604 lines
14 KiB
JavaScript
604 lines
14 KiB
JavaScript
'use strict'
|
|
|
|
let Declaration = require('./declaration')
|
|
let tokenizer = require('./tokenize')
|
|
let Comment = require('./comment')
|
|
let AtRule = require('./at-rule')
|
|
let Root = require('./root')
|
|
let Rule = require('./rule')
|
|
|
|
const SAFE_COMMENT_NEIGHBOR = {
|
|
empty: true,
|
|
space: true
|
|
}
|
|
|
|
function findLastWithPosition(tokens) {
|
|
for (let i = tokens.length - 1; i >= 0; i--) {
|
|
let token = tokens[i]
|
|
let pos = token[3] || token[2]
|
|
if (pos) return pos
|
|
}
|
|
}
|
|
|
|
class Parser {
|
|
constructor(input) {
|
|
this.input = input
|
|
|
|
this.root = new Root()
|
|
this.current = this.root
|
|
this.spaces = ''
|
|
this.semicolon = false
|
|
this.customProperty = false
|
|
|
|
this.createTokenizer()
|
|
this.root.source = { input, start: { column: 1, line: 1, offset: 0 } }
|
|
}
|
|
|
|
atrule(token) {
|
|
let node = new AtRule()
|
|
node.name = token[1].slice(1)
|
|
if (node.name === '') {
|
|
this.unnamedAtrule(node, token)
|
|
}
|
|
this.init(node, token[2])
|
|
|
|
let type
|
|
let prev
|
|
let shift
|
|
let last = false
|
|
let open = false
|
|
let params = []
|
|
let brackets = []
|
|
|
|
while (!this.tokenizer.endOfFile()) {
|
|
token = this.tokenizer.nextToken()
|
|
type = token[0]
|
|
|
|
if (type === '(' || type === '[') {
|
|
brackets.push(type === '(' ? ')' : ']')
|
|
} else if (type === '{' && brackets.length > 0) {
|
|
brackets.push('}')
|
|
} else if (type === brackets[brackets.length - 1]) {
|
|
brackets.pop()
|
|
}
|
|
|
|
if (brackets.length === 0) {
|
|
if (type === ';') {
|
|
node.source.end = this.getPosition(token[2])
|
|
this.semicolon = true
|
|
break
|
|
} else if (type === '{') {
|
|
open = true
|
|
break
|
|
} else if (type === '}') {
|
|
if (params.length > 0) {
|
|
shift = params.length - 1
|
|
prev = params[shift]
|
|
while (prev && prev[0] === 'space') {
|
|
prev = params[--shift]
|
|
}
|
|
if (prev) {
|
|
node.source.end = this.getPosition(prev[3] || prev[2])
|
|
}
|
|
}
|
|
this.end(token)
|
|
break
|
|
} else {
|
|
params.push(token)
|
|
}
|
|
} else {
|
|
params.push(token)
|
|
}
|
|
|
|
if (this.tokenizer.endOfFile()) {
|
|
last = true
|
|
break
|
|
}
|
|
}
|
|
|
|
node.raws.between = this.spacesAndCommentsFromEnd(params)
|
|
if (params.length) {
|
|
node.raws.afterName = this.spacesAndCommentsFromStart(params)
|
|
this.raw(node, 'params', params)
|
|
if (last) {
|
|
token = params[params.length - 1]
|
|
node.source.end = this.getPosition(token[3] || token[2])
|
|
this.spaces = node.raws.between
|
|
node.raws.between = ''
|
|
}
|
|
} else {
|
|
node.raws.afterName = ''
|
|
node.params = ''
|
|
}
|
|
|
|
if (open) {
|
|
node.nodes = []
|
|
this.current = node
|
|
}
|
|
}
|
|
|
|
checkMissedSemicolon(tokens) {
|
|
let colon = this.colon(tokens)
|
|
if (colon === false) return
|
|
|
|
let founded = 0
|
|
let token
|
|
for (let j = colon - 1; j >= 0; j--) {
|
|
token = tokens[j]
|
|
if (token[0] !== 'space') {
|
|
founded += 1
|
|
if (founded === 2) break
|
|
}
|
|
}
|
|
// If the token is a word, e.g. `!important`, `red` or any other valid property's value.
|
|
// Then we need to return the colon after that word token. [3] is the "end" colon of that word.
|
|
// And because we need it after that one we do +1 to get the next one.
|
|
throw this.input.error(
|
|
'Missed semicolon',
|
|
token[0] === 'word' ? token[3] + 1 : token[2]
|
|
)
|
|
}
|
|
|
|
colon(tokens) {
|
|
let brackets = 0
|
|
let token, type, prev
|
|
for (let [i, element] of tokens.entries()) {
|
|
token = element
|
|
type = token[0]
|
|
|
|
if (type === '(') {
|
|
brackets += 1
|
|
}
|
|
if (type === ')') {
|
|
brackets -= 1
|
|
}
|
|
if (brackets === 0 && type === ':') {
|
|
if (!prev) {
|
|
this.doubleColon(token)
|
|
} else if (prev[0] === 'word' && prev[1] === 'progid') {
|
|
continue
|
|
} else {
|
|
return i
|
|
}
|
|
}
|
|
|
|
prev = token
|
|
}
|
|
return false
|
|
}
|
|
|
|
comment(token) {
|
|
let node = new Comment()
|
|
this.init(node, token[2])
|
|
node.source.end = this.getPosition(token[3] || token[2])
|
|
|
|
let text = token[1].slice(2, -2)
|
|
if (/^\s*$/.test(text)) {
|
|
node.text = ''
|
|
node.raws.left = text
|
|
node.raws.right = ''
|
|
} else {
|
|
let match = text.match(/^(\s*)([^]*\S)(\s*)$/)
|
|
node.text = match[2]
|
|
node.raws.left = match[1]
|
|
node.raws.right = match[3]
|
|
}
|
|
}
|
|
|
|
createTokenizer() {
|
|
this.tokenizer = tokenizer(this.input)
|
|
}
|
|
|
|
decl(tokens, customProperty) {
|
|
let node = new Declaration()
|
|
this.init(node, tokens[0][2])
|
|
|
|
let last = tokens[tokens.length - 1]
|
|
if (last[0] === ';') {
|
|
this.semicolon = true
|
|
tokens.pop()
|
|
}
|
|
|
|
node.source.end = this.getPosition(
|
|
last[3] || last[2] || findLastWithPosition(tokens)
|
|
)
|
|
|
|
while (tokens[0][0] !== 'word') {
|
|
if (tokens.length === 1) this.unknownWord(tokens)
|
|
node.raws.before += tokens.shift()[1]
|
|
}
|
|
node.source.start = this.getPosition(tokens[0][2])
|
|
|
|
node.prop = ''
|
|
while (tokens.length) {
|
|
let type = tokens[0][0]
|
|
if (type === ':' || type === 'space' || type === 'comment') {
|
|
break
|
|
}
|
|
node.prop += tokens.shift()[1]
|
|
}
|
|
|
|
node.raws.between = ''
|
|
|
|
let token
|
|
while (tokens.length) {
|
|
token = tokens.shift()
|
|
|
|
if (token[0] === ':') {
|
|
node.raws.between += token[1]
|
|
break
|
|
} else {
|
|
if (token[0] === 'word' && /\w/.test(token[1])) {
|
|
this.unknownWord([token])
|
|
}
|
|
node.raws.between += token[1]
|
|
}
|
|
}
|
|
|
|
if (node.prop[0] === '_' || node.prop[0] === '*') {
|
|
node.raws.before += node.prop[0]
|
|
node.prop = node.prop.slice(1)
|
|
}
|
|
|
|
let firstSpaces = []
|
|
let next
|
|
while (tokens.length) {
|
|
next = tokens[0][0]
|
|
if (next !== 'space' && next !== 'comment') break
|
|
firstSpaces.push(tokens.shift())
|
|
}
|
|
|
|
this.precheckMissedSemicolon(tokens)
|
|
|
|
for (let i = tokens.length - 1; i >= 0; i--) {
|
|
token = tokens[i]
|
|
if (token[1].toLowerCase() === '!important') {
|
|
node.important = true
|
|
let string = this.stringFrom(tokens, i)
|
|
string = this.spacesFromEnd(tokens) + string
|
|
if (string !== ' !important') node.raws.important = string
|
|
break
|
|
} else if (token[1].toLowerCase() === 'important') {
|
|
let cache = tokens.slice(0)
|
|
let str = ''
|
|
for (let j = i; j > 0; j--) {
|
|
let type = cache[j][0]
|
|
if (str.trim().indexOf('!') === 0 && type !== 'space') {
|
|
break
|
|
}
|
|
str = cache.pop()[1] + str
|
|
}
|
|
if (str.trim().indexOf('!') === 0) {
|
|
node.important = true
|
|
node.raws.important = str
|
|
tokens = cache
|
|
}
|
|
}
|
|
|
|
if (token[0] !== 'space' && token[0] !== 'comment') {
|
|
break
|
|
}
|
|
}
|
|
|
|
let hasWord = tokens.some(i => i[0] !== 'space' && i[0] !== 'comment')
|
|
|
|
if (hasWord) {
|
|
node.raws.between += firstSpaces.map(i => i[1]).join('')
|
|
firstSpaces = []
|
|
}
|
|
this.raw(node, 'value', firstSpaces.concat(tokens), customProperty)
|
|
|
|
if (node.value.includes(':') && !customProperty) {
|
|
this.checkMissedSemicolon(tokens)
|
|
}
|
|
}
|
|
|
|
doubleColon(token) {
|
|
throw this.input.error(
|
|
'Double colon',
|
|
{ offset: token[2] },
|
|
{ offset: token[2] + token[1].length }
|
|
)
|
|
}
|
|
|
|
emptyRule(token) {
|
|
let node = new Rule()
|
|
this.init(node, token[2])
|
|
node.selector = ''
|
|
node.raws.between = ''
|
|
this.current = node
|
|
}
|
|
|
|
end(token) {
|
|
if (this.current.nodes && this.current.nodes.length) {
|
|
this.current.raws.semicolon = this.semicolon
|
|
}
|
|
this.semicolon = false
|
|
|
|
this.current.raws.after = (this.current.raws.after || '') + this.spaces
|
|
this.spaces = ''
|
|
|
|
if (this.current.parent) {
|
|
this.current.source.end = this.getPosition(token[2])
|
|
this.current = this.current.parent
|
|
} else {
|
|
this.unexpectedClose(token)
|
|
}
|
|
}
|
|
|
|
endFile() {
|
|
if (this.current.parent) this.unclosedBlock()
|
|
if (this.current.nodes && this.current.nodes.length) {
|
|
this.current.raws.semicolon = this.semicolon
|
|
}
|
|
this.current.raws.after = (this.current.raws.after || '') + this.spaces
|
|
this.root.source.end = this.getPosition(this.tokenizer.position())
|
|
}
|
|
|
|
freeSemicolon(token) {
|
|
this.spaces += token[1]
|
|
if (this.current.nodes) {
|
|
let prev = this.current.nodes[this.current.nodes.length - 1]
|
|
if (prev && prev.type === 'rule' && !prev.raws.ownSemicolon) {
|
|
prev.raws.ownSemicolon = this.spaces
|
|
this.spaces = ''
|
|
}
|
|
}
|
|
}
|
|
|
|
// Helpers
|
|
|
|
getPosition(offset) {
|
|
let pos = this.input.fromOffset(offset)
|
|
return {
|
|
column: pos.col,
|
|
line: pos.line,
|
|
offset
|
|
}
|
|
}
|
|
|
|
init(node, offset) {
|
|
this.current.push(node)
|
|
node.source = {
|
|
input: this.input,
|
|
start: this.getPosition(offset)
|
|
}
|
|
node.raws.before = this.spaces
|
|
this.spaces = ''
|
|
if (node.type !== 'comment') this.semicolon = false
|
|
}
|
|
|
|
other(start) {
|
|
let end = false
|
|
let type = null
|
|
let colon = false
|
|
let bracket = null
|
|
let brackets = []
|
|
let customProperty = start[1].startsWith('--')
|
|
|
|
let tokens = []
|
|
let token = start
|
|
while (token) {
|
|
type = token[0]
|
|
tokens.push(token)
|
|
|
|
if (type === '(' || type === '[') {
|
|
if (!bracket) bracket = token
|
|
brackets.push(type === '(' ? ')' : ']')
|
|
} else if (customProperty && colon && type === '{') {
|
|
if (!bracket) bracket = token
|
|
brackets.push('}')
|
|
} else if (brackets.length === 0) {
|
|
if (type === ';') {
|
|
if (colon) {
|
|
this.decl(tokens, customProperty)
|
|
return
|
|
} else {
|
|
break
|
|
}
|
|
} else if (type === '{') {
|
|
this.rule(tokens)
|
|
return
|
|
} else if (type === '}') {
|
|
this.tokenizer.back(tokens.pop())
|
|
end = true
|
|
break
|
|
} else if (type === ':') {
|
|
colon = true
|
|
}
|
|
} else if (type === brackets[brackets.length - 1]) {
|
|
brackets.pop()
|
|
if (brackets.length === 0) bracket = null
|
|
}
|
|
|
|
token = this.tokenizer.nextToken()
|
|
}
|
|
|
|
if (this.tokenizer.endOfFile()) end = true
|
|
if (brackets.length > 0) this.unclosedBracket(bracket)
|
|
|
|
if (end && colon) {
|
|
if (!customProperty) {
|
|
while (tokens.length) {
|
|
token = tokens[tokens.length - 1][0]
|
|
if (token !== 'space' && token !== 'comment') break
|
|
this.tokenizer.back(tokens.pop())
|
|
}
|
|
}
|
|
this.decl(tokens, customProperty)
|
|
} else {
|
|
this.unknownWord(tokens)
|
|
}
|
|
}
|
|
|
|
parse() {
|
|
let token
|
|
while (!this.tokenizer.endOfFile()) {
|
|
token = this.tokenizer.nextToken()
|
|
|
|
switch (token[0]) {
|
|
case 'space':
|
|
this.spaces += token[1]
|
|
break
|
|
|
|
case ';':
|
|
this.freeSemicolon(token)
|
|
break
|
|
|
|
case '}':
|
|
this.end(token)
|
|
break
|
|
|
|
case 'comment':
|
|
this.comment(token)
|
|
break
|
|
|
|
case 'at-word':
|
|
this.atrule(token)
|
|
break
|
|
|
|
case '{':
|
|
this.emptyRule(token)
|
|
break
|
|
|
|
default:
|
|
this.other(token)
|
|
break
|
|
}
|
|
}
|
|
this.endFile()
|
|
}
|
|
|
|
precheckMissedSemicolon(/* tokens */) {
|
|
// Hook for Safe Parser
|
|
}
|
|
|
|
raw(node, prop, tokens, customProperty) {
|
|
let token, type
|
|
let length = tokens.length
|
|
let value = ''
|
|
let clean = true
|
|
let next, prev
|
|
|
|
for (let i = 0; i < length; i += 1) {
|
|
token = tokens[i]
|
|
type = token[0]
|
|
if (type === 'space' && i === length - 1 && !customProperty) {
|
|
clean = false
|
|
} else if (type === 'comment') {
|
|
prev = tokens[i - 1] ? tokens[i - 1][0] : 'empty'
|
|
next = tokens[i + 1] ? tokens[i + 1][0] : 'empty'
|
|
if (!SAFE_COMMENT_NEIGHBOR[prev] && !SAFE_COMMENT_NEIGHBOR[next]) {
|
|
if (value.slice(-1) === ',') {
|
|
clean = false
|
|
} else {
|
|
value += token[1]
|
|
}
|
|
} else {
|
|
clean = false
|
|
}
|
|
} else {
|
|
value += token[1]
|
|
}
|
|
}
|
|
if (!clean) {
|
|
let raw = tokens.reduce((all, i) => all + i[1], '')
|
|
node.raws[prop] = { raw, value }
|
|
}
|
|
node[prop] = value
|
|
}
|
|
|
|
rule(tokens) {
|
|
tokens.pop()
|
|
|
|
let node = new Rule()
|
|
this.init(node, tokens[0][2])
|
|
|
|
node.raws.between = this.spacesAndCommentsFromEnd(tokens)
|
|
this.raw(node, 'selector', tokens)
|
|
this.current = node
|
|
}
|
|
|
|
spacesAndCommentsFromEnd(tokens) {
|
|
let lastTokenType
|
|
let spaces = ''
|
|
while (tokens.length) {
|
|
lastTokenType = tokens[tokens.length - 1][0]
|
|
if (lastTokenType !== 'space' && lastTokenType !== 'comment') break
|
|
spaces = tokens.pop()[1] + spaces
|
|
}
|
|
return spaces
|
|
}
|
|
|
|
// Errors
|
|
|
|
spacesAndCommentsFromStart(tokens) {
|
|
let next
|
|
let spaces = ''
|
|
while (tokens.length) {
|
|
next = tokens[0][0]
|
|
if (next !== 'space' && next !== 'comment') break
|
|
spaces += tokens.shift()[1]
|
|
}
|
|
return spaces
|
|
}
|
|
|
|
spacesFromEnd(tokens) {
|
|
let lastTokenType
|
|
let spaces = ''
|
|
while (tokens.length) {
|
|
lastTokenType = tokens[tokens.length - 1][0]
|
|
if (lastTokenType !== 'space') break
|
|
spaces = tokens.pop()[1] + spaces
|
|
}
|
|
return spaces
|
|
}
|
|
|
|
stringFrom(tokens, from) {
|
|
let result = ''
|
|
for (let i = from; i < tokens.length; i++) {
|
|
result += tokens[i][1]
|
|
}
|
|
tokens.splice(from, tokens.length - from)
|
|
return result
|
|
}
|
|
|
|
unclosedBlock() {
|
|
let pos = this.current.source.start
|
|
throw this.input.error('Unclosed block', pos.line, pos.column)
|
|
}
|
|
|
|
unclosedBracket(bracket) {
|
|
throw this.input.error(
|
|
'Unclosed bracket',
|
|
{ offset: bracket[2] },
|
|
{ offset: bracket[2] + 1 }
|
|
)
|
|
}
|
|
|
|
unexpectedClose(token) {
|
|
throw this.input.error(
|
|
'Unexpected }',
|
|
{ offset: token[2] },
|
|
{ offset: token[2] + 1 }
|
|
)
|
|
}
|
|
|
|
unknownWord(tokens) {
|
|
throw this.input.error(
|
|
'Unknown word',
|
|
{ offset: tokens[0][2] },
|
|
{ offset: tokens[0][2] + tokens[0][1].length }
|
|
)
|
|
}
|
|
|
|
unnamedAtrule(node, token) {
|
|
throw this.input.error(
|
|
'At-rule without name',
|
|
{ offset: token[2] },
|
|
{ offset: token[2] + token[1].length }
|
|
)
|
|
}
|
|
}
|
|
|
|
module.exports = Parser
|