RSS import cleanup #

Read this page if your site imports RSS feeds, or if you used the bccleanuprss extension. An RSS import adds an object for every item in the feed and never takes one away. A site importing a busy feed therefore grows without limit: the destination folder reaches tens of thousands of children, the administration interface slows down on it, the search index fills with items nobody will read again, and the nightly backup carries all of it. Since 6.0.15 the kernel can trim that for you: keep the newest N items of each active RSS import; remove the rest.

The code is ported from the bccleanuprss extension by Brookins Consulting (https://github.com/se7enxweb/bccleanuprss), GPL v2 or later. The original's selection method is kept; see What was kept, and what changed.

In short #

Set it up in five minutes #

1. Add to settings/override/content.ini.append.php:

   [RSSImportCleanupSettings]
   Enabled=true
   KeepPerFeed=200
   ClassIdentifiers[]
   ClassIdentifiers[]=article
   MoveToTrash=true

2. Clear the INI cache: php bin/php/ezcache.php --clear-tag=ini --allow-root-user

3. Dry-run and read the list: php bin/php/cleanuprss.php --dry-run --allow-root-user

4. Run it: php bin/php/cleanuprss.php --apply --allow-root-user

5. Schedule it nightly: php runcronjobs.php cleanuprss in your crontab.

It does nothing until you tell it what to remove #

This removes content. Three separate conditions must hold before it will touch anything, and none of them holds on an installation that has not been set up for it:

1. Enabled=true
2. ClassIdentifiers[] names at least one class
3. KeepPerFeed is at least 1

Any one of them unmet and the run stops and says which:

$ php bin/php/cleanuprss.php
Nothing was removed: RSSImportCleanupSettings/Enabled is not true in content.ini.

That is not an error — an installation that has not asked for this is the normal case. Both the script and the cronjob exit cleanly, so the part can be scheduled before it is configured.

Settings #

settings/content.ini, beside the import's own settings. Override per siteaccess like anything else.

[RSSImportCleanupSettings]
Enabled=false
KeepPerFeed=100
ClassIdentifiers[]
CleanupUser=admin
LogFile=rsscleanup.log
MoveToTrash=true

A working configuration:

[RSSImportCleanupSettings]
Enabled=true
KeepPerFeed=200
ClassIdentifiers[]
ClassIdentifiers[]=article
MoveToTrash=true

About MoveToTrash #

true is the default because an automated, scheduled, permanent deletion of content with no way back is not a reasonable default. Note that the trash then grows instead — pair it with the trashpurge.php cronjob, or set it to false once you trust the class list.

Running it #

By hand #

php bin/php/cleanuprss.php --dry-run --allow-root-user
php bin/php/cleanuprss.php --apply --allow-root-user
php bin/php/cleanuprss.php --apply --keep=50 --allow-root-user
php bin/php/cleanuprss.php --apply -s my_siteaccess --allow-root-user
./console exp:cleanuprss --dry-run --allow-root-user     # the same script through the console

(--allow-root-user only when you run as root. The dry run was run on this installation and, with the shipped Enabled=false, printed the "Nothing was removed" line shown above.)

Plus the ordinary script options (-s, -q, -d, --help).

Always dry-run first on a site you have not run this on before. It prints every item it would remove, by name and node id.

As a cronjob #

php runcronjobs.php cleanuprss
./console cron:cleanuprss        # the same part through the console

The part is declared in settings/cronjob.ini:

[CronjobPart-cleanuprss]
Scripts[]=cleanuprss.php

It is a named part rather than one of the default Scripts[], so it runs only when asked for. Nightly is usually right — it is the import that needs throttling, not the clock.

What it actually does #

For each active import (eZRSSImport::fetchActiveList()):

1. Take the import's destination_node_id. Skip the feed if it has none, or if the node is gone — that is reported, not fatal.
2. Ask for that node's children, of the configured classes, published status, depth 1, sorted newest first and offset by KeepPerFeed.
3. Whatever comes back is by definition the surplus; remove it in one call.

Step 2 is the whole trick, and it is the original's. Because the keep count is the query's offset, there is no second pass deciding what to drop and no window in which the newest items could be returned — the items to keep are skipped by the database.

Depth => 1 means only direct children. An item's own children go with it when it is removed, but a sub-folder of the destination is not searched.

From a template or PHP #

$cleanup = new expCleanupRSS( array( 'dry-run' => true ) );

if ( $cleanup->isEnabled() )
{
    $cleanup->cleanup();
    $counts = $cleanup->counts();   // feeds, examined, removed
    $perFeed = $cleanup->perFeed(); // name, node, surplus
}
else
{
    echo $cleanup->reason();
}

Constructor options: dry-run, keep, quiet.

What was kept, and what changed #

The selection method — newest-first sorted, offset by the keep count — is the original's and is why this port exists rather than a rewrite.

Everything around it changed, and most of the changes were not cosmetic:

The PHP 4 constructor is worth spelling out: PHP 8 removed them, so new BCCleanupRSS() no longer ran that method. Every property it set — rssLimit, rssClasses, logFile — stayed undefined, and the class could not work at all on any supported PHP version.

The default class list is worth spelling out too. RSSClasses[]=1 is folder on a stock installation. Enabled by default, against a destination containing folders, that removes folders.

Naming #

bc-prefixed names are gone, as this is no longer a vendor extension: BCCleanupRSS → expCleanupRSS, bccleanuprss.ini → a block in content.ini. Brookins Consulting's copyright lines are kept in every ported file — the branding goes, the attribution does not.

Files #

Tests #

A private functional script builds its own fixture (a folder, an active eZRSSImport pointing at it, twelve articles published an hour apart and one folder that must survive because its class is not on the list) and checks that each of the three conditions refuses on its own, that a dry run finds the surplus and removes nothing, that a real run leaves exactly the newest five, that the unlisted class is untouched, that a second run finds nothing, and that what was removed is recoverable from the trash. It is not shipped. To check the behaviour yourself, follow "Running it" with --dry-run first on a copy of the installation.

Related pages #

• RSS: feed list, podcasts and safer exports and OPML exports
• September 2026, first half: 15 September
• Changelog 6.0.15
• Cronjobs console
• Syndication specification