'use strict'; const constants = require('./constants'); const utils = require('./utils'); /** * Constants */ const { MAX_LENGTH, POSIX_REGEX_SOURCE, REGEX_NON_SPECIAL_CHARS, REGEX_SPECIAL_CHARS_BACKREF, REPLACEMENTS } = constants; /** * Helpers */ const expandRange = (args, options) => { if (typeof options.expandRange === 'function') { return options.expandRange(...args, options); } args.sort(); const value = `[${args.join('-')}]`; try { /* eslint-disable-next-line no-new */ new RegExp(value); } catch (ex) { return args.map(v => utils.escapeRegex(v)).join('..'); } return value; }; /** * Create the message for a syntax error */ const syntaxError = (type, char) => { return `Missing ${type}: "${char}" - use "\\\\${char}" to match literal characters`; }; /** * Parse the given input string. * @param {String} input * @param {Object} options * @return {Object} */ const parse = (input, options) => { if (typeof input !== 'string') { throw new TypeError('Expected a string'); } input = REPLACEMENTS[input] || input; const opts = { ...options }; const max = typeof opts.maxLength === 'number' ? Math.min(MAX_LENGTH, opts.maxLength) : MAX_LENGTH; let len = input.length; if (len > max) { throw new SyntaxError(`Input length: ${len}, exceeds maximum allowed length: ${max}`); } const bos = { type: 'bos', value: '', output: opts.prepend || '' }; const tokens = [bos]; const capture = opts.capture ? '' : '?:'; const win32 = utils.isWindows(options); // create constants based on platform, for windows or posix const PLATFORM_CHARS = constants.globChars(win32); const EXTGLOB_CHARS = constants.extglobChars(PLATFORM_CHARS); const { DOT_LITERAL, PLUS_LITERAL, SLASH_LITERAL, ONE_CHAR, DOTS_SLASH, NO_DOT, NO_DOT_SLASH, NO_DOTS_SLASH, QMARK, QMARK_NO_DOT, STAR, START_ANCHOR } = PLATFORM_CHARS; const globstar = opts => { return `(${capture}(?:(?!${START_ANCHOR}${opts.dot ? DOTS_SLASH : DOT_LITERAL}).)*?)`; }; const nodot = opts.dot ? '' : NO_DOT; const qmarkNoDot = opts.dot ? QMARK : QMARK_NO_DOT; let star = opts.bash === true ? globstar(opts) : STAR; if (opts.capture) { star = `(${star})`; } // minimatch options support if (typeof opts.noext === 'boolean') { opts.noextglob = opts.noext; } const state = { input, index: -1, start: 0, dot: opts.dot === true, consumed: '', output: '', prefix: '', backtrack: false, negated: false, brackets: 0, braces: 0, parens: 0, quotes: 0, globstar: false, tokens }; input = utils.removePrefix(input, state); len = input.length; const extglobs = []; const braces = []; const stack = []; let prev = bos; let value; /** * Tokenizing helpers */ const eos = () => state.index === len - 1; const peek = state.peek = (n = 1) => input[state.index + n]; const advance = state.advance = () => input[++state.index] || ''; const remaining = () => input.slice(state.index + 1); const consume = (value = '', num = 0) => { state.consumed += value; state.index += num; }; const append = token => { state.output += token.output != null ? token.output : token.value; consume(token.value); }; const negate = () => { let count = 1; while (peek() === '!' && (peek(2) !== '(' || peek(3) === '?')) { advance(); state.start++; count++; } if (count % 2 === 0) { return false; } state.negated = true; state.start++; return true; }; const increment = type => { state[type]++; stack.push(type); }; const decrement = type => { state[type]--; stack.pop(); }; /** * Push tokens onto the tokens array. This helper speeds up * tokenizing by 1) helping us avoid backtracking as much as possible, * and 2) helping us avoid creating extra tokens when consecutive * characters are plain text. This improves performance and simplifies * lookbehinds. */ const push = tok => { if (prev.type === 'globstar') { const isBrace = state.braces > 0 && (tok.type === 'comma' || tok.type === 'brace'); const isExtglob = tok.extglob === true || (extglobs.length && (tok.type === 'pipe' || tok.type === 'paren')); if (tok.type !== 'slash' && tok.type !== 'paren' && !isBrace && !isExtglob) { state.output = state.output.slice(0, -prev.output.length); prev.type = 'star'; prev.value = '*'; prev.output = star; state.output += prev.output; } } if (extglobs.length && tok.type !== 'paren') { extglobs[extglobs.length - 1].inner += tok.value; } if (tok.value || tok.output) append(tok); if (prev && prev.type === 'text' && tok.type === 'text') { prev.value += tok.value; prev.output = (prev.output || '') + tok.value; return; } tok.prev = prev; tokens.push(tok); prev = tok; }; const extglobOpen = (type, value) => { const token = { ...EXTGLOB_CHARS[value], conditions: 1, inner: '' }; token.prev = prev; token.parens = state.parens; token.output = state.output; const output = (opts.capture ? '(' : '') + token.open; increment('parens'); push({ type, value, output: state.output ? '' : ONE_CHAR }); push({ type: 'paren', extglob: true, value: advance(), output }); extglobs.push(token); }; const extglobClose = token => { let output = token.close + (opts.capture ? ')' : ''); let rest; if (token.type === 'negate') { let extglobStar = star; if (token.inner && token.inner.length > 1 && token.inner.includes('/')) { extglobStar = globstar(opts); } if (extglobStar !== star || eos() || /^\)+$/.test(remaining())) { output = token.close = `)$))${extglobStar}`; } if (token.inner.includes('*') && (rest = remaining()) && /^\.[^\\/.]+$/.test(rest)) { output = token.close = `)${rest})${extglobStar})`; } if (token.prev.type === 'bos') { state.negatedExtglob = true; } } push({ type: 'paren', extglob: true, value, output }); decrement('parens'); }; /** * Fast paths */ if (opts.fastpaths !== false && !/(^[*!]|[/()[\]{}"])/.test(input)) { let backslashes = false; let output = input.replace(REGEX_SPECIAL_CHARS_BACKREF, (m, esc, chars, first, rest, index) => { if (first === '\\') { backslashes = true; return m; } if (first === '?') { if (esc) { return esc + first + (rest ? QMARK.repeat(rest.length) : ''); } if (index === 0) { return qmarkNoDot + (rest ? QMARK.repeat(rest.length) : ''); } return QMARK.repeat(chars.length); } if (first === '.') { return DOT_LITERAL.repeat(chars.length); } if (first === '*') { if (esc) { return esc + first + (rest ? star : ''); } return star; } return esc ? m : `\\${m}`; }); if (backslashes === true) { if (opts.unescape === true) { output = output.replace(/\\/g, ''); } else { output = output.replace(/\\+/g, m => { return m.length % 2 === 0 ? '\\\\' : (m ? '\\' : ''); }); } } if (output === input && opts.contains === true) { state.output = input; return state; } state.output = utils.wrapOutput(output, state, options); return state; } /** * Tokenize input until we reach end-of-string */ while (!eos()) { value = advance(); if (value === '\u0000') { continue; } /** * Escaped characters */ if (value === '\\') { const next = peek(); if (next === '/' && opts.bash !== true) { continue; } if (next === '.' || next === ';') { continue; } if (!next) { value += '\\'; push({ type: 'text', value }); continue; } // collapse slashes to reduce potential for exploits const match = /^\\+/.exec(remaining()); let slashes = 0; if (match && match[0].length > 2) { slashes = match[0].length; state.index += slashes; if (slashes % 2 !== 0) { value += '\\'; } } if (opts.unescape === true) { value = advance(); } else { value += advance(); } if (state.brackets === 0) { push({ type: 'text', value }); continue; } } /** * If we're inside a regex character class, continue * until we reach the closing bracket. */ if (state.brackets > 0 && (value !== ']' || prev.value === '[' || prev.value === '[^')) { if (opts.posix !== false && value === ':') { const inner = prev.value.slice(1); if (inner.includes('[')) { prev.posix = true; if (inner.includes(':')) { const idx = prev.value.lastIndexOf('['); const pre = prev.value.slice(0, idx); const rest = prev.value.slice(idx + 2); const posix = POSIX_REGEX_SOURCE[rest]; if (posix) { prev.value = pre + posix; state.backtrack = true; advance(); if (!bos.output && tokens.indexOf(prev) === 1) { bos.output = ONE_CHAR; } continue; } } } } if ((value === '[' && peek() !== ':') || (value === '-' && peek() === ']')) { value = `\\${value}`; } if (value === ']' && (prev.value === '[' || prev.value === '[^')) { value = `\\${value}`; } if (opts.posix === true && value === '!' && prev.value === '[') { value = '^'; } prev.value += value; append({ value }); continue; } /** * If we're inside a quoted string, continue * until we reach the closing double quote. */ if (state.quotes === 1 && value !== '"') { value = utils.escapeRegex(value); prev.value += value; append({ value }); continue; } /** * Double quotes */ if (value === '"') { state.quotes = state.quotes === 1 ? 0 : 1; if (opts.keepQuotes === true) { push({ type: 'text', value }); } continue; } /** * Parentheses */ if (value === '(') { increment('parens'); push({ type: 'paren', value }); continue; } if (value === ')') { if (state.parens === 0 && opts.strictBrackets === true) { throw new SyntaxError(syntaxError('opening', '(')); } const extglob = extglobs[extglobs.length - 1]; if (extglob && state.parens === extglob.parens + 1) { extglobClose(extglobs.pop()); continue; } push({ type: 'paren', value, output: state.parens ? ')' : '\\)' }); decrement('parens'); continue; } /** * Square brackets */ if (value === '[') { if (opts.nobracket === true || !remaining().includes(']')) { if (opts.nobracket !== true && opts.strictBrackets === true) { throw new SyntaxError(syntaxError('closing', ']')); } value = `\\${value}`; } else { increment('brackets'); } push({ type: 'bracket', value }); continue; } if (value === ']') { if (opts.nobracket === true || (prev && prev.type === 'bracket' && prev.value.length === 1)) { push({ type: 'text', value, output: `\\${value}` }); continue; } if (state.brackets === 0) { if (opts.strictBrackets === true) { throw new SyntaxError(syntaxError('opening', '[')); } push({ type: 'text', value, output: `\\${value}` }); continue; } decrement('brackets'); const prevValue = prev.value.slice(1); if (prev.posix !== true && prevValue[0] === '^' && !prevValue.includes('/')) { value = `/${value}`; } prev.value += value; append({ value }); // when literal brackets are explicitly disabled // assume we should match with a regex character class if (opts.literalBrackets === false || utils.hasRegexChars(prevValue)) { continue; } const escaped = utils.escapeRegex(prev.value); state.output = state.output.slice(0, -prev.value.length); // when literal brackets are explicitly enabled // assume we should escape the brackets to match literal characters if (opts.literalBrackets === true) { state.output += escaped; prev.value = escaped; continue; } // when the user specifies nothing, try to match both prev.value = `(${capture}${escaped}|${prev.value})`; state.output += prev.value; continue; } /** * Braces */ if (value === '{' && opts.nobrace !== true) { increment('braces'); const open = { type: 'brace', value, output: '(', outputIndex: state.output.length, tokensIndex: state.tokens.length }; braces.push(open); push(open); continue; } if (value === '}') { const brace = braces[braces.length - 1]; if (opts.nobrace === true || !brace) { push({ type: 'text', value, output: value }); continue; } let output = ')'; if (brace.dots === true) { const arr = tokens.slice(); const range = []; for (let i = arr.length - 1; i >= 0; i--) { tokens.pop(); if (arr[i].type === 'brace') { break; } if (arr[i].type !== 'dots') { range.unshift(arr[i].value); } } output = expandRange(range, opts); state.backtrack = true; } if (brace.comma !== true && brace.dots !== true) { const out = state.output.slice(0, brace.outputIndex); const toks = state.tokens.slice(brace.tokensIndex); brace.value = brace.output = '\\{'; value = output = '\\}'; state.output = out; for (const t of toks) { state.output += (t.output || t.value); } } push({ type: 'brace', value, output }); decrement('braces'); braces.pop(); continue; } /** * Pipes */ if (value === '|') { if (extglobs.length > 0) { extglobs[extglobs.length - 1].conditions++; } push({ type: 'text', value }); continue; } /** * Commas */ if (value === ',') { let output = value; const brace = braces[braces.length - 1]; if (brace && stack[stack.length - 1] === 'braces') { brace.comma = true; output = '|'; } push({ type: 'comma', value, output }); continue; } /** * Slashes */ if (value === '/') { // if the beginning of the glob is "./", advance the start // to the current index, and don't add the "./" characters // to the state. This greatly simplifies lookbehinds when // checking for BOS characters like "!" and "." (not "./") if (prev.type === 'dot' && state.index === state.start + 1) { state.start = state.index + 1; state.consumed = ''; state.output = ''; tokens.pop(); prev = bos; // reset "prev" to the first token continue; } push({ type: 'slash', value, output: SLASH_LITERAL }); continue; } /** * Dots */ if (value === '.') { if (state.braces > 0 && prev.type === 'dot') { if (prev.value === '.') prev.output = DOT_LITERAL; const brace = braces[braces.length - 1]; prev.type = 'dots'; prev.output += value; prev.value += value; brace.dots = true; continue; } if ((state.braces + state.parens) === 0 && prev.type !== 'bos' && prev.type !== 'slash') { push({ type: 'text', value, output: DOT_LITERAL }); continue; } push({ type: 'dot', value, output: DOT_LITERAL }); continue; } /** * Question marks */ if (value === '?') { const isGroup = prev && prev.value === '('; if (!isGroup && opts.noextglob !== true && peek() === '(' && peek(2) !== '?') { extglobOpen('qmark', value); continue; } if (prev && prev.type === 'paren') { const next = peek(); let output = value; if (next === '<' && !utils.supportsLookbehinds()) { throw new Error('Node.js v10 or higher is required for regex lookbehinds'); } if ((prev.value === '(' && !/[!=<:]/.test(next)) || (next === '<' && !/<([!=]|\w+>)/.test(remaining()))) { output = `\\${value}`; } push({ type: 'text', value, output }); continue; } if (opts.dot !== true && (prev.type === 'slash' || prev.type === 'bos')) { push({ type: 'qmark', value, output: QMARK_NO_DOT }); continue; } push({ type: 'qmark', value, output: QMARK }); continue; } /** * Exclamation */ if (value === '!') { if (opts.noextglob !== true && peek() === '(') { if (peek(2) !== '?' || !/[!=<:]/.test(peek(3))) { extglobOpen('negate', value); continue; } } if (opts.nonegate !== true && state.index === 0) { negate(); continue; } } /** * Plus */ if (value === '+') { if (opts.noextglob !== true && peek() === '(' && peek(2) !== '?') { extglobOpen('plus', value); continue; } if ((prev && prev.value === '(') || opts.regex === false) { push({ type: 'plus', value, output: PLUS_LITERAL }); continue; } if ((prev && (prev.type === 'bracket' || prev.type === 'paren' || prev.type === 'brace')) || state.parens > 0) { push({ type: 'plus', value }); continue; } push({ type: 'plus', value: PLUS_LITERAL }); continue; } /** * Plain text */ if (value === '@') { if (opts.noextglob !== true && peek() === '(' && peek(2) !== '?') { push({ type: 'at', extglob: true, value, output: '' }); continue; } push({ type: 'text', value }); continue; } /** * Plain text */ if (value !== '*') { if (value === '$' || value === '^') { value = `\\${value}`; } const match = REGEX_NON_SPECIAL_CHARS.exec(remaining()); if (match) { value += match[0]; state.index += match[0].length; } push({ type: 'text', value }); continue; } /** * Stars */ if (prev && (prev.type === 'globstar' || prev.star === true)) { prev.type = 'star'; prev.star = true; prev.value += value; prev.output = star; state.backtrack = true; state.globstar = true; consume(value); continue; } let rest = remaining(); if (opts.noextglob !== true && /^\([^?]/.test(rest)) { extglobOpen('star', value); continue; } if (prev.type === 'star') { if (opts.noglobstar === true) { consume(value); continue; } const prior = prev.prev; const before = prior.prev; const isStart = prior.type === 'slash' || prior.type === 'bos'; const afterStar = before && (before.type === 'star' || before.type === 'globstar'); if (opts.bash === true && (!isStart || (rest[0] && rest[0] !== '/'))) { push({ type: 'star', value, output: '' }); continue; } const isBrace = state.braces > 0 && (prior.type === 'comma' || prior.type === 'brace'); const isExtglob = extglobs.length && (prior.type === 'pipe' || prior.type === 'paren'); if (!isStart && prior.type !== 'paren' && !isBrace && !isExtglob) { push({ type: 'star', value, output: '' }); continue; } // strip consecutive `/**/` while (rest.slice(0, 3) === '/**') { const after = input[state.index + 4]; if (after && after !== '/') { break; } rest = rest.slice(3); consume('/**', 3); } if (prior.type === 'bos' && eos()) { prev.type = 'globstar'; prev.value += value; prev.output = globstar(opts); state.output = prev.output; state.globstar = true; consume(value); continue; } if (prior.type === 'slash' && prior.prev.type !== 'bos' && !afterStar && eos()) { state.output = state.output.slice(0, -(prior.output + prev.output).length); prior.output = `(?:${prior.output}`; prev.type = 'globstar'; prev.output = globstar(opts) + (opts.strictSlashes ? ')' : '|$)'); prev.value += value; state.globstar = true; state.output += prior.output + prev.output; consume(value); continue; } if (prior.type === 'slash' && prior.prev.type !== 'bos' && rest[0] === '/') { const end = rest[1] !== void 0 ? '|$' : ''; state.output = state.output.slice(0, -(prior.output + prev.output).length); prior.output = `(?:${prior.output}`; prev.type = 'globstar'; prev.output = `${globstar(opts)}${SLASH_LITERAL}|${SLASH_LITERAL}${end})`; prev.value += value; state.output += prior.output + prev.output; state.globstar = true; consume(value + advance()); push({ type: 'slash', value: '/', output: '' }); continue; } if (prior.type === 'bos' && rest[0] === '/') { prev.type = 'globstar'; prev.value += value; prev.output = `(?:^|${SLASH_LITERAL}|${globstar(opts)}${SLASH_LITERAL})`; state.output = prev.output; state.globstar = true; consume(value + advance()); push({ type: 'slash', value: '/', output: '' }); continue; } // remove single star from output state.output = state.output.slice(0, -prev.output.length); // reset previous token to globstar prev.type = 'globstar'; prev.output = globstar(opts); prev.value += value; // reset output with globstar state.output += prev.output; state.globstar = true; consume(value); continue; } const token = { type: 'star', value, output: star }; if (opts.bash === true) { token.output = '.*?'; if (prev.type === 'bos' || prev.type === 'slash') { token.output = nodot + token.output; } push(token); continue; } if (prev && (prev.type === 'bracket' || prev.type === 'paren') && opts.regex === true) { token.output = value; push(token); continue; } if (state.index === state.start || prev.type === 'slash' || prev.type === 'dot') { if (prev.type === 'dot') { state.output += NO_DOT_SLASH; prev.output += NO_DOT_SLASH; } else if (opts.dot === true) { state.output += NO_DOTS_SLASH; prev.output += NO_DOTS_SLASH; } else { state.output += nodot; prev.output += nodot; } if (peek() !== '*') { state.output += ONE_CHAR; prev.output += ONE_CHAR; } } push(token); } while (state.brackets > 0) { if (opts.strictBrackets === true) throw new SyntaxError(syntaxError('closing', ']')); state.output = utils.escapeLast(state.output, '['); decrement('brackets'); } while (state.parens > 0) { if (opts.strictBrackets === true) throw new SyntaxError(syntaxError('closing', ')')); state.output = utils.escapeLast(state.output, '('); decrement('parens'); } while (state.braces > 0) { if (opts.strictBrackets === true) throw new SyntaxError(syntaxError('closing', '}')); state.output = utils.escapeLast(state.output, '{'); decrement('braces'); } if (opts.strictSlashes !== true && (prev.type === 'star' || prev.type === 'bracket')) { push({ type: 'maybe_slash', value: '', output: `${SLASH_LITERAL}?` }); } // rebuild the output if we had to backtrack at any point if (state.backtrack === true) { state.output = ''; for (const token of state.tokens) { state.output += token.output != null ? token.output : token.value; if (token.suffix) { state.output += token.suffix; } } } return state; }; /** * Fast paths for creating regular expressions for common glob patterns. * This can significantly speed up processing and has very little downside * impact when none of the fast paths match. */ parse.fastpaths = (input, options) => { const opts = { ...options }; const max = typeof opts.maxLength === 'number' ? Math.min(MAX_LENGTH, opts.maxLength) : MAX_LENGTH; const len = input.length; if (len > max) { throw new SyntaxError(`Input length: ${len}, exceeds maximum allowed length: ${max}`); } input = REPLACEMENTS[input] || input; const win32 = utils.isWindows(options); // create constants based on platform, for windows or posix const { DOT_LITERAL, SLASH_LITERAL, ONE_CHAR, DOTS_SLASH, NO_DOT, NO_DOTS, NO_DOTS_SLASH, STAR, START_ANCHOR } = constants.globChars(win32); const nodot = opts.dot ? NO_DOTS : NO_DOT; const slashDot = opts.dot ? NO_DOTS_SLASH : NO_DOT; const capture = opts.capture ? '' : '?:'; const state = { negated: false, prefix: '' }; let star = opts.bash === true ? '.*?' : STAR; if (opts.capture) { star = `(${star})`; } const globstar = opts => { if (opts.noglobstar === true) return star; return `(${capture}(?:(?!${START_ANCHOR}${opts.dot ? DOTS_SLASH : DOT_LITERAL}).)*?)`; }; const create = str => { switch (str) { case '*': return `${nodot}${ONE_CHAR}${star}`; case '.*': return `${DOT_LITERAL}${ONE_CHAR}${star}`; case '*.*': return `${nodot}${star}${DOT_LITERAL}${ONE_CHAR}${star}`; case '*/*': return `${nodot}${star}${SLASH_LITERAL}${ONE_CHAR}${slashDot}${star}`; case '**': return nodot + globstar(opts); case '**/*': return `(?:${nodot}${globstar(opts)}${SLASH_LITERAL})?${slashDot}${ONE_CHAR}${star}`; case '**/*.*': return `(?:${nodot}${globstar(opts)}${SLASH_LITERAL})?${slashDot}${star}${DOT_LITERAL}${ONE_CHAR}${star}`; case '**/.*': return `(?:${nodot}${globstar(opts)}${SLASH_LITERAL})?${DOT_LITERAL}${ONE_CHAR}${star}`; default: { const match = /^(.*?)\.(\w+)$/.exec(str); if (!match) return; const source = create(match[1]); if (!source) return; return source + DOT_LITERAL + match[2]; } } }; const output = utils.removePrefix(input, state); let source = create(output); if (source && opts.strictSlashes !== true) { source += `${SLASH_LITERAL}?`; } return source; }; module.exports = parse;