tools/node-hermes/nodelib/path.js (898 lines of code) (raw):
// @nolint
// Copyright Joyent, Inc. and other Node contributors.
//
// Permission is hereby granted, free of charge, to any person obtaining a
// copy of this software and associated documentation files (the
// "Software"), to deal in the Software without restriction, including
// without limitation the rights to use, copy, modify, merge, publish,
// distribute, sublicense, and/or sell copies of the Software, and to permit
// persons to whom the Software is furnished to do so, subject to the
// following conditions:
//
// The above copyright notice and this permission notice shall be included
// in all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
// OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
// MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN
// NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
// DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
// OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
// USE OR OTHER DEALINGS IN THE SOFTWARE.
'use strict';
var _primordials = primordials,
FunctionPrototypeBind = _primordials.FunctionPrototypeBind,
StringPrototypeCharCodeAt = _primordials.StringPrototypeCharCodeAt,
StringPrototypeIndexOf = _primordials.StringPrototypeIndexOf,
StringPrototypeLastIndexOf = _primordials.StringPrototypeLastIndexOf,
StringPrototypeReplace = _primordials.StringPrototypeReplace,
StringPrototypeSlice = _primordials.StringPrototypeSlice,
StringPrototypeToLowerCase = _primordials.StringPrototypeToLowerCase;
var _require = require('internal/constants'),
CHAR_UPPERCASE_A = _require.CHAR_UPPERCASE_A,
CHAR_LOWERCASE_A = _require.CHAR_LOWERCASE_A,
CHAR_UPPERCASE_Z = _require.CHAR_UPPERCASE_Z,
CHAR_LOWERCASE_Z = _require.CHAR_LOWERCASE_Z,
CHAR_DOT = _require.CHAR_DOT,
CHAR_FORWARD_SLASH = _require.CHAR_FORWARD_SLASH,
CHAR_BACKWARD_SLASH = _require.CHAR_BACKWARD_SLASH,
CHAR_COLON = _require.CHAR_COLON,
CHAR_QUESTION_MARK = _require.CHAR_QUESTION_MARK;
var _require2 = require('internal/validators'),
validateObject = _require2.validateObject,
validateString = _require2.validateString;
var platformIsWin32 = process.platform === 'win32';
function isPathSeparator(code) {
return code === CHAR_FORWARD_SLASH || code === CHAR_BACKWARD_SLASH;
}
function isPosixPathSeparator(code) {
return code === CHAR_FORWARD_SLASH;
}
function isWindowsDeviceRoot(code) {
return code >= CHAR_UPPERCASE_A && code <= CHAR_UPPERCASE_Z || code >= CHAR_LOWERCASE_A && code <= CHAR_LOWERCASE_Z;
} // Resolves . and .. elements in a path with directory names
function normalizeString(path, allowAboveRoot, separator, isPathSeparator) {
var res = '';
var lastSegmentLength = 0;
var lastSlash = -1;
var dots = 0;
var code = 0;
for (var i = 0; i <= path.length; ++i) {
if (i < path.length) code = StringPrototypeCharCodeAt(path, i);else if (isPathSeparator(code)) break;else code = CHAR_FORWARD_SLASH;
if (isPathSeparator(code)) {
if (lastSlash === i - 1 || dots === 1) {// NOOP
} else if (dots === 2) {
if (res.length < 2 || lastSegmentLength !== 2 || StringPrototypeCharCodeAt(res, res.length - 1) !== CHAR_DOT || StringPrototypeCharCodeAt(res, res.length - 2) !== CHAR_DOT) {
if (res.length > 2) {
var lastSlashIndex = StringPrototypeLastIndexOf(res, separator);
if (lastSlashIndex === -1) {
res = '';
lastSegmentLength = 0;
} else {
res = StringPrototypeSlice(res, 0, lastSlashIndex);
lastSegmentLength = res.length - 1 - StringPrototypeLastIndexOf(res, separator);
}
lastSlash = i;
dots = 0;
continue;
} else if (res.length !== 0) {
res = '';
lastSegmentLength = 0;
lastSlash = i;
dots = 0;
continue;
}
}
if (allowAboveRoot) {
res += res.length > 0 ? "".concat(separator, "..") : '..';
lastSegmentLength = 2;
}
} else {
if (res.length > 0) res += "".concat(separator).concat(StringPrototypeSlice(path, lastSlash + 1, i));else res = StringPrototypeSlice(path, lastSlash + 1, i);
lastSegmentLength = i - lastSlash - 1;
}
lastSlash = i;
dots = 0;
} else if (code === CHAR_DOT && dots !== -1) {
++dots;
} else {
dots = -1;
}
}
return res;
}
/**
* @param {string} sep
* @param {{
* dir?: string;
* root?: string;
* base?: string;
* name?: string;
* ext?: string;
* }} pathObject
* @returns {string}
*/
function _format(sep, pathObject) {
validateObject(pathObject, 'pathObject');
var dir = pathObject.dir || pathObject.root;
var base = pathObject.base || "".concat(pathObject.name || '').concat(pathObject.ext || '');
if (!dir) {
return base;
}
return dir === pathObject.root ? "".concat(dir).concat(base) : "".concat(dir).concat(sep).concat(base);
}
var win32 = {
/**
* path.resolve([from ...], to)
* @param {...string} args
* @returns {string}
*/
resolve: function resolve() {
var resolvedDevice = '';
var resolvedTail = '';
var resolvedAbsolute = false;
for (var i = arguments.length - 1; i >= -1; i--) {
var path = void 0;
if (i >= 0) {
path = i < 0 || arguments.length <= i ? undefined : arguments[i];
validateString(path, 'path'); // Skip empty entries
if (path.length === 0) {
continue;
}
} else if (resolvedDevice.length === 0) {
path = process.cwd();
} else {
// Windows has the concept of drive-specific current working
// directories. If we've resolved a drive letter but not yet an
// absolute path, get cwd for that drive, or the process cwd if
// the drive cwd is not available. We're sure the device is not
// a UNC path at this points, because UNC paths are always absolute.
path = process.env["=".concat(resolvedDevice)] || process.cwd(); // Verify that a cwd was found and that it actually points
// to our drive. If not, default to the drive's root.
if (path === undefined || StringPrototypeToLowerCase(StringPrototypeSlice(path, 0, 2)) !== StringPrototypeToLowerCase(resolvedDevice) && StringPrototypeCharCodeAt(path, 2) === CHAR_BACKWARD_SLASH) {
path = "".concat(resolvedDevice, "\\");
}
}
var len = path.length;
var rootEnd = 0;
var device = '';
var isAbsolute = false;
var code = StringPrototypeCharCodeAt(path, 0); // Try to match a root
if (len === 1) {
if (isPathSeparator(code)) {
// `path` contains just a path separator
rootEnd = 1;
isAbsolute = true;
}
} else if (isPathSeparator(code)) {
// Possible UNC root
// If we started with a separator, we know we at least have an
// absolute path of some kind (UNC or otherwise)
isAbsolute = true;
if (isPathSeparator(StringPrototypeCharCodeAt(path, 1))) {
// Matched double path separator at beginning
var j = 2;
var last = j; // Match 1 or more non-path separators
while (j < len && !isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j < len && j !== last) {
var firstPart = StringPrototypeSlice(path, last, j); // Matched!
last = j; // Match 1 or more path separators
while (j < len && isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j < len && j !== last) {
// Matched!
last = j; // Match 1 or more non-path separators
while (j < len && !isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j === len || j !== last) {
// We matched a UNC root
device = "\\\\".concat(firstPart, "\\").concat(StringPrototypeSlice(path, last, j));
rootEnd = j;
}
}
}
} else {
rootEnd = 1;
}
} else if (isWindowsDeviceRoot(code) && StringPrototypeCharCodeAt(path, 1) === CHAR_COLON) {
// Possible device root
device = StringPrototypeSlice(path, 0, 2);
rootEnd = 2;
if (len > 2 && isPathSeparator(StringPrototypeCharCodeAt(path, 2))) {
// Treat separator following drive name as an absolute path
// indicator
isAbsolute = true;
rootEnd = 3;
}
}
if (device.length > 0) {
if (resolvedDevice.length > 0) {
if (StringPrototypeToLowerCase(device) !== StringPrototypeToLowerCase(resolvedDevice)) // This path points to another device so it is not applicable
continue;
} else {
resolvedDevice = device;
}
}
if (resolvedAbsolute) {
if (resolvedDevice.length > 0) break;
} else {
resolvedTail = "".concat(StringPrototypeSlice(path, rootEnd), "\\").concat(resolvedTail);
resolvedAbsolute = isAbsolute;
if (isAbsolute && resolvedDevice.length > 0) {
break;
}
}
} // At this point the path should be resolved to a full absolute path,
// but handle relative paths to be safe (might happen when process.cwd()
// fails)
// Normalize the tail path
resolvedTail = normalizeString(resolvedTail, !resolvedAbsolute, '\\', isPathSeparator);
return resolvedAbsolute ? "".concat(resolvedDevice, "\\").concat(resolvedTail) : "".concat(resolvedDevice).concat(resolvedTail) || '.';
},
/**
* @param {string} path
* @returns {string}
*/
normalize: function normalize(path) {
validateString(path, 'path');
var len = path.length;
if (len === 0) return '.';
var rootEnd = 0;
var device;
var isAbsolute = false;
var code = StringPrototypeCharCodeAt(path, 0); // Try to match a root
if (len === 1) {
// `path` contains just a single char, exit early to avoid
// unnecessary work
return isPosixPathSeparator(code) ? '\\' : path;
}
if (isPathSeparator(code)) {
// Possible UNC root
// If we started with a separator, we know we at least have an absolute
// path of some kind (UNC or otherwise)
isAbsolute = true;
if (isPathSeparator(StringPrototypeCharCodeAt(path, 1))) {
// Matched double path separator at beginning
var j = 2;
var last = j; // Match 1 or more non-path separators
while (j < len && !isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j < len && j !== last) {
var firstPart = StringPrototypeSlice(path, last, j); // Matched!
last = j; // Match 1 or more path separators
while (j < len && isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j < len && j !== last) {
// Matched!
last = j; // Match 1 or more non-path separators
while (j < len && !isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j === len) {
// We matched a UNC root only
// Return the normalized version of the UNC root since there
// is nothing left to process
return "\\\\".concat(firstPart, "\\").concat(StringPrototypeSlice(path, last), "\\");
}
if (j !== last) {
// We matched a UNC root with leftovers
device = "\\\\".concat(firstPart, "\\").concat(StringPrototypeSlice(path, last, j));
rootEnd = j;
}
}
}
} else {
rootEnd = 1;
}
} else if (isWindowsDeviceRoot(code) && StringPrototypeCharCodeAt(path, 1) === CHAR_COLON) {
// Possible device root
device = StringPrototypeSlice(path, 0, 2);
rootEnd = 2;
if (len > 2 && isPathSeparator(StringPrototypeCharCodeAt(path, 2))) {
// Treat separator following drive name as an absolute path
// indicator
isAbsolute = true;
rootEnd = 3;
}
}
var tail = rootEnd < len ? normalizeString(StringPrototypeSlice(path, rootEnd), !isAbsolute, '\\', isPathSeparator) : '';
if (tail.length === 0 && !isAbsolute) tail = '.';
if (tail.length > 0 && isPathSeparator(StringPrototypeCharCodeAt(path, len - 1))) tail += '\\';
if (device === undefined) {
return isAbsolute ? "\\".concat(tail) : tail;
}
return isAbsolute ? "".concat(device, "\\").concat(tail) : "".concat(device).concat(tail);
},
/**
* @param {string} path
* @returns {boolean}
*/
isAbsolute: function isAbsolute(path) {
validateString(path, 'path');
var len = path.length;
if (len === 0) return false;
var code = StringPrototypeCharCodeAt(path, 0);
return isPathSeparator(code) || // Possible device root
len > 2 && isWindowsDeviceRoot(code) && StringPrototypeCharCodeAt(path, 1) === CHAR_COLON && isPathSeparator(StringPrototypeCharCodeAt(path, 2));
},
/**
* @param {...string} args
* @returns {string}
*/
join: function join() {
if (arguments.length === 0) return '.';
var joined;
var firstPart;
for (var i = 0; i < arguments.length; ++i) {
var arg = i < 0 || arguments.length <= i ? undefined : arguments[i];
validateString(arg, 'path');
if (arg.length > 0) {
if (joined === undefined) joined = firstPart = arg;else joined += "\\".concat(arg);
}
}
if (joined === undefined) return '.'; // Make sure that the joined path doesn't start with two slashes, because
// normalize() will mistake it for a UNC path then.
//
// This step is skipped when it is very clear that the user actually
// intended to point at a UNC path. This is assumed when the first
// non-empty string arguments starts with exactly two slashes followed by
// at least one more non-slash character.
//
// Note that for normalize() to treat a path as a UNC path it needs to
// have at least 2 components, so we don't filter for that here.
// This means that the user can use join to construct UNC paths from
// a server name and a share name; for example:
// path.join('//server', 'share') -> '\\\\server\\share\\')
var needsReplace = true;
var slashCount = 0;
if (isPathSeparator(StringPrototypeCharCodeAt(firstPart, 0))) {
++slashCount;
var firstLen = firstPart.length;
if (firstLen > 1 && isPathSeparator(StringPrototypeCharCodeAt(firstPart, 1))) {
++slashCount;
if (firstLen > 2) {
if (isPathSeparator(StringPrototypeCharCodeAt(firstPart, 2))) ++slashCount;else {
// We matched a UNC path in the first part
needsReplace = false;
}
}
}
}
if (needsReplace) {
// Find any more consecutive slashes we need to replace
while (slashCount < joined.length && isPathSeparator(StringPrototypeCharCodeAt(joined, slashCount))) {
slashCount++;
} // Replace the slashes if needed
if (slashCount >= 2) joined = "\\".concat(StringPrototypeSlice(joined, slashCount));
}
return win32.normalize(joined);
},
/**
* It will solve the relative path from `from` to `to`, for instancee
* from = 'C:\\orandea\\test\\aaa'
* to = 'C:\\orandea\\impl\\bbb'
* The output of the function should be: '..\\..\\impl\\bbb'
* @param {string} from
* @param {string} to
* @returns {string}
*/
relative: function relative(from, to) {
validateString(from, 'from');
validateString(to, 'to');
if (from === to) return '';
var fromOrig = win32.resolve(from);
var toOrig = win32.resolve(to);
if (fromOrig === toOrig) return '';
from = StringPrototypeToLowerCase(fromOrig);
to = StringPrototypeToLowerCase(toOrig);
if (from === to) return ''; // Trim any leading backslashes
var fromStart = 0;
while (fromStart < from.length && StringPrototypeCharCodeAt(from, fromStart) === CHAR_BACKWARD_SLASH) {
fromStart++;
} // Trim trailing backslashes (applicable to UNC paths only)
var fromEnd = from.length;
while (fromEnd - 1 > fromStart && StringPrototypeCharCodeAt(from, fromEnd - 1) === CHAR_BACKWARD_SLASH) {
fromEnd--;
}
var fromLen = fromEnd - fromStart; // Trim any leading backslashes
var toStart = 0;
while (toStart < to.length && StringPrototypeCharCodeAt(to, toStart) === CHAR_BACKWARD_SLASH) {
toStart++;
} // Trim trailing backslashes (applicable to UNC paths only)
var toEnd = to.length;
while (toEnd - 1 > toStart && StringPrototypeCharCodeAt(to, toEnd - 1) === CHAR_BACKWARD_SLASH) {
toEnd--;
}
var toLen = toEnd - toStart; // Compare paths to find the longest common path from root
var length = fromLen < toLen ? fromLen : toLen;
var lastCommonSep = -1;
var i = 0;
for (; i < length; i++) {
var fromCode = StringPrototypeCharCodeAt(from, fromStart + i);
if (fromCode !== StringPrototypeCharCodeAt(to, toStart + i)) break;else if (fromCode === CHAR_BACKWARD_SLASH) lastCommonSep = i;
} // We found a mismatch before the first common path separator was seen, so
// return the original `to`.
if (i !== length) {
if (lastCommonSep === -1) return toOrig;
} else {
if (toLen > length) {
if (StringPrototypeCharCodeAt(to, toStart + i) === CHAR_BACKWARD_SLASH) {
// We get here if `from` is the exact base path for `to`.
// For example: from='C:\\foo\\bar'; to='C:\\foo\\bar\\baz'
return StringPrototypeSlice(toOrig, toStart + i + 1);
}
if (i === 2) {
// We get here if `from` is the device root.
// For example: from='C:\\'; to='C:\\foo'
return StringPrototypeSlice(toOrig, toStart + i);
}
}
if (fromLen > length) {
if (StringPrototypeCharCodeAt(from, fromStart + i) === CHAR_BACKWARD_SLASH) {
// We get here if `to` is the exact base path for `from`.
// For example: from='C:\\foo\\bar'; to='C:\\foo'
lastCommonSep = i;
} else if (i === 2) {
// We get here if `to` is the device root.
// For example: from='C:\\foo\\bar'; to='C:\\'
lastCommonSep = 3;
}
}
if (lastCommonSep === -1) lastCommonSep = 0;
}
var out = ''; // Generate the relative path based on the path difference between `to` and
// `from`
for (i = fromStart + lastCommonSep + 1; i <= fromEnd; ++i) {
if (i === fromEnd || StringPrototypeCharCodeAt(from, i) === CHAR_BACKWARD_SLASH) {
out += out.length === 0 ? '..' : '\\..';
}
}
toStart += lastCommonSep; // Lastly, append the rest of the destination (`to`) path that comes after
// the common path parts
if (out.length > 0) return "".concat(out).concat(StringPrototypeSlice(toOrig, toStart, toEnd));
if (StringPrototypeCharCodeAt(toOrig, toStart) === CHAR_BACKWARD_SLASH) ++toStart;
return StringPrototypeSlice(toOrig, toStart, toEnd);
},
/**
* @param {string} path
* @returns {string}
*/
toNamespacedPath: function toNamespacedPath(path) {
// Note: this will *probably* throw somewhere.
if (typeof path !== 'string' || path.length === 0) return path;
var resolvedPath = win32.resolve(path);
if (resolvedPath.length <= 2) return path;
if (StringPrototypeCharCodeAt(resolvedPath, 0) === CHAR_BACKWARD_SLASH) {
// Possible UNC root
if (StringPrototypeCharCodeAt(resolvedPath, 1) === CHAR_BACKWARD_SLASH) {
var code = StringPrototypeCharCodeAt(resolvedPath, 2);
if (code !== CHAR_QUESTION_MARK && code !== CHAR_DOT) {
// Matched non-long UNC root, convert the path to a long UNC path
return "\\\\?\\UNC\\".concat(StringPrototypeSlice(resolvedPath, 2));
}
}
} else if (isWindowsDeviceRoot(StringPrototypeCharCodeAt(resolvedPath, 0)) && StringPrototypeCharCodeAt(resolvedPath, 1) === CHAR_COLON && StringPrototypeCharCodeAt(resolvedPath, 2) === CHAR_BACKWARD_SLASH) {
// Matched device root, convert the path to a long UNC path
return "\\\\?\\".concat(resolvedPath);
}
return path;
},
/**
* @param {string} path
* @returns {string}
*/
dirname: function dirname(path) {
validateString(path, 'path');
var len = path.length;
if (len === 0) return '.';
var rootEnd = -1;
var offset = 0;
var code = StringPrototypeCharCodeAt(path, 0);
if (len === 1) {
// `path` contains just a path separator, exit early to avoid
// unnecessary work or a dot.
return isPathSeparator(code) ? path : '.';
} // Try to match a root
if (isPathSeparator(code)) {
// Possible UNC root
rootEnd = offset = 1;
if (isPathSeparator(StringPrototypeCharCodeAt(path, 1))) {
// Matched double path separator at beginning
var j = 2;
var last = j; // Match 1 or more non-path separators
while (j < len && !isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j < len && j !== last) {
// Matched!
last = j; // Match 1 or more path separators
while (j < len && isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j < len && j !== last) {
// Matched!
last = j; // Match 1 or more non-path separators
while (j < len && !isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j === len) {
// We matched a UNC root only
return path;
}
if (j !== last) {
// We matched a UNC root with leftovers
// Offset by 1 to include the separator after the UNC root to
// treat it as a "normal root" on top of a (UNC) root
rootEnd = offset = j + 1;
}
}
}
} // Possible device root
} else if (isWindowsDeviceRoot(code) && StringPrototypeCharCodeAt(path, 1) === CHAR_COLON) {
rootEnd = len > 2 && isPathSeparator(StringPrototypeCharCodeAt(path, 2)) ? 3 : 2;
offset = rootEnd;
}
var end = -1;
var matchedSlash = true;
for (var i = len - 1; i >= offset; --i) {
if (isPathSeparator(StringPrototypeCharCodeAt(path, i))) {
if (!matchedSlash) {
end = i;
break;
}
} else {
// We saw the first non-path separator
matchedSlash = false;
}
}
if (end === -1) {
if (rootEnd === -1) return '.';
end = rootEnd;
}
return StringPrototypeSlice(path, 0, end);
},
/**
* @param {string} path
* @param {string} [ext]
* @returns {string}
*/
basename: function basename(path, ext) {
if (ext !== undefined) validateString(ext, 'ext');
validateString(path, 'path');
var start = 0;
var end = -1;
var matchedSlash = true; // Check for a drive letter prefix so as not to mistake the following
// path separator as an extra separator at the end of the path that can be
// disregarded
if (path.length >= 2 && isWindowsDeviceRoot(StringPrototypeCharCodeAt(path, 0)) && StringPrototypeCharCodeAt(path, 1) === CHAR_COLON) {
start = 2;
}
if (ext !== undefined && ext.length > 0 && ext.length <= path.length) {
if (ext === path) return '';
var extIdx = ext.length - 1;
var firstNonSlashEnd = -1;
for (var i = path.length - 1; i >= start; --i) {
var code = StringPrototypeCharCodeAt(path, i);
if (isPathSeparator(code)) {
// If we reached a path separator that was not part of a set of path
// separators at the end of the string, stop now
if (!matchedSlash) {
start = i + 1;
break;
}
} else {
if (firstNonSlashEnd === -1) {
// We saw the first non-path separator, remember this index in case
// we need it if the extension ends up not matching
matchedSlash = false;
firstNonSlashEnd = i + 1;
}
if (extIdx >= 0) {
// Try to match the explicit extension
if (code === StringPrototypeCharCodeAt(ext, extIdx)) {
if (--extIdx === -1) {
// We matched the extension, so mark this as the end of our path
// component
end = i;
}
} else {
// Extension does not match, so our result is the entire path
// component
extIdx = -1;
end = firstNonSlashEnd;
}
}
}
}
if (start === end) end = firstNonSlashEnd;else if (end === -1) end = path.length;
return StringPrototypeSlice(path, start, end);
}
for (var _i = path.length - 1; _i >= start; --_i) {
if (isPathSeparator(StringPrototypeCharCodeAt(path, _i))) {
// If we reached a path separator that was not part of a set of path
// separators at the end of the string, stop now
if (!matchedSlash) {
start = _i + 1;
break;
}
} else if (end === -1) {
// We saw the first non-path separator, mark this as the end of our
// path component
matchedSlash = false;
end = _i + 1;
}
}
if (end === -1) return '';
return StringPrototypeSlice(path, start, end);
},
/**
* @param {string} path
* @returns {string}
*/
extname: function extname(path) {
validateString(path, 'path');
var start = 0;
var startDot = -1;
var startPart = 0;
var end = -1;
var matchedSlash = true; // Track the state of characters (if any) we see before our first dot and
// after any path separator we find
var preDotState = 0; // Check for a drive letter prefix so as not to mistake the following
// path separator as an extra separator at the end of the path that can be
// disregarded
if (path.length >= 2 && StringPrototypeCharCodeAt(path, 1) === CHAR_COLON && isWindowsDeviceRoot(StringPrototypeCharCodeAt(path, 0))) {
start = startPart = 2;
}
for (var i = path.length - 1; i >= start; --i) {
var code = StringPrototypeCharCodeAt(path, i);
if (isPathSeparator(code)) {
// If we reached a path separator that was not part of a set of path
// separators at the end of the string, stop now
if (!matchedSlash) {
startPart = i + 1;
break;
}
continue;
}
if (end === -1) {
// We saw the first non-path separator, mark this as the end of our
// extension
matchedSlash = false;
end = i + 1;
}
if (code === CHAR_DOT) {
// If this is our first dot, mark it as the start of our extension
if (startDot === -1) startDot = i;else if (preDotState !== 1) preDotState = 1;
} else if (startDot !== -1) {
// We saw a non-dot and non-path separator before our dot, so we should
// have a good chance at having a non-empty extension
preDotState = -1;
}
}
if (startDot === -1 || end === -1 || // We saw a non-dot character immediately before the dot
preDotState === 0 || // The (right-most) trimmed path component is exactly '..'
preDotState === 1 && startDot === end - 1 && startDot === startPart + 1) {
return '';
}
return StringPrototypeSlice(path, startDot, end);
},
format: FunctionPrototypeBind(_format, null, '\\'),
/**
* @param {string} path
* @returns {{
* dir: string;
* root: string;
* base: string;
* name: string;
* ext: string;
* }}
*/
parse: function parse(path) {
validateString(path, 'path');
var ret = {
root: '',
dir: '',
base: '',
ext: '',
name: ''
};
if (path.length === 0) return ret;
var len = path.length;
var rootEnd = 0;
var code = StringPrototypeCharCodeAt(path, 0);
if (len === 1) {
if (isPathSeparator(code)) {
// `path` contains just a path separator, exit early to avoid
// unnecessary work
ret.root = ret.dir = path;
return ret;
}
ret.base = ret.name = path;
return ret;
} // Try to match a root
if (isPathSeparator(code)) {
// Possible UNC root
rootEnd = 1;
if (isPathSeparator(StringPrototypeCharCodeAt(path, 1))) {
// Matched double path separator at beginning
var j = 2;
var last = j; // Match 1 or more non-path separators
while (j < len && !isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j < len && j !== last) {
// Matched!
last = j; // Match 1 or more path separators
while (j < len && isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j < len && j !== last) {
// Matched!
last = j; // Match 1 or more non-path separators
while (j < len && !isPathSeparator(StringPrototypeCharCodeAt(path, j))) {
j++;
}
if (j === len) {
// We matched a UNC root only
rootEnd = j;
} else if (j !== last) {
// We matched a UNC root with leftovers
rootEnd = j + 1;
}
}
}
}
} else if (isWindowsDeviceRoot(code) && StringPrototypeCharCodeAt(path, 1) === CHAR_COLON) {
// Possible device root
if (len <= 2) {
// `path` contains just a drive root, exit early to avoid
// unnecessary work
ret.root = ret.dir = path;
return ret;
}
rootEnd = 2;
if (isPathSeparator(StringPrototypeCharCodeAt(path, 2))) {
if (len === 3) {
// `path` contains just a drive root, exit early to avoid
// unnecessary work
ret.root = ret.dir = path;
return ret;
}
rootEnd = 3;
}
}
if (rootEnd > 0) ret.root = StringPrototypeSlice(path, 0, rootEnd);
var startDot = -1;
var startPart = rootEnd;
var end = -1;
var matchedSlash = true;
var i = path.length - 1; // Track the state of characters (if any) we see before our first dot and
// after any path separator we find
var preDotState = 0; // Get non-dir info
for (; i >= rootEnd; --i) {
code = StringPrototypeCharCodeAt(path, i);
if (isPathSeparator(code)) {
// If we reached a path separator that was not part of a set of path
// separators at the end of the string, stop now
if (!matchedSlash) {
startPart = i + 1;
break;
}
continue;
}
if (end === -1) {
// We saw the first non-path separator, mark this as the end of our
// extension
matchedSlash = false;
end = i + 1;
}
if (code === CHAR_DOT) {
// If this is our first dot, mark it as the start of our extension
if (startDot === -1) startDot = i;else if (preDotState !== 1) preDotState = 1;
} else if (startDot !== -1) {
// We saw a non-dot and non-path separator before our dot, so we should
// have a good chance at having a non-empty extension
preDotState = -1;
}
}
if (end !== -1) {
if (startDot === -1 || // We saw a non-dot character immediately before the dot
preDotState === 0 || // The (right-most) trimmed path component is exactly '..'
preDotState === 1 && startDot === end - 1 && startDot === startPart + 1) {
ret.base = ret.name = StringPrototypeSlice(path, startPart, end);
} else {
ret.name = StringPrototypeSlice(path, startPart, startDot);
ret.base = StringPrototypeSlice(path, startPart, end);
ret.ext = StringPrototypeSlice(path, startDot, end);
}
} // If the directory is the root, use the entire root as the `dir` including
// the trailing slash if any (`C:\abc` -> `C:\`). Otherwise, strip out the
// trailing slash (`C:\abc\def` -> `C:\abc`).
if (startPart > 0 && startPart !== rootEnd) ret.dir = StringPrototypeSlice(path, 0, startPart - 1);else ret.dir = ret.root;
return ret;
},
sep: '\\',
delimiter: ';',
win32: null,
posix: null
};
var posixCwd = function () {
if (platformIsWin32) {
// Converts Windows' backslash path separators to POSIX forward slashes
// and truncates any drive indicator
var regexp = /\\/g;
return function () {
var cwd = StringPrototypeReplace(process.cwd(), regexp, '/');
return StringPrototypeSlice(cwd, StringPrototypeIndexOf(cwd, '/'));
};
} // We're already on POSIX, no need for any transformations
return function () {
return process.cwd();
};
}();
var posix = {
/**
* path.resolve([from ...], to)
* @param {...string} args
* @returns {string}
*/
resolve: function resolve() {
var resolvedPath = '';
var resolvedAbsolute = false;
for (var i = arguments.length - 1; i >= -1 && !resolvedAbsolute; i--) {
var path = i >= 0 ? i < 0 || arguments.length <= i ? undefined : arguments[i] : posixCwd();
validateString(path, 'path'); // Skip empty entries
if (path.length === 0) {
continue;
}
resolvedPath = "".concat(path, "/").concat(resolvedPath);
resolvedAbsolute = StringPrototypeCharCodeAt(path, 0) === CHAR_FORWARD_SLASH;
} // At this point the path should be resolved to a full absolute path, but
// handle relative paths to be safe (might happen when process.cwd() fails)
// Normalize the path
resolvedPath = normalizeString(resolvedPath, !resolvedAbsolute, '/', isPosixPathSeparator);
if (resolvedAbsolute) {
return "/".concat(resolvedPath);
}
return resolvedPath.length > 0 ? resolvedPath : '.';
},
/**
* @param {string} path
* @returns {string}
*/
normalize: function normalize(path) {
validateString(path, 'path');
if (path.length === 0) return '.';
var isAbsolute = StringPrototypeCharCodeAt(path, 0) === CHAR_FORWARD_SLASH;
var trailingSeparator = StringPrototypeCharCodeAt(path, path.length - 1) === CHAR_FORWARD_SLASH; // Normalize the path
path = normalizeString(path, !isAbsolute, '/', isPosixPathSeparator);
if (path.length === 0) {
if (isAbsolute) return '/';
return trailingSeparator ? './' : '.';
}
if (trailingSeparator) path += '/';
return isAbsolute ? "/".concat(path) : path;
},
/**
* @param {string} path
* @returns {boolean}
*/
isAbsolute: function isAbsolute(path) {
validateString(path, 'path');
return path.length > 0 && StringPrototypeCharCodeAt(path, 0) === CHAR_FORWARD_SLASH;
},
/**
* @param {...string} args
* @returns {string}
*/
join: function join() {
if (arguments.length === 0) return '.';
var joined;
for (var i = 0; i < arguments.length; ++i) {
var arg = i < 0 || arguments.length <= i ? undefined : arguments[i];
validateString(arg, 'path');
if (arg.length > 0) {
if (joined === undefined) joined = arg;else joined += "/".concat(arg);
}
}
if (joined === undefined) return '.';
return posix.normalize(joined);
},
/**
* @param {string} from
* @param {string} to
* @returns {string}
*/
relative: function relative(from, to) {
validateString(from, 'from');
validateString(to, 'to');
if (from === to) return ''; // Trim leading forward slashes.
from = posix.resolve(from);
to = posix.resolve(to);
if (from === to) return '';
var fromStart = 1;
var fromEnd = from.length;
var fromLen = fromEnd - fromStart;
var toStart = 1;
var toLen = to.length - toStart; // Compare paths to find the longest common path from root
var length = fromLen < toLen ? fromLen : toLen;
var lastCommonSep = -1;
var i = 0;
for (; i < length; i++) {
var fromCode = StringPrototypeCharCodeAt(from, fromStart + i);
if (fromCode !== StringPrototypeCharCodeAt(to, toStart + i)) break;else if (fromCode === CHAR_FORWARD_SLASH) lastCommonSep = i;
}
if (i === length) {
if (toLen > length) {
if (StringPrototypeCharCodeAt(to, toStart + i) === CHAR_FORWARD_SLASH) {
// We get here if `from` is the exact base path for `to`.
// For example: from='/foo/bar'; to='/foo/bar/baz'
return StringPrototypeSlice(to, toStart + i + 1);
}
if (i === 0) {
// We get here if `from` is the root
// For example: from='/'; to='/foo'
return StringPrototypeSlice(to, toStart + i);
}
} else if (fromLen > length) {
if (StringPrototypeCharCodeAt(from, fromStart + i) === CHAR_FORWARD_SLASH) {
// We get here if `to` is the exact base path for `from`.
// For example: from='/foo/bar/baz'; to='/foo/bar'
lastCommonSep = i;
} else if (i === 0) {
// We get here if `to` is the root.
// For example: from='/foo/bar'; to='/'
lastCommonSep = 0;
}
}
}
var out = ''; // Generate the relative path based on the path difference between `to`
// and `from`.
for (i = fromStart + lastCommonSep + 1; i <= fromEnd; ++i) {
if (i === fromEnd || StringPrototypeCharCodeAt(from, i) === CHAR_FORWARD_SLASH) {
out += out.length === 0 ? '..' : '/..';
}
} // Lastly, append the rest of the destination (`to`) path that comes after
// the common path parts.
return "".concat(out).concat(StringPrototypeSlice(to, toStart + lastCommonSep));
},
/**
* @param {string} path
* @returns {string}
*/
toNamespacedPath: function toNamespacedPath(path) {
// Non-op on posix systems
return path;
},
/**
* @param {string} path
* @returns {string}
*/
dirname: function dirname(path) {
validateString(path, 'path');
if (path.length === 0) return '.';
var hasRoot = StringPrototypeCharCodeAt(path, 0) === CHAR_FORWARD_SLASH;
var end = -1;
var matchedSlash = true;
for (var i = path.length - 1; i >= 1; --i) {
if (StringPrototypeCharCodeAt(path, i) === CHAR_FORWARD_SLASH) {
if (!matchedSlash) {
end = i;
break;
}
} else {
// We saw the first non-path separator
matchedSlash = false;
}
}
if (end === -1) return hasRoot ? '/' : '.';
if (hasRoot && end === 1) return '//';
return StringPrototypeSlice(path, 0, end);
},
/**
* @param {string} path
* @param {string} [ext]
* @returns {string}
*/
basename: function basename(path, ext) {
if (ext !== undefined) validateString(ext, 'ext');
validateString(path, 'path');
var start = 0;
var end = -1;
var matchedSlash = true;
if (ext !== undefined && ext.length > 0 && ext.length <= path.length) {
if (ext === path) return '';
var extIdx = ext.length - 1;
var firstNonSlashEnd = -1;
for (var i = path.length - 1; i >= 0; --i) {
var code = StringPrototypeCharCodeAt(path, i);
if (code === CHAR_FORWARD_SLASH) {
// If we reached a path separator that was not part of a set of path
// separators at the end of the string, stop now
if (!matchedSlash) {
start = i + 1;
break;
}
} else {
if (firstNonSlashEnd === -1) {
// We saw the first non-path separator, remember this index in case
// we need it if the extension ends up not matching
matchedSlash = false;
firstNonSlashEnd = i + 1;
}
if (extIdx >= 0) {
// Try to match the explicit extension
if (code === StringPrototypeCharCodeAt(ext, extIdx)) {
if (--extIdx === -1) {
// We matched the extension, so mark this as the end of our path
// component
end = i;
}
} else {
// Extension does not match, so our result is the entire path
// component
extIdx = -1;
end = firstNonSlashEnd;
}
}
}
}
if (start === end) end = firstNonSlashEnd;else if (end === -1) end = path.length;
return StringPrototypeSlice(path, start, end);
}
for (var _i2 = path.length - 1; _i2 >= 0; --_i2) {
if (StringPrototypeCharCodeAt(path, _i2) === CHAR_FORWARD_SLASH) {
// If we reached a path separator that was not part of a set of path
// separators at the end of the string, stop now
if (!matchedSlash) {
start = _i2 + 1;
break;
}
} else if (end === -1) {
// We saw the first non-path separator, mark this as the end of our
// path component
matchedSlash = false;
end = _i2 + 1;
}
}
if (end === -1) return '';
return StringPrototypeSlice(path, start, end);
},
/**
* @param {string} path
* @returns {string}
*/
extname: function extname(path) {
validateString(path, 'path');
var startDot = -1;
var startPart = 0;
var end = -1;
var matchedSlash = true; // Track the state of characters (if any) we see before our first dot and
// after any path separator we find
var preDotState = 0;
for (var i = path.length - 1; i >= 0; --i) {
var code = StringPrototypeCharCodeAt(path, i);
if (code === CHAR_FORWARD_SLASH) {
// If we reached a path separator that was not part of a set of path
// separators at the end of the string, stop now
if (!matchedSlash) {
startPart = i + 1;
break;
}
continue;
}
if (end === -1) {
// We saw the first non-path separator, mark this as the end of our
// extension
matchedSlash = false;
end = i + 1;
}
if (code === CHAR_DOT) {
// If this is our first dot, mark it as the start of our extension
if (startDot === -1) startDot = i;else if (preDotState !== 1) preDotState = 1;
} else if (startDot !== -1) {
// We saw a non-dot and non-path separator before our dot, so we should
// have a good chance at having a non-empty extension
preDotState = -1;
}
}
if (startDot === -1 || end === -1 || // We saw a non-dot character immediately before the dot
preDotState === 0 || // The (right-most) trimmed path component is exactly '..'
preDotState === 1 && startDot === end - 1 && startDot === startPart + 1) {
return '';
}
return StringPrototypeSlice(path, startDot, end);
},
format: FunctionPrototypeBind(_format, null, '/'),
/**
* @param {string} path
* @returns {{
* dir: string;
* root: string;
* base: string;
* name: string;
* ext: string;
* }}
*/
parse: function parse(path) {
validateString(path, 'path');
var ret = {
root: '',
dir: '',
base: '',
ext: '',
name: ''
};
if (path.length === 0) return ret;
var isAbsolute = StringPrototypeCharCodeAt(path, 0) === CHAR_FORWARD_SLASH;
var start;
if (isAbsolute) {
ret.root = '/';
start = 1;
} else {
start = 0;
}
var startDot = -1;
var startPart = 0;
var end = -1;
var matchedSlash = true;
var i = path.length - 1; // Track the state of characters (if any) we see before our first dot and
// after any path separator we find
var preDotState = 0; // Get non-dir info
for (; i >= start; --i) {
var code = StringPrototypeCharCodeAt(path, i);
if (code === CHAR_FORWARD_SLASH) {
// If we reached a path separator that was not part of a set of path
// separators at the end of the string, stop now
if (!matchedSlash) {
startPart = i + 1;
break;
}
continue;
}
if (end === -1) {
// We saw the first non-path separator, mark this as the end of our
// extension
matchedSlash = false;
end = i + 1;
}
if (code === CHAR_DOT) {
// If this is our first dot, mark it as the start of our extension
if (startDot === -1) startDot = i;else if (preDotState !== 1) preDotState = 1;
} else if (startDot !== -1) {
// We saw a non-dot and non-path separator before our dot, so we should
// have a good chance at having a non-empty extension
preDotState = -1;
}
}
if (end !== -1) {
var _start = startPart === 0 && isAbsolute ? 1 : startPart;
if (startDot === -1 || // We saw a non-dot character immediately before the dot
preDotState === 0 || // The (right-most) trimmed path component is exactly '..'
preDotState === 1 && startDot === end - 1 && startDot === startPart + 1) {
ret.base = ret.name = StringPrototypeSlice(path, _start, end);
} else {
ret.name = StringPrototypeSlice(path, _start, startDot);
ret.base = StringPrototypeSlice(path, _start, end);
ret.ext = StringPrototypeSlice(path, startDot, end);
}
}
if (startPart > 0) ret.dir = StringPrototypeSlice(path, 0, startPart - 1);else if (isAbsolute) ret.dir = '/';
return ret;
},
sep: '/',
delimiter: ':',
win32: null,
posix: null
};
posix.win32 = win32.win32 = win32;
posix.posix = win32.posix = posix; // Legacy internal API, docs-only deprecated: DEP0080
win32._makeLong = win32.toNamespacedPath;
posix._makeLong = posix.toNamespacedPath;
module.exports = platformIsWin32 ? win32 : posix;