Ask HN: What do you need in HN search?

66 points by jlemoine ↗ HN
Based on your feedback we've come up with a new search UI that should better stick to HN's style: https://new-hn.algolia.com

Tell us what you think about it and what you'd like to see next (we're thinking for instance about adding search analytics on frequent queries in the last 24h/month/....)

Talking about next evolutions, we actually had something in mind and wanted to run it through HN's community. It's kind of crazy different so we hope we'll get tons of feedback on this one. This is an experimental skin with more search features and you can try on https://new-hn.algolia.com/?experimental

Here is a blog post explaining our reasoning about this experimental style: http://blog.algolia.com/try-new-experimental-version-hn-search/

We can't wait to get your comments and ideas!

60 comments

[ 2.7 ms ] story [ 111 ms ] thread
Perhaps this is too far outside of your existing scope of functionality but I'll throw it out there: I would really love a "Hacker News Alerts" feature. I make the same searches every week on a few key terms but it's a manual process.

If you could push new search results directly to my inbox it would be convenient.

I ran into a problem today where I'm trying to find something I saw a really long time ago - so sorting by date from oldest to newest would be great.

EDIT - If I can filter by date ranges like others have suggested, I would much prefer that to my sort direction request!

I do this all the time. I need to go to google and hope I can find it. I usually only remember some small thing about it... like "flights" or "data set". Sorting by date it was on the front page of HN would be best for me.
Well, my large corporate employer uses Websense which seems to categorize hn.algolia.com as ("Category: hacking"), and blocks all queries (HN itself is fine). So I guess I'd say I need...search. Maybe you can get off the blacklist/on a whitelist?
Oh, interesting :/ Is it the same on "new-hn.algolia.com"? (we removed the HTTP OPTION query that was performed to deal with extra headers & CORS -> maybe it didn't help).
I can access new-hn though I can't say whether it's due to your changes. The https redirect adds an extra step because of a click through "this site could be evil" that Websense adds--I think it stores the acceptance for the http site, but then the https redirect happens which isn't registered on the "possibly evil" click through acceptance list so it initially fails. Manually going to the https version fixes this problem, though.

Man, I hate proxies.

1. A more powerful time filter would be very useful (Last week, last month, last year, custom range where dates can be picked).

Without this search is not very useful to me at the moment. So I use Google to search HN since they do offer a filter

2. Points filter and comments filter would also be great (> 10 points and/or > 5 comments). I get lot of results which have no activity

1. Makes sense, good point!

2. Yes, there is actually a trick in the query syntax: if you use "points>10" you'll get all posts that have more than 10 points. I'll plug the filter on the number of comments as well (like comments>X). Do you think we should rather have sliders?

Sliders will probably make the UI a bit busy in the current version. If the syntax is easy, just linking to the syntax should be sufficient.

In your experimental UI, the sidebar can expose all these filters without cluttering up the main UI

I agree, my biggest annoyance with HN search by far is the time filter. If I am looking for a post I saw a few weeks ago I have to search through a years worth of posts with my keywords in it (if it's between a year and a month ago).

I think a good feature to add to HN Search would be some kind of clustering of the same link. So if a link has been reposted many times you can see all the discussions of it in one result.

How do you search HN with Google? My Google-fu is lacking.
inurl:news.ycombinator.com
"site:" will work better than "inurl:" I think "inurl:" limits the search to urls with "news.ycombinator.com" in their url while "site:" will only search the url "news.ycombinator.com" thus limiting the search to HN. So the search would be "site:news.ycombinator.com" then you just use search tools and limit the date.
Notifications like Google Alerts. Whenever I add some certain topics, I want to be notified if someone adds a comment with those words.
Would be cool if it crawled the search terms of the target link. Often I try to find some HN post but I have problems thinking of the words in it.
Do you mean that some automatic tags would help you to remember the keywords for a future search?
I like the additional filtering in the experimental interface, for example to switch between "All" and "Hot". I could see that being quite useful. Though the experimental page has too much whitespace for my liking - I can see 5-6 stories on a page vs. 10 or so on the HN style page.

It would be great if it saved my preferences (Sort by, date range) since I usually end up flipping those values to the same thing.

One minor annoyance is "JavaScript" being picked up when I search for "java". Would be nice to have the option to search for whole words.

A very minor annoyance is usernames are treated the same as text in stories. If I search for 'IBM' for example, I'll get all stories submitted by user 'IBM'.

Yes "prefixed" search is enabled by default as-you-type. Just press "Enter", it will disable it. Being able to totally remove the "prefixed" search makes sense however.
A search URL that does not change, preferably hosted on ycombinator.com. A search URL that does not redirect to Octopart, bless their hearts.
I'd like something that searches the remote contents of the posted urls, sorted by date.
How about making the site responsive/mobile-friendly first before tackling search?
it's planned but we wanted to validate the layout before making all responsive.
I think he means the main HN site, not the search site.
The ability to search only my saved articles. I'm guessing HN does no link click tracking, but searching viewed articles would be nice. These would be especially useful if the text on the target page were indexed as well.
You can star articles and limit your search to those articles by clicking on the 'Starred' tab from the left sidebar. Does it help?
(comment deleted)
The option to only search in titles, ignoring the full text of text posts.
By default the title is better ranked but then it will also fetch results in the text, you're right. It could actually makes sense. Just need to figure out the best way to use it (without flooding the UI). Maybe in the dropdown menu before the input or with a search syntax.
I mostly search ordered by date (because I often try to find things again and have a rough idea when they were submitted), so I obviously don't see the ranking effect.
A real customized algorithm based on vector space.
Do you mean representing each user as a vector in an N-dimensional space for N = number of possible votes, then using the distance between them as indication of similarity to weigh their votes relative to your preferences?

If so, +1.

Um, yes distance being calculated by a smarter calculation than Pearsons or Euclidean, a variation of RV or something that closely mimicks the way our minds compare for similarity. No bag of words or naive Bayes. True probability.
Search over only articles I have read - I often want to find an article I read last week
I don't. I use Google site search. It's never failed me.
I don't. I use Google site search. It's never failed me.
The one thing I do like about the new search site is that it doesn't open everything up in a new tab... something that is an annoyance with the current search. Please keep it that way! If I want something in a new tab, I'll just Ctrl+Click on the link.

Some other issues:

* Stories which have no external URL (text submission only) have titles which are not clickable.

* Comment results link to the submitted URL of the story. I was expecting the link would go to the comment itself, though this is likely a personal preference. When I view an individual comment, I either want to see all the other comments on the story or see the context by navigating to the parent if the comment is a reply. Seldom would I expect to view a comment several replies deep and then want to read the story.

* Clicking on the time reference (eg: "an hour ago") takes you to the anchor for the comment within all comments in the story, whereas clicking the "comments" link takes you to the specific individual comment. This is backwards.

* The up arrow shown in results seems to have no function. While it preserves the HN feel, it breaks expectations by having a different behavior. As it's unlikely upvoting from an external site is going to ever work, this should be changed to some other visual indicator.

Thank you for the detailed feedback; we'll work on that!
I used to use HN Search until the Algolia switch. Now I use Google site search. Why? Algolia just doesn't help me find what I'm looking for.
Could you share with us some failing queries, I would love to fix them.
(comment deleted)