Importing files from a URI with a fid - D6 to D7 migration

So it looks like core file_save() sort of prevents this possibility, since it uses drupal_write_record to do either an update or insert based upon whether or not $file->fid is set.

If we set $file->fid, file_save() will try to update a row and we get nothing inserted.

For completeness I guess, eg there might be a token in a text area linking to a specific file id, or a custom module might provide zipped downloads based upon a path including a file id eg /downloadzipped/$fid

So core file_save is quite a short function and what I've done for my case is provide an alternate version in the class to save with the file id.

  /**
   * Clone of core file_save allowing us to upsert based on file id.
   *
   * @param $file
   * @return $file
   */
  protected  function file_save($file) {
    $file->timestamp = REQUEST_TIME;
    $file->filesize = filesize($file->uri);

    // Load the stored entity, if any.
    if (!empty($file->fid) && !isset($file->original)) {
      $file->original = entity_load_unchanged('file', $file->fid);
    }

    module_invoke_all('file_presave', $file);
    module_invoke_all('entity_presave', $file, 'file');

    // Convert the $file object to an array for db_merge
    $file_fields = array();
    $file_schema = drupal_get_schema('file_managed');
    foreach ($file_schema['fields'] as $name => $properties) {
      if (isset($file->{$name})) $file_fields[$name] = $file->{$name};
    }
    $fid = (isset($file_fields['fid']))? $file_fields['fid'] : null;

    // Upsert the file record
    $status = db_merge('file_managed')
      ->key(array('fid' => $fid ))
      ->fields( $file_fields)->execute();
    switch ($status) {
      // Record inserted
      case (STATUS_INSERT):
        $file->fid = db_query('SELECT LAST_INSERT_ID();')->fetchField();
        module_invoke_all('file_insert', $file);
        module_invoke_all('entity_insert', $file, 'file');
        break;
      // Record updated
      case (STATUS_UPDATE):
        module_invoke_all('file_update', $file);
        module_invoke_all('entity_update', $file, 'file');
        break;
      default:
        watchdog('file', 'Error upserting file record for uri', $file_fields);
    }

    unset($file->original);
    return $file;
  }

Ugly but working. I've also altered $source->processFile($file->value, $file->uid) to pass in the file object as well of course. Don't have a patch for this to hand, figured it would be too brittle to get in.