[postgis-users] Speed issues with MakeLine

[Kurt Schwehr]

Hi All,

I am working on building lines out of groups of points using  
MakeLine.  Works pretty well as long as each group has less than 20K  
points.  However, I have a number of lines that contain more that  
170K points.  MakeLine on these is taking more than half an hour on  
each (1.5 GHz G4 Mac with 1 GB of RAM).  I thought it might be the  
shared memory configuration, but I upped that and still no luck for a  
speedup....

sysctl -a | grep shm
kern.sysv.shmmax: 33554432
kern.sysv.shmmin: 1
kern.sysv.shmmni: 64
kern.sysv.shmseg: 32
kern.sysv.shmall: 32768

I am currently writing up code to generate the WKB in python (using  
PCL-core) to see if it is faster to pull all the points over via  
psycopg2 and do the work outside of the db, but while I'm doing that  
I figured I'd ask if there is something else that I should be doing  
that would help.

Doing a count of the points in the 188K line is super fast, so I that  
I've got my indexing done right.

fink list postgis82 postgresql82

i   postgis82                        1.2.1-1024                
PostgreSQL geographic object support
i   postgresql82                     8.2.4-1021                
PostgreSQL open-source database
i   postgresql82-dev                 8.2.4-1021                
PostgreSQL development headers and libraries
i   postgresql82-shlibs              8.2.4-1021                
PostgreSQL shared libraries


Thanks,
-kurt