[FEATURE] Ability to stream to an Amazon S3 bucket

[ayacoo]

Related #2249

This is:

- [ ] a bugfix
- [x] a new feature

Requirements
composer required aws/aws-sdk-php

Example

require '../vendor/autoload.php';

use PhpOffice\PhpSpreadsheet\Spreadsheet;
use PhpOffice\PhpSpreadsheet\Writer\Xlsx;

$client = new Aws\S3\S3Client([
    'credentials' => [
        'key' => 'KEY',
        'secret' => 'SECRET',
    ],
    'default_cache_config' => '',
    'certificate_authority' => true,
    'region' => 'REGION',
    'version' => 'latest'
]);

// Register the stream wrapper from an S3Client object
$client->registerStreamWrapper();

$spreadsheet = new Spreadsheet();
$sheet = $spreadsheet->getActiveSheet();
$sheet->setCellValue('A1', 'Hello World !');

$writer = new Xlsx($spreadsheet);
$writer->save('s3://BUCKETNAME/myfile.xlsx');

[oleibman]

I usually like to see a unit test demonstrating that the change works. However, such a test seems impractical here - it requires access to s3, which probably introduces security and availability challenges. It seems clear to me that the change will have no adverse effect for non-s3 access.

[ayacoo]

Of course, you could refactor a little more. Example: There is a function setModeForFileHandle() There one could test also without S3 package dependency simply whether the scheme changes.

For future changes e.g. further PHP wrapper this would be surely also more suitable. What do you think?

[alexanderkraemer]

This change makes a lot of sense! Looking forward for it getting merged!

[oleibman]

Thank you for your contribution.

[warespace-silas-winter]

Hi everyone, sorry for reopening the pull request.

I see that the S3 protocol works for the writer. Could s3:// be a workaround for #1147 and #1931? At least for me it would be great.

[oleibman]

I think s3:// would work in much the same way as http://. Unfortunately, http:// doesn't work for read. It will fail in Shared\File::assertFile because is_file will return false. And, even if I change the code to bypass that check, it will will fail in Reader\Xlsx::loadSpreadsheetFromFile because $zip->open will fail and return ZipArchive::ER_READ.

However, the following code may be useful to you (substitute your own s3 filename - I have no way to test that):

        $filename = 'http://localhost:8000/filename.xlsx';
        $contents = file_get_contents($filename);
        $tmpfile = tmpfile();
        $path = stream_get_meta_data($tmpfile)['uri'];
        echo "path is $path\n";
        fwrite($tmpfile, $contents);
        fseek($tmpfile, 0);
        $reader = new \PhpOffice\PhpSpreadsheet\Reader\Xlsx();
        $spreadsheet = $reader->load($path);
        fclose($tmpfile);

The temporary file should be deleted by the fclose, and you can always write some sort of handler to force its deletion even if the spreadsheet load fails.