Bash is a command interpreter used in system scripts for calling external commands, so I wanted to compare the speed of Bash to the speed of other interpreted languages by doing a simple test.

For the test, the scripts call the same external command one hundred times and then the times are compared to see which interpreter executed the commands in the shortest time.

The results show that Bash, in average, executed the commands in the shortest amount of time. It also shows that the slowest language, Python, was 34% slower than Bash. This is surprising considering that Bash is exceptionally slow at recursion and Bash is slow at arithmetic.

Python and JavaScript, despite being slower at this specific task, provide:

  • Support for object-oriented code
  • Package managers
  • Rich standard libraries
  • Development tools like linters

This means that you can write scripts that call external commands in your preferred language instead of in Bash if you find a performance loss of 34% acceptable. For example, I prefer to write Git hooks in JavaScript when working in JavaScript projects instead of writing the hooks in Bash.

Execution times for calling an external command one hundred times:

Command Interpreter Mean [s] Min [s] Max [s] Relative
./test.bash bash 5.2.26 3.230 ± 0.285 2.712 3.670 1.00
./test.pl perl 5.38.2 4.079 ± 0.265 3.652 4.529 1.26 ± 0.14
./test.mjs node 22.2.0 4.252 ± 0.106 4.109 4.427 1.32 ± 0.12
./test.py python 3.11.9 4.341 ± 0.190 4.058 4.630 1.34 ± 0.13

Implementations

Bash 
#!/usr/bin/env bash
for _ in {1..100}; do
    nix eval --expr 1+1;
done
Perl 
#!/usr/bin/env perl
system qw(nix eval --expr 1+1) for 1 .. 100;
JavaScript 
#!/usr/bin/env node
import {execSync} from "node:child_process";
for (let i = 0; i < 100; i++) {
    process.stdout.write(execSync("nix eval --expr 1+1", {encoding: "utf-8"}));
}
Python 
#!/usr/bin/env python3
from subprocess import call
for _ in range(100):
    call(["nix", "eval", "--expr", "1+1"])