Anyone else noticed the Discourse search feature is fairly poor?

Hello everyone!

I keep trying to find specific posts on this forum, but I cannot find them. Recently, we had a post about cellular providers that work with the Librem 5 and I know this has been discussed here.

However, it does not show in the search.
Usually, for posts like these I would help provide users with the previous discussions, but I cannot seem to find them.

Does anyone else have this issue with Discourse? I wanted to create the discussion here.
Thank you all so much for your time and for helping create such a wonderful community.

—Best
Myrcy

2 Likes

Hi!

One thing that I have seen is that ā€œSort by Relevanceā€ in Discourse and in some other apps I’ve used for my work recently never seems to make sense to me. The data metrics are not smart and don’t know what’s relevant. But if I sort by ā€œLatest Postā€ then I get a chronological history which makes more sense to me.

Are you looking for this thread:

They link to this (official band list):

And this (wiki of stuff people tried):

1 Like

What I meant is that this specific topic has been discussed multiple times on the forum.
However, I believe search only shows one of these topics.

Similarly, I know AT&T has many posts on the forum, but none actually show up when searching:

Perhaps, these are just outliers?

1 Like

I think it is generally accepted that the built-in search is not as good as a dedicated search.

This forum is about to upgrade to Discourse 3.5.0 and, who knows, maybe there will be some search improvements.

It helps if you can remember anything else about the post (who? when?) and thereby use the Advanced Filters.

AT&T may be a bad example because it may be treated as two words, both of which are too short to search for.

2 Likes

I usually sort the results by most recent, and sometimes add an @(username) if I remember who posted it.

2 Likes

PS It has already been complained about that the default external search that is offered is Google! :face_vomiting:

2 Likes

To be fully honest, I thought perhaps I wrote it.
Regardless, there used to be a more thorough discussion on AT&T/ATT’s support for the Librem 5.

I was hoping to add the post to the discussion as, unless I recall improperly, some staff members member helped provide details as to why AT&T is not fully supported.

1 Like

What is the difference between built-in and dedicated search?

1 Like

1 Like

Here is the post:

It is just weird that it does not show up under any search result (at least that I could think of).

2 Likes

I mean .. a company or project whose sole function on earth is to do good search.

Discourse has built-in search but project resources are going to be spread across a range of functional areas and search may not be the highest priority.

3 Likes

Just to elaborate on that … dedicated search functions allow boolean operators and allow careful overrides so that ā€œAT&Tā€ is not misinterpreted, and support stemming/synonyms, and may allow some control over what is a word delimiter, and allow proximity matching … yada yada yada.

The bottom line is that you can always use an external search provider (of your choice) - for those parts of the forum that are public - and then you get whatever better functionality is offered by that provider. And, yes, you are even free to use Google.

1 Like

I think that it is the ā€˜&’ messing things up, and ATT works better as long as the post or title also contain att (in any case), but things like attack also match. " ATT " works better, but if there is any punctuation instead of " " before or after ATT there will not be a match.

2 Likes

Or, in short, good searching is harder than it looks e.g. the ampersand character is usually a word delimiter but not in this one forum-specific case. (By forum-specific I mean that AT&T will be mentioned in this forum far more than in an ā€œaverageā€ forum, and with much greater significance. So we would like an exception even if the average forum wouldn’t bother.)

1 Like

It would be nice if something like AT\&T or regular expressions like AT?T or AT.T would work, but no such luck. We could just start calling them Southwestern Bell, which is their real name. (In the US in the 1980’s all the regional operating companies (Baby Bells) got forcibly spun off from AT&T by the government with reason. (ā€œWe don’t care, we don’t have to, we’re the phone company.ā€ - Lilly Tomlin’s Ernestine on Rowan and Martin’s Laugh In) Over time SBC bought some of the other Baby Bells and then eventually the desiccated husk of the former AT&T to get the name.)

1 Like

That’s really really hard to do efficiently, in the general case, when you may have gigabytes of post text to examine. (With general REs as a basis of comparison, AT.T actually wouldn’t be too bad.)

You don’t want the search function to be a DoS attack. Some web sites (some forums) only allow (advanced) search by logged-in users, for this reason.

As a recent real-world example of the difficulties of roll-your-own search functionality … go to IMDB and try to find the film whose title is 1. (Just my opinion, but it was a disappointing film, so you may be better off not finding it. :wink:)

There are some challenging trade-offs between simplicity/usability and richness of functionality.

But you don’t need REs to find AT&T. All you need is for both the indexing code and the client-interface code to recognise it as a word (which it is).

2 Likes