How to emulate lookbehind within a regex with non-fixed width - php

Question

I'm currently encountering a problem with Regex in given circumstances : I need to parse PHP source files (class especially) to look for constants that are defined within those files and to retrieve them back to the output.
Those constants can have some documentation (and that's why I left the idea of Reflection since retrieving constants via Reflection only returns their name and their value) that may be shipped within comments tags.
I did manage to build the two separate parts of the regex (1 being the comment tag, the other being the const declaration) but I can't manage to link them both successfully : it seems that the very first constant within the file will also contain all the previously declared elements until it reaches the very first comment block.
My regex is as follows (I'm not a regex God so feel free to bring any criticism) :
((\t\ )*(/\*+(.|\n)*\*/)\R+)?([\t| ]*(?|(public|protected|private)\s*)?const\s+([a-zA-Z_\x7f-\xff][a-zA-Z0-9_\x7f-\xff]*)\s*=\s*(.*);)
There goes the sample test : Regex101
In case the initial code disappears :
/**
*
*/
class Test {
/**
*
*/
public const LOL = "damn";
/**
*
*/
private const TEST = 5;
public const plop = "dong";
}
I did look there and there for tips and I've learnt about positive lookbehind but from what I understood, it only works with fixed-width patterns.
I'm running out of ideas.

Answer 1

I would favor a multi-step approach: separate every class, then look for comments (eventually) and for the constants. In terms of regex, this can be achieved via
class\h*(?P<classname>\w+)[^{}]* # look for class literally and capture the name
(\{
(?:[^{}]*|(?2))* # the whole block matches the class content
\})
See a demo on regex101.com.
Now, to the comments and constants
^\h*
(?:(?P<comment>\Q/*\E(?s:.*?)\Q*/\E)(?s:.*?))?
(?:public|private)\h*const\h*
(?P<key>\w+)\h*=\h*(?P<value>[^;]+)
See a demo for this step on regex101.com as well.
The last step would be to clean the comments:
^\h*/?\*+\h*/?
See a demo for the cleansing on regex101.com.
Lastly, you'll need two loops:
preg_match_all($regex_class, $source, $matches, PREG_SET_ORDER);
foreach ($matches as $match) {
preg_match_all($const_class, $match[0], $constants, PREG_SET_ORDER);
foreach ($constants as $constant) {
$comment = preg_replace($clean_comment, '', $constant["comment"]);
# find the actual values here
echo "Class: {$match["classname"]}, Constant Name: {$constant["key"]}, Constant Value: {$constant["value"]}, Comment: $comment\n";
}
}
An overall demo can be found on ideone.com.
Mind the individual regex modifiers in the demo and source code (especially verbose and multiline !).
You can do it in an array as well:
$result = [];
preg_match_all($regex_class, $source, $matches, PREG_SET_ORDER);
foreach ($matches as $match) {
preg_match_all($const_class, $match[0], $constants, PREG_SET_ORDER);
foreach ($constants as $constant) {
$comment = trim(preg_replace($clean_comment, '', $constant["comment"]));
$result[$match["classname"]][] = array('name' => $constant["key"], 'value' => $constant['value'], 'comment' => $comment);
}
}
print_r($result);

Answer 2

You can do it without positive lookbehind:
You have to match a comment, immediately followed by a const declaration:
(?:(?:^/\*\*$\s+)((?:^ ?\*.*$\s*?)+)(?:\s+^\*/$\s+))?^\s+(public|protected|private) const (\S+)\s+= ([^;]+);
The first group will allow you to retrieve the documentation:
Comment part
(?:^/\*\*$\s+) finds the beginning of a block comment
((?:^ ?\*.*$\s*?)+) represents the group containing the content of your comments
(?:\s+^\*/$\s+) end of the comment
Declaration part:
^\s+ to skip the whitespace at the beginning of the line
(public|protected|private) const a group to determine visibility
(\S+)\s+= ([^;]+); groups for the name and value