clap_lex/lib.rs
1//! Minimal, flexible command-line parser
2//!
3//! As opposed to a declarative parser, this processes arguments as a stream of tokens. As lexing
4//! a command-line is not context-free, we rely on the caller to decide how to interpret the
5//! arguments.
6//!
7//! # Examples
8//!
9//! ```rust
10//! use std::path::PathBuf;
11//! use std::ffi::OsStr;
12//!
13//! type BoxedError = Box<dyn std::error::Error + Send + Sync>;
14//!
15//! #[derive(Debug)]
16//! struct Args {
17//! paths: Vec<PathBuf>,
18//! color: Color,
19//! verbosity: usize,
20//! }
21//!
22//! #[derive(Debug)]
23//! enum Color {
24//! Always,
25//! Auto,
26//! Never,
27//! }
28//!
29//! impl Color {
30//! fn parse(s: Option<&OsStr>) -> Result<Self, BoxedError> {
31//! let s = s.map(|s| s.to_str().ok_or(s));
32//! match s {
33//! Some(Ok("always")) | Some(Ok("")) | None => {
34//! Ok(Color::Always)
35//! }
36//! Some(Ok("auto")) => {
37//! Ok(Color::Auto)
38//! }
39//! Some(Ok("never")) => {
40//! Ok(Color::Never)
41//! }
42//! Some(invalid) => {
43//! Err(format!("Invalid value for `--color`, {invalid:?}").into())
44//! }
45//! }
46//! }
47//! }
48//!
49//! fn parse_args(
50//! raw: impl IntoIterator<Item=impl Into<std::ffi::OsString>>
51//! ) -> Result<Args, BoxedError> {
52//! let mut args = Args {
53//! paths: Vec::new(),
54//! color: Color::Auto,
55//! verbosity: 0,
56//! };
57//!
58//! let raw = clap_lex::RawArgs::new(raw);
59//! let mut cursor = raw.cursor();
60//! raw.next(&mut cursor); // Skip the bin
61//! while let Some(arg) = raw.next(&mut cursor) {
62//! if arg.is_escape() {
63//! args.paths.extend(raw.remaining(&mut cursor).map(PathBuf::from));
64//! } else if arg.is_stdio() {
65//! args.paths.push(PathBuf::from("-"));
66//! } else if let Some((long, value)) = arg.to_long() {
67//! match long {
68//! Ok("verbose") => {
69//! if let Some(value) = value {
70//! return Err(format!("`--verbose` does not take a value, got `{value:?}`").into());
71//! }
72//! args.verbosity += 1;
73//! }
74//! Ok("color") => {
75//! args.color = Color::parse(value)?;
76//! }
77//! _ => {
78//! return Err(
79//! format!("Unexpected flag: --{}", arg.display()).into()
80//! );
81//! }
82//! }
83//! } else if let Some(mut shorts) = arg.to_short() {
84//! while let Some(short) = shorts.next_flag() {
85//! match short {
86//! Ok('v') => {
87//! args.verbosity += 1;
88//! }
89//! Ok('c') => {
90//! let value = shorts.next_value_os();
91//! args.color = Color::parse(value)?;
92//! }
93//! Ok(c) => {
94//! return Err(format!("Unexpected flag: -{c}").into());
95//! }
96//! Err(e) => {
97//! return Err(format!("Unexpected flag: -{}", e.to_string_lossy()).into());
98//! }
99//! }
100//! }
101//! } else {
102//! args.paths.push(PathBuf::from(arg.to_value_os().to_owned()));
103//! }
104//! }
105//!
106//! Ok(args)
107//! }
108//!
109//! let args = parse_args(["bin", "--hello", "world"]);
110//! println!("{args:?}");
111//! ```
112
113#![cfg_attr(docsrs, feature(doc_auto_cfg))]
114#![warn(missing_docs)]
115#![warn(clippy::print_stderr)]
116#![warn(clippy::print_stdout)]
117
118mod ext;
119
120use std::ffi::OsStr;
121use std::ffi::OsString;
122
123pub use std::io::SeekFrom;
124
125pub use ext::OsStrExt;
126
127/// Command-line arguments
128#[derive(Default, Clone, Debug, PartialEq, Eq)]
129pub struct RawArgs {
130 items: Vec<OsString>,
131}
132
133impl RawArgs {
134 //// Create an argument list to parse
135 ///
136 /// <div class="warning">
137 ///
138 /// **NOTE:** The argument returned will be the current binary.
139 ///
140 /// </div>
141 ///
142 /// # Example
143 ///
144 /// ```rust,no_run
145 /// # use std::path::PathBuf;
146 /// let raw = clap_lex::RawArgs::from_args();
147 /// let mut cursor = raw.cursor();
148 /// let _bin = raw.next_os(&mut cursor);
149 ///
150 /// let mut paths = raw.remaining(&mut cursor).map(PathBuf::from).collect::<Vec<_>>();
151 /// println!("{paths:?}");
152 /// ```
153 pub fn from_args() -> Self {
154 Self::new(std::env::args_os())
155 }
156
157 //// Create an argument list to parse
158 ///
159 /// # Example
160 ///
161 /// ```rust,no_run
162 /// # use std::path::PathBuf;
163 /// let raw = clap_lex::RawArgs::new(["bin", "foo.txt"]);
164 /// let mut cursor = raw.cursor();
165 /// let _bin = raw.next_os(&mut cursor);
166 ///
167 /// let mut paths = raw.remaining(&mut cursor).map(PathBuf::from).collect::<Vec<_>>();
168 /// println!("{paths:?}");
169 /// ```
170 pub fn new(iter: impl IntoIterator<Item = impl Into<OsString>>) -> Self {
171 let iter = iter.into_iter();
172 Self::from(iter)
173 }
174
175 /// Create a cursor for walking the arguments
176 ///
177 /// # Example
178 ///
179 /// ```rust,no_run
180 /// # use std::path::PathBuf;
181 /// let raw = clap_lex::RawArgs::new(["bin", "foo.txt"]);
182 /// let mut cursor = raw.cursor();
183 /// let _bin = raw.next_os(&mut cursor);
184 ///
185 /// let mut paths = raw.remaining(&mut cursor).map(PathBuf::from).collect::<Vec<_>>();
186 /// println!("{paths:?}");
187 /// ```
188 pub fn cursor(&self) -> ArgCursor {
189 ArgCursor::new()
190 }
191
192 /// Advance the cursor, returning the next [`ParsedArg`]
193 pub fn next(&self, cursor: &mut ArgCursor) -> Option<ParsedArg<'_>> {
194 self.next_os(cursor).map(ParsedArg::new)
195 }
196
197 /// Advance the cursor, returning a raw argument value.
198 pub fn next_os(&self, cursor: &mut ArgCursor) -> Option<&OsStr> {
199 let next = self.items.get(cursor.cursor).map(|s| s.as_os_str());
200 cursor.cursor = cursor.cursor.saturating_add(1);
201 next
202 }
203
204 /// Return the next [`ParsedArg`]
205 pub fn peek(&self, cursor: &ArgCursor) -> Option<ParsedArg<'_>> {
206 self.peek_os(cursor).map(ParsedArg::new)
207 }
208
209 /// Return a raw argument value.
210 pub fn peek_os(&self, cursor: &ArgCursor) -> Option<&OsStr> {
211 self.items.get(cursor.cursor).map(|s| s.as_os_str())
212 }
213
214 /// Return all remaining raw arguments, advancing the cursor to the end
215 ///
216 /// # Example
217 ///
218 /// ```rust,no_run
219 /// # use std::path::PathBuf;
220 /// let raw = clap_lex::RawArgs::new(["bin", "foo.txt"]);
221 /// let mut cursor = raw.cursor();
222 /// let _bin = raw.next_os(&mut cursor);
223 ///
224 /// let mut paths = raw.remaining(&mut cursor).map(PathBuf::from).collect::<Vec<_>>();
225 /// println!("{paths:?}");
226 /// ```
227 pub fn remaining(&self, cursor: &mut ArgCursor) -> impl Iterator<Item = &OsStr> {
228 let remaining = self.items[cursor.cursor..].iter().map(|s| s.as_os_str());
229 cursor.cursor = self.items.len();
230 remaining
231 }
232
233 /// Adjust the cursor's position
234 pub fn seek(&self, cursor: &mut ArgCursor, pos: SeekFrom) {
235 let pos = match pos {
236 SeekFrom::Start(pos) => pos,
237 SeekFrom::End(pos) => (self.items.len() as i64).saturating_add(pos).max(0) as u64,
238 SeekFrom::Current(pos) => (cursor.cursor as i64).saturating_add(pos).max(0) as u64,
239 };
240 let pos = (pos as usize).min(self.items.len());
241 cursor.cursor = pos;
242 }
243
244 /// Inject arguments before the [`RawArgs::next`]
245 pub fn insert(
246 &mut self,
247 cursor: &ArgCursor,
248 insert_items: impl IntoIterator<Item = impl Into<OsString>>,
249 ) {
250 self.items.splice(
251 cursor.cursor..cursor.cursor,
252 insert_items.into_iter().map(Into::into),
253 );
254 }
255
256 /// Any remaining args?
257 pub fn is_end(&self, cursor: &ArgCursor) -> bool {
258 self.peek_os(cursor).is_none()
259 }
260}
261
262impl<I, T> From<I> for RawArgs
263where
264 I: Iterator<Item = T>,
265 T: Into<OsString>,
266{
267 fn from(val: I) -> Self {
268 Self {
269 items: val.map(|x| x.into()).collect(),
270 }
271 }
272}
273
274/// Position within [`RawArgs`]
275#[derive(Clone, Debug, PartialEq, Eq, PartialOrd, Ord)]
276pub struct ArgCursor {
277 cursor: usize,
278}
279
280impl ArgCursor {
281 fn new() -> Self {
282 Self { cursor: 0 }
283 }
284}
285
286/// Command-line Argument
287#[derive(Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash)]
288pub struct ParsedArg<'s> {
289 inner: &'s OsStr,
290}
291
292impl<'s> ParsedArg<'s> {
293 fn new(inner: &'s OsStr) -> Self {
294 Self { inner }
295 }
296
297 /// Argument is length of 0
298 pub fn is_empty(&self) -> bool {
299 self.inner.is_empty()
300 }
301
302 /// Does the argument look like a stdio argument (`-`)
303 pub fn is_stdio(&self) -> bool {
304 self.inner == "-"
305 }
306
307 /// Does the argument look like an argument escape (`--`)
308 pub fn is_escape(&self) -> bool {
309 self.inner == "--"
310 }
311
312 /// Does the argument look like a negative number?
313 ///
314 /// This won't parse the number in full but attempts to see if this looks
315 /// like something along the lines of `-3`, `-0.3`, or `-33.03`
316 pub fn is_negative_number(&self) -> bool {
317 self.to_value()
318 .ok()
319 .and_then(|s| Some(is_number(s.strip_prefix('-')?)))
320 .unwrap_or_default()
321 }
322
323 /// Treat as a long-flag
324 pub fn to_long(&self) -> Option<(Result<&str, &OsStr>, Option<&OsStr>)> {
325 let raw = self.inner;
326 let remainder = raw.strip_prefix("--")?;
327 if remainder.is_empty() {
328 debug_assert!(self.is_escape());
329 return None;
330 }
331
332 let (flag, value) = if let Some((p0, p1)) = remainder.split_once("=") {
333 (p0, Some(p1))
334 } else {
335 (remainder, None)
336 };
337 let flag = flag.to_str().ok_or(flag);
338 Some((flag, value))
339 }
340
341 /// Can treat as a long-flag
342 pub fn is_long(&self) -> bool {
343 self.inner.starts_with("--") && !self.is_escape()
344 }
345
346 /// Treat as a short-flag
347 pub fn to_short(&self) -> Option<ShortFlags<'_>> {
348 if let Some(remainder_os) = self.inner.strip_prefix("-") {
349 if remainder_os.starts_with("-") {
350 None
351 } else if remainder_os.is_empty() {
352 debug_assert!(self.is_stdio());
353 None
354 } else {
355 Some(ShortFlags::new(remainder_os))
356 }
357 } else {
358 None
359 }
360 }
361
362 /// Can treat as a short-flag
363 pub fn is_short(&self) -> bool {
364 self.inner.starts_with("-") && !self.is_stdio() && !self.inner.starts_with("--")
365 }
366
367 /// Treat as a value
368 ///
369 /// <div class="warning">
370 ///
371 /// **NOTE:** May return a flag or an escape.
372 ///
373 /// </div>
374 pub fn to_value_os(&self) -> &OsStr {
375 self.inner
376 }
377
378 /// Treat as a value
379 ///
380 /// <div class="warning">
381 ///
382 /// **NOTE:** May return a flag or an escape.
383 ///
384 /// </div>
385 pub fn to_value(&self) -> Result<&str, &OsStr> {
386 self.inner.to_str().ok_or(self.inner)
387 }
388
389 /// Safely print an argument that may contain non-UTF8 content
390 ///
391 /// This may perform lossy conversion, depending on the platform. If you would like an implementation which escapes the path please use Debug instead.
392 pub fn display(&self) -> impl std::fmt::Display + '_ {
393 self.inner.to_string_lossy()
394 }
395}
396
397/// Walk through short flags within a [`ParsedArg`]
398#[derive(Clone, Debug)]
399pub struct ShortFlags<'s> {
400 inner: &'s OsStr,
401 utf8_prefix: std::str::CharIndices<'s>,
402 invalid_suffix: Option<&'s OsStr>,
403}
404
405impl<'s> ShortFlags<'s> {
406 fn new(inner: &'s OsStr) -> Self {
407 let (utf8_prefix, invalid_suffix) = split_nonutf8_once(inner);
408 let utf8_prefix = utf8_prefix.char_indices();
409 Self {
410 inner,
411 utf8_prefix,
412 invalid_suffix,
413 }
414 }
415
416 /// Move the iterator forward by `n` short flags
417 pub fn advance_by(&mut self, n: usize) -> Result<(), usize> {
418 for i in 0..n {
419 self.next().ok_or(i)?.map_err(|_| i)?;
420 }
421 Ok(())
422 }
423
424 /// No short flags left
425 pub fn is_empty(&self) -> bool {
426 self.invalid_suffix.is_none() && self.utf8_prefix.as_str().is_empty()
427 }
428
429 /// Does the short flag look like a number
430 ///
431 /// Ideally call this before doing any iterator
432 pub fn is_negative_number(&self) -> bool {
433 self.invalid_suffix.is_none() && is_number(self.utf8_prefix.as_str())
434 }
435
436 /// Advance the iterator, returning the next short flag on success
437 ///
438 /// On error, returns the invalid-UTF8 value
439 pub fn next_flag(&mut self) -> Option<Result<char, &'s OsStr>> {
440 if let Some((_, flag)) = self.utf8_prefix.next() {
441 return Some(Ok(flag));
442 }
443
444 if let Some(suffix) = self.invalid_suffix {
445 self.invalid_suffix = None;
446 return Some(Err(suffix));
447 }
448
449 None
450 }
451
452 /// Advance the iterator, returning everything left as a value
453 pub fn next_value_os(&mut self) -> Option<&'s OsStr> {
454 if let Some((index, _)) = self.utf8_prefix.next() {
455 self.utf8_prefix = "".char_indices();
456 self.invalid_suffix = None;
457 // SAFETY: `char_indices` ensures `index` is at a valid UTF-8 boundary
458 let remainder = unsafe { ext::split_at(self.inner, index).1 };
459 return Some(remainder);
460 }
461
462 if let Some(suffix) = self.invalid_suffix {
463 self.invalid_suffix = None;
464 return Some(suffix);
465 }
466
467 None
468 }
469}
470
471impl<'s> Iterator for ShortFlags<'s> {
472 type Item = Result<char, &'s OsStr>;
473
474 fn next(&mut self) -> Option<Self::Item> {
475 self.next_flag()
476 }
477}
478
479fn split_nonutf8_once(b: &OsStr) -> (&str, Option<&OsStr>) {
480 match b.try_str() {
481 Ok(s) => (s, None),
482 Err(err) => {
483 // SAFETY: `err.valid_up_to()`, which came from str::from_utf8(), is guaranteed
484 // to be a valid UTF8 boundary
485 let (valid, after_valid) = unsafe { ext::split_at(b, err.valid_up_to()) };
486 let valid = valid.try_str().unwrap();
487 (valid, Some(after_valid))
488 }
489 }
490}
491
492fn is_number(arg: &str) -> bool {
493 // Return true if this looks like an integer or a float where it's all
494 // digits plus an optional single dot after some digits.
495 //
496 // For floats allow forms such as `1.`, `1.2`, `1.2e10`, etc.
497 let mut seen_dot = false;
498 let mut position_of_e = None;
499 for (i, c) in arg.as_bytes().iter().enumerate() {
500 match c {
501 // Digits are always valid
502 b'0'..=b'9' => {}
503
504 // Allow a `.`, but only one, only if it comes before an
505 // optional exponent, and only if it's not the first character.
506 b'.' if !seen_dot && position_of_e.is_none() && i > 0 => seen_dot = true,
507
508 // Allow an exponent `e`/`E` but only at most one after the first
509 // character.
510 b'e' | b'E' if position_of_e.is_none() && i > 0 => position_of_e = Some(i),
511
512 _ => return false,
513 }
514 }
515
516 // Disallow `-1e` which isn't a valid float since it doesn't actually have
517 // an exponent.
518 match position_of_e {
519 Some(i) => i != arg.len() - 1,
520 None => true,
521 }
522}