search api attachments does not appear to support s3fs

@SomebodySysop are you seeing anything in the logs?

Just looking through the code, TextExtractorPluginBase::getRealpath() is of interest to me:

  /**
   * Helper method to get the real path from an uri.
   *
   * @param string $uri
   *   The URI of the file, e.g. public://directory/file.jpg.
   *
   * @return mixed
   *   The real path to the file if it is a local file. An URL otherwise.
   */
  public function getRealpath($uri) {
    $wrapper = $this->streamWrapperManager->getViaUri($uri);
    if($wrapper != FALSE){
      $scheme = $this->streamWrapperManager->getScheme($uri);
      $local_wrappers = $this->streamWrapperManager->getWrappers(StreamWrapperInterface::LOCAL);
      if (in_array($scheme, array_keys($local_wrappers))) {
        return $wrapper->realpath();
      }
      else {
        return $wrapper->getExternalUrl();
      }
    }
  }

I'm wondering if the method is failing to return a usable value here.

Thanks for the response. I have not looked at it in over a year since I don't know enough about the module to develop a patch. It was my hope that someone with much more knowledge would stumble upon this and figure out a solution.

This patch will resolve the issue, but please ensure that your site's domain has access to the S3 bucket.

We've been using this module with s3fs for a long time with no additional patches or code changes needed but from memory solr needs to be configured to allow remote streaming

Went back to have a look at the project we were using for this.

Originally when using Solr 8.x we had enableRemoteStreaming set to true through some custom request dispatcher config - something like:

search_api_solr.solr_request_dispatcher.request_dispatcher_remote_streaming.yml:

uuid: ....
langcode: en
status: true
id: request_dispatcher_remote_streaming
label: 'Remote Steaming'
minimum_solr_version: 7.0.0
environments: {  }
recommended: true
request_dispatcher:
  name: requestParsers
  enableRemoteStreaming: true
  multipartUploadLimitInKB: -1
  formdataUploadLimitInKB: -1
  addHttpRequestToContext: true

In later solr version this changed to being enabled by an environment var - so now we just have an environment variable:

SOLR_OPTS: "-Dsolr.enableRemoteStreaming=true"

But before fill you with false hope - we were using S3FS module but with the public file takeover, meaning the bucket is publicly accessible and the external URL is used. I haven't tested with the non public wrapper.