Compute label position in map/coordinate space rather than pixel space

[ignatz]

Computing polylabel positions is insanely expensive eating up some 70% of my UI thread render times.

[TesteurManiak]

LGTM, I'm wondering if we should consider spawning an isolate to compute all label positions in one go to further improve performances 🤔

[ignatz]

One could. Sparse in and output seem well suited given the comms overhead. That said, it would make Polygon construction async and thus add extra complexity for the user unless you make labels best-effort. I'd rather add a label position art to the Polygon constructor letting users do whatever they want. I also have a patch out to polylabel to reduce computational cost by another 70℅. In it's current shape I wouldn't move to isolates for my use case. With this change already the frame to frame cost for computing the label's pixel position has gone to virtually zero.

…

On Thu, Sep 7, 2023, 08:31 Guillaume Roux ***@***.***> wrote: ***@***.**** approved this pull request. LGTM, I'm wondering if we should consider spawning an isolate to compute all label positions in one go to further improve performances 🤔 — Reply to this email directly, view it on GitHub <#1641 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAA3K4UWOX6422Z7P3ZCKQ3XZFS2PANCNFSM6AAAAAA4NTZ5O4> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[JaffaKetchup]

LGTM. I think using Isolates would be too complex for now.