'Order by' should remember choice (Seeders, Size, Name, etc.)

[dominicirving]

When searching in the web UI, the results are always ordered by 'Relevance'. I personally like to always see them ordered by Seeders. I have to manually set the ordering after every individual search.

I'd like to have the sorting order to be user-settable and remembered as a default, in the same way as the Theme is remembered.

I've tried to dig into the source code myself but unfortunately the web UI seems to be obfuscated/pre-compiled and not clearly editable.

[mgdigital]

I see what you're saying, I sometimes end up doing the same - but only for certain searches. The behaviour you're describing is too fussy/confusing to surface for most users. When you search you expect relevant items at the top. Any other default ordering will give worse results for many searches. Relevance is only applicable when search is populated, so you propose it remembers a separate order when search is populated Vs not? That's too complicated and would result in its own issues being opened. Manually setting the order you want takes 1 second and is probably the best compromise.

The frontend code isn't obfuscated, I assume you're looking at the build output, you're free to edit and rebuild as you like if you wanna customize the behaviour.

[SjoerdV]

I really don’t understand people advocating something as subjective and fuzzy as ‘Relevance sorting’. Result are relevant for who, for what, with which criteria? It’s just always wrong for me…

So I already updated my pipeline to only show results with at least 3 seeders.

Now I just want the results to present themselves chronologically by default, everywhere at all times.

I would like to advocate adding some flexibility for the UI and adding a ‘default sort order setting’ or some cookie storing the search order would be fine as well. Seems all very basic to me.

[mgdigital]

I really don’t understand people advocating something as subjective and fuzzy as ‘Relevance sorting’

Relevance sorting is absolutely not subjective or fuzzy. See a precise explanation of what it means here: https://www.postgresql.org/docs/current/textsearch-controls.html#TEXTSEARCH-RANKING - in comparison, crawl recency (the time your client happened to first see the torrent) and number of seeders (an arbitrary number one node on the DHT told you) are both fuzzier than the relevance score.

Now I just want the results to present themselves chronologically by default, everywhere at all times. I would like to advocate adding some flexibility for the UI and adding a ‘default sort order setting’ or some cookie storing the search order would be fine as well. Seems all very basic to me ...

The problem is that there is no one default sort order that's the best for all searches. That's why a "default sort order" setting doesn't make much sense IMO. If you're looking for something old then "most recent" is a bad sort; if you're looking for something rare then "most seeders" is probably wrong; sorting by relevance is predictable and gives a high chance to have the best result on page 1 for any given search. For the reasons I previously explained I think anything else would be confusing and create more problems than it solves. Switching to your desired sort order is 1 second of work. I'm not shutting down the argument but that's my view and nothing so far has persuaded me otherwise...

[SjoerdV]

Well I read the link. But this is something that ‘works the same as the predominant AI slop we see today’. Fuzzy, fuzzy, fuzzy… And I have a history with enterprise search, so I know a bit.

Such an algo will just never know what I find relevant… not with any complicated non-deterministic math in the world! For me it just keeps suggesting ‘Casablanca (1942)’.. like ALWAYS 😂

If you don’t want to do it (or at least fix the nasty bug that changes the selected sort order drop down, if you switch to search with or without a search term in the the search box) then so be it.

[mgdigital]

fix the nasty bug

It's not a bug it's deliberate. Whether it's the right decision is a different question but it isn't a bug.

Such an algo will just never know what I find relevant

Relevance is the ranking score Postgres assigned the search term you typed in, without which you wouldn't be seeing any search results at all. It doesn't aim to be what you "find" relevant.

[SjoerdV]

It's not a bug it's deliberate. Whether it's the right decision is a different question but it isn't a bug.

O... well, I might suggest it being a wrong decision 😉

"It doesn't aim to be what you "find" relevant."

Well, then why do I need to see that option at all? If it is all automatic under the hood (and not meant for me anyways) it should not be a UI element.

I have the feeling we have to agree to disagree on this, no prob!