denoland-deno/ext/node/polyfills/path/glob.ts

// Copyright 2018-2023 the Deno authors. All rights reserved. MIT license.

import { isWindows, osType } from "internal:deno_node/polyfills/_util/os.ts";
import {
  SEP,
  SEP_PATTERN,
} from "internal:deno_node/polyfills/path/separator.ts";
import * as _win32 from "internal:deno_node/polyfills/path/win32.ts";
import * as _posix from "internal:deno_node/polyfills/path/posix.ts";
import type { OSType } from "internal:deno_node/polyfills/_util/os.ts";

const path = isWindows ? _win32 : _posix;
const { join, normalize } = path;

export interface GlobOptions {
  /** Extended glob syntax.
   * See https://www.linuxjournal.com/content/bash-extended-globbing.
   *
   * @default {true}
   */
  extended?: boolean;
  /** Globstar syntax.
   * See https://www.linuxjournal.com/content/globstar-new-bash-globbing-option.
   * If false, `**` is treated like `*`.
   *
   * @default {true}
   */
  globstar?: boolean;
  /** Whether globstar should be case-insensitive. */
  caseInsensitive?: boolean;
  /** Operating system. Defaults to the native OS. */
  os?: OSType;
}

export type GlobToRegExpOptions = GlobOptions;

const regExpEscapeChars = [
  "!",
  "$",
  "(",
  ")",
  "*",
  "+",
  ".",
  "=",
  "?",
  "[",
  "\\",
  "^",
  "{",
  "|",
];
const rangeEscapeChars = ["-", "\\", "]"];

/** Convert a glob string to a regular expression.
 *
 * Tries to match bash glob expansion as closely as possible.
 *
 * Basic glob syntax:
 * - `*` - Matches everything without leaving the path segment.
 * - `?` - Matches any single character.
 * - `{foo,bar}` - Matches `foo` or `bar`.
 * - `[abcd]` - Matches `a`, `b`, `c` or `d`.
 * - `[a-d]` - Matches `a`, `b`, `c` or `d`.
 * - `[!abcd]` - Matches any single character besides `a`, `b`, `c` or `d`.
 * - `[[:<class>:]]` - Matches any character belonging to `<class>`.
 *     - `[[:alnum:]]` - Matches any digit or letter.
 *     - `[[:digit:]abc]` - Matches any digit, `a`, `b` or `c`.
 *     - See https://facelessuser.github.io/wcmatch/glob/#posix-character-classes
 *       for a complete list of supported character classes.
 * - `\` - Escapes the next character for an `os` other than `"windows"`.
 * - \` - Escapes the next character for `os` set to `"windows"`.
 * - `/` - Path separator.
 * - `\` - Additional path separator only for `os` set to `"windows"`.
 *
 * Extended syntax:
 * - Requires `{ extended: true }`.
 * - `?(foo|bar)` - Matches 0 or 1 instance of `{foo,bar}`.
 * - `@(foo|bar)` - Matches 1 instance of `{foo,bar}`. They behave the same.
 * - `*(foo|bar)` - Matches _n_ instances of `{foo,bar}`.
 * - `+(foo|bar)` - Matches _n > 0_ instances of `{foo,bar}`.
 * - `!(foo|bar)` - Matches anything other than `{foo,bar}`.
 * - See https://www.linuxjournal.com/content/bash-extended-globbing.
 *
 * Globstar syntax:
 * - Requires `{ globstar: true }`.
 * - `**` - Matches any number of any path segments.
 *     - Must comprise its entire path segment in the provided glob.
 * - See https://www.linuxjournal.com/content/globstar-new-bash-globbing-option.
 *
 * Note the following properties:
 * - The generated `RegExp` is anchored at both start and end.
 * - Repeating and trailing separators are tolerated. Trailing separators in the
 *   provided glob have no meaning and are discarded.
 * - Absolute globs will only match absolute paths, etc.
 * - Empty globs will match nothing.
 * - Any special glob syntax must be contained to one path segment. For example,
 *   `?(foo|bar/baz)` is invalid. The separator will take precedence and the
 *   first segment ends with an unclosed group.
 * - If a path segment ends with unclosed groups or a dangling escape prefix, a
 *   parse error has occurred. Every character for that segment is taken
 *   literally in this event.
 *
 * Limitations:
 * - A negative group like `!(foo|bar)` will wrongly be converted to a negative
 *   look-ahead followed by a wildcard. This means that `!(foo).js` will wrongly
 *   fail to match `foobar.js`, even though `foobar` is not `foo`. Effectively,
 *   `!(foo|bar)` is treated like `!(@(foo|bar)*)`. This will work correctly if
 *   the group occurs not nested at the end of the segment. */
export function globToRegExp(
  glob: string,
  {
    extended = true,
    globstar: globstarOption = true,
    os = osType,
    caseInsensitive = false,
  }: GlobToRegExpOptions = {},
): RegExp {
  if (glob == "") {
    return /(?!)/;
  }

  const sep = os == "windows" ? "(?:\\\\|/)+" : "/+";
  const sepMaybe = os == "windows" ? "(?:\\\\|/)*" : "/*";
  const seps = os == "windows" ? ["\\", "/"] : ["/"];
  const globstar = os == "windows"
    ? "(?:[^\\\\/]*(?:\\\\|/|$)+)*"
    : "(?:[^/]*(?:/|$)+)*";
  const wildcard = os == "windows" ? "[^\\\\/]*" : "[^/]*";
  const escapePrefix = os == "windows" ? "`" : "\\";

  // Remove trailing separators.
  let newLength = glob.length;
  for (; newLength > 1 && seps.includes(glob[newLength - 1]); newLength--);
  glob = glob.slice(0, newLength);

  let regExpString = "";

  // Terminates correctly. Trust that `j` is incremented every iteration.
  for (let j = 0; j < glob.length;) {
    let segment = "";
    const groupStack: string[] = [];
    let inRange = false;
    let inEscape = false;
    let endsWithSep = false;
    let i = j;

    // Terminates with `i` at the non-inclusive end of the current segment.
    for (; i < glob.length && !seps.includes(glob[i]); i++) {
      if (inEscape) {
        inEscape = false;
        const escapeChars = inRange ? rangeEscapeChars : regExpEscapeChars;
        segment += escapeChars.includes(glob[i]) ? `\\${glob[i]}` : glob[i];
        continue;
      }

      if (glob[i] == escapePrefix) {
        inEscape = true;
        continue;
      }

      if (glob[i] == "[") {
        if (!inRange) {
          inRange = true;
          segment += "[";
          if (glob[i + 1] == "!") {
            i++;
            segment += "^";
          } else if (glob[i + 1] == "^") {
            i++;
            segment += "\\^";
          }
          continue;
        } else if (glob[i + 1] == ":") {
          let k = i + 1;
          let value = "";
          while (glob[k + 1] != null && glob[k + 1] != ":") {
            value += glob[k + 1];
            k++;
          }
          if (glob[k + 1] == ":" && glob[k + 2] == "]") {
            i = k + 2;
            if (value == "alnum") segment += "\\dA-Za-z";
            else if (value == "alpha") segment += "A-Za-z";
            else if (value == "ascii") segment += "\x00-\x7F";
            else if (value == "blank") segment += "\t ";
            else if (value == "cntrl") segment += "\x00-\x1F\x7F";
            else if (value == "digit") segment += "\\d";
            else if (value == "graph") segment += "\x21-\x7E";
            else if (value == "lower") segment += "a-z";
            else if (value == "print") segment += "\x20-\x7E";
            else if (value == "punct") {
              segment += "!\"#$%&'()*+,\\-./:;<=>?@[\\\\\\]^_‘{|}~";
            } else if (value == "space") segment += "\\s\v";
            else if (value == "upper") segment += "A-Z";
            else if (value == "word") segment += "\\w";
            else if (value == "xdigit") segment += "\\dA-Fa-f";
            continue;
          }
        }
      }

      if (glob[i] == "]" && inRange) {
        inRange = false;
        segment += "]";
        continue;
      }

      if (inRange) {
        if (glob[i] == "\\") {
          segment += `\\\\`;
        } else {
          segment += glob[i];
        }
        continue;
      }

      if (
        glob[i] == ")" && groupStack.length > 0 &&
        groupStack[groupStack.length - 1] != "BRACE"
      ) {
        segment += ")";
        const type = groupStack.pop()!;
        if (type == "!") {
          segment += wildcard;
        } else if (type != "@") {
          segment += type;
        }
        continue;
      }

      if (
        glob[i] == "|" && groupStack.length > 0 &&
        groupStack[groupStack.length - 1] != "BRACE"
      ) {
        segment += "|";
        continue;
      }

      if (glob[i] == "+" && extended && glob[i + 1] == "(") {
        i++;
        groupStack.push("+");
        segment += "(?:";
        continue;
      }

      if (glob[i] == "@" && extended && glob[i + 1] == "(") {
        i++;
        groupStack.push("@");
        segment += "(?:";
        continue;
      }

      if (glob[i] == "?") {
        if (extended && glob[i + 1] == "(") {
          i++;
          groupStack.push("?");
          segment += "(?:";
        } else {
          segment += ".";
        }
        continue;
      }

      if (glob[i] == "!" && extended && glob[i + 1] == "(") {
        i++;
        groupStack.push("!");
        segment += "(?!";
        continue;
      }

      if (glob[i] == "{") {
        groupStack.push("BRACE");
        segment += "(?:";
        continue;
      }

      if (glob[i] == "}" && groupStack[groupStack.length - 1] == "BRACE") {
        groupStack.pop();
        segment += ")";
        continue;
      }

      if (glob[i] == "," && groupStack[groupStack.length - 1] == "BRACE") {
        segment += "|";
        continue;
      }

      if (glob[i] == "*") {
        if (extended && glob[i + 1] == "(") {
          i++;
          groupStack.push("*");
          segment += "(?:";
        } else {
          const prevChar = glob[i - 1];
          let numStars = 1;
          while (glob[i + 1] == "*") {
            i++;
            numStars++;
          }
          const nextChar = glob[i + 1];
          if (
            globstarOption && numStars == 2 &&
            [...seps, undefined].includes(prevChar) &&
            [...seps, undefined].includes(nextChar)
          ) {
            segment += globstar;
            endsWithSep = true;
          } else {
            segment += wildcard;
          }
        }
        continue;
      }

      segment += regExpEscapeChars.includes(glob[i]) ? `\\${glob[i]}` : glob[i];
    }

    // Check for unclosed groups or a dangling backslash.
    if (groupStack.length > 0 || inRange || inEscape) {
      // Parse failure. Take all characters from this segment literally.
      segment = "";
      for (const c of glob.slice(j, i)) {
        segment += regExpEscapeChars.includes(c) ? `\\${c}` : c;
        endsWithSep = false;
      }
    }

    regExpString += segment;
    if (!endsWithSep) {
      regExpString += i < glob.length ? sep : sepMaybe;
      endsWithSep = true;
    }

    // Terminates with `i` at the start of the next segment.
    while (seps.includes(glob[i])) i++;

    // Check that the next value of `j` is indeed higher than the current value.
    if (!(i > j)) {
      throw new Error("Assertion failure: i > j (potential infinite loop)");
    }
    j = i;
  }

  regExpString = `^${regExpString}$`;
  return new RegExp(regExpString, caseInsensitive ? "i" : "");
}

/** Test whether the given string is a glob */
export function isGlob(str: string): boolean {
  const chars: Record<string, string> = { "{": "}", "(": ")", "[": "]" };
  const regex =
    /\\(.)|(^!|\*|\?|[\].+)]\?|\[[^\\\]]+\]|\{[^\\}]+\}|\(\?[:!=][^\\)]+\)|\([^|]+\|[^\\)]+\))/;

  if (str === "") {
    return false;
  }

  let match: RegExpExecArray | null;

  while ((match = regex.exec(str))) {
    if (match[2]) return true;
    let idx = match.index + match[0].length;

    // if an open bracket/brace/paren is escaped,
    // set the index to the next closing character
    const open = match[1];
    const close = open ? chars[open] : null;
    if (open && close) {
      const n = str.indexOf(close, idx);
      if (n !== -1) {
        idx = n + 1;
      }
    }

    str = str.slice(idx);
  }

  return false;
}

/** Like normalize(), but doesn't collapse "**\/.." when `globstar` is true. */
export function normalizeGlob(
  glob: string,
  { globstar = false }: GlobOptions = {},
): string {
  if (glob.match(/\0/g)) {
    throw new Error(`Glob contains invalid characters: "${glob}"`);
  }
  if (!globstar) {
    return normalize(glob);
  }
  const s = SEP_PATTERN.source;
  const badParentPattern = new RegExp(
    `(?<=(${s}|^)\\*\\*${s})\\.\\.(?=${s}|$)`,
    "g",
  );
  return normalize(glob.replace(badParentPattern, "\0")).replace(/\0/g, "..");
}

/** Like join(), but doesn't collapse "**\/.." when `globstar` is true. */
export function joinGlobs(
  globs: string[],
  { extended = true, globstar = false }: GlobOptions = {},
): string {
  if (!globstar || globs.length == 0) {
    return join(...globs);
  }
  if (globs.length === 0) return ".";
  let joined: string | undefined;
  for (const glob of globs) {
    const path = glob;
    if (path.length > 0) {
      if (!joined) joined = path;
      else joined += `${SEP}${path}`;
    }
  }
  if (!joined) return ".";
  return normalizeGlob(joined, { extended, globstar });
}