Open GoogleCodeExporter opened 9 years ago
Original comment by schmi...@gmail.com
on 28 Jan 2011 at 1:55
Original comment by schmi...@gmail.com
on 19 May 2011 at 5:01
Original comment by sjsrey
on 6 Aug 2011 at 12:15
shapefile profiling/refactoring...
read all polygons from us census tracts shapefile.
Baseline: ~9.559 seconds
refactored to use arrays for homogeneous data (point arrays):
~9.069 seconds, ~5% improvement over baseline
refactored to reduce number of function calls by grouping heterogeneous data
into single struct.unpack calls:
~8.063 seconds, ~11% over arrays, ~15.7% improvement over baseline
committed in r1021
~37% of cumulative time is spent on is_clockwise function, which accounts for
only 5.5% of function calls...
ncalls tottime percall cumtime percall filename:lineno(function)
128472 2.980 0.000 3.004 0.000 standalone.py:604(is_clockwise)
is_clockwise is currently invoked twice per ring, once in the shapefile reader
and once in the Polygon constructor. removing this duplication will provide
another ~%19 speedup over the refactored reader for a total improvement of ~%31
over baseline.
The only other obvious route to a more efficient reader is to drastically
reduce the number of function calls. At over 2.3 million functions for ~60k
polygons, we're using ~38 function calls per polygon. function call overhead
could account for a large percentage of running time.
Original comment by schmi...@gmail.com
on 20 Oct 2011 at 8:33
Original comment by schmi...@gmail.com
on 9 Jan 2012 at 9:10
Original comment by schmi...@gmail.com
on 13 Jul 2012 at 6:09
Original issue reported on code.google.com by
schmi...@gmail.com
on 5 May 2010 at 7:18