update the regex for strip()

while our solution works just fine for the majority of the cases, there still appear to be a few outliers that manage to break the matching. note that the algorithm may be unoptimised too, so thats a thing to look out for as well.

for reference, heres the (fully commented) presumably working version of the function, courtesy of xymph (WARNING PHP):

// (markdown is a bit trippy here, everything preceded by // is a comment despite the incorrect highlighting)
function stripFormatting($input, $for_tm = true) {
    return
        // Replace all occurrences of a null character back with a pair of dollar signs for displaying in TM, or a single dollar for log messages etc.
        str_replace("\0", ($for_tm ? '$$' : '$'),
            // Replace links (introduced in TMU)
            preg_replace(
                '/
                // Strip TMF H, L & P links by stripping everything between each square bracket pair until another $H, $L or $P sequence (or EoS) is found, this allows a $H to close a $L and vice versa, as does the game
                \\$[hlp](.*?)(?:\\[.*?\\](.*?))*(?:\\$[hlp]|$)
                /ixu',
                // Keep the first and third capturing groups if present
                '$1$2',
                // Replace various patterns beginning with an unescaped dollar
                preg_replace(
                    '/
                    // Match a single dollar sign and any of the following:
                    \\$
                    (?:
                        // Strip color codes by matching any hexadecimal character and any other two characters following it (except $)
                        [0-9a-f][^$][^$]
                        // Strip any incomplete color codes by matching any hexadecimal character followed by another character (except $)
                        |[0-9a-f][^$]
                        // Strip any single style code (including an invisible UTF8 char) that is not an H, L or P link or a bracket ($[ and $])
                        |[^][hlp]
                        // Strip the dollar sign if it is followed by [ or ], but do not strip the brackets themselves
                        |(?=[][])
                        // Strip the dollar sign if it is at the end of the string
                        |$
                    )
                    // Ignore alphabet case, ignore whitespace in pattern & use UTF-8 mode
                    /ixu',
                    // Replace any matches with nothing (i.e. strip matches)
                    '',
                    // Replace all occurrences of dollar sign pairs with a null character
                    str_replace('$$', "\0", $input)
                )
            )
        )
    ;
}

ignoring the comments and """fancy""" format (aka i love wasting space in my editor):

function stripColours($input, $forTM = true) {
    return 
    str_replace("\0", ($forTM ? '$$' : '$'), 
    preg_replace('/\\$[hlp](.*?)(?:\\[.*?\\](.*?))*(?:\\$[hlp]|$)/ixu', '$1$2', 
    preg_replace('/\\$(?:[0-9a-f][^$][^$]|[0-9a-f][^$]|[^][hlp]|(?=[][])|$)/ixu', '', 
    str_replace('$$', "\0", $input))));
}

forTM can obviously be ignored. it is unused even in xaseco itself XD.

NOTE that the function above does not remove the other formatting codes, for that \$[SHWIPLONGTZ]\ should be used (usually as the last OR statement)

here's a list of entries that should be tested against the regex to ensure its working (use regexr or something similar, dont overcomplicate your life lol):

$0CFωα$0CFғ$0CFα$i$000$i | $3CFNexogg GF
$L[goo.gl/UJy69u]$fffмσтι$0C6$fffи|ғ
$H[manialinker]$Hheythere$P[manialinker]
$000[$5f6CMC$000]$b$f60GuessWho
$B$FFFUber Bug
TM$5$6/$0$ARU DES
|$fadǵ$faeƒ$faf.$i$fffsabø$f|
$i$adf$wP$fffazeh
$<thefunnyname$>
dollar$$dollar

expected output:

ωαғα | Nexogg GF
мσтιи|ғ
heythere
[CMC]GuessWho
Uber Bug
TM DES
ǵƒ.sabø
Pazeh
thefunnyname
dollar$$dollar

current output:

ωαғα | Nexogg GF
мσтιи|ғ
heythere
[CMC]$bGuessWho
$BUber Bug
TM$5$0 DES
|ǵƒ.sabø
Pazeh
$<thefunnyname$>
dollar$lar

lythx / trakman

update the regex for strip() #133