Setup

Author: niladrimcn

README.md

@@ -0,0 +1,47 @@
+## PHP Textract
+A laravel 9 package to extract text from files. This work is ispired by ["textract node"](https://www.npmjs.com/package/textract).
+
+### Supported file formats
+Following file formats is supported currently. You need to install proper extensions
+to your server to work with all the following extension related files. The package will 
+check file content MIME type before execute. So with the extension you have maintain
+the current content type to work this package-
+- HTML
+- TEXT
+- DOC
+- DOCX
+- XLS, XLSX, XLSM, XLTX, XLTM, XLT
+- CSV
+- PDF
+- Image
+  - Jpeg
+  - Pdf
+- ODT
+- ODS
+- RTF
+
+### Install
+``` 
+composer require nilgems/textract:^0.1
+```
+### Configuration
+You don't need to anything special for your laravel application to work with this
+package.
+
+### Example
+Use the use ```Nilge\Textract\Textract``` facade to run the extractor. 
+
+Example 1: 
+```
+........
+use Nilge\Textract\Textract;
+
+Route::get('/textract', function(){
+    return Textract::run({file_path});
+});
+........
+```
+
+Example 2:
+
+### Dependencies

composer.json

@@ -0,0 +1,40 @@
+{
+    "name": "nilgems/php-textract",
+    "description": "A text extraction package for laravel",
+    "type": "library",
+    "require": {
+        "php": "^8.0",
+        "ext-fileinfo": "*",
+        "ext-zip": "*",
+        "symfony/process": "6.1",
+        "illuminate/console": "^9",
+        "phpoffice/phpspreadsheet": "^1.23",
+        "phpoffice/phpword": "^0.18"
+    },
+    "require-dev": {
+        "squizlabs/php_codesniffer": "^3.6"
+    },
+    "license": "MIT",
+    "autoload": {
+        "psr-4": {
+            "Nilgems\\PhpTextract\\": "src/"
+        }
+    },
+    "extra": {
+        "laravel": {
+            "providers": [
+                "Nilgems\\PhpTextract\\Providers\\ServiceProvider"
+            ],
+            "aliases": {
+                "Textract":"Nilgems\\PhpTextract\\Textract"
+            }
+        }
+    },
+    "authors": [
+        {
+            "name": "Niladri Shekhar Mondal",
+            "email": "[email protected]"
+        }
+    ],
+    "minimum-stability": "dev"
+}

config/textract.php

@@ -0,0 +1,3 @@
+<?php
+
+return [];

src/Concerns/AbstractExtractor.php

@@ -0,0 +1,110 @@
+<?php
+
+namespace Nilgems\PhpTextract\Concerns;
+
+use Illuminate\Support\Collection;
+use Nilgems\PhpTextract\Exceptions\TextractException;
+use Nilgems\PhpTextract\Services\UtilsService;
+
+abstract class AbstractExtractor
+{
+    /**
+     * @var string $file_path
+     */
+    protected string $file_path = "";
+    /**
+     * @var string $error_message
+     */
+    protected string $error_message = "The extractor plugin is not installed in the system. Please install and try again.";
+    /**
+     * Extractor name
+     * @var string $extractor_name
+     */
+    protected string $extractor_name = 'The extractor';
+
+    protected array $extractor_supported_extension = [];
+    /**
+     * @var array $mime_accepts
+     */
+    protected array $mime_accepts = [];
+    /**
+     * @var string $current_mime_type
+     */
+    protected string $current_mime_type = "";
+    /**
+     * @var Collection $data
+     */
+    protected Collection $data;
+
+    public function __construct() {
+        $this->data = new Collection([]);
+    }
+
+    /**
+     * Set data
+     * @param $key
+     * @param $value
+     * @return $this
+     */
+    public function setData($key, $value): self {
+        $this->data->put($key, $value);
+        return $this;
+    }
+    /**
+     * Get accept mime types
+     * @return array
+     */
+    public function getAcceptMimeTypes(): array {
+        if(method_exists($this, 'mimeAccepts')) {
+            return $this->mimeAccepts();
+        }
+        return $this->mime_accepts;
+    }
+
+    /**
+     * Get acceptable extensions
+     * @return array
+     */
+    public function getAcceptExtensions(): array {
+        return $this->extractor_supported_extension;
+    }
+    /**
+     * Has match mime type
+     * @param string $mime_type
+     * @return bool
+     */
+    public function hasMatchMimeType(string $mime_type): bool
+    {
+        $acceptable_mime_type = $this->getAcceptMimeTypes();
+        if(empty($acceptable_mime_type)) {
+            return true;
+        }
+        return in_array(strtolower($mime_type), $acceptable_mime_type, true);
+    }
+
+    /**
+     * @param string $file_path
+     * @param array $data
+     * @return string|null
+     * @throws TextractException
+     */
+    public function boot(string $file_path, array $data = []): ?string {
+        $this->file_path = $file_path;
+        $this->data = $this->data->merge($data);
+        $utilsService = app(UtilsService::class)->setFilePath($file_path);
+        $utilsService->setFilePath($file_path);
+        $this->current_mime_type = $utilsService->getFileMimeType();
+        if(!$this->hasMatchMimeType($this->current_mime_type)) {
+            throw new TextractException($this->extractor_name . ' unable to process the file. Please ensure the content of file is a ' . implode('/', $this->extractor_supported_extension) . 'file.');
+        }
+        $has_valid = $this->checkHaveProviderPackage();
+        if($has_valid) {
+            return $this->getTextFromFile();
+        }
+        throw new TextractException($this->error_message);
+    }
+
+    abstract protected function checkHaveProviderPackage();
+
+    abstract protected function getTextFromFile();
+}

src/Concerns/MustHaveResponse.php

@@ -0,0 +1,8 @@
+<?php
+
+namespace Nilgems\PhpTextract\Concerns;
+
+interface MustHaveResponse
+{
+    public function __construct(string $job_id, string $file_path, string $output = null, string $error = null);
+}

src/Concerns/TextractOutput.php

@@ -0,0 +1,48 @@
+<?php
+
+namespace Nilgems\PhpTextract\Concerns;
+
+use Illuminate\Contracts\Support\Arrayable;
+use Illuminate\Support\Collection;
+
+/**
+ * @property-read string $text
+ * @property-read string $word_count
+ */
+class TextractOutput implements Arrayable
+{
+    protected Collection $collection;
+
+    /**
+     * @param string $raw_output
+     */
+    public function __construct(string $raw_output)
+    {
+        $this->collection = new Collection([
+            'text' => $raw_output,
+            'word_count' => str_word_count($raw_output, 0)
+        ]);
+    }
+
+    /**
+     * To array
+     * @return array
+     */
+    public function toArray(): array
+    {
+        return $this->collection->toArray();
+    }
+
+    public function __get(string $key) {
+        return $this->collection->get($key);
+    }
+
+    /**
+     * To string
+     * @return string
+     */
+    public function __toString(): string
+    {
+        return $this->collection->get('text');
+    }
+}

src/Console/Contracts/MustHaveExtractCommand.php

@@ -0,0 +1,8 @@
+<?php
+
+namespace Nilgems\PhpTextract\Console\Contracts;
+
+interface MustHaveExtractCommand
+{
+    public function handle(): int;
+}

src/Console/TextractCommand.php

@@ -0,0 +1,24 @@
+<?php
+
+namespace Nilgems\PhpTextract\Console;
+
+use Illuminate\Console\Command;
+use Nilgems\PhpTextract\Console\Contracts\MustHaveExtractCommand;
+use Nilgems\PhpTextract\Services\ConsoleExtractionService;
+
+class TextractCommand extends Command implements MustHaveExtractCommand
+{
+    protected $name = "Textract command";
+
+    protected $signature = 'textract:run {file_path} {job_id}';
+
+    protected $description = 'Extract text from the supported file';
+
+    public function handle(): int {
+        $file_path = $this->argument('file_path');
+        $job_id = $this->argument('job_id');
+        $output = app()->get('textract')->run($file_path, $job_id);
+        info($output);
+        return 1;
+    }
+}

src/Exceptions/TextractException.php

@@ -0,0 +1,8 @@
+<?php
+
+namespace Nilgems\PhpTextract\Exceptions;
+
+class TextractException extends \Exception
+{
+
+}

src/Extractor/Contracts/HasPhpWord.php

@@ -0,0 +1,53 @@
+<?php
+
+namespace Nilgems\PhpTextract\Extractor\Contracts;
+
+use PhpOffice\PhpWord\Element\Cell;
+use PhpOffice\PhpWord\Element\Text as PhpWordElementText;
+use PhpOffice\PhpWord\Element\TextRun as PhpWordElementTextRun;
+use PhpOffice\PhpWord\IOFactory;
+
+trait HasPhpWord
+{
+    protected function getSectionsText(string $file_path, string $readerName = 'Word2007'): string
+    {
+        $data = [];
+        $phpWord = IOFactory::load($file_path, $readerName);
+        foreach ($phpWord->getSections() as $section) {
+            $elements = $section->getElements();
+            $data  = [...$data, ...$this->getElementText($elements)];
+        }
+        return implode(" ", array_filter($data));
+    }
+
+    protected function getElementText(array $elements): array {
+        $docs = [];
+        foreach ($elements as $element) {
+            if($element instanceof PhpWordElementText) {
+                $docs[] = trim($element->getText());
+            }
+            if($element instanceof PhpWordElementTextRun) {
+                $nested_data = $this->getElementText($element->getElements());
+                $docs = [...$docs, ...$nested_data];
+            }
+//            if($element instanceof PhpWordElementTable) {
+//                $nested_data = $this->getTableRowText($element->getRows());
+//                $docs = [...$docs, ...$nested_data];
+//            }
+        }
+        return $docs;
+    }
+
+    protected function getTableRowText(array $rows): array
+    {
+        $data = [];
+        foreach ($rows as $row) {
+            foreach ($row->getCells() as $cell) {
+                if($cell instanceof Cell) {
+                    $data[] = $this->getElementText($cell->getElements());
+                }
+            }
+        }
+        return $data;
+    }
+}