You can clone with
No one assigned
Test file: http://cl.ly/2y3P2b2g0x2h1V3b1l1x
With the Mapnik shapefile driver, only one point appears and one point's attributes are available. After resaving in QGIS, all points render.
Output (blue dot before, red after): http://dl.dropbox.com/u/68059/Screenshots/y~~rfu-c0jwy.png
Usually a shapefile that displays unpredictably with mapnik, but that works perfectly when re-saved with QGIS/OGR, indicates a file that was originally created with an old version of QGIS that inserted invalid binary data (or otherwise is corrup somehow). Could this be the case?
This report is via @cageyjames who was using FME to export the shapefile. The resaved-via-QGIS and saved-via-FME files are definitely different via Hex Fiend.
Oh, interesting. I'll take a closer look then - thanks. I cannot imagine FME is authoring an invalid shapefile, so something must be up on the mapnik side. I'll post back when I have more info.
I can replicate the single rendered point. What appears to be happening is that the shape type is properly detected as pointz (3d point type) but the m dimension is not getting skipped here such that on the second pass of the shapefile the pos is invalid.
shape_.reclength_ here is 18 with the FME shapefile and 14 with a pure copy using ogr2ogr
http://trac.osgeo.org/gdal/ticket/2124 (and linked issues) seems to indicate that its unclear whether the M dimension is optional for pointz. Its explicitly optional in the spec for multipointz.
hit another instance of this bug with this 3d shapefile (this time from arcinfo): http://cl.ly/1C3o000M1s2w. Downstream issue: http://support.mapbox.com/discussions/tilemill/3502-problem-adding-point-shapefiles
+ fix record length check to account for OGR not writing PointZ accor…
shapefile spec - PointZ: type,x,y,z,measure (#1193, #504)
@springmeyer @tmcw - fixed in da0d745
@springmeyer - was wrong record length check to skip M (OGR bug) in 8ac189c
record length is in 16-bit and not including record header which is 8 bytes (or 4 16-bit words)
Add more shapefile point type tests - one should be failing due to #1193