audiobookshelf/server/libs/archiver/archiverUtils/minimatch/index.js
2022-07-08 18:11:09 -05:00

909 lines
27 KiB
JavaScript

const minimatch = module.exports = (p, pattern, options = {}) => {
assertValidPattern(pattern)
// shortcut: comments match nothing.
if (!options.nocomment && pattern.charAt(0) === '#') {
return false
}
return new Minimatch(pattern, options).match(p)
}
module.exports = minimatch
const isWindows = typeof process === 'object' &&
process &&
process.platform === 'win32'
const path = isWindows ? { sep: '\\' } : { sep: '/' }
minimatch.sep = path.sep
const GLOBSTAR = Symbol('globstar **')
minimatch.GLOBSTAR = GLOBSTAR
const expand = require('../braceExpansion')
const plTypes = {
'!': { open: '(?:(?!(?:', close: '))[^/]*?)' },
'?': { open: '(?:', close: ')?' },
'+': { open: '(?:', close: ')+' },
'*': { open: '(?:', close: ')*' },
'@': { open: '(?:', close: ')' }
}
// any single thing other than /
// don't need to escape / when using new RegExp()
const qmark = '[^/]'
// * => any number of characters
const star = qmark + '*?'
// ** when dots are allowed. Anything goes, except .. and .
// not (^ or / followed by one or two dots followed by $ or /),
// followed by anything, any number of times.
const twoStarDot = '(?:(?!(?:\\\/|^)(?:\\.{1,2})($|\\\/)).)*?'
// not a ^ or / followed by a dot,
// followed by anything, any number of times.
const twoStarNoDot = '(?:(?!(?:\\\/|^)\\.).)*?'
// "abc" -> { a:true, b:true, c:true }
const charSet = s => s.split('').reduce((set, c) => {
set[c] = true
return set
}, {})
// characters that need to be escaped in RegExp.
const reSpecials = charSet('().*{}+?[]^$\\!')
// characters that indicate we have to add the pattern start
const addPatternStartSet = charSet('[.(')
// normalizes slashes.
const slashSplit = /\/+/
minimatch.filter = (pattern, options = {}) =>
(p, i, list) => minimatch(p, pattern, options)
const ext = (a, b = {}) => {
const t = {}
Object.keys(a).forEach(k => t[k] = a[k])
Object.keys(b).forEach(k => t[k] = b[k])
return t
}
minimatch.defaults = def => {
if (!def || typeof def !== 'object' || !Object.keys(def).length) {
return minimatch
}
const orig = minimatch
const m = (p, pattern, options) => orig(p, pattern, ext(def, options))
m.Minimatch = class Minimatch extends orig.Minimatch {
constructor(pattern, options) {
super(pattern, ext(def, options))
}
}
m.Minimatch.defaults = options => orig.defaults(ext(def, options)).Minimatch
m.filter = (pattern, options) => orig.filter(pattern, ext(def, options))
m.defaults = options => orig.defaults(ext(def, options))
m.makeRe = (pattern, options) => orig.makeRe(pattern, ext(def, options))
m.braceExpand = (pattern, options) => orig.braceExpand(pattern, ext(def, options))
m.match = (list, pattern, options) => orig.match(list, pattern, ext(def, options))
return m
}
// Brace expansion:
// a{b,c}d -> abd acd
// a{b,}c -> abc ac
// a{0..3}d -> a0d a1d a2d a3d
// a{b,c{d,e}f}g -> abg acdfg acefg
// a{b,c}d{e,f}g -> abdeg acdeg abdeg abdfg
//
// Invalid sets are not expanded.
// a{2..}b -> a{2..}b
// a{b}c -> a{b}c
minimatch.braceExpand = (pattern, options) => braceExpand(pattern, options)
const braceExpand = (pattern, options = {}) => {
assertValidPattern(pattern)
// Thanks to Yeting Li <https://github.com/yetingli> for
// improving this regexp to avoid a ReDOS vulnerability.
if (options.nobrace || !/\{(?:(?!\{).)*\}/.test(pattern)) {
// shortcut. no need to expand.
return [pattern]
}
return expand(pattern)
}
const MAX_PATTERN_LENGTH = 1024 * 64
const assertValidPattern = pattern => {
if (typeof pattern !== 'string') {
throw new TypeError('invalid pattern')
}
if (pattern.length > MAX_PATTERN_LENGTH) {
throw new TypeError('pattern is too long')
}
}
// parse a component of the expanded set.
// At this point, no pattern may contain "/" in it
// so we're going to return a 2d array, where each entry is the full
// pattern, split on '/', and then turned into a regular expression.
// A regexp is made at the end which joins each array with an
// escaped /, and another full one which joins each regexp with |.
//
// Following the lead of Bash 4.1, note that "**" only has special meaning
// when it is the *only* thing in a path portion. Otherwise, any series
// of * is equivalent to a single *. Globstar behavior is enabled by
// default, and can be disabled by setting options.noglobstar.
const SUBPARSE = Symbol('subparse')
minimatch.makeRe = (pattern, options) =>
new Minimatch(pattern, options || {}).makeRe()
minimatch.match = (list, pattern, options = {}) => {
const mm = new Minimatch(pattern, options)
list = list.filter(f => mm.match(f))
if (mm.options.nonull && !list.length) {
list.push(pattern)
}
return list
}
// replace stuff like \* with *
const globUnescape = s => s.replace(/\\(.)/g, '$1')
const regExpEscape = s => s.replace(/[-[\]{}()*+?.,\\^$|#\s]/g, '\\$&')
class Minimatch {
constructor(pattern, options) {
assertValidPattern(pattern)
if (!options) options = {}
this.options = options
this.set = []
this.pattern = pattern
this.windowsPathsNoEscape = !!options.windowsPathsNoEscape ||
options.allowWindowsEscape === false
if (this.windowsPathsNoEscape) {
this.pattern = this.pattern.replace(/\\/g, '/')
}
this.regexp = null
this.negate = false
this.comment = false
this.empty = false
this.partial = !!options.partial
// make the set of regexps etc.
this.make()
}
debug() { }
make() {
const pattern = this.pattern
const options = this.options
// empty patterns and comments match nothing.
if (!options.nocomment && pattern.charAt(0) === '#') {
this.comment = true
return
}
if (!pattern) {
this.empty = true
return
}
// step 1: figure out negation, etc.
this.parseNegate()
// step 2: expand braces
let set = this.globSet = this.braceExpand()
if (options.debug) this.debug = (...args) => console.error(...args)
this.debug(this.pattern, set)
// step 3: now we have a set, so turn each one into a series of path-portion
// matching patterns.
// These will be regexps, except in the case of "**", which is
// set to the GLOBSTAR object for globstar behavior,
// and will not contain any / characters
set = this.globParts = set.map(s => s.split(slashSplit))
this.debug(this.pattern, set)
// glob --> regexps
set = set.map((s, si, set) => s.map(this.parse, this))
this.debug(this.pattern, set)
// filter out everything that didn't compile properly.
set = set.filter(s => s.indexOf(false) === -1)
this.debug(this.pattern, set)
this.set = set
}
parseNegate() {
if (this.options.nonegate) return
const pattern = this.pattern
let negate = false
let negateOffset = 0
for (let i = 0; i < pattern.length && pattern.charAt(i) === '!'; i++) {
negate = !negate
negateOffset++
}
if (negateOffset) this.pattern = pattern.substr(negateOffset)
this.negate = negate
}
// set partial to true to test if, for example,
// "/a/b" matches the start of "/*/b/*/d"
// Partial means, if you run out of file before you run
// out of pattern, then that's fine, as long as all
// the parts match.
matchOne(file, pattern, partial) {
var options = this.options
this.debug('matchOne',
{ 'this': this, file: file, pattern: pattern })
this.debug('matchOne', file.length, pattern.length)
for (var fi = 0,
pi = 0,
fl = file.length,
pl = pattern.length
; (fi < fl) && (pi < pl)
; fi++, pi++) {
this.debug('matchOne loop')
var p = pattern[pi]
var f = file[fi]
this.debug(pattern, p, f)
// should be impossible.
// some invalid regexp stuff in the set.
/* istanbul ignore if */
if (p === false) return false
if (p === GLOBSTAR) {
this.debug('GLOBSTAR', [pattern, p, f])
// "**"
// a/**/b/**/c would match the following:
// a/b/x/y/z/c
// a/x/y/z/b/c
// a/b/x/b/x/c
// a/b/c
// To do this, take the rest of the pattern after
// the **, and see if it would match the file remainder.
// If so, return success.
// If not, the ** "swallows" a segment, and try again.
// This is recursively awful.
//
// a/**/b/**/c matching a/b/x/y/z/c
// - a matches a
// - doublestar
// - matchOne(b/x/y/z/c, b/**/c)
// - b matches b
// - doublestar
// - matchOne(x/y/z/c, c) -> no
// - matchOne(y/z/c, c) -> no
// - matchOne(z/c, c) -> no
// - matchOne(c, c) yes, hit
var fr = fi
var pr = pi + 1
if (pr === pl) {
this.debug('** at the end')
// a ** at the end will just swallow the rest.
// We have found a match.
// however, it will not swallow /.x, unless
// options.dot is set.
// . and .. are *never* matched by **, for explosively
// exponential reasons.
for (; fi < fl; fi++) {
if (file[fi] === '.' || file[fi] === '..' ||
(!options.dot && file[fi].charAt(0) === '.')) return false
}
return true
}
// ok, let's see if we can swallow whatever we can.
while (fr < fl) {
var swallowee = file[fr]
this.debug('\nglobstar while', file, fr, pattern, pr, swallowee)
// XXX remove this slice. Just pass the start index.
if (this.matchOne(file.slice(fr), pattern.slice(pr), partial)) {
this.debug('globstar found match!', fr, fl, swallowee)
// found a match.
return true
} else {
// can't swallow "." or ".." ever.
// can only swallow ".foo" when explicitly asked.
if (swallowee === '.' || swallowee === '..' ||
(!options.dot && swallowee.charAt(0) === '.')) {
this.debug('dot detected!', file, fr, pattern, pr)
break
}
// ** swallows a segment, and continue.
this.debug('globstar swallow a segment, and continue')
fr++
}
}
// no match was found.
// However, in partial mode, we can't say this is necessarily over.
// If there's more *pattern* left, then
/* istanbul ignore if */
if (partial) {
// ran out of file
this.debug('\n>>> no match, partial?', file, fr, pattern, pr)
if (fr === fl) return true
}
return false
}
// something other than **
// non-magic patterns just have to match exactly
// patterns with magic have been turned into regexps.
var hit
if (typeof p === 'string') {
hit = f === p
this.debug('string match', p, f, hit)
} else {
hit = f.match(p)
this.debug('pattern match', p, f, hit)
}
if (!hit) return false
}
// Note: ending in / means that we'll get a final ""
// at the end of the pattern. This can only match a
// corresponding "" at the end of the file.
// If the file ends in /, then it can only match a
// a pattern that ends in /, unless the pattern just
// doesn't have any more for it. But, a/b/ should *not*
// match "a/b/*", even though "" matches against the
// [^/]*? pattern, except in partial mode, where it might
// simply not be reached yet.
// However, a/b/ should still satisfy a/*
// now either we fell off the end of the pattern, or we're done.
if (fi === fl && pi === pl) {
// ran out of pattern and filename at the same time.
// an exact hit!
return true
} else if (fi === fl) {
// ran out of file, but still had pattern left.
// this is ok if we're doing the match as part of
// a glob fs traversal.
return partial
} else /* istanbul ignore else */ if (pi === pl) {
// ran out of pattern, still have file left.
// this is only acceptable if we're on the very last
// empty segment of a file with a trailing slash.
// a/* should match a/b/
return (fi === fl - 1) && (file[fi] === '')
}
// should be unreachable.
/* istanbul ignore next */
throw new Error('wtf?')
}
braceExpand() {
return braceExpand(this.pattern, this.options)
}
parse(pattern, isSub) {
assertValidPattern(pattern)
const options = this.options
// shortcuts
if (pattern === '**') {
if (!options.noglobstar)
return GLOBSTAR
else
pattern = '*'
}
if (pattern === '') return ''
let re = ''
let hasMagic = !!options.nocase
let escaping = false
// ? => one single character
const patternListStack = []
const negativeLists = []
let stateChar
let inClass = false
let reClassStart = -1
let classStart = -1
let cs
let pl
let sp
// . and .. never match anything that doesn't start with .,
// even when options.dot is set.
const patternStart = pattern.charAt(0) === '.' ? '' // anything
// not (start or / followed by . or .. followed by / or end)
: options.dot ? '(?!(?:^|\\\/)\\.{1,2}(?:$|\\\/))'
: '(?!\\.)'
const clearStateChar = () => {
if (stateChar) {
// we had some state-tracking character
// that wasn't consumed by this pass.
switch (stateChar) {
case '*':
re += star
hasMagic = true
break
case '?':
re += qmark
hasMagic = true
break
default:
re += '\\' + stateChar
break
}
this.debug('clearStateChar %j %j', stateChar, re)
stateChar = false
}
}
for (let i = 0, c; (i < pattern.length) && (c = pattern.charAt(i)); i++) {
this.debug('%s\t%s %s %j', pattern, i, re, c)
// skip over any that are escaped.
if (escaping) {
/* istanbul ignore next - completely not allowed, even escaped. */
if (c === '/') {
return false
}
if (reSpecials[c]) {
re += '\\'
}
re += c
escaping = false
continue
}
switch (c) {
/* istanbul ignore next */
case '/': {
// Should already be path-split by now.
return false
}
case '\\':
clearStateChar()
escaping = true
continue
// the various stateChar values
// for the "extglob" stuff.
case '?':
case '*':
case '+':
case '@':
case '!':
this.debug('%s\t%s %s %j <-- stateChar', pattern, i, re, c)
// all of those are literals inside a class, except that
// the glob [!a] means [^a] in regexp
if (inClass) {
this.debug(' in class')
if (c === '!' && i === classStart + 1) c = '^'
re += c
continue
}
// if we already have a stateChar, then it means
// that there was something like ** or +? in there.
// Handle the stateChar, then proceed with this one.
this.debug('call clearStateChar %j', stateChar)
clearStateChar()
stateChar = c
// if extglob is disabled, then +(asdf|foo) isn't a thing.
// just clear the statechar *now*, rather than even diving into
// the patternList stuff.
if (options.noext) clearStateChar()
continue
case '(':
if (inClass) {
re += '('
continue
}
if (!stateChar) {
re += '\\('
continue
}
patternListStack.push({
type: stateChar,
start: i - 1,
reStart: re.length,
open: plTypes[stateChar].open,
close: plTypes[stateChar].close
})
// negation is (?:(?!js)[^/]*)
re += stateChar === '!' ? '(?:(?!(?:' : '(?:'
this.debug('plType %j %j', stateChar, re)
stateChar = false
continue
case ')':
if (inClass || !patternListStack.length) {
re += '\\)'
continue
}
clearStateChar()
hasMagic = true
pl = patternListStack.pop()
// negation is (?:(?!js)[^/]*)
// The others are (?:<pattern>)<type>
re += pl.close
if (pl.type === '!') {
negativeLists.push(pl)
}
pl.reEnd = re.length
continue
case '|':
if (inClass || !patternListStack.length) {
re += '\\|'
continue
}
clearStateChar()
re += '|'
continue
// these are mostly the same in regexp and glob
case '[':
// swallow any state-tracking char before the [
clearStateChar()
if (inClass) {
re += '\\' + c
continue
}
inClass = true
classStart = i
reClassStart = re.length
re += c
continue
case ']':
// a right bracket shall lose its special
// meaning and represent itself in
// a bracket expression if it occurs
// first in the list. -- POSIX.2 2.8.3.2
if (i === classStart + 1 || !inClass) {
re += '\\' + c
continue
}
// handle the case where we left a class open.
// "[z-a]" is valid, equivalent to "\[z-a\]"
// split where the last [ was, make sure we don't have
// an invalid re. if so, re-walk the contents of the
// would-be class to re-translate any characters that
// were passed through as-is
// TODO: It would probably be faster to determine this
// without a try/catch and a new RegExp, but it's tricky
// to do safely. For now, this is safe and works.
cs = pattern.substring(classStart + 1, i)
try {
RegExp('[' + cs + ']')
} catch (er) {
// not a valid class!
sp = this.parse(cs, SUBPARSE)
re = re.substr(0, reClassStart) + '\\[' + sp[0] + '\\]'
hasMagic = hasMagic || sp[1]
inClass = false
continue
}
// finish up the class.
hasMagic = true
inClass = false
re += c
continue
default:
// swallow any state char that wasn't consumed
clearStateChar()
if (reSpecials[c] && !(c === '^' && inClass)) {
re += '\\'
}
re += c
break
} // switch
} // for
// handle the case where we left a class open.
// "[abc" is valid, equivalent to "\[abc"
if (inClass) {
// split where the last [ was, and escape it
// this is a huge pita. We now have to re-walk
// the contents of the would-be class to re-translate
// any characters that were passed through as-is
cs = pattern.substr(classStart + 1)
sp = this.parse(cs, SUBPARSE)
re = re.substr(0, reClassStart) + '\\[' + sp[0]
hasMagic = hasMagic || sp[1]
}
// handle the case where we had a +( thing at the *end*
// of the pattern.
// each pattern list stack adds 3 chars, and we need to go through
// and escape any | chars that were passed through as-is for the regexp.
// Go through and escape them, taking care not to double-escape any
// | chars that were already escaped.
for (pl = patternListStack.pop(); pl; pl = patternListStack.pop()) {
let tail
tail = re.slice(pl.reStart + pl.open.length)
this.debug('setting tail', re, pl)
// maybe some even number of \, then maybe 1 \, followed by a |
tail = tail.replace(/((?:\\{2}){0,64})(\\?)\|/g, (_, $1, $2) => {
/* istanbul ignore else - should already be done */
if (!$2) {
// the | isn't already escaped, so escape it.
$2 = '\\'
}
// need to escape all those slashes *again*, without escaping the
// one that we need for escaping the | character. As it works out,
// escaping an even number of slashes can be done by simply repeating
// it exactly after itself. That's why this trick works.
//
// I am sorry that you have to see this.
return $1 + $1 + $2 + '|'
})
this.debug('tail=%j\n %s', tail, tail, pl, re)
const t = pl.type === '*' ? star
: pl.type === '?' ? qmark
: '\\' + pl.type
hasMagic = true
re = re.slice(0, pl.reStart) + t + '\\(' + tail
}
// handle trailing things that only matter at the very end.
clearStateChar()
if (escaping) {
// trailing \\
re += '\\\\'
}
// only need to apply the nodot start if the re starts with
// something that could conceivably capture a dot
const addPatternStart = addPatternStartSet[re.charAt(0)]
// Hack to work around lack of negative lookbehind in JS
// A pattern like: *.!(x).!(y|z) needs to ensure that a name
// like 'a.xyz.yz' doesn't match. So, the first negative
// lookahead, has to look ALL the way ahead, to the end of
// the pattern.
for (let n = negativeLists.length - 1; n > -1; n--) {
const nl = negativeLists[n]
const nlBefore = re.slice(0, nl.reStart)
const nlFirst = re.slice(nl.reStart, nl.reEnd - 8)
let nlAfter = re.slice(nl.reEnd)
const nlLast = re.slice(nl.reEnd - 8, nl.reEnd) + nlAfter
// Handle nested stuff like *(*.js|!(*.json)), where open parens
// mean that we should *not* include the ) in the bit that is considered
// "after" the negated section.
const openParensBefore = nlBefore.split('(').length - 1
let cleanAfter = nlAfter
for (let i = 0; i < openParensBefore; i++) {
cleanAfter = cleanAfter.replace(/\)[+*?]?/, '')
}
nlAfter = cleanAfter
const dollar = nlAfter === '' && isSub !== SUBPARSE ? '$' : ''
re = nlBefore + nlFirst + nlAfter + dollar + nlLast
}
// if the re is not "" at this point, then we need to make sure
// it doesn't match against an empty path part.
// Otherwise a/* will match a/, which it should not.
if (re !== '' && hasMagic) {
re = '(?=.)' + re
}
if (addPatternStart) {
re = patternStart + re
}
// parsing just a piece of a larger pattern.
if (isSub === SUBPARSE) {
return [re, hasMagic]
}
// skip the regexp for non-magical patterns
// unescape anything in it, though, so that it'll be
// an exact match against a file etc.
if (!hasMagic) {
return globUnescape(pattern)
}
const flags = options.nocase ? 'i' : ''
try {
return Object.assign(new RegExp('^' + re + '$', flags), {
_glob: pattern,
_src: re,
})
} catch (er) /* istanbul ignore next - should be impossible */ {
// If it was an invalid regular expression, then it can't match
// anything. This trick looks for a character after the end of
// the string, which is of course impossible, except in multi-line
// mode, but it's not a /m regex.
return new RegExp('$.')
}
}
makeRe() {
if (this.regexp || this.regexp === false) return this.regexp
// at this point, this.set is a 2d array of partial
// pattern strings, or "**".
//
// It's better to use .match(). This function shouldn't
// be used, really, but it's pretty convenient sometimes,
// when you just want to work with a regex.
const set = this.set
if (!set.length) {
this.regexp = false
return this.regexp
}
const options = this.options
const twoStar = options.noglobstar ? star
: options.dot ? twoStarDot
: twoStarNoDot
const flags = options.nocase ? 'i' : ''
// coalesce globstars and regexpify non-globstar patterns
// if it's the only item, then we just do one twoStar
// if it's the first, and there are more, prepend (\/|twoStar\/)? to next
// if it's the last, append (\/twoStar|) to previous
// if it's in the middle, append (\/|\/twoStar\/) to previous
// then filter out GLOBSTAR symbols
let re = set.map(pattern => {
pattern = pattern.map(p =>
typeof p === 'string' ? regExpEscape(p)
: p === GLOBSTAR ? GLOBSTAR
: p._src
).reduce((set, p) => {
if (!(set[set.length - 1] === GLOBSTAR && p === GLOBSTAR)) {
set.push(p)
}
return set
}, [])
pattern.forEach((p, i) => {
if (p !== GLOBSTAR || pattern[i - 1] === GLOBSTAR) {
return
}
if (i === 0) {
if (pattern.length > 1) {
pattern[i + 1] = '(?:\\\/|' + twoStar + '\\\/)?' + pattern[i + 1]
} else {
pattern[i] = twoStar
}
} else if (i === pattern.length - 1) {
pattern[i - 1] += '(?:\\\/|' + twoStar + ')?'
} else {
pattern[i - 1] += '(?:\\\/|\\\/' + twoStar + '\\\/)' + pattern[i + 1]
pattern[i + 1] = GLOBSTAR
}
})
return pattern.filter(p => p !== GLOBSTAR).join('/')
}).join('|')
// must match entire pattern
// ending in a * or ** will make it less strict.
re = '^(?:' + re + ')$'
// can match anything, as long as it's not this.
if (this.negate) re = '^(?!' + re + ').*$'
try {
this.regexp = new RegExp(re, flags)
} catch (ex) /* istanbul ignore next - should be impossible */ {
this.regexp = false
}
return this.regexp
}
match(f, partial = this.partial) {
this.debug('match', f, this.pattern)
// short-circuit in the case of busted things.
// comments, etc.
if (this.comment) return false
if (this.empty) return f === ''
if (f === '/' && partial) return true
const options = this.options
// windows: need to use /, not \
if (path.sep !== '/') {
f = f.split(path.sep).join('/')
}
// treat the test path as a set of pathparts.
f = f.split(slashSplit)
this.debug(this.pattern, 'split', f)
// just ONE of the pattern sets in this.set needs to match
// in order for it to be valid. If negating, then just one
// match means that we have failed.
// Either way, return on the first hit.
const set = this.set
this.debug(this.pattern, 'set', set)
// Find the basename of the path by looking for the last non-empty segment
let filename
for (let i = f.length - 1; i >= 0; i--) {
filename = f[i]
if (filename) break
}
for (let i = 0; i < set.length; i++) {
const pattern = set[i]
let file = f
if (options.matchBase && pattern.length === 1) {
file = [filename]
}
const hit = this.matchOne(file, pattern, partial)
if (hit) {
if (options.flipNegate) return true
return !this.negate
}
}
// didn't get any hits. this is success if it's a negative
// pattern, failure otherwise.
if (options.flipNegate) return false
return this.negate
}
static defaults(def) {
return minimatch.defaults(def).Minimatch
}
}
minimatch.Minimatch = Minimatch