Fastest way to transform WKT in python

Question

Is it possible to use pyproj to transform WKT without creating geometry objects such shapely geometry, GEOS geometry or OGR geometry.

If not what is the fastest way to do it?

Answer 1

Sure. In that case, you need to transform each point separately. First parse WKT and extract point coordinates, then loop over coordinates and perform transformation like:

import pyproj
srcProj = pyproj.Proj(init='epsg:%i' % epsg_in, preserve_units=True)
dstProj = pyproj.Proj(init='epsg:%i' % epsg_out, preserve_units=True)
x_out,y_out = pyproj.transform(srcProj, dstProj, x_in, y_in)

• Below the Radar: alright, but parsing and reconstructing WKT is not that easy for multipolygon for instance
• Matej: you might want to take a look at shapely toblerity.org/shapely/manual.html#well-known-formats

Answer 2

Also check out Lars Butler's https://github.com/larsbutler/geomet, a pure Python WKT/WKB <-> GeoJSON converter.

• Below the Radar: Yes, I am using it to convert fiona f['geometry'] to WKT. Thank you @sgillies

Answer 3

If you cannot or don't want to use Shapely, you could use those functions:

import re
from pyproj import CRS
from pyproj import Transformer
wgs84_wkt = "POLYGON((9.175287 48.780916, 9.185501 48.777522, 9.181467 48.773704, 9.174429 48.768472, 9.168807 48.773902, 9.175287 48.780916))"
COORDINATES_REGEX = re.compile(r"(\-?\d+\.\d*) (\-?\d+\.\d*)")
TO_LOCAL_CRS = Transformer.from_crs(CRS.from_epsg(4326),
 CRS.from_epsg(32632),
 always_xy=True)
def convert_coordinates(match):
 longitude, latitude = match.groups()
 x, y = TO_LOCAL_CRS.transform(longitude, latitude)
 return f"{x} {y}"
def convert_wkt_to_local(wkt):
 return COORDINATES_REGEX.sub(convert_coordinates, wkt)
print(convert_wkt_to_local(wgs84_wkt))
# POLYGON((512877.0688394928 5403116.534101295, 513628.3360295398 5402741.031035185, 513332.97838651366 5402315.905104928, 512817.2051751098 5401733.104146439, 512402.75806575816 5402335.776478222, 512877.0688394928 5403116.534101295))

It simply looks for longitude and latitude via regex, and converts the coordinates. It should work fine with holes or multipolygons too.