Python To PostgreSQL - ST_GeometryFromText and NaN

Question

I am trying to import a shapefile into Postgre using python and psycopg2. I have run into a few issues.

I am using ST_GeometryFromText() of a WKT string.

First, the shape@WKT returns MULTIPOLYGON ZM at the beginning of the string. I believe that it is only supposed to be MULTIPOLYGON correct?

I fixed this by doing a replace on the string but this does not seem like what should be done. What do I do to make this correct?

Second, the shape@WKT contains X, Y, 0, NAN as values. I am expecting only 2 values, x and y. Why am I getting 0 and NAN?

with arcpy.da.SearchCursor(input_fc, ["FID", "Shape@WKT"]) as da_cursor:
 for row in da_cursor:
 geoms = row[1].replace(' ZM ','')
 print geoms
 pg_cursor.execute("INSERT INTO %s (fid_p, geom) VALUES (%s, ST_GeometryFromText('%s', %s))" %(dataset_name, int(row[0]), geoms, int(coordinate_system)))
pg_connection.commit()

RESULT of print geoms:

MULTIPOLYGON ZM (((440185.43178632273 6350644.499023444 0 NAN, 440189.09859540447 6350575.940429694 0 NAN, .........

PYTHON ERROR:

InternalError: parse error - invalid geometry HINT: "...85.43178632273 6350644.499023444 0 NA" <-- parse error at position 60 within geometry

• Get Spatial: I think there might be some confusion between what is actually encoded in the shapefile, which is the MULTIPOLYGON ZM, and what you want, which is just MULTIPOLYGON. You are receiving a python error because there are values for the Z and M coordinates in the shapefile geometry. You need to either resave the shapefile before running this script, to strip out the ZM values, or specify a parameter on this function to only look at the 2D elements.
• Vince: It's actually trivial to just copy a shapefile to PostgreSQL using ArcGIS 10.4 or higher (a dozen lines of Python, with extensive documentation).
• D_C: please share Vince

Answer 1

The last two values are Z and M values for each vertex. Looks like your dataset is Z/M enabled, but the features you're importing don't have values set, thus you see 0/NaN.

If you copy the features into a non ZM-enabled feature class, you won't have to worry about the ZM replace or parsing the 0/NaNs in the WKT string.