ko1nksm / sh-base64

Base64 encoder/decoder implementation for portable shell scripts

sh-base64

Base64 encoder/decoder implementation for portable shell scripts

This is not an extremely slow pure shell script implementation. sh-base64 uses some basic POSIX commands that are portable to improve performance (but still slower than native commands).

Requirements

• POSIX shell (dash, bash, ksh, mksh, yash, zsh, etc)
• Basic POSIX commands (od, tr, fold, xargs, and awk)

Usage

The base64encode/base64decode argument can be a 2 or 3 character string. If any other number of characters is specified, the behavior is undefined.

. ./base64.sh

# Base64 encode
base64encode       # Basic usage (base64 standard)
base64encode "+/"  # Use + and / for the 62nd and 63rd characters, no padding
base64encode "-_"  # Use - and _ for the 62nd and 63rd characters, no padding (base64url)
base64encode "+/=" # Use + and / for the 62nd and 63rd characters and = for padding

# Base64 decode
base64decode       # Basic usage (base64 standard)
base64decode "+/"  # Use + and / for the 62nd and 63rd characters
base64decode "-_"  # Use - and _ for the 62nd and 63rd characters (base64url)

Example

Included ./base64 is for testing purposes.

$ echo abc | ./base64
YWJjCg==

$ echo abc | ./base64 | ./base64 -d
abc

$ echo abc | AWK=mawk ./base64 | AWK=mawk ./base64 -d
abc

$ echo ">>>???" | ./base64      # standard
Pj4+Pz8/Cg==

$ echo ">>>???" | ./base64 "+/" # no padding
Pj4+Pz8/Cg

$ echo ">>>???" | ./base64 "-_" # base64url
Pj4-Pz8_Cg

Performance

Preparation

Created a 10 MB file for performance testing

$ head -c 10m /dev/urandom > data.bin
$ base64 < data.bin > data.base64

$ md5sum data.bin data.base64
a740084d87aaed16c1f4ea40a6acf1b0  data.bin
8096a63dc595bae96d6dfcf797c9429d  data.base64

Case 1: Ubuntu

Test environment: Ubuntu (CPU: 3.4 GHz quad-core)

Base64 encoder comparison with base64 command

$ time base64 < data.bin | md5sum
35419688d09e4be616e6bd95b1f029f5  -

real    0m0.035s
user    0m0.034s
sys     0m0.023s

Different hash values due to different fold back
$ base64 < data.bin | paste -s -d '' | md5sum
8096a63dc595bae96d6dfcf797c9429d  -

$ time ./base64  < data.bin | md5sum
8096a63dc595bae96d6dfcf797c9429d  -

real    0m4.218s
user    0m6.745s
sys     0m0.142s

Base64 decoder comparison with base64 command

$ time base64 -d < data.base64 | md5sum
a740084d87aaed16c1f4ea40a6acf1b0  -

real    0m0.041s
user    0m0.044s
sys     0m0.022s

$ time ./base64 -d < data.base64 | md5sum
a740084d87aaed16c1f4ea40a6acf1b0  -

real    0m4.141s
user    0m4.575s
sys     0m0.216s

Case 2: macOS

Test environment: macOS (CPU: 2.4 GHz dual-core)

Base64 encoder comparison with base64 command

$ time base64 < data.bin | md5sum
8096a63dc595bae96d6dfcf797c9429d  -

real	0m1.932s
user	0m1.930s
sys 	0m0.067s

$ time ./base64 < data.bin | md5sum
8096a63dc595bae96d6dfcf797c9429d  -

real	0m20.466s
user	0m30.385s
sys 	0m0.319s

Base64 decoder comparison with base64 command

$ time base64 -d < data.base64 | md5sum
a740084d87aaed16c1f4ea40a6acf1b0  -

real	0m0.247s
user	0m0.380s
sys 	0m0.102s

$ time ./base64 -d < data.base64 | md5sum
a740084d87aaed16c1f4ea40a6acf1b0  -

real	0m16.181s
user	0m19.320s
sys 	0m1.074s