使用Amazon Textract处理PDF时遇UnsupportedDocumentException错误求助

问题概述 #

使用Amazon Textract分析PDF时收到400 Bad Request错误，提示"Request has unsupported document format"。已验证文件MIME类型为PDF、符合大小限制，且已编写PDF拆分逻辑但未调用。

解决方案 #

1. 直接使用S3对象路径而非本地文件字节 #

Textract支持直接引用S3上的文件，避免本地文件读写可能导致的字节损坏问题。修改processTextract方法：

protected function processTextract($s3FilePath)
{
    $textractClient = new TextractClient([
        'region' => env('AWS_DEFAULT_REGION'),
        'version' => 'latest',
        'credentials' => [
            'key' => env('AWS_ACCESS_KEY_ID'),
            'secret' => env('AWS_SECRET_ACCESS_KEY'),
        ],
        'use_path_style_endpoint' => env('AWS_USE_PATH_STYLE_ENDPOINT'),
    ]);

    try {
        $result = $textractClient->analyzeDocument([
            'Document' => [
                'S3Object' => [
                    'Bucket' => env('AWS_BUCKET'),
                    'Name' => $s3FilePath
                ]
            ],
            'FeatureTypes' => ['TABLES', 'FORMS'],
        ]);

        return $result;
    } catch (AwsException $e) {
        return $e->getMessage();
    }
}

2. 启用PDF拆分逻辑 #

Textract的AnalyzeDocument接口单PDF最多支持11页，必须调用已编写的splitPdf方法拆分文件后逐个处理。修改uploadFile方法：

public function uploadFile(Request $request)
{
    $request->validate([
        'file' => 'required|mimes:pdf|max:2048',
        'validity' => 'required|date',
        'name' => 'required|string',
        'logo' => 'nullable|image|mimes:jpeg,png,jpg,gif,svg|max:2048',
    ]);

    // Upload do arquivo PDF
    $file = $request->file('file');
    $path = $file->store('pdf_tabelas_precos', 's3');

    // URL do arquivo PDF armazenado no S3
    $urlFile = Storage::disk('s3')->url($path);

    // Upload do logo, se fornecido
    $logoPath = null;
    $urlLogo = '';
    if ($request->hasFile('logo')) {
        $logo = $request->file('logo');
        $logoPath = $logo->store('logos', 's3');
        $urlLogo = Storage::disk('s3')->url($logoPath);
    }

    // 拆分PDF并处理每个子文件
    list($splitFiles, $downloadLinks) = $this->splitPdf($path);
    $allExtractedData = [];

    foreach ($splitFiles as $splitFile) {
        $extracted = $this->processTextractLocalFile($splitFile);
        $allExtractedData[] = $extracted;
        unlink($splitFile); // 清理本地拆分文件
    }

    // Salvar detalhes no banco de dados
    $operator = new Operator();
    $operator->name = $request->input('name');
    $operator->validity = $request->input('validity');
    $operator->url = $urlLogo;
    $operator->url_file = $urlFile;
    $operator->extracted_data = json_encode($allExtractedData);
    $operator->save();

    // Retornar links de download dos arquivos divididos
    return back()->with('success', 'Arquivo enviado e processado com sucesso!')
        ->with('file', $urlFile)
        ->with('downloadLinks', $downloadLinks);
}

// 新增处理本地拆分文件的方法
protected function processTextractLocalFile($localFilePath)
{
    $textractClient = new TextractClient([
        'region' => env('AWS_DEFAULT_REGION'),
        'version' => 'latest',
        'credentials' => [
            'key' => env('AWS_ACCESS_KEY_ID'),
            'secret' => env('AWS_SECRET_ACCESS_KEY'),
        ],
        'use_path_style_endpoint' => env('AWS_USE_PATH_STYLE_ENDPOINT'),
    ]);

    try {
        $result = $textractClient->analyzeDocument([
            'Document' => [
                'Bytes' => file_get_contents($localFilePath),
            ],
            'FeatureTypes' => ['TABLES', 'FORMS'],
        ]);

        return $result;
    } catch (AwsException $e) {
        return $e->getMessage();
    }
}

3. 验证PDF文件完整性 #

• 检查上传的PDF是否为可搜索PDF（非扫描件图片PDF）：若为扫描件，需使用Textract的DetectDocumentText接口
• 验证本地临时文件的字节完整性，添加哈希校验：

// 在原processTextract方法中添加哈希验证
$fileContent = Storage::disk('s3')->get($s3FilePath);
$localPath = tempnam(sys_get_temp_dir(), 'pdf_') . '.pdf';
file_put_contents($localPath, $fileContent);

// 验证哈希一致性
$s3Hash = md5($fileContent);
$localHash = md5_file($localPath);
if ($s3Hash !== $localHash) {
    unlink($localPath);
    throw new \Exception('Arquivo local corrompido durante o download do S3.');
}

关键注意事项 #

• Textract的AnalyzeDocument仅支持文本可搜索的PDF，扫描件需用DetectDocumentText接口处理
• 单PDF文件页数限制为11，必须拆分后处理
• 直接使用S3对象路径能避免本地文件读写的字节错误

内容的提问来源于stack exchange，提问作者Marcello Pato