如何在PHP中使用正则表达式匹配嵌套大括号？

Question

I have an LaTeX document I want to match. And I need a RegEx match that matches the following:

\ # the backslash in the beginning
[a-zA-Z]+ #a word
(\{.+\})* # any amount of {something}

However, and her is the catch;

In the last line, it 1. needs to be greedy and 2. needs to have a matching number of {} inside itself.

Meaning if I have the string \\test{something\\somthing{9}} it would match the whole. And it needs to be in that order ( {} ). So that it doesn't match the following:

\\LaTeX{} is a document preparation system for the \\TeX{}

just

\\LaTeX{}

and

\\TeX{}

Help anyone? Maybe someone have an better idea for matching? Should I not use regular expressions?

Answer 1

This can be done with recursion:

$input = "\LaTeX{} is a document preparation system for the \TeX{}
\latex{something\somthing{9}}";

preg_match_all('~(?<token>
        \\\\ # the slash in the beginning
        [a-zA-Z]+ #a word
        (\{[^{}]*((?P>token)[^{}]*)?\}) # {something}
)~x', $input, $matches);

This correctly matches \\LaTeX{} , \\TeX{} , and \\latex{something\\somthing{9}}

Answer 2

PHP could be used since it supports recursive regex-matching. But , as I said, if you have comments in your LaTeX-like strings that can have { or } in them, this will fail.

A demo:

$text = 'This is a \LaTeX{ foo { bar { ... } baz test {} done } } document
preparation system for the \TeX{a{b{c}d}e{f}g{h}i}-y people out there';
preg_match_all('/\\\\[A-Za-z]+(\{(?:[^{}]|(?1))*})/', $text, $matches, PREG_SET_ORDER);
print_r($matches);

which produces:

Array
(
    [0] => Array
        (
            [0] => \LaTeX{ foo { bar { ... } baz test {} done } }
            [1] => { foo { bar { ... } baz test {} done } }
        )

    [1] => Array
        (
            [0] => \TeX{a{b{c}d}e{f}g{h}i}
            [1] => {a{b{c}d}e{f}g{h}i}
        )

)

A quick explanation:

\\\\         # the literal '\'
[A-Za-z]+    # one or more letters
(            # start capture group 1   <-----------------+
  \{         #   the literal '{'                         |
  (?:        #   start non-capture group A               |
    [^{}]    #     any character other than '{' and '}'  |
    |        #     OR                                    |
    (?1)     #     recursively match capture group 1  ---+
  )          #   end non-capture group A
  *          #   non-capture group A zero or more times
  }          #   the literal '}'
)            # end capture group 1 

Answer 3

Unfortunately, I believe this is impossible. Bracket matching (detecting properly paired, nested brackets) is commonly used as an example of a problem that cannot be solved with a finite state machine, such as a regular expression parser. You could do it with a context free grammar, but that's just not how regex works. Your best solution is to use a regex like {*[^{}]*}* for the initial check, and then another short script to check whether it's an even number.

In conclusion: don't try and do it with only regex. This is not a problem that can be solved with regex alone.