将 shell 引用的字符串解析为与 execv 兼容的参数向量

Question

假设我有如下字符串

echo "foo" "bar\"blub""baz" "'" "\"" foo\ bar "\" '\'' """"       Lots" "of\ whitespace

现在我想将上面的字符串作为命令执行，就好像它是通过调用 Unix.execv 输入到 shell 中一样。如果我没有犯任何错误，那么 shell 会将上面的内容解析为以下 ocaml 列表：

["echo"; "foo"; "bar\"blubbaz"; "'"; "\""; "foo bar"; "\"; "'", "", "Lots of whitespace"]

哪个库让我从原始字符串到解析后的列表？

最终我想把结果列表交给Unix.execvpe。还有 Unix.open_process_full 可以使用 /bin/sh 处理我的原始字符串，但我发现在不使用 /bin/sh 的情况下直接调用外部程序时，我的应用程序快了 16%。现在我希望能够接受更多的输入字符串，包括引号和转义符。

我必须推出自己的解析器吗？

存在 POSIX 函数 wordexp 但包装该函数并不能解决我的问题，因为 wordexp 做的比我想要的更多（命令替换、计算 glob、替换tilda 和环境变量）。

我只想解决引号和转义问题。

Answer 1

我使用 ocamllex 制定了解决此问题的方法。将它张贴在这里以防其他人想要做类似的事情。它应该很容易扩展，以包含超出当前支持的转义字符和其他 shell 功能范围的功能。

{
  exception UnknownShellEscape of string
  exception UnmatchedChar of char
  let buf_from_str str =
    let buf = Buffer.create 16 in
    Buffer.add_string buf str;
    buf
}

let safechars = [^ '"' ''' '\' ' ' '\t']+
let space = [ ' ' '\t' ]+

rule shell_command argv = parse
 | space         { shell_command argv lexbuf }
 | safechars     { uquote argv (buf_from_str (Lexing.lexeme lexbuf)) lexbuf }
 | '\' '"'      { uquote argv (buf_from_str "\"") lexbuf }
 | '\' '''      { uquote argv (buf_from_str "'") lexbuf }
 | '\' '\'     { uquote argv (buf_from_str "\") lexbuf }
 | '\' ' '      { uquote argv (buf_from_str " ") lexbuf }
 | '\' _ as c   { raise (UnknownShellEscape c) }
 | '"'           { dquote argv (Buffer.create 16) lexbuf }
 | '''           { squote argv (Buffer.create 16) lexbuf }
 | _ as c        { raise (UnmatchedChar c) }
 | eof { List.rev argv }
and uquote argv buf = parse
 | (space|eof) { shell_command ((Buffer.contents buf)::argv) lexbuf }
 | '\' '"'    { Buffer.add_string buf "\""; uquote argv buf lexbuf }
 | '\' '''    { Buffer.add_string buf "'"; uquote argv buf lexbuf }
 | '\' '\'   { Buffer.add_string buf "\"; uquote argv buf lexbuf }
 | '\' ' '    { Buffer.add_string buf " "; uquote argv buf lexbuf }
 | '\' _ as c { raise (UnknownShellEscape c) }
 | '"'         { dquote argv buf lexbuf }
 | '''         { squote argv buf lexbuf }
 | safechars   { Buffer.add_string buf (Lexing.lexeme lexbuf); uquote argv buf lexbuf }
 | _ as c      { raise (UnmatchedChar c) }
and dquote argv buf = parse
 | '"' (space|eof) { shell_command ((Buffer.contents buf)::argv) lexbuf }
 | '"' '"'         { dquote argv buf lexbuf }
 | '"' '''         { squote argv buf lexbuf }
 | '"'             { uquote argv buf lexbuf }
 | '\' '"'        { Buffer.add_string buf "\""; dquote argv buf lexbuf }
 | '\' '\'       { Buffer.add_string buf "\"; dquote argv buf lexbuf }
 | '\' _ as c     { raise (UnknownShellEscape c) }
 | [^ '"' '\' ]+  { Buffer.add_string buf (Lexing.lexeme lexbuf); dquote argv buf lexbuf }
 | _ as c          { raise (UnmatchedChar c) }
and squote argv buf = parse
 | ''' (space|eof) { shell_command ((Buffer.contents buf)::argv) lexbuf }
 | ''' '''         { squote argv buf lexbuf }
 | ''' '"'         { dquote argv buf lexbuf }
 | '''             { uquote argv buf lexbuf }
 | [^ ''' ]+       { Buffer.add_string buf (Lexing.lexeme lexbuf); squote argv buf lexbuf }
 | _ as c          { raise (UnmatchedChar c) }

{
  let main () =
    let cin =
      if Array.length Sys.argv > 1
      then open_in Sys.argv.(1)
      else stdin
    in
    let lexbuf = Lexing.from_channel cin in
    let argv = shell_command [] lexbuf in
    List.iter (Printf.printf "%s\n") argv

  let _ = Printexc.print main ()
}

试一试运行:

$ ocamllex test.mll
$ echo 'echo "foo" "bar\"blub""baz" "'\''" "\"" foo\ bar '\
> '"\\" """"'\'''\'''\'''\''""       Lots" "of\ whitespace' \
> | ocaml test.ml
echo
foo
bar"blubbaz
'
"
foo bar
\

Lots of whitespace

成功！ \o/

将 shell 引用的字符串解析为与 execv 兼容的参数向量

Parse shell quoted string into execv compatible argument vector

shell

ocaml

argv