解析类似正则表达式的字符串

Question

I'm trying to parse a Regex-like string, with format like the following:

• The only characters that need to be escaped are: [ , \\ , and - ;
• A valid string can be a sequence of:
  • "regular characters", eg a , b
  • "escaped special characters", eg \\\\ , \\[
  • Sequences containing above two, wrapped by a pair of brackets, eg [abc] , [a\\]]

For example, abc[def]g , abc\\-\\[[def\\]]gh\\\\ are both valid strings.

Is there some way that I can get the character / character class (in the third case above) at each index? Using pure regex / sed or some Python library works for me.

Answer 1

Usually, you can't parse it character by character, but have to parse it
construct by construct.

Knowing the group that matched tells you what the construct is.
When the class construct matches, you have to parse it's contents
separate from the main regex.

You'd check this all in (pseudo-code)

while( regex find )
{
if group 1 matched // the character escaped
else
if group 2 matched // a non-class start or non-escaped char
// Check if it should be escaped, or is a metachar
else
if group 3 matched // class contents
// parse the class contents here
else
if group 4 matched // error
}

For example purposes ..

(?s)(?:\\\\(.)|([^\\[])|\\[((?:\\\\.|[^\\]])*)\\]|(.))

Expanded

 (?s)                # Dot all modifier
 (?:
      \\                  # Escape anything
      ( . )               # (1)
   |                    # or,
      ( [^\[] )           # (2), Anything that does not start a char class
   |                    # or,
      \[                  # Start of char class
      (                   # (3 start)
           (?:                 # ----------
                \\ .                # Escape anything
             |                    # or,
                [^\]]               # Anthing that does not end a char class
           )*                  # ----------
      )                   # (3 end)
      \]                  # End of char class
   |                    # or,
      ( . )               # (4), Error, probably an unbalanced '['     
 )