(Attempt to) stop jumping markers for duplicate wiki pages #1922
Conversation
Always use the oldest node (/relation/way if nodes don't exist)
| @@ -26,44 +26,34 @@ | |||
| sql10 = """ | |||
| SELECT | |||
| (array_agg(tid))[1:10], | |||
| ST_AsText(any_locate((array_agg(type))[1], (array_agg(id))[1])), | |||
| ST_AsText(any_locate(substring(MIN(tid), 1, 1), substring(MIN(tid), 2)::bigint)), | |||
There was a problem hiding this comment.
It is not the min as expected, because it compare on strings.
There was a problem hiding this comment.
Ah, true, n56789 would be after n100000. I've updated the initial post.
However, my goal was just to have a stable output (as long as no object is added/removed) to get rid of the jumping markers, so whether it's alphabetically the first or numerically the first should not make a difference, right?
There was a problem hiding this comment.
I think older is more stable.
It just about concatenate type later.
There was a problem hiding this comment.
Not sure how to get the oldest element.
If I add a new way today, the ID will be around 1185000000. A node with the same ID would be 12 years old. And a relation with that ID doesn't exist yet. So comparison by id will not retrieve the oldest element
There was a problem hiding this comment.
Just order by type and id will do the job. The older by type is sufficient, isn't it ?
There was a problem hiding this comment.
I'll have another look at a later moment.
Just calling ORDER BY won't work, since I can't add id or type to GROUP BY or it'll be filtered out. Probably have to modify the COUNT
There was a problem hiding this comment.
Nope, also with COUNT(w) it doesn't work. Probably I misunderstood what you meant @frodrigo, where would I have to call the ORDER BY type, id?
There was a problem hiding this comment.
Actually, the issue ID changes anyway if elements are added or deleted, right? So is it really needed to be the same element every time, also when elements are added? (Of course, the jumping when there's no changes must be fixed)
|
Closed in favour of #1958 |
Always use the same element to show the marker: #1912
Also removes the need to propagate
id, typetogether withtidHowever, seeing if this stops the 'jumping markers' probably has to be observed "live".