[postgis-tickets] [PostGIS] #3321: Performance regression in topology load

[PostGIS]

#3321: Performance regression in topology load
-----------------------+---------------------------
  Reporter:  strk      |      Owner:  strk
      Type:  defect    |     Status:  new
  Priority:  medium    |  Milestone:  PostGIS 2.2.1
 Component:  topology  |    Version:  trunk
Resolution:            |   Keywords:
-----------------------+---------------------------

Comment (by strk):

 Ok here's the thing. The query for updating edges bounding the old face in
 2.1 is a complex one involving both a geographical filter (edges that are
 contained [or not] in the new face) and an integer filter (edges that have
 left or right face matching). Doing so lets the planner pick the most
 selective of the two.

 In 2.2 the current callbacks only allow for simpler queries so we're
 fetching _all_ edges having the face on a side and then doing the
 geographical filter "manually".

 For this case, all polygons are disjoint and so they split the universe
 face.
 The query in 2.2 returns all previously added edges as they all have the
 universe face on the side.
 This explains the raising time problem.

--
Ticket URL: <https://trac.osgeo.org/postgis/ticket/3321#comment:8>
PostGIS <http://trac.osgeo.org/postgis/>
The PostGIS Trac is used for bug, enhancement & task tracking, a user and developer wiki, and a view into the subversion code repository of PostGIS project.