Running this query:
https://stacker.news/search?q=Antonopoulos&sort=recent
... results in what seems like a random sort order. Testing a few other similar searches, they all seem to result in a similar situation - the output is not sorted from most recent to least recent.
It's a gaussian notion of recent - meaning it's biases toward more recent content but is not ordered strictly.
We did this because it's hard to maintain relevance and recency simultaneously. I'll create a github issue if we don't have one.
In the meantime clamping it by time, e.g.
1 weekshould help.This issue has come up a number of times now and makes it clear that user expectations for
recentis that it'll be a strict recency sort order.I wonder if
recentsearches should be keyword only, sorting all keyword matches strictly by date.I vaguely remember that being the case before, but there was some other bug that made it too restrict thus often returning null results.
Yeah, I’d noticed this was happening before. I think your solution’s the best one. One rule in software is: don’t do stuff the user didn’t ask for.
Your memory matches mine. In cases where there are otherwise no results, people want
recentto be less strict. When there are many results, people wantrecentto be strict.I am for having that stricter search from the past. Makes searching a post about a certain thing on a common word harder for me. Also made me take extra steps when I need to do help/support replies about GrapheneOS as I didn't always see who was discussing at first.