Add an option to delete all nodes when the feed source is empty

Thanks for spotting my duplicate issue. https://www.drupal.org/node/2553331

I wonder if there are others having the same issue, it's kind of hard to find the exact cause because it isn't consistent (everything works as expected until the feed is completely empty).

Also, I was thinking about it more, and it would be nice to not even have the "Delete nodes even if feed is empty" setting. I'm not familiar with the code, but to me it makes sense that if the request to get the feed is an HTTP 200, and the parser doesn't throw an exception, all content that was imported should be deleted/unpublished. The UI is already complicated, I say default to removing the nodes unless there's an obvious error.

For now, if anyone is interested in the checkbox solution for their project, this should do the trick.

Setting to "Needs review" so the testbot will evaluate the patch.

I have tweaked the patch above to work with the latest dev release. Please review and test accordingly.

Setting to "Needs review" so the testbot will evaluate the patch.

It would be great to have an automated test for this feature.

Quoting MegaChriz https://www.drupal.org/node/1470530#comment-9120655

@AndrackOSnack
I noticed back in #181 too that no items are deleted/unpublished when the source is completely empty, but I would say this is intentional. I think that it is rare to be wanting to import a blank source. For example, it could happen that something went wrong during fetching the source, ultimately resulting in a blank source. You don't want all items to be deleted in such case.

It depends.
If the fetcher has no results due to an error, then I think this is true. Note that the Fetcher plugins will throw Exceptions in case of a failed download.
If the fetcher is working properly but there are no results, shouldn't deleting/unpublishing be default behaviour?

So my proposal would be to drop "delete_if_empty" setting (or make it default) and trust on a catched Exception in FeedsFetcher. That should stop the FeedsProcessor from removing/unpublishing nodes in case of an error.

We definitely need this functionality. We use it to show alerts published from an incident management tool. When all alerts are cleared, we need the alerts to be unpublished without manual intervention. We have other checks in place to ensure the feed is not broken.

@lauren.moriarty
If you can provide an automated test for this feature, I'll add it to the module.

@ndf
If a download fails, it *could* be that the import stops and nothing gets deleted. But in that case that would also require tests to ensure unwanted deletions do not happen. We would for example need to test what happens if the source returns a 404. And what happens if the source has a WSOD and returns a complete blank page.
A setting will be safer.

#8 worked for me but the title "Delete items even if source feed is empty" is a bit misleading.

Something like "Execute action if the source is empty" or something like that should be more clear, as it actually executes the configuration you chose in "update-non-existent".

Updated #8 to make wording more generic.

Reverse the checkbox default so that process empty source is off by default. When using this option, it is important to validate that you feeds is actually empty. not because of network or source change or other factors.

e.g One way would be use hook_feeds_after_parse().

function my_module_bedework_feeds_after_parse(FeedsSource $source, FeedsParserResult $result) {
  if ($source->id == 'my_events') {
    $events = $result->items;
    if (empty($events)) {
      $raw_response = $source->fetcher_result->getRaw();
      $json = json_decode($raw_response);
      if (!isset($json->keytocheck)) {
        $config = $source->importer->processor->getConfig();
        $config['process_empty'] = FALSE;
        $source->importer->processor->setConfig($config);
      }
    }

Re-rolled patch for the latest 7.x-2.x release

This does still require an automated test.

Is anyone working on this feature for Drupal 8?

Hi
I suggest that the FeedsResult class shoud have an attribute to store error codes at both fetching and parsing stages so they can be passed on to the processor, so that a processor can make an "informed" decision on where it is appropriate to delete all items when the feed is empty, e.g. if the httpfetcher downloads an empty rss xml document with http status code 200 then the parser correctly parses it as empty (so error code is "no error") then the processor could know that the feed is actually empty with no errors and delete all items (if user selected the option), if either the fetcher or the parser returned an error then it would instead keep all items and log a warning to alert the admin that something was wrong with that feed.

Thank you for the fantastic module and regards.

Crossposting from https://www.drupal.org/project/feeds/issues/3337811#comment-15145862:

If looking for a work around so that feeds will unpublish missing nodes even when the response is empty...

Set your importer to include the published status.

Then use the following code to ensure there's always at least one node in your response.

...I did not realize that feeds do not import nodes when the response is empty or there are no items.

For future reference, I added the following code for my REST Export to ensure there is always at least one item in my response.

use Drupal\views\ViewExecutable;
use Drupal\views\ResultRow;
use Drupal\node\Entity\Node;

/**
 * Implements hook_views_post_execute().
 */
function HOOK_views_post_execute(ViewExecutable $view) {
  if ($view->id() == 'MY_VIEW_ID') {
    // Feeds does not import an empty response. Therefore, add a dummy feeds item.
    // @see https://www.drupal.org/project/feeds/issues/2333667#comment-15145873
    if ($view->total_rows === 0) {
      $dummyNode = Node::create([
        'type' => 'job_posting',
        'nid'  => -1, // Our GUID. This is valid because we never save the node
        'title' => 'PLACEHOLDER NODE',
        'status' => 0, // Be unpublished, so it doesn't appear on the importing website
      ]);
      $view->result[] = new ResultRow([
        '_entity' => $dummyNode,
      ]);
    }
  }
}