2 use super::strcursor::StrCursor as Cur;
3 use rustc_span::InnerSpan;
5 /// Represents a single `printf`-style substitution.
6 #[derive(Clone, PartialEq, Debug)]
7 pub enum Substitution<'a> {
8 /// A formatted output substitution with its internal byte offset.
10 /// A literal `%%` escape.
14 impl<'a> Substitution<'a> {
15 pub fn as_str(&self) -> &str {
17 Substitution::Format(ref fmt) => fmt.span,
18 Substitution::Escape => "%%",
22 pub fn position(&self) -> Option<InnerSpan> {
24 Substitution::Format(ref fmt) => Some(fmt.position),
29 pub fn set_position(&mut self, start: usize, end: usize) {
30 if let Substitution::Format(ref mut fmt) = self {
31 fmt.position = InnerSpan::new(start, end);
35 /// Translate this substitution into an equivalent Rust formatting directive.
37 /// This ignores cases where the substitution does not have an exact equivalent, or where
38 /// the substitution would be unnecessary.
39 pub fn translate(&self) -> Option<String> {
41 Substitution::Format(ref fmt) => fmt.translate(),
42 Substitution::Escape => None,
47 #[derive(Clone, PartialEq, Debug)]
48 /// A single `printf`-style formatting directive.
49 pub struct Format<'a> {
50 /// The entire original formatting directive.
52 /// The (1-based) parameter to be converted.
53 pub parameter: Option<u16>,
56 /// Minimum width of the output.
57 pub width: Option<Num>,
58 /// Precision of the conversion.
59 pub precision: Option<Num>,
60 /// Length modifier for the conversion.
61 pub length: Option<&'a str>,
62 /// Type of parameter being converted.
64 /// Byte offset for the start and end of this formatting directive.
65 pub position: InnerSpan,
69 /// Translate this directive into an equivalent Rust formatting directive.
71 /// Returns `None` in cases where the `printf` directive does not have an exact Rust
72 /// equivalent, rather than guessing.
73 pub fn translate(&self) -> Option<String> {
76 let (c_alt, c_zero, c_left, c_plus) = {
77 let mut c_alt = false;
78 let mut c_zero = false;
79 let mut c_left = false;
80 let mut c_plus = false;
81 for c in self.flags.chars() {
90 (c_alt, c_zero, c_left, c_plus)
93 // Has a special form in Rust for numbers.
94 let fill = c_zero.then_some("0");
96 let align = c_left.then_some("<");
98 // Rust doesn't have an equivalent to the `' '` flag.
99 let sign = c_plus.then_some("+");
101 // Not *quite* the same, depending on the type...
104 let width = match self.width {
106 // NOTE: Rust doesn't support this.
109 w @ Some(Num::Arg(_)) => w,
110 w @ Some(Num::Num(_)) => w,
114 let precision = self.precision;
116 // NOTE: although length *can* have an effect, we can't duplicate the effect in Rust, so
117 // we just ignore it.
119 let (type_, use_zero_fill, is_int) = match self.type_ {
120 "d" | "i" | "u" => (None, true, true),
121 "f" | "F" => (None, false, false),
122 "s" | "c" => (None, false, false),
123 "e" | "E" => (Some(self.type_), true, false),
124 "x" | "X" | "o" => (Some(self.type_), true, true),
125 "p" => (Some(self.type_), false, true),
126 "g" => (Some("e"), true, false),
127 "G" => (Some("E"), true, false),
131 let (fill, width, precision) = match (is_int, width, precision) {
132 (true, Some(_), Some(_)) => {
133 // Rust can't duplicate this insanity.
136 (true, None, Some(p)) => (Some("0"), Some(p), None),
137 (true, w, None) => (fill, w, None),
138 (false, w, p) => (fill, w, p),
141 let align = match (self.type_, width.is_some(), align.is_some()) {
142 ("s", true, false) => Some(">"),
146 let (fill, zero_fill) = match (fill, use_zero_fill) {
147 (Some("0"), true) => (None, true),
148 (fill, _) => (fill, false),
151 let alt = match type_ {
152 Some("x" | "X") => alt,
156 let has_options = fill.is_some()
162 || precision.is_some()
165 // Initialise with a rough guess.
166 let cap = self.span.len() + if has_options { 2 } else { 0 };
167 let mut s = String::with_capacity(cap);
171 if let Some(arg) = self.parameter {
172 write!(s, "{}", arg.checked_sub(1)?).ok()?;
178 let align = if let Some(fill) = fill {
185 if let Some(align) = align {
189 if let Some(sign) = sign {
201 if let Some(width) = width {
202 width.translate(&mut s).ok()?;
205 if let Some(precision) = precision {
207 precision.translate(&mut s).ok()?;
210 if let Some(type_) = type_ {
220 /// A general number used in a `printf` formatting directive.
221 #[derive(Copy, Clone, PartialEq, Debug)]
223 // The range of these values is technically bounded by `NL_ARGMAX`... but, at least for GNU
224 // libc, it apparently has no real fixed limit. A `u16` is used here on the basis that it
225 // is *vanishingly* unlikely that *anyone* is going to try formatting something wider, or
226 // with more precision, than 32 thousand positions which is so wide it couldn't possibly fit
228 /// A specific, fixed value.
230 /// The value is derived from a positional argument.
232 /// The value is derived from the "next" unconverted argument.
237 fn from_str(s: &str, arg: Option<&str>) -> Self {
238 if let Some(arg) = arg {
239 Num::Arg(arg.parse().unwrap_or_else(|_| panic!("invalid format arg `{:?}`", arg)))
243 Num::Num(s.parse().unwrap_or_else(|_| panic!("invalid format num `{:?}`", s)))
247 fn translate(&self, s: &mut String) -> std::fmt::Result {
250 Num::Num(n) => write!(s, "{}", n),
252 let n = n.checked_sub(1).ok_or(std::fmt::Error)?;
255 Num::Next => write!(s, "*"),
260 /// Returns an iterator over all substitutions in a given string.
261 pub fn iter_subs(s: &str, start_pos: usize) -> Substitutions<'_> {
262 Substitutions { s, pos: start_pos }
265 /// Iterator over substitutions in a string.
266 pub struct Substitutions<'a> {
271 impl<'a> Iterator for Substitutions<'a> {
272 type Item = Substitution<'a>;
273 fn next(&mut self) -> Option<Self::Item> {
274 let (mut sub, tail) = parse_next_substitution(self.s)?;
277 Substitution::Format(_) => {
278 if let Some(inner_span) = sub.position() {
279 sub.set_position(inner_span.start + self.pos, inner_span.end + self.pos);
280 self.pos += inner_span.end;
283 Substitution::Escape => self.pos += 2,
288 fn size_hint(&self) -> (usize, Option<usize>) {
289 // Substitutions are at least 2 characters long.
290 (0, Some(self.s.len() / 2))
305 /// Parse the next substitution from the input string.
306 pub fn parse_next_substitution(s: &str) -> Option<(Substitution<'_>, &str)> {
310 let start = s.find('%')?;
311 if let '%' = s[start + 1..].chars().next()? {
312 return Some((Substitution::Escape, &s[start + 2..]));
315 Cur::new_at(&s[..], start)
318 // This is meant to be a translation of the following regex:
323 // (?: (?P<parameter> \d+) \$ )?
324 // (?P<flags> [-+ 0\#']* )
325 // (?P<width> \d+ | \* (?: (?P<widtha> \d+) \$ )? )?
326 // (?: \. (?P<precision> \d+ | \* (?: (?P<precisiona> \d+) \$ )? ) )?
329 // hh | h | ll | l | L | z | j | t
332 // | I32 | I64 | I | q
337 // Used to establish the full span at the end.
339 // The current position within the string.
340 let mut at = at.at_next_cp()?;
341 // `c` is the next codepoint, `next` is a cursor after it.
342 let (mut c, mut next) = at.next_cp()?;
344 // Update `at`, `c`, and `next`, exiting if we're out of input.
345 macro_rules! move_to {
348 let (c_, next_) = at.next_cp()?;
354 // Constructs a result when parsing fails.
356 // Note: `move` used to capture copies of the cursors as they are *now*.
357 let fallback = move || {
359 Substitution::Format(Format {
360 span: start.slice_between(next).unwrap(),
366 type_: at.slice_between(next).unwrap(),
367 position: InnerSpan::new(start.at, next.at),
373 // Next parsing state.
374 let mut state = Start;
376 // Sadly, Rust isn't *quite* smart enough to know these *must* be initialised by the end.
377 let mut parameter: Option<u16> = None;
378 let mut flags: &str = "";
379 let mut width: Option<Num> = None;
380 let mut precision: Option<Num> = None;
381 let mut length: Option<&str> = None;
382 let mut type_: &str = "";
385 if let Start = state {
388 let end = at_next_cp_while(next, is_digit);
389 match end.next_cp() {
390 // Yes, this *is* the parameter.
391 Some(('$', end2)) => {
393 parameter = Some(at.slice_between(end).unwrap().parse().unwrap());
396 // Wait, no, actually, it's the width.
401 width = Some(Num::from_str(at.slice_between(end).unwrap(), None));
404 // It's invalid, is what it is.
405 None => return fallback(),
416 if let Flags = state {
417 let end = at_next_cp_while(at, is_flag);
419 flags = at.slice_between(end).unwrap();
423 if let Width = state {
430 let end = at_next_cp_while(next, is_digit);
432 width = Some(Num::from_str(at.slice_between(end).unwrap(), None));
443 if let WidthArg = state {
444 let end = at_next_cp_while(at, is_digit);
445 match end.next_cp() {
446 Some(('$', end2)) => {
448 width = Some(Num::from_str("", Some(at.slice_between(end).unwrap())));
453 width = Some(Num::Next);
459 if let Prec = state {
473 if let PrecInner = state {
476 let end = at_next_cp_while(next, is_digit);
477 match end.next_cp() {
478 Some(('$', end2)) => {
480 precision = Some(Num::from_str("*", next.slice_between(end)));
485 precision = Some(Num::Next);
491 let end = at_next_cp_while(next, is_digit);
493 precision = Some(Num::from_str(at.slice_between(end).unwrap(), None));
496 _ => return fallback(),
500 if let Length = state {
501 let c1_next1 = next.next_cp();
502 match (c, c1_next1) {
503 ('h', Some(('h', next1))) | ('l', Some(('l', next1))) => {
505 length = Some(at.slice_between(next1).unwrap());
509 ('h' | 'l' | 'L' | 'z' | 'j' | 't' | 'q', _) => {
511 length = Some(at.slice_between(next).unwrap());
518 .and_then(|end| end.at_next_cp())
519 .map(|end| (next.slice_between(end).unwrap(), end));
520 let end = match end {
521 Some(("32", end)) => end,
522 Some(("64", end)) => end,
526 length = Some(at.slice_between(end).unwrap());
538 if let Type = state {
540 type_ = at.slice_between(next).unwrap();
542 // Don't use `move_to!` here, as we *can* be at the end of the input.
550 let position = InnerSpan::new(start.at, end.at);
553 span: start.slice_between(end).unwrap(),
562 Some((Substitution::Format(f), end.slice_after()))
565 fn at_next_cp_while<F>(mut cur: Cur<'_>, mut pred: F) -> Cur<'_>
567 F: FnMut(char) -> bool,
570 match cur.next_cp() {
583 fn is_digit(c: char) -> bool {
590 fn is_flag(c: char) -> bool {
592 '0' | '-' | '+' | ' ' | '#' | '\'' => true,
602 use super::strcursor::StrCursor as Cur;
603 use rustc_span::InnerSpan;
605 #[derive(Clone, PartialEq, Debug)]
606 pub enum Substitution<'a> {
607 Ordinal(u8, (usize, usize)),
608 Name(&'a str, (usize, usize)),
609 Escape((usize, usize)),
612 impl Substitution<'_> {
613 pub fn as_str(&self) -> String {
615 Substitution::Ordinal(n, _) => format!("${}", n),
616 Substitution::Name(n, _) => format!("${}", n),
617 Substitution::Escape(_) => "$$".into(),
621 pub fn position(&self) -> Option<InnerSpan> {
623 Substitution::Ordinal(_, pos)
624 | Substitution::Name(_, pos)
625 | Substitution::Escape(pos) => Some(InnerSpan::new(pos.0, pos.1)),
629 pub fn set_position(&mut self, start: usize, end: usize) {
631 Substitution::Ordinal(_, ref mut pos)
632 | Substitution::Name(_, ref mut pos)
633 | Substitution::Escape(ref mut pos) => *pos = (start, end),
637 pub fn translate(&self) -> Option<String> {
639 Substitution::Ordinal(n, _) => Some(format!("{{{}}}", n)),
640 Substitution::Name(n, _) => Some(format!("{{{}}}", n)),
641 Substitution::Escape(_) => None,
646 /// Returns an iterator over all substitutions in a given string.
647 pub fn iter_subs(s: &str, start_pos: usize) -> Substitutions<'_> {
648 Substitutions { s, pos: start_pos }
651 /// Iterator over substitutions in a string.
652 pub struct Substitutions<'a> {
657 impl<'a> Iterator for Substitutions<'a> {
658 type Item = Substitution<'a>;
659 fn next(&mut self) -> Option<Self::Item> {
660 match parse_next_substitution(self.s) {
661 Some((mut sub, tail)) => {
663 if let Some(InnerSpan { start, end }) = sub.position() {
664 sub.set_position(start + self.pos, end + self.pos);
673 fn size_hint(&self) -> (usize, Option<usize>) {
674 (0, Some(self.s.len()))
678 /// Parse the next substitution from the input string.
679 pub fn parse_next_substitution(s: &str) -> Option<(Substitution<'_>, &str)> {
681 let start = s.find('$')?;
682 match s[start + 1..].chars().next()? {
683 '$' => return Some((Substitution::Escape((start, start + 2)), &s[start + 2..])),
685 let n = (c as u8) - b'0';
686 return Some((Substitution::Ordinal(n, (start, start + 2)), &s[start + 2..]));
688 _ => { /* fall-through */ }
691 Cur::new_at(&s[..], start)
694 let at = at.at_next_cp()?;
695 let (c, inner) = at.next_cp()?;
697 if !is_ident_head(c) {
700 let end = at_next_cp_while(inner, is_ident_tail);
701 let slice = at.slice_between(end).unwrap();
702 let start = at.at - 1;
703 let end_pos = at.at + slice.len();
704 Some((Substitution::Name(slice, (start, end_pos)), end.slice_after()))
708 fn at_next_cp_while<F>(mut cur: Cur<'_>, mut pred: F) -> Cur<'_>
710 F: FnMut(char) -> bool,
713 match cur.next_cp() {
726 fn is_ident_head(c: char) -> bool {
728 'a'..='z' | 'A'..='Z' | '_' => true,
733 fn is_ident_tail(c: char) -> bool {
736 c => is_ident_head(c),
745 pub struct StrCursor<'a> {
750 impl<'a> StrCursor<'a> {
751 pub fn new_at(s: &'a str, at: usize) -> StrCursor<'a> {
755 pub fn at_next_cp(mut self) -> Option<StrCursor<'a>> {
756 match self.try_seek_right_cp() {
762 pub fn next_cp(mut self) -> Option<(char, StrCursor<'a>)> {
763 let cp = self.cp_after()?;
764 self.seek_right(cp.len_utf8());
768 fn slice_before(&self) -> &'a str {
772 pub fn slice_after(&self) -> &'a str {
776 pub fn slice_between(&self, until: StrCursor<'a>) -> Option<&'a str> {
777 if !str_eq_literal(self.s, until.s) {
780 use std::cmp::{max, min};
781 let beg = min(self.at, until.at);
782 let end = max(self.at, until.at);
783 Some(&self.s[beg..end])
787 fn cp_after(&self) -> Option<char> {
788 self.slice_after().chars().next()
791 fn try_seek_right_cp(&mut self) -> bool {
792 match self.slice_after().chars().next() {
794 self.at += c.len_utf8();
801 fn seek_right(&mut self, bytes: usize) {
806 impl Copy for StrCursor<'_> {}
808 impl<'a> Clone for StrCursor<'a> {
809 fn clone(&self) -> StrCursor<'a> {
814 impl std::fmt::Debug for StrCursor<'_> {
815 fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
816 write!(fmt, "StrCursor({:?} | {:?})", self.slice_before(), self.slice_after())
820 fn str_eq_literal(a: &str, b: &str) -> bool {
821 a.as_bytes().as_ptr() == b.as_bytes().as_ptr() && a.len() == b.len()