessio\TesseractOCR;
+
+use thiagoalessio\TesseractOCR\Command;
+use thiagoalessio\TesseractOCR\Option;
+use thiagoalessio\TesseractOCR\FriendlyErrors;
+
+class TesseractOCR
+{
+	public $command;
+	private $outputFile = null;
+
+	public function __construct($image=null, $command=null)
+	{
+		$this->command = $command ?: new Command;
+		$this->image("$image");
+	}
+
+	public function run($timeout = 0)
+	{
+		try {
+			if ($this->outputFile !== null) {
+				FriendlyErrors::checkWritePermissions($this->outputFile);
+				$this->command->useFileAsOutput = true;
+			}
+
+			FriendlyErrors::checkTesseractPresence($this->command->executable);
+			if ($this->command->useFileAsInput) {
+				FriendlyErrors::checkImagePath($this->command->image);
+			}
+
+			$process = new Process("{$this->command}");
+
+			if (!$this->command->useFileAsInput) {
+				$process->write($this->command->image, $this->command->imageSize);
+				$process->closeStdin();
+			}
+			$output = $process->wait($timeout);
+
+			FriendlyErrors::checkCommandExecution($this->command, $output["out"], $output["err"]);
+		}
+		catch (TesseractOcrException $e) {
+			if ($this->command->useFileAsOutput) $this->cleanTempFiles();
+			throw $e;
+		}
+
+		if ($this->command->useFileAsOutput) {
+			$text = file_get_contents($this->command->getOutputFile());
+
+			if ($this->outputFile !== null) {
+				rename($this->command->getOutputFile(), $this->outputFile);
+			}
+
+			$this->cleanTempFiles();
+		}
+		else
+			$text = $output["out"];
+
+		return trim($text, " \t\n\r\0\x0A\x0B\x0C");
+	}
+
+	public function imageData($image, $size)
+	{
+		FriendlyErrors::checkTesseractVersion("3.03-rc1", "Reading image data from stdin", $this->command);
+		$this->command->useFileAsInput = false;
+		$this->command->image = $image;
+		$this->command->imageSize = $size;
+		return $this;
+	}
+
+	public function withoutTempFiles()
+	{
+		FriendlyErrors::checkTesseractVersion("3.03-rc1", "Writing to stdout (without using temp files)", $this->command);
+		$this->command->useFileAsOutput = false;
+		return $this;
+	}
+
+	public function image($image)
+	{
+		$this->command->image = $image;
+		return $this;
+	}
+
+	public function executable($executable)
+	{
+		FriendlyErrors::checkTesseractPresence($executable);
+		$this->command->executable = $executable;
+		return $this;
+	}
+
+	public function configFile($configFile)
+	{
+		$this->command->configFile = $configFile;
+		return $this;
+	}
+
+	public function tempDir($tempDir)
+	{
+		$this->command->tempDir = $tempDir;
+		return $this;
+	}
+
+	public function threadLimit($limit)
+	{
+		$this->command->threadLimit = $limit;
+		return $this;
+	}
+
+	// @deprecated
+	public function format($fmt) { return $this->configFile($fmt); }
+
+	public function setOutputFile($path) {
+		$this->outputFile = $path;
+		return $this;
+	}
+
+	public function allowlist()
+	{
+		$concat = function ($arg) { return is_array($arg) ? join('', $arg) : $arg; };
+		$allowlist = join('', array_map($concat, func_get_args()));
+		$this->command->options[] = Option::config('tessedit_char_whitelist', $allowlist);
+		return $this;
+	}
+
+	public function whitelist()
+	{
+		$warningMsg = 'Notice: whitelist is deprecated, use allowlist instead.';
+		trigger_error($warningMsg, E_USER_NOTICE);
+
+		$concat = function ($arg) { return is_array($arg) ? join('', $arg) : $arg; };
+		$allowlist = join('', array_map($concat, func_get_args()));
+		return $this->allowlist($allowlist);
+	}
+
+	public function version()
+	{
+		return $this->command->getTesseractVersion();
+	}
+
+	public function availableLanguages()
+	{
+		return $this->command->getAvailableLanguages();
+	}
+
+	public function __call($method, $args)
+	{
+		if ($this->isConfigFile($method)) return $this->configFile($method);
+		if ($this->isOption($method)) {
+			$option = $this->getOptionClassName().'::'.$method;
+			$this->command->options[] = call_user_func_array($option, $args);
+			return $this;
+		}
+		$arg = empty($args) ? null : $args[0];
+		$this->command->options[] = Option::config($method, $arg);
+		return $this;
+	}
+
+	private function isConfigFile($name)
+	{
+		return in_array($name, array('digits', 'hocr', 'pdf', 'quiet', 'tsv', 'txt'));
+	}
+
+	private function isOption($name)
+	{
+		return in_array($name, get_class_methods($this->getOptionClassName()));
+	}
+
+	private function getOptionClassName()
+	{
+		return __NAMESPACE__.'\\Option';
+	}
+
+	private function cleanTempFiles()
+	{
+		if (file_exists($this->command->getOutputFile(false))) {
+			unlink($this->command->getOutputFile(false));
+		}
+		if (file_exists($this->command->getOutputFile(true))) {
+			unlink($this->command->getOutputFile(true));
+		}
+	}
+}

tesseract/vendor/thiagoalessio/tesseract_ocr/src/TesseractOcrException.php

@@ -0,0 +1,7 @@
+<?php
+
+namespace thiagoalessio\TesseractOCR;
+
+abstract class TesseractOcrException extends \Exception
+{
+}

tesseract/vendor/thiagoalessio/tesseract_ocr/src/UnsuccessfulCommandException.php

@@ -0,0 +1,7 @@
+<?php
+
+namespace thiagoalessio\TesseractOCR;
+
+class UnsuccessfulCommandException extends TesseractOcrException
+{
+}