Question

I would like to use regex to look inside a string for the smallest sequence that fits a starting and ending delimiter (taking an escape character into account). For example, if I had the following string I would want to locate the lowest matches [ two ] and [ four \[ five \] ] while ignoring the match they are contained in [ one ... three ... six].

zero [ one [ two ] three [ four \[ five \] ] six ] seven

So far I have the following regex which uses negative look-behinds to check and isn’t quite caching the last ] in the second match.

(\[)(?:(?!(?:[^\\])\1|\]).)*]

My goal is to have a simple parser I can use to process simple nested command blocks.

Answer 1

The following works:

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

See it working: http://www.rubular.com/r/cAajtm2wxw

Explanation:

\[                # opening bracket
(?:               # start of non-capturing group (repeat zero or more times)
   \\[\[\]]         # backslash followed by [ or ]
   |                # OR
   [^\[\]]          # any character except [ or ]
)*                # end of non-capturing group
]                 # closing bracket

Note that this isn’t quite safe because in a string like [ one \\[ two ] three ] the backslash is escaped, so the backslash before the [ should not escape it.

To fix this you could use the following:

\[(?:(?<!\\)(?:\\\\)*\\[\[\]]|[^\[\]])*]

This changes the single escaped backslash \\ in the original regex to the following regex which checks for an odd number of backslashes:

(?<!\\)           # fail if previous character is a backslash
(?:\\\\)*         # some even number of backslashes
\\                # one more backslash to make it odd

http://www.rubular.com/r/BhQzLQpyB9