Tag Parser

<?php

// Parsed as per W3's HTML Standard

function parseBBCode($HTMLSource)


    $PARSER_POSITION = 1;
    $$DATA_STATE = ""; // State of byte consumption
    $TEXT_COLLECTION; // Used to collect character tokens. Instead of emitting them as a single character
    $DOCUMENT = array(); // The final, tokenized, document structure.
    $OPEN_ELEMENTS = array(); // Track elements that have not been closed. Used to determine parents and children. (If a tag is open, and{ another is opened, we've found a family)
    $TEXT_COLLECTION = ""; // Collection of unprocessed unicode. Dumped into a textnode if left unused without an opening tag encountered.
    $ACTIVE_TOKEN; // Reference to the currently active token
    $ACTIVE_ATTRIBUTE; // Reference to the currently active attribute
    $TOKENS = array();
    $VALID_TAGS = array(
        'spoiler' => "spoiler",
    };

    function CreateToken($Type){
        $Token = array(
            'TagName' => $Type,
            'SelfClosingFlag' => false,
            'IsClosingTag' => false, // Key element to making the tree.
            'Data' => "",
            'Attributes' => array(),
            'Style' => array(),
        );
        array_push($OPEN_ELEMENTS, $Token);
        array_push($TOKENS, $Token);
        if ($ACTIVE_TOKEN){
            // I believe this is an impossible parser error. Never had it happen.
            if (in_array($ACTIVE_TOKEN, $OPEN_ELEMENTS)){
                array_push($ACTIVE_TOKEN['Children'], $Token);
            }
        }
        return $Token;
    }

    function AddAttribute($TOKEN){
        $Attr = array(
            'AttributeName' => "",
            'Value' => ""
        );
        array_push($TOKEN['Attributes'], $Attr);
        return $Attr;
    }

    function EmitCharToOpenToken($Character){
        $OpenToken = $ACTIVE_TOKEN;
        $OpenToken['Data'] .= $Character;
    }

    function EmitToken($TOKEN){
        // Why do we add the token to the DOCUMENT without knowing its children?
        // Because closing tags are ALSO added to the document.
        // The closing tags will tell us when a nest ends.
        // EXAMPLE:
        /*
            OPEN TAG A
            OPEN TAG B
            CLOSE TAG B
            CLOSE TAG A
        */
        // Because closing tags are there, we will know when one nest ends. We know OPEN TAG B is inside of A because it comes after A and has a valid close tag.

        // Verify the tag name is a valid tag from VALID_TAGS
        // You may remove this check to accept arbitrary HTML tag names not supported via ROBLOX GUIs
        if ($VALID_TAGS[$TOKEN['TagName']] || strtolower($TOKEN['TagName']) == "comment" || strtolower($TOKEN['TagName']) == "__text"){
            if ($VALID_TAGS[$TOKEN['TagName']]){
                $TOKEN['TagName'] = $VALID_TAGS[$TOKEN['TagName']];
            }
            array_push($DOCUMENT, $ACTIVE_TOKEN);
            $key = array_search($ACTIVE_TOKEN, $TOKENS);
            unset($TOKENS[$key]);
            if ($TOKEN['SelfClosingFlag']){
                // If it is self-closing, remove it from the open list. Otherwise, it is still open.
                $key = array_search($OPEN_ELEMENTS, $ACTIVE_TOKEN);
                unsert($OPEN_ELEMENTS[$key]);
            }
        }
        $ACTIVE_TOKEN = null;
        $ACTIVE_ATTRIBUTE = null;
    }

    // Begin parsing
    while ($PARSER_POSITION < strlen($HTMLSource)) do
        $CurrentChar = substr($HTMLSource, $PARSER_POSITION, 1);
        if ($DATA_STATE == ""){
            if ($CurrentChar == "<"){
                // Emit text collection
                if (TEXT_COLLECTION){
                    $ACTIVE_TOKEN = CreateToken("__Text")
                    $ACTIVE_TOKEN.Data = TEXT_COLLECTION
                    $ACTIVE_TOKEN.SelfClosingFlag = true
                    EmitToken(ACTIVE_TOKEN)
                    TEXT_COLLECTION = nil
                }else{
                    EmitToken(ACTIVE_TOKEN)
                }
                $DATA_STATE = "TAG_OPEN"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                // Emit character token
                if (TEXT_COLLECTION){
                    TEXT_COLLECTION = TEXT_COLLECTION .. CurrentChar
                }else{
                    TEXT_COLLECTION = CurrentChar
            }
            $PARSER_POSITION = $PARSER_POSITION + 1
        }
        }elseif{} ($DATA_STATE == "TAG_OPEN"){
            if ($CurrentChar == "!"){
                $DATA_STATE = "MARKUP_DECLARATION_OPEN"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == "/"){
                $DATA_STATE = "END_TAG_OPEN"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                $DATA_STATE = "TAG_NAME"
                $ACTIVE_TOKEN = CreateToken("")
                // DO NOT CONSUME
        }
        }elseif{ ($DATA_STATE == "END_TAG_OPEN"){
            if ($CurrentChar == ">"){
                // Parse error
                $DATA_STATE = ""
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                $DATA_STATE = "TAG_NAME"
                $ACTIVE_TOKEN = CreateToken("")
                $ACTIVE_TOKEN.IsClosingTag = true
                // DO NOT CONSUME
        }
        }elseif{ ($DATA_STATE == "TAG_NAME"){
            if ($CurrentChar == " "){
                $DATA_STATE = "BEFORE_ATTRIBUTE_NAME"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == "/"){
                $DATA_STATE = "SELF_CLOSING_START_TAG"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == ">"){
                $DATA_STATE = ""
                EmitToken(ACTIVE_TOKEN)
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                $ACTIVE_TOKEN.TagName = ACTIVE_TOKEN.TagName .. CurrentChar
                $PARSER_POSITION = $PARSER_POSITION + 1
        }
        }elseif{ ($DATA_STATE == "BEFORE_ATTRIBUTE_NAME"){
            if ($CurrentChar == " "){
                // Ignore
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == "/"){
                $DATA_STATE = "SELF_CLOSING_START_TAG"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == '"' or CurrentChar == "'" or CurrentChar == "<"){
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == ">"){
                $DATA_STATE = ""
                EmitToken(ACTIVE_TOKEN)
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                $DATA_STATE = "ATTRIBUTE_NAME"
                ACTIVE_ATTRIBUTE = AddAttribute(ACTIVE_TOKEN)
                // DO NOT CONSUME
        }
        }elseif{ ($DATA_STATE == "ATTRIBUTE_NAME"){
            if ($CurrentChar == " "){
                $DATA_STATE = "AFTER_ATTRIBUTE_NAME"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == "/"){
                $DATA_STATE = "SELF_CLOSING_START_TAG"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == '"' or CurrentChar == "'"){
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == "="){
                $DATA_STATE = "BEFORE_ATTRIBUTE_VALUE"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == ">"){
                $DATA_STATE = ""
                EmitToken(ACTIVE_TOKEN)
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                $PARSER_POSITION = $PARSER_POSITION + 1
                ACTIVE_ATTRIBUTE.AttributeName = ACTIVE_ATTRIBUTE.AttributeName .. CurrentChar
        }
        }elseif{ ($DATA_STATE == "BEFORE_ATTRIBUTE_VALUE"){
            if ($CurrentChar == " "){
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == '"'){
                $DATA_STATE = "ATTRIBUTE_VALUE_DOUBLE_QUOTE"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == '&'){
                $DATA_STATE = "ATTRIBUTE_VALUE_UNQUOTED"
                // DO NOT CONSUME THE CHARACTER
            }elseif{ ($CurrentChar == "'"){
                $DATA_STATE = "ATTRIBUTE_VALUE_SINGLE_QUOTE"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == ">"){
                $DATA_STATE = ""
                EmitToken(ACTIVE_TOKEN)
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == "<" or CurrentChar == "="){
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                ACTIVE_ATTRIBUTE.Value = ACTIVE_ATTRIBUTE.Value .. CurrentChar
            }
        }elseif{ ($DATA_STATE == "ATTRIBUTE_VALUE_DOUBLE_QUOTE"){
            if ($CurrentChar == '"'){
                $DATA_STATE = "AFTER_ATTRIBUTE_VALUE_QUOTED"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                ACTIVE_ATTRIBUTE.Value = ACTIVE_ATTRIBUTE.Value .. CurrentChar
                $PARSER_POSITION = $PARSER_POSITION + 1
        }elseif{ ($DATA_STATE == "ATTRIBUTE_VALUE_SINGLE_QUOTE"){
            if ($CurrentChar == "'"){
                $DATA_STATE = "AFTER_ATTRIBUTE_VALUE_QUOTED"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                ACTIVE_ATTRIBUTE.Value = ACTIVE_ATTRIBUTE.Value .. CurrentChar
                $PARSER_POSITION = $PARSER_POSITION + 1
        }elseif{ ($DATA_STATE == "AFTER_ATTRIBUTE_VALUE_QUOTED"){
            if ($CurrentChar == " "){
                $DATA_STATE = "BEFORE_ATTRIBUTE_NAME"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == "/"){
                $DATA_STATE = "SELF_CLOSING_START_TAG"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                $DATA_STATE = "BEFORE_ATTRIBUTE_NAME"
                // DO NOT CONSUME CHARACTER
        }elseif{ ($DATA_STATE == "SELF_CLOSING_START_TAG"){
            if ($CurrentChar == ">"){
                $DATA_STATE = ""
                $ACTIVE_TOKEN.SelfClosingFlag = true
                EmitToken(ACTIVE_TOKEN)
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                $DATA_STATE = "BEFORE_ATTRIBUTE_NAME"
                $PARSER_POSITION = $PARSER_POSITION + 1
        }elseif{ ($DATA_STATE == "MARKUP_DECLARATION_OPEN"){
            // Are next two bytes '//' ?
            if ($CurrentChar == "-" and HTMLSource:sub$PARSER_POSITION + 1,$PARSER_POSITION + 1) == "-"){
                $DATA_STATE = "COMMENT_START"
                $PARSER_POSITION = $PARSER_POSITION + 2
                $ACTIVE_TOKEN = CreateToken("Comment")
                // Comments are self-closing
                $ACTIVE_TOKEN.SelfClosingFlag = true
            }else{
                $DATA_STATE = "BOGUS_COMMENT"
                $PARSER_POSITION = $PARSER_POSITION + 1
        }elseif{ ($DATA_STATE == "BOGUS_COMMENT"){
            if ($CurrentChar ~= ">"){
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                $DATA_STATE = "";
                $PARSER_POSITION = $PARSER_POSITION + 1
        }elseif{ ($DATA_STATE == "COMMENT_START"){
            if ($CurrentChar == "-"){
                $DATA_STATE = "COMMENT_START_DASH"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                EmitCharToOpenToken($CurrentChar)
                $PARSER_POSITION = $PARSER_POSITION + 1
        }elseif{ ($DATA_STATE == "COMMENT_START_DASH"){
            if ($CurrentChar == "-"){
                $DATA_STATE = "COMMENT_END"
                $PARSER_POSITION = $PARSER_POSITION + 1
            }elseif{ ($CurrentChar == ">"){
                $DATA_STATE = ""
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                $ACTIVE_TOKEN.Data = ACTIVE_TOKEN.Data .. CurrentChar
                $PARSER_POSITION = $PARSER_POSITION + 1
        }elseif{ ($DATA_STATE == "COMMENT_END"){
            if ($CurrentChar == ">"){
                $DATA_STATE = ""
                EmitToken(ACTIVE_TOKEN)
                $PARSER_POSITION = $PARSER_POSITION + 1
            }else{
                $DATA_STATE = "COMMENT_START"
            }
        }
    }
}