History

refactor (#198)

Sebastian Thiel committed 5 years ago

Commit ef3fc6d1 parent 53243914 files · +303 −309Browse files

Showing the changes to git-object/src/commit/message (3 of 4 files). Show all

Unified · Split

3 files
  1. git-object/src/commit/message/body.rs +123 −0
  2. git-object/src/commit/message/decode.rs +57 −0
  3. git-object/src/commit/message/mod.rs +123 −0
git-object/src/commit/message/body.rs+123−0
@@ -0,0 +1,123 @@
1+use nom::{
2+ bytes::complete::{tag, take_until1},
3+ combinator::all_consuming,
4+ error::{ErrorKind, ParseError},
5+ sequence::terminated,
6+ IResult,
7+};
8+
9+use crate::{
10+ bstr::{BStr, ByteSlice},
11+ commit::message::BodyRef,
12+};
13+
14+/// An iterator over trailers as parsed from a commit message body.
15+///
16+/// lines with parsing failures will be skipped
17+pub struct Trailers<'a> {
18+ pub(crate) cursor: &'a [u8],
19+}
20+
21+/// A trailer as parsed from the commit message body.
22+#[derive(PartialEq, Eq, Debug, Hash, Ord, PartialOrd, Clone, Copy)]
23+#[cfg_attr(feature = "serde1", derive(serde::Serialize, serde::Deserialize))]
24+pub struct TrailerRef<'a> {
25+ /// The name of the trailer, like "Signed-off-by", up to the separator ": "
26+ #[cfg_attr(feature = "serde1", serde(borrow))]
27+ pub token: &'a BStr,
28+ /// The value right after the separator ": ", with leading and trailing whitespace trimmed.
29+ /// Note that multi-line values aren't currently supported.
30+ pub value: &'a BStr,
31+}
32+
33+fn parse_single_line_trailer<'a, E: ParseError<&'a [u8]>>(i: &'a [u8]) -> IResult<&'a [u8], (&'a BStr, &'a BStr), E> {
34+ let (value, token) = terminated(take_until1(b":".as_ref()), tag(b": "))(i.trim_end())?;
35+ if token.trim_end().len() != token.len() || value.trim_start().len() != value.len() {
36+ Err(nom::Err::Failure(E::from_error_kind(i, ErrorKind::Fail)))
37+ } else {
38+ Ok((&[], (token.as_bstr(), value.as_bstr())))
39+ }
40+}
41+
42+#[cfg(test)]
43+mod test_parse_trailer {
44+ use super::*;
45+
46+ fn parse(input: &str) -> (&BStr, &BStr) {
47+ parse_single_line_trailer::<()>(input.as_bytes()).unwrap().1
48+ }
49+
50+ #[test]
51+ fn simple_newline() {
52+ assert_eq!(parse("foo: bar\n"), ("foo".into(), "bar".into()));
53+ }
54+
55+ #[test]
56+ fn simple_non_ascii_no_newline() {
57+ assert_eq!(parse("🤗: 🎉"), ("🤗".into(), "🎉".into()));
58+ }
59+
60+ #[test]
61+ fn with_lots_of_whitespace_newline() {
62+ assert_eq!(
63+ parse("hello foo: bar there \n"),
64+ ("hello foo".into(), "bar there".into())
65+ );
66+ }
67+
68+ #[test]
69+ fn extra_whitespace_before_token_or_value_is_error() {
70+ assert!(parse_single_line_trailer::<()>(b"foo : bar").is_err());
71+ assert!(parse_single_line_trailer::<()>(b"foo: bar").is_err())
72+ }
73+
74+ #[test]
75+ fn simple_newline_windows() {
76+ assert_eq!(parse("foo: bar\r\n"), ("foo".into(), "bar".into()));
77+ }
78+}
79+
80+impl<'a> Iterator for Trailers<'a> {
81+ type Item = TrailerRef<'a>;
82+
83+ fn next(&mut self) -> Option<Self::Item> {
84+ if self.cursor.is_empty() {
85+ return None;
86+ }
87+ for line in self.cursor.lines_with_terminator() {
88+ self.cursor = &self.cursor[line.len()..];
89+ if let Some(trailer) =
90+ all_consuming(parse_single_line_trailer::<()>)(line)
91+ .ok()
92+ .map(|(_, (token, value))| TrailerRef {
93+ token: token.trim().as_bstr(),
94+ value: value.trim().as_bstr(),
95+ })
96+ {
97+ return Some(trailer);
98+ }
99+ }
100+ None
101+ }
102+}
103+
104+impl<'a> BodyRef<'a> {
105+ /// Parse `body` bytes into the trailer and the actual body.
106+ pub fn from_bytes(body: &'a [u8]) -> Self {
107+ body.rfind(b"\n\n")
108+ .map(|pos| (2, pos))
109+ .or_else(|| body.rfind(b"\r\n\r\n").map(|pos| (4, pos)))
110+ .and_then(|(sep_len, pos)| {
111+ let trailer = &body[pos + sep_len..];
112+ let body = &body[..pos];
113+ Trailers { cursor: trailer }.next().map(|_| BodyRef {
114+ body_without_trailer: body.as_bstr(),
115+ start_of_trailer: trailer,
116+ })
117+ })
118+ .unwrap_or_else(|| BodyRef {
119+ body_without_trailer: body.as_bstr(),
120+ start_of_trailer: &[],
121+ })
122+ }
123+}
git-object/src/commit/message/decode.rs+57−0
@@ -0,0 +1,57 @@
1+use nom::{
2+ branch::alt,
3+ bytes::complete::{tag, take_till1},
4+ combinator::all_consuming,
5+ error::ParseError,
6+ sequence::pair,
7+ IResult,
8+};
9+
10+use crate::bstr::{BStr, ByteSlice};
11+
12+pub(crate) fn newline<'a, E: ParseError<&'a [u8]>>(i: &'a [u8]) -> IResult<&'a [u8], &'a [u8], E> {
13+ alt((tag(b"\r\n"), tag(b"\n")))(i)
14+}
15+
16+fn subject_and_body<'a, E: ParseError<&'a [u8]>>(i: &'a [u8]) -> IResult<&'a [u8], (&'a BStr, Option<&'a BStr>), E> {
17+ let mut c = i;
18+ let mut consumed_bytes = 0;
19+ while !c.is_empty() {
20+ c = match take_till1::<_, _, E>(|c| c == b'\n' || c == b'\r')(c) {
21+ Ok((i1, segment)) => {
22+ consumed_bytes += segment.len();
23+ match pair::<_, _, _, E, _, _>(newline, newline)(i1) {
24+ Ok((body, _)) => {
25+ return Ok((
26+ &[],
27+ (
28+ i[0usize..consumed_bytes].as_bstr(),
29+ (!body.is_empty()).then(|| body.as_bstr()),
30+ ),
31+ ));
32+ }
33+ Err(_) => match i1.get(1..) {
34+ Some(next) => {
35+ consumed_bytes += 1;
36+ next
37+ }
38+ None => break,
39+ },
40+ }
41+ }
42+ Err(_) => match c.get(1..) {
43+ Some(next) => {
44+ consumed_bytes += 1;
45+ next
46+ }
47+ None => break,
48+ },
49+ };
50+ }
51+ Ok((&[], (i.as_bstr(), None)))
52+}
53+
54+/// Returns title and body, without separator
55+pub fn message(input: &[u8]) -> (&BStr, Option<&BStr>) {
56+ all_consuming(subject_and_body::<()>)(input).expect("cannot fail").1
57+}
git-object/src/commit/message/mod.rs+123−0
@@ -0,0 +1,123 @@
1+use std::{borrow::Cow, ops::Deref};
2+
3+use crate::{
4+ bstr::{BStr, BString, ByteSlice, ByteVec},
5+ commit::MessageRef,
6+ CommitRef,
7+};
8+
9+///
10+pub mod body;
11+mod decode;
12+
13+impl<'a> CommitRef<'a> {
14+ /// Return exactly the same message as [`MessageRef::summary()`].
15+ pub fn message_summary(&self) -> Cow<'a, BStr> {
16+ summary(self.message)
17+ }
18+
19+ /// Return an iterator over message trailers as obtained from the last paragraph of the commit message.
20+ /// May be empty.
21+ pub fn message_trailers(&self) -> body::Trailers<'a> {
22+ BodyRef::from_bytes(&self.message).trailers()
23+ }
24+}
25+
26+impl<'a> MessageRef<'a> {
27+ /// Parse the given `input` as message.
28+ ///
29+ /// Note that this cannot fail as everything will be interpreted as title if there is no body separator.
30+ pub fn from_bytes(input: &'a [u8]) -> Self {
31+ let (title, body) = decode::message(input);
32+ MessageRef { title, body }
33+ }
34+
35+ /// Produce a short commit summary for the message title.
36+ ///
37+ /// This means the following
38+ ///
39+ /// * Take the subject line which is delimited by two newlines (\n\n)
40+ /// * transform intermediate consecutive whitespace including \r into one space
41+ ///
42+ /// The resulting summary will have folded whitespace before a newline into spaces and stopped that process
43+ /// once two consecutive newlines are encountered.
44+ pub fn summary(&self) -> Cow<'a, BStr> {
45+ summary(self.title)
46+ }
47+
48+ /// Further parse the body into into non-trailer and trailers, which can be iterated from the returned [`BodyRef`].
49+ pub fn body(&self) -> Option<BodyRef<'a>> {
50+ self.body.map(|b| BodyRef::from_bytes(b))
51+ }
52+}
53+
54+pub(crate) fn summary(message: &BStr) -> Cow<'_, BStr> {
55+ let message = message.trim();
56+ match message.find_byte(b'\n') {
57+ Some(mut pos) => {
58+ let mut out = BString::default();
59+ let mut previous_pos = None;
60+ loop {
61+ if let Some(previous_pos) = previous_pos {
62+ if previous_pos + 1 == pos {
63+ let len_after_trim = out.trim_end().len();
64+ out.resize(len_after_trim, 0);
65+ break out.into();
66+ }
67+ }
68+ let message_to_newline = &message[previous_pos.map(|p| p + 1).unwrap_or(0)..pos];
69+
70+ if let Some(pos_before_whitespace) = message_to_newline.rfind_not_byteset(b"\t\n\x0C\r ") {
71+ out.extend_from_slice(&message_to_newline[..pos_before_whitespace + 1]);
72+ }
73+ out.push_byte(b' ');
74+ previous_pos = Some(pos);
75+ match message.get(pos + 1..).and_then(|i| i.find_byte(b'\n')) {
76+ Some(next_nl_pos) => pos += next_nl_pos + 1,
77+ None => {
78+ if let Some(slice) = message.get((pos + 1)..) {
79+ out.extend_from_slice(slice);
80+ }
81+ break out.into();
82+ }
83+ }
84+ }
85+ }
86+ None => message.as_bstr().into(),
87+ }
88+}
89+
90+/// A reference to a message body, further parsed to only contain the non-trailer parts.
91+///
92+/// See [git-interpret-trailers](https://git-scm.com/docs/git-interpret-trailers) for more information
93+/// on what constitutes trailers and not that this implementation is only good for typical sign-off footer or key-value parsing.
94+///
95+/// Note that we only parse trailers from the bottom of the body.
96+#[derive(PartialEq, Eq, Debug, Hash, Ord, PartialOrd, Clone, Copy)]
97+pub struct BodyRef<'a> {
98+ body_without_trailer: &'a BStr,
99+ start_of_trailer: &'a [u8],
100+}
101+
102+impl<'a> BodyRef<'a> {
103+ /// Return an iterator over the trailers parsed from the last paragraph of the body. May be empty.
104+ pub fn trailers(&self) -> body::Trailers<'a> {
105+ body::Trailers {
106+ cursor: self.start_of_trailer,
107+ }
108+ }
109+}
110+
111+impl<'a> AsRef<BStr> for BodyRef<'a> {
112+ fn as_ref(&self) -> &BStr {
113+ self.body_without_trailer
114+ }
115+}
116+
117+impl<'a> Deref for BodyRef<'a> {
118+ type Target = BStr;
119+
120+ fn deref(&self) -> &Self::Target {
121+ self.body_without_trailer
122+ }
123+}