What’s the best search engine for lemmy? (sh.itjust.works)

submitted 11 months ago by oak00@sh.itjust.works to c/fediverse@lemmy.world

12 comments fedilink hide all child comments

top 12 comments

sorted by: hot top controversial new old

[-] Blaze@sopuli.xyz 53 points 11 months ago

[-] marsara9@lemmy.world 45 points 11 months ago

Thanks for the shout-out.

But FYI I've run into some bugs that's preventing new content from being indexed. So you won't see anything new (from about a week ago) until I can find a new method to fetch new posts.

[-] leo@lemmy.linuxuserspace.show 11 points 11 months ago

Just dropping by to say “thanks”

So, thanks for putting in the work. I know it isn’t easy. And can easily be overlooked. ❤️

[-] redcalcium@lemmy.institute 6 points 11 months ago* (last edited 11 months ago)

Would it be better if your backend federates with popular instances so they push new posts and comments to your search engine? That way, you don't need to scrape those instances to index their contents (because they'll voluntarily send the contents to you via activitypub).

[-] marsara9@lemmy.world 8 points 11 months ago

Yep that's the new idea. The sad part is that with this method there's no way to get historical data. Only new posts. So if a server goes down, gets DDOSd etc... I'll lose posts forever.

Also building an ActivityPub implementation from scratch isn't trivial either. So that'll take some time.

I've got a few other ideas I'm playing with as well. Like just assuming that internal post IDs are all sequential and literally fetching them one by one. Or maybe some combination of both?

[-] Die4Ever@programming.dev 5 points 11 months ago* (last edited 11 months ago)

Instead of building a new ActivityPub implementation, you could just run a regular instance of Lemmy and pull data from its database directly? Or use its API for searches?

[-] marsara9@lemmy.world 6 points 11 months ago

I was using it's APIs. But new restrictions have effectively been put in place that prevent me from using them for what I need. Similar API calls were being made that were causing DDOS attacks on lemmy.world.

As for running a lemmy instance itself. That's a thought but I need the data in a different format to do efficient searches. It's a tricky problem.

[-] redcalcium@lemmy.institute 2 points 11 months ago* (last edited 11 months ago)

Why not talk to the instance admins directly and ask for their database dumps (minus the user accounts table and DMs) so you can ingest it into your search index? You're doing this for the benefit of the fediverse, right? I'm sure most instance admins would help you if you ask (and it's easier on their servers too because your scraper aren't bombarding their server), I know I would, tough my instance only have 2 users right now so it'll probably useless to index. This should take care the historical data problem, and you can use activitypub for obtaining new data going forward without scraping the instances.

[-] Gamey@lemmy.world 12 points 11 months ago

The issue is that you can't search a ton of servers at once unless you index all their posts. To get a proper search engine someone would not only have to spend the time to code it but also the money to host that beast! :/