Instantiating CRS objects is slow

[craigds]

Hi there

We're finding pyproj CRS objects very slow compared to GDAL/OSRSpatialReference objects, for some reason.
In the following script, I load 100 CRS objects using pyproj and OSR, and pyproj is 20 to 40 times slower:

$ ./bench_pyproj.py
pyproj succeeded in 3.862s
OSR succeeded in 0.137s
pyproj was 28.2x slower than OSR

script follows:

#!/usr/bin/env python3
import time
import pyproj
from osgeo import osr

codes = [2000, 20004, 20005, 20006, 20007, 20008, 20009, 2001, 20010, 20011, 20012, 20013, 20014, 20015, 20016, 20017, 20018, 20019, 2002, 20020, 20021, 20022, 20023, 20024, 20025, 20026, 20027, 20028, 20029, 2003, 20030, 20031, 20032, 2004, 2005, 2006, 20064, 20065, 20066, 20067, 20068, 20069, 2007, 20070, 20071, 20072, 20073, 20074, 20075, 20076, 20077, 20078, 20079, 2008, 20080, 20081, 20082, 20083, 20084, 20085, 20086, 20087, 20088, 20089, 2009, 20090, 20091, 20092, 2010, 2011, 2012, 2013, 20135, 20136, 20137, 20138, 2014, 2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023, 2024, 20248, 20249, 2025, 20250, 20251, 20252, 20253, 20254, 20255, 20256, 20257, 20258, 2026]

t0 = time.monotonic()
proj_crses = [pyproj.CRS.from_string(f"EPSG:{code}") for code in codes]
proj_duration = time.monotonic() - t0
print(f"pyproj succeeded in {proj_duration:.3f}s")

t1 = time.monotonic()
osr_crses = []
for code in codes:
    s = osr.SpatialReference()
    s.SetFromUserInput(f"EPSG:{code}")
    osr_crses.append(s)
osr_duration = time.monotonic() - t1
print(f"OSR succeeded in {osr_duration:.3f}s")

mult = proj_duration / osr_duration
print(f"pyproj was {mult:.1f}x slower than OSR")

Are there any known ways to speed this up?

[snowman2]

Potentially helpful references:

• https://pyproj4.github.io/pyproj/stable/advanced_examples.html#optimizing-single-threaded-applications
• https://pyproj4.github.io/pyproj/stable/api/global_context.html#global-context