Need to dynamically read optional fields?

[cchderrick]

As it seems from shapelib implementation, the reader will need to check if the number of Content Length (in bytes) matches Number of points for these record types:
MultiPointM
PolyLineM
PolygonM
MultiPointZ
PolyLineZ
PolygonZ
MultiPatch

For example:
PolyLineM record with 1 part and 10 points, without M, the content length would be 44 + (4 * 1) + (16 * 10), with M, the content length becomes 44 + (4 * 1) + (16 * 10) + the optional 8 + 8 + (8 * 10)

[visr]

I don't think I fully understand. A PolyLineM always has an M coordinate right? If it doesn't it is a PolyLine.

We can of course add a check on the content length, but isn't how much we read currently already decided by numparts, numpoints, etc?

[cchderrick]

The ESRI Shapefile spec allows Z or M parts to be optional for some of the type.

And looking at the shapelib C implementation in GDAL, it seems like it also do some sort of size check to determine if the optional fields are presented.

Ref:
https://github.com/OSGeo/gdal/blob/master/gdal/ogr/ogrsf_frmts/shape/shpopen.c#L2523-L2529

[visr]

For me it's much easier to read the technical description: https://www.esri.com/library/whitepapers/pdfs/shapefile.pdf

There it says that for measures only, floats smaller than -1e38 are considered "no data".

[cchderrick]

Okay, I think I have misunderstood the word "Optional" in the spec. What I thought it meant the data array could be omitted from the file. Upon some more research, what the spec really meant for "Optional" is the M array can all be filled with no-data values.

[visr]

Ah yeah I see, it is quite confusing. Since there is still a possible improvement for explicitly handling missing measures, I opened up #28. Thanks!