A proposed edition to geopandas.io for reading SpatiaLite data via wkt

read_spatialite_patch.py

"""Addition to geopandas.io.sql.py
by Garin Wally / Windfall Spatial, Nov 10, 2015
"""

# Place with the other imports
import shapely.wkt

def read_spatialite(sql, con, geom_col, crs=None, index_col=None,
                 coerce_float=True, params=None):
    """
    Returns a GeoDataFrame corresponding to the result of the query 
    string, which must contain a geometry column.
    
    Warning: the connection parameter must have extensions enabled,
      and the 'mod_spatialite.dll' extension loaded.

    Examples:
    sql = "SELECT geom, kind FROM polygons;"
    df = geopandas.read_spatialite(sql, con)

    Parameters
    ----------
    sql: string
    con: DB connection object or SQLAlchemy engine
    geom_col: string (removed default "geom")
        column name to convert to shapely geometries
    crs: optional
        CRS to use for the returned GeoDataFrame

    See the documentation for pandas.read_sql for further explanation 
    of the following parameters:
    index_col, coerce_float, params

    """
    # Read the full table as a DataFrame
    df = read_sql(sql, con, index_col=index_col, coerce_float=coerce_float,
                  params=params)

    # Ensure geometry column was specified correctly
    if geom_col not in df:
        raise ValueError("Geometry column specified does not exist: "
                          "'{0}'".format(geom_col))
                          
                          
    # Ensure there are no Null geometries. I have bad data...
    # prevents 'Windows Error: exception: access violation reading 0x0000000'
    assert True not in set(df[geom_col].apply(lambda x: x is None)), \
        ("NULL geometries found! "
        "Try adding 'WHERE {0} IS NOT NULL;'".format(geom_col))
            
    # Collect just the geometry column as a separate DataFrame
    from_sql = sql[sql.lower().index("from"):]
    as_text = "ST_AsText({0})".format(geom_col)
    geom_sql = "SELECT {0} {1}".format(as_text, from_sql)
    wkt_geoms = read_sql(geom_sql, con)
    
    # Convert to shapely geometries
    
    shapely_geoms = wkt_geoms[as_text].apply(lambda x: shapely.wkt.loads(x.encode()))
    
    # Combine the geometry and the rest of the data
    df[geom_col] = GeoSeries(shapely_geoms)
    
    return GeoDataFrame(df, crs=crs, geometry=geom_col)

Working version.