@cplieger/deadset-ts 8.0.0-dev.11 → 8.0.0-dev.13

package/src/text/shell.ts

@@ -1,161 +0,0 @@
-/**
- * Shell command lines read as text, never run: the commands a line runs, the tokens the
- * Contract splits a command into, and the words POSIX token recognition reads.
- */
-
-/**
- * The shell's list and pipeline operators, each of which ends one command. An `&` beside a
- * redirection's `<` or `>` (`2>&1`, and bash's `&>log`) is part of the redirection.
- */
-const COMMAND_OPERATORS = /&&|\|\||;|\||(?<![<>])&(?!>)/u;
-
-/** Where a command splits into tokens: white space and the operators the Contract names. */
-const TOKEN_SEPARATORS = /\s+|&&|\|\||;|\|/u;
-
-/** A text written inside one pair of quotes. */
-const QUOTED = /^(["'])(.*)\1$/su;
-
-/** The commands one line runs, the text between two of the shell's list and pipeline operators. */
-export function commandsOf(line: string): readonly string[] {
-  return line.split(COMMAND_OPERATORS);
-}
-
-/** One text with one pair of enclosing quotes removed. */
-export function unquoted(text: string): string {
-  return text.replace(QUOTED, "$2");
-}
-
-/**
- * Every token one script's command line holds, each with one pair of enclosing quotes
- * removed. A token holding a wildcard is a pattern rather than a path, and names no
- * entry point.
- */
-export function scriptTokens(command: string): readonly string[] {
-  return command
-    .split(TOKEN_SEPARATORS)
-    .map(unquoted)
-    .filter((token) => token !== "" && !token.includes("*"));
-}
-
-/** One shell line up to its comment: a `#` that starts a word outside quotes. */
-export function withoutComment(line: string): string {
-  let quote = "";
-  for (let at = 0; at < line.length; at += 1) {
-    const char = line[at];
-    if (quote !== "") {
-      if (char === quote) {
-        quote = "";
-      } else if (char === "\\" && quote === '"') {
-        at += 1;
-      }
-    } else if (char === "'" || char === '"') {
-      quote = char;
-    } else if (char === "\\") {
-      at += 1;
-    } else if (char === "#" && (at === 0 || /\s/u.test(line[at - 1] ?? ""))) {
-      return line.slice(0, at);
-    }
-  }
-  return line;
-}
-
-/** The characters that end a word outside quotes and separate nothing: blanks. */
-// Stryker disable next-line Regex: its \v to \V mutant is no valid escape under the u flag, so the whole module fails to parse
-const BLANK = /[ \t\r\f\v]/u;
-
-/** The characters that end a word outside quotes and begin an operator: newline included. */
-const OPERATOR = /[\n;&|()<>]/u;
-
-/** The characters a backslash escapes inside double quotes; before any other it stands for itself. */
-const DOUBLE_QUOTE_ESCAPES = new Set(["$", "`", '"', "\\"]);
-
-/**
- * The words of one command line, as POSIX token recognition reads quotes, backslashes,
- * line continuations, comments and operators. A word an expansion may change (an unquoted
- * `$`, backquote, glob character or leading `~`, or a `$` or backquote in double quotes)
- * is computed, and so is the rest of a line whose quote is never closed. A computed word
- * and each operator are undefined, so a flag's file is never read across them.
- */
-export function shellWords(command: string): readonly (string | undefined)[] {
-  const words: (string | undefined)[] = [];
-  let word: string | undefined;
-  let computed = false;
-  const end = (): void => {
-    if (word !== undefined) {
-      words.push(computed ? undefined : word);
-    }
-    word = undefined;
-    computed = false;
-  };
-  for (let at = 0; at < command.length; at += 1) {
-    const char = command.charAt(at);
-    if (char === "\\" && command.charAt(at + 1) === "\n") {
-      at += 1;
-    } else if (BLANK.test(char)) {
-      end();
-    } else if (OPERATOR.test(char)) {
-      end();
-      words.push(undefined);
-    } else if (char === "#" && word === undefined) {
-      const newline = command.indexOf("\n", at);
-      at = newline < 0 ? command.length : newline - 1;
-    } else if (char === "\\") {
-      at += 1;
-      word = (word ?? "") + command.charAt(at);
-    } else if (char === "'" || char === '"') {
-      const close = closingQuote(command, at);
-      if (close < 0) {
-        word ??= "";
-        computed = true;
-        break;
-      }
-      const quoted =
-        char === "'"
-          ? { text: command.slice(at + 1, close), computed: false }
-          : doubleQuoted(command.slice(at + 1, close));
-      word = (word ?? "") + quoted.text;
-      computed ||= quoted.computed;
-      at = close;
-    } else {
-      computed ||= /[$`*?[]/u.test(char) || (char === "~" && word === undefined);
-      word = (word ?? "") + char;
-    }
-  }
-  end();
-  return words;
-}
-
-/** The text between one pair of double quotes, unescaped, and whether an expansion may change it. */
-function doubleQuoted(inner: string): { readonly text: string; readonly computed: boolean } {
-  let text = "";
-  let computed = false;
-  for (let at = 0; at < inner.length; at += 1) {
-    const char = inner.charAt(at);
-    const next = inner.charAt(at + 1);
-    if (char === "\\" && next === "\n") {
-      at += 1;
-    } else if (char === "\\" && DOUBLE_QUOTE_ESCAPES.has(next)) {
-      text += next;
-      at += 1;
-    } else {
-      computed ||= char === "$" || char === "`";
-      text += char;
-    }
-  }
-  return { text, computed };
-}
-
-/** The index of the quote that closes the one at `open`, or -1 where none does. */
-function closingQuote(command: string, open: number): number {
-  const quote = command.charAt(open);
-  for (let at = open + 1; at < command.length; at += 1) {
-    const char = command.charAt(at);
-    if (char === quote) {
-      return at;
-    }
-    if (quote === '"' && char === "\\") {
-      at += 1;
-    }
-  }
-  return -1;
-}