In PHP, how would I split a string where the start and end delimiter are dissimilar

Question

Intro

In PHP, how would I split a line with this syntax:

<As's\\as'dsd> asqwedasd <sa sdasd> [a sadasd] [<asdsad> [as ddsd]] 'asdsad assd'

into this?

array(5) {
  [0]=>
  string(14) "<As's\\as'dsd>"
  [1]=>
  string(9) "asqwedasd"
  [2]=>
  string(10) "<sa sdasd>"
  [3]=>
  string(10) "[a sadasd]"
  [4]=>
  string(20) "[<asdsad> [as ddsd]]"
  [5]=>
  string(13) "'asdsad assd'"
}

More detailed explanation

Now I'm not the best at explaining, so I hope that the above example explains my situation well enough that you won't need my explanation but here it is anyway:

I want to split this string by every space except some specific ones:

• If the space is inside angle brackets or square brackets it should NOT split that line. See number 2 and 3.
• There could potentially be a bracket inside a bracket. This should just be returned as one whole string. See number 4.
• There could potentially be items that are not in brackets. See number 1.
• Items not wrapped in brackets will NOT contain spaces UNLESS quoted by apostrophes. See number 5.
• The items can contain all UTF-8 characters EXCEPT for [ ] < >

Sources that could possibly help

Explode string except where surrounded by parentheses?

Thank you in advance! I know this is a humongous task but I have absolutely no idea how to do this myself.

Answer 1

With all the disclaimers about using regex to parse html... And only if you're ready for some recursive beauty...

Matching What you Want vs. Splitting on What You Don't Want

If you're going to use regex, in this case, to get your array, matching what you do want will be easier than splitting on what you don't want. Here is a starting place, which we can refine:

(\[(?:[^[\]]++|(?1))*\])|<[^>]*>|'[^']*'|[!-~]+

See demo.

How it works:

• We match several possibilities, separated by the alternation operator |
• The first match option (\[(?:[^[\]]++|(?1))*\]) recursively matches all [sets of [brackets]]
• The <[^>]*> matches `'
• The '[^']*' matches 'complete quotes'. If needed, it could be improved to account for potential escaped quotes \'
• The [!-~]+ matches any non-space printable characters that remain. It is a guess, based on the lone word asqwedasd in your input, and that too could be refined. For instance, if you want to specify, for validation purposes, that the leftover strings have no <>[] characters, you can use this instead (suggested by @CasimiretHippolyte) \s*\K[^[<]+(?<!\s)

Sample code

See this output of this demo. The array $m[0] contains the "splits" you wanted.

$regex = "%(\[(?:[^[\]]++|(?1))*\])|<[^>]*>|'[^']*'|[!-~]+%";
$string = "<As's\\as'dsd> asqwedasd <sa sdasd> [a sadasd] [<asdsad> [as ddsd]] 'asdsad assd'";
$count = preg_match_all($regex,$string,$m);
print_r($m[0]);

Another Solution

@HamZa came up with another solution which I find quite beautiful. He didn't want to post it himself, but was happy for me to add it here for completion.

How does it work? The idea is to match the right space characters, and to split on them. The base principle for this is explained in detail in this question about "regex-matching a pattern unless...". First, in a similar fashion to my regex (but with more checks and recursion), he defines all the groups we want to match, and matches them. Then, he uses (*SKIP)(*F) to make the regex fail if these groups are matched, after which the engine skips to the position in the string that follows the last character that was matched. On the other side of the alternation, he matches the space characters we will split on, and we know these are the right space characters because they were not matched by the expression on the left. At this stage, we can use preg_split.

A further refinement is the use of what I call the HRRT, which stands for the HamZa Regex Refactoring Technique. To make the regex digestible, he breaks it down into smaller named patterns: singlequotes, brackets and so on. This lets him define another name: skippable, for all these groups. After the definitions, the matching begins. If we can match the skippable pattern, the regex fails with (*SKIP)(*F) and the engine skips to the next position in the string.

That is the gist of it.

Here's the demo.

(?(DEFINE)
   (?P<signs>
      <
         (?:
            [^<>]
            |
            (?&signs)
         )*
      >
   )

   (?P<brackets>
      \[
         (?:
            [^][]
            |
            (?&brackets)
         )*
      \]
   )

   (?P<singlequotes>
      (?<!\\)'(?:[^\\]|\\.)*?'
   )

   (?P<doublequotes>
      (?<!\\)"(?:[^\\]|\\.)*?"
   )

   (?P<quotes>
      (?&singlequotes)|(?&doublequotes)
   )

   (?P<skippable>
      (?&brackets)|(?&signs)|(?&quotes)
   )
)

(?&skippable)(*SKIP)(*FAIL)
|
[ ]+

Answer 2

Updated:
this pattern also worked for me
(\[(?:[^\[\]]*?|(?R))*\])|(<.*?>)|\G\s([^<>\[\]]+)
Demo