Files
less.js/lib/less/parser.js
cloudhead 2ebbf37c4b More accurate parse error messages
Because of the backtracking, we lose the previous point of failure.
The solution is to store the furthest point the parser has parsed to,
when backtracking, and using it when showing parse errors.
2010-04-25 00:06:45 -04:00

895 lines
32 KiB
JavaScript

if (typeof(require) !== 'undefined') {
var less = exports;
var tree = require('less/tree');
} else {
var less = tree = {};
}
//
// less.js - parser
//
// A relatively straight-forward recursive-descent parser.
// There is no tokenization/lexing stage, the input is parsed
// in one sweep.
//
// To make the parser fast enough to run in the browser, several
// optimization had to be made:
//
// - Instead of the more commonly used technique of slicing the
// input string on every match, we use global regexps (/g),
// and move the `lastIndex` pointer on match, foregoing `slice()`
// completely. This gives us a 3x speed-up.
//
// - Matching on a huge input is often cause of slowdowns,
// especially with the /g flag. The solution to that is to
// chunkify the input: we split it by /\n\n/, just to be on
// the safe side. The chunks are stored in the `chunks` var,
// `j` holds the current chunk index, and `current` holds
// the index of the current chunk in relation to `input`.
// This gives us an almost 4x speed-up.
//
// - In many cases, we don't need to match individual tokens;
// for example, if a value doesn't hold any variables, operations
// or dynamic references, the parser can effectively 'skip' it,
// treating it as a literal.
// An example would be '1px solid #000' - which evaluates to itself,
// we don't need to know what the individual components are.
// The drawback, of course is that you don't get the benefits of
// syntax-checking on the CSS. This gives us a 50% speed-up in the parser,
// and a smaller speed-up in the code-gen.
//
//
// Token matching is done with the `$` function, which either takes
// a terminal string or regexp, or a non-terminal function to call.
// It also takes care of moving all the indices forwards.
//
//
less.Parser = function Parser(env) {
var input, // LeSS input string
i, // current index in `input`
j, // current chunk
furthest, // furthest index the parser has gone to
chunks, // chunkified input
current, // index of current chunk, in `input`
inputLength,
parser;
var that = this;
// This function is called after all files
// have been imported through `@import`.
var finish = function () {};
var imports = this.imports = {
paths: env && env.paths || [], // Search paths, when importing
queue: [], // Files which haven't been imported yet
files: {}, // Holds the imported parse trees
push: function (path, callback) {
var that = this;
this.queue.push(path);
//
// Import a file asynchronously
//
less.Parser.importer(path, this.paths, function (root) {
that.queue.splice(that.queue.indexOf(path), 1); // Remove the path from the queue
that.files[path] = root; // Store the root
callback(root);
if (that.queue.length === 0) { finish() } // Call `finish` if we're done importing
});
}
};
//
// Parse from a token, regexp or string, and move forward if match
//
function $(tok) {
var match, args, length, c, index, endIndex;
//
// Non-terminal
//
if (tok instanceof Function) {
return tok.call(parser.parsers);
//
// Terminal
//
// Either match a single character in the input,
// or match a regexp in the current chunk (chunk[j]).
//
} else if (typeof(tok) === 'string') {
match = input[i] === tok ? tok : null;
length = 1;
// 1. We move to the next chunk, if necessary.
// 2. Set the `lastIndex` to be relative
// to the current chunk, and try to match in it.
// 3. Make sure we matched at `index`. Because we use
// the /g flag, the match could be anywhere in the
// chunk. We have to make sure it's at our previous
// index, which we stored in [2].
//
} else {
if (i >= current + chunks[j].length &&
j < chunks.length - 1) { // 1.
current += chunks[j++].length;
}
tok.lastIndex = index = i - current; // 2.
match = tok.exec(chunks[j]);
if (match) {
length = match[0].length;
if (tok.lastIndex - length !== index) { return } // 3.
}
}
// The match is confirmed, add the match length to `i`,
// and consume any extra white-space characters (' ' || '\n')
// which come after that. The reason for this is that LeSS's
// grammar is mostly white-space insensitive.
//
if (match) {
i += length;
endIndex = current + chunks[j].length;
while (i <= endIndex) {
c = input.charCodeAt(i);
if (! (c === 32 || c === 10 || c === 9)) { break }
i++;
}
return match.length === 1 ? match[0] : match;
}
}
// Same as $(), but don't change the state of the parser,
// just return the match.
function peek(tok) {
var match;
if (typeof(tok) === 'string') {
return input[i] === tok;
} else {
tok.lastIndex = i;
if ((match = tok.exec(input)) &&
(tok.lastIndex - match[0].length === i)) {
return match;
}
}
}
this.env = env || {};
// The optimization level dictates the thoroughness of the parser,
// the lower the number, the less nodes it will create in the tree.
// This could matter for debugging, or if you want to access
// the individual nodes in the tree.
this.optimization = this.env.optimization || 2;
//
// The Parser
//
return parser = {
imports: imports,
//
// Parse an input string into an abstract syntax tree,
// call `callback` when done.
//
parse: function (str, callback) {
var root, start, end, zone, line, buff = [], c, error = null;
i = j = current = furthest = 0;
chunks = [];
input = str.replace(/\r\n/g, '\n');
inputLength = input.length;
// Split the input into chunks,
// Either delimited by /\n\n/ or
// delmited by '\n}' (see rationale above),
// depending on the level of optimization.
if (that.optimization > 0) {
if (that.optimization > 2) {
input = input.replace(/\/\*(?:[^*]|\*+[^\/*])*\*+\//g, '');
chunks = input.split(/^(?=\n)/mg);
} else {
for (var k = 0; k < input.length; k++) {
if ((c = input.charAt(k)) === '}' && input.charCodeAt(k - 1) === 10) {
chunks.push(buff.concat('}').join(''));
buff = [];
} else {
buff.push(c);
}
}
chunks.push(buff.join(''));
}
} else {
chunks = [input];
}
// Start with the primary rule.
// The whole syntax tree is held under a Ruleset node,
// with the `root` property set to true, so no `{}` are
// output. The callback is called when the input is parsed.
root = new(tree.Ruleset)([], $(this.parsers.primary));
root.root = true;
// If `i` is smaller than the `input.length - 1`,
// it means the parser wasn't able to parse the whole
// string, so we've got a parsing error.
//
// We try to extract a \n delimited string,
// showing the line where the parse error occured.
// We split it up into two parts (the part which parsed,
// and the part which didn't), so we can color them differently.
if (i < input.length - 1) {
i = furthest;
start = (function () {
for (var n = i; n > 0; n--) {
if (input[n] === '\n') { break }
}
return n;
})() + 1;
line = (input.slice(0, i).match(/\n/g) || "").length + 1;
end = input.slice(i).indexOf('\n') + i;
zone = stylize(input.slice(start, i), 'green') +
stylize(stylize(input[i], 'inverse') +
input.slice(i + 1, end),'yellow') + '\033[0m\n';
error = { name: "ParseError", message: "Parse Error on line " + line + ":\n" + zone };
}
if (this.imports.queue.length > 0) {
finish = function () { callback(error, root) };
} else {
callback(error, root);
}
},
//
// Here in, the parsing rules/functions
//
// The basic structure of the syntax tree generated is as follows:
//
// Ruleset -> Rule -> Value -> Expression -> Entity
//
// Here's some LESS code:
//
// .class {
// color: #fff;
// border: 1px solid #000;
// width: @w + 4px;
// > .child {...}
// }
//
// And here's what the parse tree might look like:
//
// Ruleset (Selector '.class', [
// Rule ("color", Value ([Expression [Color #fff]]))
// Rule ("border", Value ([Expression [Dimension 1px][Keyword "solid"][Color #000]]))
// Rule ("width", Value ([Expression [Operation "+" [Variable "@w"][Dimension 4px]]]))
// Ruleset (Selector [Element '>', '.child'], [...])
// ])
//
// In general, most rules will try to parse a token with the `$()` function, and if the return
// value is truly, will return a new node, of the relevant type. Sometimes, we need to check
// first, before parsing, that's when we use `peek()`.
//
parsers: {
//
// The `primary` rule is the *entry* and *exit* point of the parser.
// The rules here can appear at any level of the parse tree.
//
// The recursive nature of the grammar is an interplay between the `block`
// rule, which represents `{ ... }`, the `ruleset` rule, and this `primary` rule,
// as represented by this simplified grammar:
//
// primary → (ruleset | rule)+
// ruleset → selector+ block
// block → '{' primary '}'
//
// Only at one point is the primary rule not called from the
// block rule: at the root level.
//
primary: function () {
var node, root = [];
while (node = $(this.mixin.definition) || $(this.rule) || $(this.ruleset) ||
$(this.mixin.call) || $(this.comment) ||
$(/[\n\s]+/g) || $(this.directive)) {
root.push(node);
}
return root;
},
// We create a Comment node for CSS comments `/* */`,
// but keep the LeSS comments `//` silent, by just skipping
// over them.
comment: function () {
var comment;
if (input[i] !== '/') return;
if (comment = $(/\/\*(?:[^*]|\*+[^\/*])*\*+\/\n?/g)) {
return new(tree.Comment)(comment);
} else {
return $(/\/\/.*/g);
}
},
//
// Entities are tokens which can be found inside an Expression
//
entities: {
//
// A string, which supports escaping " and '
//
// "milky way" 'he\'s the one!'
//
quoted: function () {
var str;
if (input[i] !== '"' && input[i] !== "'") return;
if (str = $(/"((?:[^"\\\r\n]|\\.)*)"|'((?:[^'\\\r\n]|\\.)*)'/g)) {
return new(tree.Quoted)(str[0], str[1] || str[2]);
}
},
//
// A catch-all word, such as:
//
// black border-collapse
//
keyword: function () {
var k;
if (k = $(/[A-Za-z-]+/g)) { return new(tree.Keyword)(k) }
},
//
// A function call
//
// rgb(255, 0, 255)
//
// We also try to catch IE's `alpha()`, but let the `alpha` parser
// deal with the details.
//
// The arguments are parsed with the `entities.arguments` parser.
//
call: function () {
var name, args;
if (! (name = $(/([a-zA-Z0-9_-]+)\(/g))) return;
if (name[1].toLowerCase() === 'alpha') { return $(this.alpha) }
args = $(this.entities.arguments);
if (! $(')')) return;
if (name) { return new(tree.Call)(name[1], args) }
},
arguments: function () {
var args = [], arg;
while (arg = $(this.expression)) {
args.push(arg);
if (! $(',')) { break }
}
return args;
},
literal: function () {
return $(this.entities.dimension) ||
$(this.entities.color) ||
$(this.entities.quoted);
},
//
// Parse url() tokens
//
// We use a specific rule for urls, because they don't really behave like
// standard function calls. The difference is that the argument doesn't have
// to be enclosed within a string, so it can't be parsed as an Expression.
//
url: function () {
var value;
if (input[i] !== 'u' || !$(/url\(/g)) return;
value = $(this.entities.quoted) || $(/[-a-zA-Z0-9_%@$\/.&=:;#+?]+/g);
if (! $(')')) throw new(Error)("missing closing ) for url()");
return new(tree.URL)(value);
},
//
// A Variable entity, such as `@fink`, in
//
// width: @fink + 2px
//
// We use a different parser for variable definitions,
// see `parsers.variable`.
//
variable: function () {
var name;
if (input[i] === '@' && (name = $(/@[a-zA-Z0-9_-]+/g))) {
return new(tree.Variable)(name);
}
},
//
// A Hexadecimal color
//
// #4F3C2F
//
// `rgb` and `hsl` colors are parsed through the `entities.call` parser.
//
color: function () {
var rgb;
if (input[i] === '#' && (rgb = $(/#([a-fA-F0-9]{6}|[a-fA-F0-9]{3})/g))) {
return new(tree.Color)(rgb[1]);
}
},
//
// A Dimension, that is, a number and a unit
//
// 0.5em 95%
//
dimension: function () {
var value, c = input.charCodeAt(i);
if ((c > 57 || c < 45) || c === 47) return;
if (value = $(/(-?[0-9]*\.?[0-9]+)(px|%|em|pc|ex|in|deg|s|ms|pt|cm|mm)?/g)) {
return new(tree.Dimension)(value[1], value[2]);
}
}
},
//
// The variable part of a variable definition. Used in the `rule` parser
//
// @fink:
//
variable: function () {
var name;
if (input[i] === '@' && (name = $(/(@[a-zA-Z0-9_-]+)\s*:/g))) { return name[1] }
},
//
// A font size/line-height shorthand
//
// small/12px
//
// We need to peek first, or we'll match on keywords and dimensions
//
shorthand: function () {
var a, b;
if (! peek(/[@\w.-]+\/[@\w.-]+/g)) return;
if ((a = $(this.entity)) && $('/') && (b = $(this.entity))) {
return new(tree.Shorthand)(a, b);
}
},
//
// Mixins
//
mixin: {
//
// A Mixin call, with an optional argument list
//
// #mixins > .square(#fff);
// .rounded(4px, black);
// .button;
//
// The `while` loop is there because mixins can be
// namespaced, but we only support the child and descendant
// selector for now.
//
call: function () {
var elements = [], e, c, args;
while (e = $(/[#.]?[a-zA-Z0-9_-]+/g)) {
elements.push(new(tree.Element)(c, e));
c = $('>');
}
$('(') && (args = $(this.entities.arguments)) && $(')');
if (elements.length > 0 && ($(';') || peek('}'))) {
return new(tree.mixin.Call)(elements, args);
}
},
//
// A Mixin definition, with a list of parameters
//
// .rounded (@radius: 2px, @color) {
// ...
// }
//
// Until we have a finer grained state-machine, we have to
// do a look-ahead, to make sure we don't have a mixin call.
// See the `rule` function for more information.
//
// We start by matching `.rounded (`, and then proceed on to
// the argument list, which has optional default values.
// We store the parameters in `params`, with a `value` key,
// if there is a value, such as in the case of `@radius`.
//
// Once we've got our params list, and a closing `)`, we parse
// the `{...}` block.
//
definition: function () {
var name, params = [], match, ruleset, param, value;
if (input[i] !== '.' || peek(/[^{]*(;|})/g)) return;
if (match = $(/([#.][a-zA-Z0-9_-]+)\s*\(/g)) {
name = match[1];
while (param = $(/@[\w-]+/g)) {
if ($(':')) {
if (value = $(this.expression)) {
params.push({ name: param, value: value });
} else {
throw new(Error)("Expected value");
}
} else {
params.push({ name: param });
}
if (! $(',')) { break }
}
if (! $(')')) throw new(Error)("Expected )");
ruleset = $(this.block);
if (ruleset) {
return new(tree.mixin.Definition)(name, params, ruleset);
}
}
}
},
//
// Entities are the smallest recognized token,
// and can be found inside a rule's value.
//
entity: function () {
return $(this.entities.literal) || $(this.entities.variable) || $(this.entities.url) ||
$(this.entities.call) || $(this.entities.keyword);
},
//
// A Rule terminator. Note that we use `peek()` to check for '}',
// because the `block` rule will be expecting it, but we still need to make sure
// it's there, if ';' was ommitted.
//
end: function () {
return $(';') || peek('}');
},
//
// IE's alpha function
//
// alpha(opacity=88)
//
alpha: function () {
var value;
if (! $(/opacity=/gi)) return;
if (value = $(/[0-9]+/g) || $(this.entities.variable)) {
if (! $(')')) throw new(Error)("missing closing ) for alpha()");
return new(tree.Alpha)(value);
}
},
//
// A Selector Element
//
// div
// + h1
// #socks
// input[type="text"]
//
// Elements are the building blocks for Selectors,
// they are made out of a `Combinator` (see combinator rule),
// and an element name, such as a tag a class, or `*`.
//
element: function () {
var e, t;
c = $(this.combinator);
e = $(/[.#:]?[a-zA-Z0-9_-]+/g) || $('*') || $(this.attribute) || $(/\([^)@]+\)/g);
if (e) { return new(tree.Element)(c, e) }
},
//
// Combinators combine elements together, in a Selector.
//
// Because our parser isn't white-space sensitive, special care
// has to be taken, when parsing the descendant combinator, ` `,
// as it's an empty space. We have to check the previous character
// in the input, to see if it's a ` ` character. More info on how
// we deal with this in *combinator.js*.
//
combinator: function () {
var match;
if (match = $(/[+>~]/g) || $('&') || $(/::/g)) {
return new(tree.Combinator)(match);
} else {
return new(tree.Combinator)(input[i - 1] === " " ? " " : null);
}
},
//
// A CSS Selector
//
// .class > div + h1
// li a:hover
//
// Selectors are made out of one or more Elements, see above.
//
selector: function () {
var sel, e, elements = [], match;
while (e = $(this.element)) { elements.push(e) }
if (elements.length > 0) { return new(tree.Selector)(elements) }
},
tag: function () {
return $(/[a-zA-Z][a-zA-Z-]*[0-9]?/g) || $('*');
},
attribute: function () {
var attr = '', key, val, op;
if (! $('[')) return;
if (key = $(/[a-z]+/g) || $(this.entities.quoted)) {
if ((op = $(/[|~*$^]?=/g)) &&
(val = $(this.entities.quoted) || $(/[\w-]+/g))) {
attr = [key, op, val.toCSS ? val.toCSS() : val].join('');
} else { attr = key }
}
if (! $(']')) return;
if (attr) { return "[" + attr + "]" }
},
//
// The `block` rule is used by `ruleset` and `mixin.definition`.
// It's a wrapper around the `primary` rule, with added `{}`.
//
block: function () {
var content;
if ($('{') && (content = $(this.primary)) && $('}')) {
return content;
}
},
//
// div, .class, body > p {...}
//
ruleset: function () {
var selectors = [], s, rules, match, memo = i;
if (match = peek(/([a-z.#: _-]+)[\s\n]*\{/g)) {
i += match[0].length - 1;
selectors = [new(tree.Selector)([new(tree.Element)(null, match[1])])];
} else {
while (s = $(this.selector)) {
selectors.push(s);
if (! $(',')) { break }
}
if (s) $(this.comment);
}
if (selectors.length > 0 && (rules = $(this.block))) {
return new(tree.Ruleset)(selectors, rules);
} else {
// Backtrack
furthest = i;
i = memo;
}
},
rule: function () {
var value;
var memo = i;
if (name = $(this.property) || $(this.variable)) {
if ((name[0] != '@') && (match = peek(/([^@+\/*(;{}-]*);/g))) {
i += match[0].length - 1;
value = match[1];
} else if (name === "font") {
value = $(this.font);
} else {
value = $(this.value);
}
if ($(this.end)) {
return new(tree.Rule)(name, value);
} else {
furthest = i;
i = memo;
}
}
},
//
// An @import directive
//
// @import "lib";
//
// Depending on our environemnt, importing is done differently:
// In the browser, it's an XHR request, in Node, it would be a
// file-system operation. The function used for importing is
// stored in `import`, which we pass to the Import constructor.
//
"import": function () {
var path;
if ($(/@import\s+/g) &&
(path = $(this.entities.quoted) || $(this.entities.url)) &&
$(';')) {
return new(tree.Import)(path, imports);
}
},
//
// A CSS Directive
//
// @charset "utf-8";
//
directive: function () {
var name, value, rules, types;
if (input[i] !== '@') return;
if (value = $(this['import'])) {
return value;
} else if (name = $(/@media|@page/g)) {
types = $(/[a-z:, ]+/g).trim();
if (rules = $(this.block)) {
return new(tree.Directive)(name + " " + types, rules);
}
} else if (name = $(/@[-a-z]+/g)) {
if (name === '@font-face') {
if (rules = $(this.block)) {
return new(tree.Directive)(name, rules);
}
} else if ((value = $(this.entity)) && $(';')) {
return new(tree.Directive)(name, value);
}
}
},
font: function () {
var value = [], expression = [], weight, shorthand, font, e;
while (e = $(this.shorthand) || $(this.entity)) {
expression.push(e);
}
value.push(new(tree.Expression)(expression));
if ($(',')) {
while (e = $(this.expression)) {
value.push(e);
if (! $(',')) { break }
}
}
return new(tree.Value)(value, $(this.important));
},
//
// A Value is a comma-delimited list of Expressions
//
// font-family: Baskerville, Georgia, serif;
//
// In a Rule, a Value represents everything after the `:`,
// and before the `;`.
//
value: function () {
var e, expressions = [], important;
while (e = $(this.expression)) {
expressions.push(e);
if (! $(',')) { break }
}
important = $(this.important);
if (expressions.length > 0) {
return new(tree.Value)(expressions, important);
}
},
important: function () {
return $(/!\s*important/g);
},
sub: function () {
var e;
if ($('(') && (e = $(this.expression)) && $(')')) {
return e;
}
},
multiplication: function () {
var m, a, op;
if (m = $(this.operand)) {
if ((op = $(/[\/*]/g)) && (a = $(this.multiplication))) {
return new(tree.Operation)(op, [m, a]);
} else {
return m;
}
}
},
addition: function () {
var m, a, op;
if (m = $(this.multiplication)) {
if ((op = $(/[-+]\s+/g) || (input[i - 1] != ' ' && $(/[-+]/g))) &&
(a = $(this.addition))) {
return new(tree.Operation)(op, [m, a]);
} else {
return m;
}
}
},
//
// An operand is anything that can be part of an operation,
// such as a Color, or a Variable
//
operand: function () {
return $(this.sub) || $(this.entities.dimension) ||
$(this.entities.color) || $(this.entities.variable);
},
//
// Expressions either represent mathematical operations,
// or white-space delimited Entities.
//
// 1px solid black
// @var * 2
//
expression: function () {
var e, delim, entities = [], d;
while (e = $(this.addition) || $(this.entity)) {
entities.push(e);
}
if (entities.length > 0) {
return new(tree.Expression)(entities);
}
},
property: function () {
var name;
if (name = $(/(\*?-?[-a-z]+)\s*:/g)) {
return name[1];
}
}
}
};
};
less.Parser.importer = null;
// Stylize a string
function stylize(str, style) {
var styles = {
'bold' : [1, 22],
'inverse' : [7, 27],
'underline' : [4, 24],
'yellow' : [33, 39],
'green' : [32, 39],
'red' : [31, 39]
};
return '\033[' + styles[style][0] + 'm' + str +
'\033[' + styles[style][1] + 'm';
}