YAYUI Compressor (Yet Another YUI Compressor)

What this appliation does:

• Take your source code,removes the whitespaces and shorten the variables names
• Delivers the minified/compressed source code (which theoretically should be fine :-)

How it works:

• I use no external library/engine/parser (eg. Rhino) whatsoever; everything is done with the aid of plain regular expressions (aka Regex)
  • there are 7 different regex's for trimming the whitespaces (see Minify options)
  • there are few different regex's that together work in cascade helping to capture and replace the functions arguments/variables with shorten names, a-z (see Obfuscate options)

Caveats:

• This application is experimental. I made it just for fun so use it in production but with care!
• Normally a minifier/obfuscator should be done with the aid of a parser and not rely only on few regex patterns. Although I tested it against few hundred KB JavaScript source code I cannot guarantee that I've implemented all the possible patterns.
• By checking all possible options (minify/obfuscate) your code will still work but will not be validated by some code quality tools (eg. JSLint)
• There is still room for optimization. Perhaps some patterns can be writen even shorter.

Patterns that it uses and the algorithm description:

• for minification:

  1. remove JavaScript hints : /['"]use\s+strict['"]\s*;/ (see the complete hint list)
  2. remove dispensable semicolons : replace /;[\s|\n]*;/m with ;
  3. remove single line comments : replace /(?<![\'"])\/\/.*/m with '<empty-string>'
  4. remove block comments : replace /\/*([^\/]|[\w\d\s]\/[\w\d\s])**\//m with '<empty-string>'
  5. match these blocks /\b(if|else|while)\b.[^;{]({)([^{}]+)(})/m then:
     • replace the curly braces /[{}]/ with '<blank>' if the code block doesn't match:
       • 2+ /;/ OR 1+ /:/ OR 1+ /\b(function|try|catch)\b/
  6. remove the whitespaces /\s([^\w\d\s])\s|[ ]{2,}/' with '$1'
  7. remove the CR/LF:
     • first replace /(?<=[\W])\n|[ ]{2,}/m with '$1'
     • second replace /\n/m with '<blank>'

• for obfuscation:

  1. first backup all strings (we don't want to change them) : /([\'"])(\\\1|.)*?\1/m
  2. capture all functions: /\bfunction\b(([\n]|.)?)(?=[^(]\bfunction\b\s[\w\d]+?)/m then for each:
     • capture the function declaration: /\bfunction\b(.*?){/m and split the arguments by comma
     • for each argument (eg. myVar):
       • take the left part /[^\w\d]/ and then replace all its occurences with your obfuscated variable name: replace /([^\w\d\'".])(\bmyVar\b)([^\w\d\'"])/ with '$1myNewVar$3'
     • capture the function inner block codes: /{[^{}]*}/m then:
       • capture the var declaration: /\bvar\b[^;]+/m , split variables by comma and replace /([\s([{+-*\/|<>;&%!,])(\bmyVar\b)([\s)]}+-*\/|<>;&!%,.=]?)/ with '$1myNewVar$3'
  3. restore the original strings; there is no risk in changing them now

FAQ

• Does it work for other languages, such as CSS, PHP, Java, C++, etc? + The minification might work but the obfuscation is not compatible yet
• Can I get the source code? In what language is it written? + the source code has been made available at its bitbucket.org respository + it is written in PHP and encapsuled in a standalone class