the-forest/client/node_modules/webpack/lib/css/walkCssTokens.js
2024-09-17 20:35:18 -04:00

776 lines
20 KiB
JavaScript

/*
MIT License http://www.opensource.org/licenses/mit-license.php
Author Tobias Koppers @sokra
*/
"use strict";
/**
* @typedef {object} CssTokenCallbacks
* @property {function(string, number): boolean=} isSelector
* @property {function(string, number, number, number, number): number=} url
* @property {function(string, number, number): number=} string
* @property {function(string, number, number): number=} leftParenthesis
* @property {function(string, number, number): number=} rightParenthesis
* @property {function(string, number, number): number=} pseudoFunction
* @property {function(string, number, number): number=} function
* @property {function(string, number, number): number=} pseudoClass
* @property {function(string, number, number): number=} atKeyword
* @property {function(string, number, number): number=} class
* @property {function(string, number, number): number=} identifier
* @property {function(string, number, number): number=} id
* @property {function(string, number, number): number=} leftCurlyBracket
* @property {function(string, number, number): number=} rightCurlyBracket
* @property {function(string, number, number): number=} semicolon
* @property {function(string, number, number): number=} comma
*/
/** @typedef {function(string, number, CssTokenCallbacks): number} CharHandler */
// spec: https://drafts.csswg.org/css-syntax/
const CC_LINE_FEED = "\n".charCodeAt(0);
const CC_CARRIAGE_RETURN = "\r".charCodeAt(0);
const CC_FORM_FEED = "\f".charCodeAt(0);
const CC_TAB = "\t".charCodeAt(0);
const CC_SPACE = " ".charCodeAt(0);
const CC_SOLIDUS = "/".charCodeAt(0);
const CC_REVERSE_SOLIDUS = "\\".charCodeAt(0);
const CC_ASTERISK = "*".charCodeAt(0);
const CC_LEFT_PARENTHESIS = "(".charCodeAt(0);
const CC_RIGHT_PARENTHESIS = ")".charCodeAt(0);
const CC_LEFT_CURLY = "{".charCodeAt(0);
const CC_RIGHT_CURLY = "}".charCodeAt(0);
const CC_LEFT_SQUARE = "[".charCodeAt(0);
const CC_RIGHT_SQUARE = "]".charCodeAt(0);
const CC_QUOTATION_MARK = '"'.charCodeAt(0);
const CC_APOSTROPHE = "'".charCodeAt(0);
const CC_FULL_STOP = ".".charCodeAt(0);
const CC_COLON = ":".charCodeAt(0);
const CC_SEMICOLON = ";".charCodeAt(0);
const CC_COMMA = ",".charCodeAt(0);
const CC_PERCENTAGE = "%".charCodeAt(0);
const CC_AT_SIGN = "@".charCodeAt(0);
const CC_LOW_LINE = "_".charCodeAt(0);
const CC_LOWER_A = "a".charCodeAt(0);
const CC_LOWER_U = "u".charCodeAt(0);
const CC_LOWER_E = "e".charCodeAt(0);
const CC_LOWER_Z = "z".charCodeAt(0);
const CC_UPPER_A = "A".charCodeAt(0);
const CC_UPPER_E = "E".charCodeAt(0);
const CC_UPPER_U = "U".charCodeAt(0);
const CC_UPPER_Z = "Z".charCodeAt(0);
const CC_0 = "0".charCodeAt(0);
const CC_9 = "9".charCodeAt(0);
const CC_NUMBER_SIGN = "#".charCodeAt(0);
const CC_PLUS_SIGN = "+".charCodeAt(0);
const CC_HYPHEN_MINUS = "-".charCodeAt(0);
const CC_LESS_THAN_SIGN = "<".charCodeAt(0);
const CC_GREATER_THAN_SIGN = ">".charCodeAt(0);
/**
* @param {number} cc char code
* @returns {boolean} true, if cc is a newline
*/
const _isNewLine = cc =>
cc === CC_LINE_FEED || cc === CC_CARRIAGE_RETURN || cc === CC_FORM_FEED;
/** @type {CharHandler} */
const consumeSpace = (input, pos, _callbacks) => {
/** @type {number} */
let cc;
do {
pos++;
cc = input.charCodeAt(pos);
} while (_isWhiteSpace(cc));
return pos;
};
/**
* @param {number} cc char code
* @returns {boolean} true, if cc is a newline
*/
const _isNewline = cc =>
cc === CC_LINE_FEED || cc === CC_CARRIAGE_RETURN || cc === CC_FORM_FEED;
/**
* @param {number} cc char code
* @returns {boolean} true, if cc is a space (U+0009 CHARACTER TABULATION or U+0020 SPACE)
*/
const _isSpace = cc => cc === CC_TAB || cc === CC_SPACE;
/**
* @param {number} cc char code
* @returns {boolean} true, if cc is a whitespace
*/
const _isWhiteSpace = cc => _isNewline(cc) || _isSpace(cc);
/**
* ident-start code point
*
* A letter, a non-ASCII code point, or U+005F LOW LINE (_).
* @param {number} cc char code
* @returns {boolean} true, if cc is a start code point of an identifier
*/
const isIdentStartCodePoint = cc =>
(cc >= CC_LOWER_A && cc <= CC_LOWER_Z) ||
(cc >= CC_UPPER_A && cc <= CC_UPPER_Z) ||
cc === CC_LOW_LINE ||
cc >= 0x80;
/** @type {CharHandler} */
const consumeDelimToken = (input, pos, _callbacks) => pos + 1;
/** @type {CharHandler} */
const consumeComments = (input, pos, _callbacks) => {
// If the next two input code point are U+002F SOLIDUS (/) followed by a U+002A
// ASTERISK (*), consume them and all following code points up to and including
// the first U+002A ASTERISK (*) followed by a U+002F SOLIDUS (/), or up to an
// EOF code point. Return to the start of this step.
//
// If the preceding paragraph ended by consuming an EOF code point, this is a parse error.
// But we are silent on errors.
if (
input.charCodeAt(pos) === CC_SOLIDUS &&
input.charCodeAt(pos + 1) === CC_ASTERISK
) {
pos += 1;
while (pos < input.length) {
if (
input.charCodeAt(pos) === CC_ASTERISK &&
input.charCodeAt(pos + 1) === CC_SOLIDUS
) {
pos += 2;
break;
}
pos++;
}
}
return pos;
};
/** @type {function(number): CharHandler} */
const consumeString = quoteCc => (input, pos, callbacks) => {
const start = pos;
pos = _consumeString(input, pos, quoteCc);
if (callbacks.string !== undefined) {
pos = callbacks.string(input, start, pos);
}
return pos;
};
/**
* @param {string} input input
* @param {number} pos position
* @param {number} quoteCc quote char code
* @returns {number} new position
*/
const _consumeString = (input, pos, quoteCc) => {
pos++;
for (;;) {
if (pos === input.length) return pos;
const cc = input.charCodeAt(pos);
if (cc === quoteCc) return pos + 1;
if (_isNewLine(cc)) {
// bad string
return pos;
}
if (cc === CC_REVERSE_SOLIDUS) {
// we don't need to fully parse the escaped code point
// just skip over a potential new line
pos++;
if (pos === input.length) return pos;
pos++;
} else {
pos++;
}
}
};
/**
* @param {number} cc char code
* @returns {boolean} is identifier start code
*/
const _isIdentifierStartCode = cc =>
cc === CC_LOW_LINE ||
(cc >= CC_LOWER_A && cc <= CC_LOWER_Z) ||
(cc >= CC_UPPER_A && cc <= CC_UPPER_Z) ||
cc > 0x80;
/**
* @param {number} first first code point
* @param {number} second second code point
* @returns {boolean} true if two code points are a valid escape
*/
const _isTwoCodePointsAreValidEscape = (first, second) => {
if (first !== CC_REVERSE_SOLIDUS) return false;
if (_isNewLine(second)) return false;
return true;
};
/**
* @param {number} cc char code
* @returns {boolean} is digit
*/
const _isDigit = cc => cc >= CC_0 && cc <= CC_9;
/**
* @param {string} input input
* @param {number} pos position
* @returns {boolean} true, if input at pos starts an identifier
*/
const _startsIdentifier = (input, pos) => {
const cc = input.charCodeAt(pos);
if (cc === CC_HYPHEN_MINUS) {
if (pos === input.length) return false;
const cc = input.charCodeAt(pos + 1);
if (cc === CC_HYPHEN_MINUS) return true;
if (cc === CC_REVERSE_SOLIDUS) {
const cc = input.charCodeAt(pos + 2);
return !_isNewLine(cc);
}
return _isIdentifierStartCode(cc);
}
if (cc === CC_REVERSE_SOLIDUS) {
const cc = input.charCodeAt(pos + 1);
return !_isNewLine(cc);
}
return _isIdentifierStartCode(cc);
};
/** @type {CharHandler} */
const consumeNumberSign = (input, pos, callbacks) => {
const start = pos;
pos++;
if (pos === input.length) return pos;
if (
callbacks.isSelector &&
callbacks.isSelector(input, pos) &&
_startsIdentifier(input, pos)
) {
pos = _consumeIdentifier(input, pos, callbacks);
if (callbacks.id !== undefined) {
return callbacks.id(input, start, pos);
}
}
return pos;
};
/** @type {CharHandler} */
const consumeMinus = (input, pos, callbacks) => {
const start = pos;
pos++;
if (pos === input.length) return pos;
const cc = input.charCodeAt(pos);
// If the input stream starts with a number, reconsume the current input code point, consume a numeric token, and return it.
if (cc === CC_FULL_STOP || _isDigit(cc)) {
return consumeNumericToken(input, pos, callbacks);
} else if (cc === CC_HYPHEN_MINUS) {
pos++;
if (pos === input.length) return pos;
const cc = input.charCodeAt(pos);
if (cc === CC_GREATER_THAN_SIGN) {
return pos + 1;
}
pos = _consumeIdentifier(input, pos, callbacks);
if (callbacks.identifier !== undefined) {
return callbacks.identifier(input, start, pos);
}
} else if (cc === CC_REVERSE_SOLIDUS) {
if (pos + 1 === input.length) return pos;
const cc = input.charCodeAt(pos + 1);
if (_isNewLine(cc)) return pos;
pos = _consumeIdentifier(input, pos, callbacks);
if (callbacks.identifier !== undefined) {
return callbacks.identifier(input, start, pos);
}
} else if (_isIdentifierStartCode(cc)) {
pos = consumeOtherIdentifier(input, pos - 1, callbacks);
}
return pos;
};
/** @type {CharHandler} */
const consumeDot = (input, pos, callbacks) => {
const start = pos;
pos++;
if (pos === input.length) return pos;
const cc = input.charCodeAt(pos);
if (_isDigit(cc)) return consumeNumericToken(input, pos - 2, callbacks);
if (
(callbacks.isSelector && !callbacks.isSelector(input, pos)) ||
!_startsIdentifier(input, pos)
)
return pos;
pos = _consumeIdentifier(input, pos, callbacks);
if (callbacks.class !== undefined) return callbacks.class(input, start, pos);
return pos;
};
/** @type {CharHandler} */
const consumeNumericToken = (input, pos, callbacks) => {
pos = _consumeNumber(input, pos, callbacks);
if (pos === input.length) return pos;
if (_startsIdentifier(input, pos))
return _consumeIdentifier(input, pos, callbacks);
const cc = input.charCodeAt(pos);
if (cc === CC_PERCENTAGE) return pos + 1;
return pos;
};
/** @type {CharHandler} */
const consumeOtherIdentifier = (input, pos, callbacks) => {
const start = pos;
pos = _consumeIdentifier(input, pos, callbacks);
if (pos !== input.length && input.charCodeAt(pos) === CC_LEFT_PARENTHESIS) {
pos++;
if (callbacks.function !== undefined) {
return callbacks.function(input, start, pos);
}
} else if (callbacks.identifier !== undefined) {
return callbacks.identifier(input, start, pos);
}
return pos;
};
/** @type {CharHandler} */
const consumePotentialUrl = (input, pos, callbacks) => {
const start = pos;
pos = _consumeIdentifier(input, pos, callbacks);
const nextPos = pos + 1;
if (
pos === start + 3 &&
input.slice(start, nextPos).toLowerCase() === "url("
) {
pos++;
let cc = input.charCodeAt(pos);
while (_isWhiteSpace(cc)) {
pos++;
if (pos === input.length) return pos;
cc = input.charCodeAt(pos);
}
if (cc === CC_QUOTATION_MARK || cc === CC_APOSTROPHE) {
if (callbacks.function !== undefined) {
return callbacks.function(input, start, nextPos);
}
return nextPos;
}
const contentStart = pos;
/** @type {number} */
let contentEnd;
for (;;) {
if (cc === CC_REVERSE_SOLIDUS) {
pos++;
if (pos === input.length) return pos;
pos++;
} else if (_isWhiteSpace(cc)) {
contentEnd = pos;
do {
pos++;
if (pos === input.length) return pos;
cc = input.charCodeAt(pos);
} while (_isWhiteSpace(cc));
if (cc !== CC_RIGHT_PARENTHESIS) return pos;
pos++;
if (callbacks.url !== undefined) {
return callbacks.url(input, start, pos, contentStart, contentEnd);
}
return pos;
} else if (cc === CC_RIGHT_PARENTHESIS) {
contentEnd = pos;
pos++;
if (callbacks.url !== undefined) {
return callbacks.url(input, start, pos, contentStart, contentEnd);
}
return pos;
} else if (cc === CC_LEFT_PARENTHESIS) {
return pos;
} else {
pos++;
}
if (pos === input.length) return pos;
cc = input.charCodeAt(pos);
}
} else {
if (callbacks.identifier !== undefined) {
return callbacks.identifier(input, start, pos);
}
return pos;
}
};
/** @type {CharHandler} */
const consumePotentialPseudo = (input, pos, callbacks) => {
const start = pos;
pos++;
if (
(callbacks.isSelector && !callbacks.isSelector(input, pos)) ||
!_startsIdentifier(input, pos)
)
return pos;
pos = _consumeIdentifier(input, pos, callbacks);
const cc = input.charCodeAt(pos);
if (cc === CC_LEFT_PARENTHESIS) {
pos++;
if (callbacks.pseudoFunction !== undefined) {
return callbacks.pseudoFunction(input, start, pos);
}
return pos;
}
if (callbacks.pseudoClass !== undefined) {
return callbacks.pseudoClass(input, start, pos);
}
return pos;
};
/** @type {CharHandler} */
const consumeLeftParenthesis = (input, pos, callbacks) => {
pos++;
if (callbacks.leftParenthesis !== undefined) {
return callbacks.leftParenthesis(input, pos - 1, pos);
}
return pos;
};
/** @type {CharHandler} */
const consumeRightParenthesis = (input, pos, callbacks) => {
pos++;
if (callbacks.rightParenthesis !== undefined) {
return callbacks.rightParenthesis(input, pos - 1, pos);
}
return pos;
};
/** @type {CharHandler} */
const consumeLeftCurlyBracket = (input, pos, callbacks) => {
pos++;
if (callbacks.leftCurlyBracket !== undefined) {
return callbacks.leftCurlyBracket(input, pos - 1, pos);
}
return pos;
};
/** @type {CharHandler} */
const consumeRightCurlyBracket = (input, pos, callbacks) => {
pos++;
if (callbacks.rightCurlyBracket !== undefined) {
return callbacks.rightCurlyBracket(input, pos - 1, pos);
}
return pos;
};
/** @type {CharHandler} */
const consumeSemicolon = (input, pos, callbacks) => {
pos++;
if (callbacks.semicolon !== undefined) {
return callbacks.semicolon(input, pos - 1, pos);
}
return pos;
};
/** @type {CharHandler} */
const consumeComma = (input, pos, callbacks) => {
pos++;
if (callbacks.comma !== undefined) {
return callbacks.comma(input, pos - 1, pos);
}
return pos;
};
/** @type {CharHandler} */
const _consumeIdentifier = (input, pos) => {
for (;;) {
const cc = input.charCodeAt(pos);
if (cc === CC_REVERSE_SOLIDUS) {
pos++;
if (pos === input.length) return pos;
pos++;
} else if (
_isIdentifierStartCode(cc) ||
_isDigit(cc) ||
cc === CC_HYPHEN_MINUS
) {
pos++;
} else {
return pos;
}
}
};
/** @type {CharHandler} */
const _consumeNumber = (input, pos) => {
pos++;
if (pos === input.length) return pos;
let cc = input.charCodeAt(pos);
while (_isDigit(cc)) {
pos++;
if (pos === input.length) return pos;
cc = input.charCodeAt(pos);
}
if (cc === CC_FULL_STOP && pos + 1 !== input.length) {
const next = input.charCodeAt(pos + 1);
if (_isDigit(next)) {
pos += 2;
cc = input.charCodeAt(pos);
while (_isDigit(cc)) {
pos++;
if (pos === input.length) return pos;
cc = input.charCodeAt(pos);
}
}
}
if (cc === CC_LOWER_E || cc === CC_UPPER_E) {
if (pos + 1 !== input.length) {
const next = input.charCodeAt(pos + 2);
if (_isDigit(next)) {
pos += 2;
} else if (
(next === CC_HYPHEN_MINUS || next === CC_PLUS_SIGN) &&
pos + 2 !== input.length
) {
const next = input.charCodeAt(pos + 2);
if (_isDigit(next)) {
pos += 3;
} else {
return pos;
}
} else {
return pos;
}
}
} else {
return pos;
}
cc = input.charCodeAt(pos);
while (_isDigit(cc)) {
pos++;
if (pos === input.length) return pos;
cc = input.charCodeAt(pos);
}
return pos;
};
/** @type {CharHandler} */
const consumeLessThan = (input, pos, _callbacks) => {
if (input.slice(pos + 1, pos + 4) === "!--") return pos + 4;
return pos + 1;
};
/** @type {CharHandler} */
const consumeAt = (input, pos, callbacks) => {
const start = pos;
pos++;
if (pos === input.length) return pos;
if (_startsIdentifier(input, pos)) {
pos = _consumeIdentifier(input, pos, callbacks);
if (callbacks.atKeyword !== undefined) {
pos = callbacks.atKeyword(input, start, pos);
}
}
return pos;
};
/** @type {CharHandler} */
const consumeReverseSolidus = (input, pos, callbacks) => {
const start = pos;
pos++;
if (pos === input.length) return pos;
// If the input stream starts with a valid escape, reconsume the current input code point, consume an ident-like token, and return it.
if (
_isTwoCodePointsAreValidEscape(
input.charCodeAt(start),
input.charCodeAt(pos)
)
) {
return consumeOtherIdentifier(input, pos - 1, callbacks);
}
// Otherwise, this is a parse error. Return a <delim-token> with its value set to the current input code point.
return pos;
};
const CHAR_MAP = Array.from({ length: 0x80 }, (_, cc) => {
// https://drafts.csswg.org/css-syntax/#consume-token
switch (cc) {
// whitespace
case CC_LINE_FEED:
case CC_CARRIAGE_RETURN:
case CC_FORM_FEED:
case CC_TAB:
case CC_SPACE:
return consumeSpace;
// U+0022 QUOTATION MARK (")
case CC_QUOTATION_MARK:
return consumeString(cc);
// U+0023 NUMBER SIGN (#)
case CC_NUMBER_SIGN:
return consumeNumberSign;
// U+0027 APOSTROPHE (')
case CC_APOSTROPHE:
return consumeString(cc);
// U+0028 LEFT PARENTHESIS (()
case CC_LEFT_PARENTHESIS:
return consumeLeftParenthesis;
// U+0029 RIGHT PARENTHESIS ())
case CC_RIGHT_PARENTHESIS:
return consumeRightParenthesis;
// U+002B PLUS SIGN (+)
case CC_PLUS_SIGN:
return consumeNumericToken;
// U+002C COMMA (,)
case CC_COMMA:
return consumeComma;
// U+002D HYPHEN-MINUS (-)
case CC_HYPHEN_MINUS:
return consumeMinus;
// U+002E FULL STOP (.)
case CC_FULL_STOP:
return consumeDot;
// U+003A COLON (:)
case CC_COLON:
return consumePotentialPseudo;
// U+003B SEMICOLON (;)
case CC_SEMICOLON:
return consumeSemicolon;
// U+003C LESS-THAN SIGN (<)
case CC_LESS_THAN_SIGN:
return consumeLessThan;
// U+0040 COMMERCIAL AT (@)
case CC_AT_SIGN:
return consumeAt;
// U+005B LEFT SQUARE BRACKET ([)
case CC_LEFT_SQUARE:
return consumeDelimToken;
// U+005C REVERSE SOLIDUS (\)
case CC_REVERSE_SOLIDUS:
return consumeReverseSolidus;
// U+005D RIGHT SQUARE BRACKET (])
case CC_RIGHT_SQUARE:
return consumeDelimToken;
// U+007B LEFT CURLY BRACKET ({)
case CC_LEFT_CURLY:
return consumeLeftCurlyBracket;
// U+007D RIGHT CURLY BRACKET (})
case CC_RIGHT_CURLY:
return consumeRightCurlyBracket;
// Optimization
case CC_LOWER_U:
case CC_UPPER_U:
return consumePotentialUrl;
default:
// digit
if (_isDigit(cc)) return consumeNumericToken;
// ident-start code point
if (isIdentStartCodePoint(cc)) {
return consumeOtherIdentifier;
}
// EOF, but we don't have it
// anything else
return consumeDelimToken;
}
});
/**
* @param {string} input input css
* @param {CssTokenCallbacks} callbacks callbacks
* @returns {void}
*/
module.exports = (input, callbacks) => {
// This section describes how to consume a token from a stream of code points. It will return a single token of any type.
let pos = 0;
while (pos < input.length) {
// Consume comments.
pos = consumeComments(input, pos, callbacks);
const cc = input.charCodeAt(pos);
// Consume the next input code point.
if (cc < 0x80) {
pos = CHAR_MAP[cc](input, pos, callbacks);
} else {
pos++;
}
}
};
module.exports.isIdentStartCodePoint = isIdentStartCodePoint;
/**
* @param {string} input input
* @param {number} pos position
* @returns {number} position after comments
*/
module.exports.eatComments = (input, pos) => {
for (;;) {
const originalPos = pos;
pos = consumeComments(input, pos, {});
if (originalPos === pos) {
break;
}
}
return pos;
};
/**
* @param {string} input input
* @param {number} pos position
* @returns {number} position after whitespace
*/
module.exports.eatWhitespace = (input, pos) => {
while (_isWhiteSpace(input.charCodeAt(pos))) {
pos++;
}
return pos;
};
/**
* @param {string} input input
* @param {number} pos position
* @returns {number} position after whitespace and comments
*/
module.exports.eatWhitespaceAndComments = (input, pos) => {
for (;;) {
const originalPos = pos;
pos = consumeComments(input, pos, {});
while (_isWhiteSpace(input.charCodeAt(pos))) {
pos++;
}
if (originalPos === pos) {
break;
}
}
return pos;
};
/**
* @param {string} input input
* @param {number} pos position
* @returns {number} position after whitespace
*/
module.exports.eatWhiteLine = (input, pos) => {
for (;;) {
const cc = input.charCodeAt(pos);
if (_isSpace(cc)) {
pos++;
continue;
}
if (_isNewLine(cc)) pos++;
// For `\r\n`
if (cc === CC_CARRIAGE_RETURN && input.charCodeAt(pos + 1) === CC_LINE_FEED)
pos++;
break;
}
return pos;
};