Feature: Import related tags to articles (Categories)

[LinkJim]

Authors : Jimmy HUYNH (@LinkJim) - Marco NASSABAIN (@mnassabain)

🚀 Feature

Import related tags for each article

• Could be useful to sort articles by tags (in future...)
• Could be integrated in front-end to show keywords
• Could be used to generate hashtags for sharing on social media (e.g.: Twitter)
• UX / UI Friendly !

We are already working on sharing articles with users and on social media. (pr existants)

In case of sharing articles on social medias, we thought that sharing posts, would be more relevant if hashtags were automatically generated. Being that generating tags based on a text is a complex problem, we would use the tags retrieved from the RSS feeds instead.

This is an example of usage that the newly added "tags" field would offer. But we think that there are surely more advantages.

💻 Implementation

• We added a migration that inserts a "tags" column for news_items.
• We updated the Item model to include the "tags" field.
• In the FeedFetcher, we simply import the tags using the "categories" tag from the RSS feed.

💡 Questions

• In our current implementation the "tags" column is a "text" field. We thought that it is more appropriate than a string with a defined length?
• This current implementation only adds tags for newly imported articles. Already existing articles in the database will not be updated.

✅ To do

• Add tests
• Update already existing items to fetch their tags? Would this be possible?

---

Any recommandation would be appreciated, thanks. ;-)

[SMillerDev]

Code looks fine to me. But it will need tests.

[anoymouserver]

✅ To do

• Update already existing items to fetch their tags? Would this be possible?

If an article has been changed at the actual feed provider, it is automatically updated (old items, which no longer exist in the feed, are not updated for obvious reasons). But since this is 'only' a transitional problem, I honestly wouldn't do anything to force-refetch it.

---

💡 Questions

• In our current implementation the "tags" column is a "text" field. We thought that it is more appropriate than a string with a defined length?

Using text seems reasonable, as it's unknown how long the content will be.

---

I'm not sure if 'tag' is the best name for it, since I personally associate it with a tagging feature for the user for sorting and filtering (NC does the same with 'Collaborative tags').
Maybe using the feed's naming of 'category' would be clearer.

The fetched categories should then also be added into generateSearchIndex.

[mnassabain]

To do

• Rename "tags" to "categories"
• Return categories in jsonSerialize (we think this is right?)
• jsonSerialize should return categories as an array
• Add categories to generateSearchIndex
• Add tests
• Use json to implement categories
• Update changelog

[LinkJim]

Thanks for your quick replies and support.

Features and related tests have been successfully implemented according to your guidance.
One thing we're not sure about for the search index. We simply dump the json of the categories into it.. Does it seem correct or is there an eventual convention applied ?

Check it out and let us know about what you think !

[SMillerDev]

The searchindex is a concatenated string of all search terms. So adding something like join('', $this->categories()) would be better.

[SMillerDev]

Thanks @LinkJim!

[anoymouserver]

I just wanted to try it on my test instance but found no feed with categories included. Any suggestions?

[mnassabain]

Since it is an optional element (specification), not all feeds have it.

The easiest example of a feed that contains categories is the Nextcloud News feed. You can find it in the explore page, or you can subscribe manually to it via this url: https://nextcloud.com/blogfeed

[anoymouserver]

Yes I know .. I've tried multiple news sites and blogs, but found none with a category tag.
But I didn't even think about the NC feed .. thanks 👍