Trying to understand filters and missing something...

[ehcloninger]

I'm trying FreshRSS self-hosted and like what I've seen so far. Coming from 10 years of paying for Newsblur. I specifically want better control over filtering than Newsblur's "learning" method.

Here's one example...

https://boingboing.net/feed

In this feed, I want to suppress the articles that are things for sale. Not that I have anything against them monetizing their feed, but I never buy anything from them, so why waste my time reading the headlines?

I started with a filter for author:Boing Boing's Shop and waited a day for the next batch of articles to arrive. The articles still included the stuff for sale. So, I tried #'BoingBoing Shop' as that is one of the tags in the article. Still not working. So, I pulled the feed manually and looked at the text. Here's the relevant XML.

		<dc:creator><![CDATA[Boing Boing's Shop]]></dc:creator>
		<category><![CDATA[BoingBoing Shop]]></category>
		<category><![CDATA[shop]]></category>

Does dc:creator equate to the author: filter or do I need to specify some other filter keyword to get this?

The category filter appears to want a numeric value rather than CDATA. Am I missing something about how filters work?

On a usability note, I'd say it would be useful to have a "preview filter" function to see the filter results on the feed with articles that are already in the database, rather than waiting for the feed to update and remembering to check the filter.

My next attempt will be to filter articles from technology feeds that includes the name of a certain car company and their CEO, but somehow leave intact the articles relating to the 19th/20th century Serbian inventor.

Thanks!

[Alkarex]

Hello,
In that case, searching on #shop is probably sufficient.

A little tip: just click on an existing tag to get the syntax right:

Or click on an author:

To test that your filters are working, you can already try to see whether using them to search finds the articles you want.

By the way, an alternative to automatically marking such articles as read is to construct a complex search with many search terms you are not interested in, and save that search as a User query, and click on it to have a quick look before marking all as read, as one could do with a spam folder.

Something like intitle:ABC OR intitle:"Hello World" OR intitle:🔥 OR #shop OR #Deals etc. (regex is also supported)

In our current rolling release (edge), there is a new advanced search form to help with the syntax:

• Implement search form #8103

[ehcloninger]

Ah, I think I see the disconnect on my part. The filters are working, but they apply only to the reader UI, not articles that are delivered via the API. I tend to do my reading on mobile and use a reader app on Android. Those articles don't appear on the FreshRSS web UI, but they get delivered to third-party clients. Your tips are interesting, I need to figure out more about that because I feel that's what Newsblur has been missing. I don't need AI to do things for me, I just need to find ways to do it for myself.

I saw there's a pending PR to merge in a checkbox to apply the filters to delivered articles. That should help as well.

Great project and I'm glad it's got such an active community. Thanks for your help @Alkarex

[Alkarex]

When articles are marked as read, this information is conveyed through the API as well

[ehcloninger]

Hmmm. OK, maybe I have something else going on. Appreciate the info!