Polygon Search

This simple program uses Geospatial Partitioning to find the postal code of a specific geographical point in Germany. I use this program to compare the performance of four popular programming languages. There are some points regarding this benchmark:

• I didn't use any third-party library for this benchmark, and all the implementations remain based on standard data structures and libraries available in each programming language.
• The benchmark result is based on sample data containing polygons that cover all the german region.
• To keep the system as simple as possible, I am using a custom file format for importing the polygons data. you can download the original data from https://www.suche-postleitzahl.org/plz-karte-erstellen as a Geojson file
• the algorithm contains three operation modes:
  • HIGH
  • MEDIUM
  • LOW

Compiler version

• The results are related to the following compiler versions :
  • gcc 9.3.0
  • clang 10.0.0
  • jdk-17
  • go version 1.17.1
  • rustc 1.55.0

Build

C++ gcc

cd cppgeo
mkdir __buid
cmake ..
make -j

C++ clang

uncomment set(CMAKE_C_COMPILER "clang") set(CMAKE_CXX_COMPILER "clang++") in the CMakeLists.txt file
cd cppgeo
mkdir __buid
cmake ..
make -j

JAVA

cd java
export PATH=$PATH:{JDKPATH}
export PATH=$PATH:{GRADLEPATH}/bin
gradle build

GO

cd go
go build

RUST

cd rust/polygon_search
cargo build --release

Usage

For viewing the available options for each version, use -h switch

Test Scenario

The results are based on 1000000 times lookup of the specified point and are in Microsecond. To consider the effect of the java jit loop predicate, I executed the tests in two modes, In the first mode I searched a constant point and in the second mode I added some random noises to it

CASE 1

The first Scenario considers the location of the Freiburg central church.

Freiburger Münster Lookup results (High and low accuracy)

Language	High Accuracy	Medium Accuracy	Low Accuracy
C++ (gcc)	1394029 μs	354843 μs	271316 μs
C++ (clang)	1654293 μs	384636 μs	320392 μs
JAVA	1969147 μs	595850 μs	514363 μs
GO	2130709 μs	719377 μs	652770 μs
RUST	2200332 μs	504338 μs	433640 μs

Random points around Freiburger Münster

Language	High Accuracy	Medium Accuracy	Low Accuracy
C++ (gcc)	1501170 μs	355831 μs	273120 μs
C++ (clang)	1724351 μs	387201 μs	323969 μs
JAVA	2377246 μs	630499 μs	539974 μs
GO	2216277 μs	725739 μs	668134 μs
RUST	2283935 μs	508129 μs	443500 μs

A location in a big polygon (High and low accuracy)

Language	High Accuracy	Medium Accuracy	Low Accuracy
C++ (gcc)	5187784 μs	378546 μs	321742 μs
C++ (clang)	5187583 μs	380561 μs	335955 μs
JAVA	7507826 μs	566826 μs	504545 μs
GO	6492508 μs	708009 μs	643355 μs
RUST	9399210 μs	508060 μs	430104 μs

A location in a big polygon ( Random Points )

Language	High Accuracy	Medium Accuracy	Low Accuracy
C++ (gcc)	5246180 μs	386938 μs	326808 μs
C++ (clang)	5261288 μs	388207 μs	336071 μs
JAVA	7867935 μs	604901 μs	574479 μs
GO	6531627 μs	712063 μs	650678 μs
RUST	9396629 μs	506499 μs	444764 μs