forked from proxmox-mirrors/proxmox
164 lines
5.6 KiB
Rust
164 lines
5.6 KiB
Rust
use anyhow::*;
|
|
|
|
/// Shell quote type
|
|
pub use rustyline::completion::Quote;
|
|
|
|
#[derive(PartialEq)]
|
|
enum ParseMode {
|
|
Space,
|
|
DoubleQuote,
|
|
EscapeNormal,
|
|
EscapeInDoubleQuote,
|
|
Normal,
|
|
SingleQuote,
|
|
}
|
|
|
|
/// Parsing strings as they would be interpreted by the UNIX Bourne shell.
|
|
///
|
|
/// - ``finalize``: assume this is a complete command line. Set this
|
|
/// to false for the 'completion' helper, which needs to get
|
|
/// information about the last unfinished parameter.
|
|
///
|
|
/// Returns the list of fully parsed words (unescaped and quotes
|
|
/// removed). If there are unclosed quotes, the start of that
|
|
/// parameter, the parameter value (unescaped and quotes removed), and
|
|
/// the quote type are returned.
|
|
pub fn shellword_split_unclosed(
|
|
s: &str,
|
|
finalize: bool,
|
|
) -> (Vec<String>, Option<(usize, String, Quote)>) {
|
|
let char_indices = s.char_indices();
|
|
let mut args: Vec<String> = Vec::new();
|
|
let mut field_start = None;
|
|
let mut field = String::new();
|
|
let mut mode = ParseMode::Space;
|
|
|
|
let space_chars = [' ', '\t', '\n'];
|
|
|
|
for (index, c) in char_indices {
|
|
match mode {
|
|
ParseMode::Space => match c {
|
|
'"' => {
|
|
mode = ParseMode::DoubleQuote;
|
|
field_start = Some((index, Quote::Double));
|
|
}
|
|
'\\' => {
|
|
mode = ParseMode::EscapeNormal;
|
|
field_start = Some((index, Quote::None));
|
|
}
|
|
'\'' => {
|
|
mode = ParseMode::SingleQuote;
|
|
field_start = Some((index, Quote::Single));
|
|
}
|
|
c if space_chars.contains(&c) => (), // skip space
|
|
c => {
|
|
mode = ParseMode::Normal;
|
|
field_start = Some((index, Quote::None));
|
|
field.push(c);
|
|
}
|
|
},
|
|
ParseMode::EscapeNormal => {
|
|
mode = ParseMode::Normal;
|
|
field.push(c);
|
|
}
|
|
ParseMode::EscapeInDoubleQuote => {
|
|
// Within double quoted strings, backslashes are only
|
|
// treated as metacharacters when followed by one of
|
|
// the following characters: $ ' " \ newline
|
|
match c {
|
|
'$' | '\'' | '"' | '\\' | '\n' => (),
|
|
_ => field.push('\\'),
|
|
}
|
|
field.push(c);
|
|
mode = ParseMode::DoubleQuote;
|
|
}
|
|
ParseMode::Normal => match c {
|
|
'"' => mode = ParseMode::DoubleQuote,
|
|
'\'' => mode = ParseMode::SingleQuote,
|
|
'\\' => mode = ParseMode::EscapeNormal,
|
|
c if space_chars.contains(&c) => {
|
|
mode = ParseMode::Space;
|
|
let (_start, _quote) = field_start.take().unwrap();
|
|
args.push(field.split_off(0));
|
|
}
|
|
c => field.push(c), // continue
|
|
},
|
|
ParseMode::DoubleQuote => match c {
|
|
'"' => mode = ParseMode::Normal,
|
|
'\\' => mode = ParseMode::EscapeInDoubleQuote,
|
|
c => field.push(c), // continue
|
|
},
|
|
ParseMode::SingleQuote => match c {
|
|
// Note: no escape in single quotes
|
|
'\'' => mode = ParseMode::Normal,
|
|
c => field.push(c), // continue
|
|
},
|
|
}
|
|
}
|
|
|
|
if finalize && mode == ParseMode::Normal {
|
|
let (_start, _quote) = field_start.take().unwrap();
|
|
args.push(field.split_off(0));
|
|
}
|
|
|
|
match field_start {
|
|
Some((start, quote)) => (args, Some((start, field, quote))),
|
|
None => (args, None),
|
|
}
|
|
}
|
|
|
|
/// Splits a string into a vector of words in the same way the UNIX Bourne shell does.
|
|
///
|
|
/// Return words unescaped and without quotes.
|
|
pub fn shellword_split(s: &str) -> Result<Vec<String>, Error> {
|
|
let (args, unclosed_field) = shellword_split_unclosed(s, true);
|
|
if unclosed_field.is_some() {
|
|
bail!("shellword split failed - found unclosed quote.");
|
|
}
|
|
Ok(args)
|
|
}
|
|
|
|
#[test]
|
|
fn test_shellword_split() {
|
|
let expect = ["ls", "/etc"];
|
|
let expect: Vec<String> = expect.iter().map(|v| v.to_string()).collect();
|
|
|
|
assert_eq!(expect, shellword_split("ls /etc").unwrap());
|
|
assert_eq!(expect, shellword_split("ls \"/etc\"").unwrap());
|
|
assert_eq!(expect, shellword_split("ls '/etc'").unwrap());
|
|
assert_eq!(expect, shellword_split("ls '/etc'").unwrap());
|
|
|
|
assert_eq!(expect, shellword_split("ls /e\"t\"c").unwrap());
|
|
assert_eq!(expect, shellword_split("ls /e'tc'").unwrap());
|
|
assert_eq!(expect, shellword_split("ls /e't''c'").unwrap());
|
|
|
|
let expect = ["ls", "/etc 08x"];
|
|
let expect: Vec<String> = expect.iter().map(|v| v.to_string()).collect();
|
|
assert_eq!(expect, shellword_split("ls /etc\\ \\08x").unwrap());
|
|
|
|
let expect = ["ls", "/etc \\08x"];
|
|
let expect: Vec<String> = expect.iter().map(|v| v.to_string()).collect();
|
|
assert_eq!(expect, shellword_split("ls \"/etc \\08x\"").unwrap());
|
|
}
|
|
|
|
#[test]
|
|
fn test_shellword_split_unclosed() {
|
|
let expect = ["ls".to_string()].to_vec();
|
|
assert_eq!(
|
|
(
|
|
expect,
|
|
Some((3, "./File1 name with spaces".to_string(), Quote::Single))
|
|
),
|
|
shellword_split_unclosed("ls './File1 name with spaces", false)
|
|
);
|
|
|
|
let expect = ["ls".to_string()].to_vec();
|
|
assert_eq!(
|
|
(
|
|
expect,
|
|
Some((3, "./File2 name with spaces".to_string(), Quote::Double))
|
|
),
|
|
shellword_split_unclosed("ls \"./File2 \"name\" with spaces", false)
|
|
);
|
|
}
|