Ticket #1725 (new defect)

Opened 14 months ago

Last modified 7 months ago

Handle Empty Geometry more correctly in 2d Index

Reported by: pramsey Owned by: pramsey
Priority: medium Milestone: PostGIS Future
Component: postgis Version: 2.0.x
Keywords: Cc:

Description (last modified by pramsey) (diff)

The problems that caused a crash in the ND index (#1697) could also cause some garbage results in the case of the 2D index (empty geometries being returned by the index when they shouldn't be).

Change History

Changed 14 months ago by pramsey

This is not going to happen without changing the structure of the BOX2DF, an index key, which means PostGIS 3.0 since it would force dump/restore (or, really, drop index, rebuild index).

Changed 14 months ago by robe

  • milestone changed from PostGIS 2.0.1 to PostGIS 2.1.0

If it is just a drop index / rebuild index. That is not as big of a deal since we can provide a script for that. I think that is legal to squeeze in a minor.

Heck remember PostgreSQL in 8.3 release (or was it 9.0) had a bug like that that required a reindexing (hmm might have required a full restore in fact) and we aren't even as formal as they are and they pushed it in a micro.

Changed 7 months ago by pramsey

  • description modified (diff)
  • milestone changed from PostGIS 2.1.0 to PostGIS Future

And yet, not sure I want to touch this. It's a problem that has existed since Time Immemorial.

Note: See TracTickets for help on using tickets.